担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

担当区分：最終著者,　責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

担当区分：筆頭著者,　最終著者,　責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

Liquid biopsy allows non-invasive collection of circulating tumor cells (CTCs) in blood without the need for sampling and can clearly demonstrate their presence in many types of cancer. In this study, we propose a method to automatically identify CTCs from fluorescence microscopy images and enable quantitative analysis based on convolutional neural networks (CNN). In this paper, a cell nucleus region cropping algorithm is applied in addition to a filtering process centered on a selective enhancement filter. Next, identification by SqueezeNet is performed. We performed the proposed method to 5,040 images of 6 samples and conducted experiments to identify CTCs. The number of detected CTCs was 148 (TPR = 100%), and the number of over-detected non-CTCs was 925. For the identification, TPR = 88.51% and FPR = 5.102% for the CNN model using SqueezeNet. The proposed method successfully reduced the number of detections by about 71.4% without missing any correct answers, but the proposed method did not show good results in any of the evaluation metrics.

DOI： 10.24507/ijicic.19.01.1

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85148695370&origin=inward

担当区分：筆頭著者,　責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Lung cancer progresses rapidly, and early detection and treatment are important. Computed tomography (CT) equipment is mainly used for the examination. However, the number of CT images is huge and places a heavy burden on physicians. Therefore, a computer aided diagnosis (CAD) system is necessary to overcome these problem. One of the CAD is temporal subtraction technique. The technique is that performs a subtraction operation using the current and past images of the same patient to remove normal structures and emphasize newly appearing areas that have changed over time. However, because two images taken at different dates are used, artifacts due to misalignment often occur. Therefore, it is necessary to classify partial images with high false-positive rates and enhanced temporal changes into lesions and normal tissue. In this paper, the temporal subtraction image generation method is used to extract candidate areas of abnormal shadows as initial shadows. The proposed method was applied to 22 cases of chest CT images, and the results showed that the proposed method achieved a discrimination accuracy of Accuracy=74.54, TPR=72.65, FPR=22.59, and AUC=79.16 in XGBoost, which is a kind of machine learning method.

DOI： 10.23919/ICCAS59377.2023.10316767

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85179180349&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

When lung cancer is suspected, a CT scan of the chest is widely used as a method of precise examination. However, the number of CT images obtained in a single examination is enormous, placing a heavy burden on the physician who reads the images. In addition, there is concern about the possibility of undetected lesions due to differences in the skills and experience of the reader. Therefore, computer-aided diagnosis systems have been introduced to reduce their workload and undetected lesions. When physicians make a diagnosis, they consider not only CT images but also information about the patient. Therefore, attempts are being made to improve the accuracy of diagnosis by mimicking this process using artificial intelligence. In this paper, we propose a model for identifying nodular shadows using deep learning, aiming to improve the accuracy of diagnosis by introducing medical record information in addition to image information. Normal tissue is divided into three classes: branched vessels, thin vessels, and round vessels, and a total of four classes are classified, including abnormal tissue. Experimental results show that the accuracy of nodular shade discrimination is improved when medical record information is added.

DOI： 10.23919/ICCAS59377.2023.10316787

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85179182295&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Root resorption is a disease in which the cells of the root of the tooth are destroyed. This disease may be harmless, depending on the cause. When the root resorption has progressed, early detection and treatment is necessary. If root resorption is left untreated, the lesion can progress to pulp exposure, necrosis, sinus and periodontal defects, and perforation. Diagnosis of root resorption is made by clinical examination and a panoramic x-ray machine, however accurate diagnosis requires computed tomography (CT). CT allows full-plane examination of the area of interest and provides an accurate diagnosis. However, CT is less widely used than panoramic radiography in private dental office. Therefore, there is a need for a method to detect root resorption from panoramic radiographs with a high reliability comparable to that of CT images. In this paper, we propose an image analysis method to identify the presence or absence of root resorption from panoramic radiographs. In particular, we propose a method that uses EfficientNet to extract features with fine-tuning and learns them using L2-softmax loss as an error function. We applied the proposed method to panoramic X-ray images of 59 actual cases and obtained TPR: 71% and FPR: 43% respectively.

DOI： 10.23919/ICCAS59377.2023.10316940

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85179183951&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

In this paper, we focus on the manpower shortage in factories and conduct an experiment to try to automate the process. Factory automation has many advantages, such as reducing labor costs and improving production efficiency. Among them, we focused on the sorting of plastic bottles at a waste disposal plant. Currently, garbage is sent to the landfill without being sorted, and it is sorted by hand. Unlike cans, plastic bottles cannot be easily automated because they cannot be separated using magnets. Therefore, we propose an image processing technique to recognize where plastic bottles are located, and to use a robot arm to sort them, thereby realizing automation. In this paper, we focus on image processing technique based on deep learning. As basic research, we conducted an experiment to see how well the robot can identify a single plastic bottle in an image. We attempted to use semantic segmentation methods for detection, using DeepLabv3+ as the basic model, and improved it. scSE Block and fine-tuning were introduced to improve the accuracy significantly compared to previous studies.

DOI： 10.23919/ICCAS59377.2023.10316854

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85179179143&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Environmental change detection is one of the uses of satellite images. This process is performed by subtracting image pairs obtained by different time series or sensors. Therefore, image registration is an important pre-processing step in detection of environmental changes. Currently, image registration methods based on deep learning are gaining attention. In general, higher satellite image resolution results in more accurate registration. However, the increase in image size leads to higher computational costs during training and estimation of deep learning models. Then, we propose a method that reduces the number of parameters of the model to lower the computational cost while maintaining the accuracy. This method makes it easier to handle high-resolution satellite images. The proposed method modified the GMCNN (Geometric-matching CNN) architecture by adding CSA (Cosine Similarity Attention) and SE (Squeeze-and-Excitation) layers to enhance the feature map, and point-wise convolution to reduce the number of parameters. The improved GMCNN decreases the grid MSE by 0.0037 compared to the conventional GMCNN. It also reduces the number of parameters by 49.6%.

DOI： 10.23919/ICCAS59377.2023.10316818

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85179178272&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Cleft lip is one of the most common birth defects. Several operations are performed to form a natural lip. A problem with these operations is that the criteria for the facial symmetry are unclear. Based on this background, we propose a method for evaluating the facial symmetry using 4D point cloud data. We evaluate the facial symmetry in two ways. One is based on the temporal changes in the face landmarks. Corresponding points are searched using two local geometric descriptors. The results are approximated to the 3D lines and these slopes are compared to obtain the left-right difference of the movement. The other is based on the center of gravity. The lip part is extracted from the point cloud and divided into subregions. The centroid coordinates are then calculated for each subregion and compared to obtain the left-right difference of the facial structure. The experiment is performed using artificially generated 4D data. As an experimental result, it is shown that our method can find point correspondences with smaller error than comparative methods.

DOI： 10.23919/ICCAS59377.2023.10316967

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85179178953&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

6D pose estimation of arbitrary objects is a crucial topic for intelligent transportation infrastructure measurement. However, some external environmental factors and the characteristics of the object itself impact the accuracy of the object’s pose estimation in practical applications. In this paper, we propose a new multi-class dataset ICD-4 (Industrial car Components Dataset) for 6D object pose estimation, which mainly includes four component categories, and every category takes 20,000 different scenarios. ICD-4 dataset delivers quite a few research challenges involving the range of object pose transformations and has significant research value for small-scale pose estimation tasks. We also propose an innovative method PoseMLP, a pose estimation network that uses residual MLP (multilayer perceptron) modules to predict the 6D pose estimation directly. Simultaneously, the experimental results demonstrate the effectiveness and reliability of the proposed method.

DOI： 10.1109/TITS.2023.3250604

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85153361050&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Deep learning-based methods have taken center stage in image recognition, such as AlexNet and deep learning-based method. At present, Image recognition based on deep learning has been widely used in agriculture, factory automation, automated driving, medical fields and so on. In the fields of automated driving and medical care, the accuracy of the image recognition directly affects human lives. For these reasons, the importance of improving the accuracy of image recognition is clear. In this paper, we focus on instance segmentation tasks. The method used is Mask R-CNN, which is the basis of current state-of-the-art methods. The network structure based on ResNet, and we tried to improve the accuracy by adding Squeeze-and-Excitation Block (SE Block). According to the result of experiments, it is proved that this method has certain advantages for object instance segmentation.

DOI： 10.1007/978-981-99-2789-0_5

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85163323022&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

In Japan, increasing amounts of waste are becoming a social problem. One of approaches to solve the problem is recycle of the plastic bottles. However, they are thrown away with their caps still attached, and it should be removed by hand. To solve this problem, we developed a method for automatic identification of plastic bottles and caps using deep learning technique. In this paper, we propose a method that combines different numbers of Efficient blocks and adds an attention structure and verify its usefulness through experiments.

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85149463467&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85177460604&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Japan has the highest number of CT scanners per million people among developed countries, and the annual medical radiation dose is about six times higher than the world average. Screening of high-risk populations by means of low-dose CT has been verified to be able to reduce lung cancer mortality. However, the low-dose CT image noise should be accurately estimated in order to achieve good image quality as normal dose CT. In this paper, we apply a denoising method using convolutional neural networks without use of the clean images during learning phase. The CNN model is based on SRResNet, which has achieved high performance in super-resolution tasks, and has two attention mechanisms, Spatial Attention and Channel Attention, in the Residual Block. Experiments using whole body slice CT images of pigs showed that the proposed method is useful by comparison with normal dose CT images and by evaluation of image quality using peak signal-to-noise ratio (PSNR).

DOI： 10.23919/ICCAS59377.2023.10317004

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85179178164&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Radiation is widely used in medicine to diagnose and treat disease. CT (Computed Tomography) scans allow early detection of externally invisible diseases and appropriate treatment. However, radiation exposure from the examination may result in a future risk of cancer. Efforts are therefore being made to reduce radiation exposure. During the examination, noise is generated in the image when the dose is reduced. Noise reduces the visibility of the image and may cause lesions to be missed. In this study, we focus on Convolutional Neural Networks (CNNs), a type of deep learning model that has recorded high accuracy in image processing tasks. The proportion of frequency components in an image has more low-frequency components and fewer high-frequency components. Since image features are treated equally across channels, important information such as noise and edges are easily lost. To solve this problem, we propose CNN with channel attention module. In addition, we employ MAE as the loss function to enable effective learning. Using whole body slice CT images of pigs, we evaluate the image quality by Peak Signal-to-Noise Ratio (PSNR) and show that the proposed method is effective.

DOI： 10.23919/ICCAS59377.2023.10317050

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85179178606&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

In this paper, we focus on brain atrophy in Alzheimer's disease and propose a 3D image registration method of brain MR images for the purpose of calculating the atrophy rate by temporal subtraction technique. The nonlinear 3D image registration model of Stergios et al. is recombined into a linear image registration model, and image registration is performed with isotropic affine and rigid body transformations. In addition, the accuracy is improved by the Instance-specific Optimization, which performs the optimization process for each case. For the training of the model, we used pseudo-data obtained by arbitrarily linear transformation of the ADNI brain MR images. Synthetic data and actual ADNI time series data (AD, MCI) were used for the test. The experimental results show that the addition of the optimization process improves the normalized cross-correlation (NCC) by more than 0.1. The loss function of the model is compared between Grid loss and MSE, and it is confirmed that MSE is effective for NCC. Furthermore, the output images confirm the usefulness of the image registration of the skull in the linear transformation.

DOI： 10.23919/ICCAS59377.2023.10316755

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85179178497&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

In Japan, it is serious problems such as increasing nursing care of handicapped people requiring and the aging of caregivers which is caused in a super-aging society. One of the main causes of being classified as persons requiring nursing care include joint diseases such as rheumatoid arthritis and bone fractures due to osteoporosis. Early detection and treatment of these diseases are considered important. Rheumatoid arthritis and osteoporosis are generally diagnosed by simple X-ray examination. However, there are problems with radiographic diagnosis by physicians, such as lack of objectivity and reproducibility of diagnosis, and increased workload on the radiologists. To solve these problems, a Computer-Aided Diagnosis (CAD) system is being developed. Because the CAD system may use the results of quantitative computer analysis, it is expected to improve the reproducibility and accuracy of diagnosis and reduce the burden on physicians. Therefore, this paper proposes a segmentation method of phalanx region from CR images to develop a CAD system. The proposed method is based on HRNet + JPU + U-Net. The proposed method was applied to 101 cases of X-ray images, and mIoU=0.897 was obtained. Experiments for segmentation from images confirmed the usefulness of the proposed method by improving the extraction accuracy of the boundary of the phalanx region.

DOI： 10.23919/ICCAS59377.2023.10317032

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85179183712&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Cancer is a leading cause of death worldwide, accounting for nearly 10 million deaths in 2020. Lung cancer is the most common cause of death from this cancer. Diagnosis is made mainly by biopsy, which checks for mutations in the driver genes of lung cancer. If there is a mutation, molecularly targeted drugs with higher therapeutic efficacy can be used. However, it is difficult for physicians to make a decision and places a heavy burden on patients. To solve this problem, a computer aided diagnosis (CAD) system is needed to identify the presence or absence of driver gene mutations from CT images. In this paper, Radiomics is used to extract features from 3D lung cancer in CT images and analyze them using machine learning to classify the presence or absence of mutations. Because the features obtained is huge, dimensionality reduction is performed using Null Importance. Furthermore, the accuracy is improved by adding gender as clinical information. The proposed method was applied to a dataset consisting of 175 cases. As a result, we obtained an AUC (Area Under the Curve) of 0.985, accuracy of 92.0%, true positive rate of 85.7%, and false positive rate of 2.20%.

DOI： 10.23919/ICCAS59377.2023.10316810

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85179182718&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Identification of primary lung cancer is very important because it influences the course of treatment, especially for small cell carcinomas, which metastasize rapidly and must be detected at an early stage. In addition to imaging, clinical information is often used in CAD (computer aided diagnosis) systems. In addition to images, clinical information is often used in CAD systems, especially information on smoking history, which is considered to be important in the diagnosis of lung cancer. In this paper, we propose a method to identify primary lung cancer by adding clinical information from medical records in addition to images in order to improve the accuracy of diagnosis. We use tumor images surrounded by rectangular regions from CT images in an open dataset as input images and train the method by deep learning. We evaluate the proposed method by discriminating tumors from unknown data. In our experiments, we found that the accuracy was improved by about 5% when clinical information was added to 655 images, which included four classes of cancer: adenocarcinoma, small cell carcinoma, squamous cell carcinoma, and large cell carcinoma.

DOI： 10.23919/ICCAS59377.2023.10316865

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85179178408&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Lung cancer currently kills the largest number of people in the world, with lung cancer accounting for the highest number of deaths among men and women. Therefore, early detection and treatment are important issues to reduce the number of lung cancer deaths. Genetic testing can confirm the presence or absence of driver gene mutations involved in cancer cell growth and other factors. In the process of identifying the presence or absence of driver gene mutations, lung cancer regions are extracted manually in cooperation with a radiologist. One of the concerns is that manual lung cancer extraction places a heavy burden on the physician who reads the images. Therefore, we propose an automatic extraction of lung tumor regions from chest CT images using deep learning, with the goal of developing a CAD system that reduces the workload of the physician and prevents lesions from being overlooked. We constructed a new model based on U-Net with the addition of CBAM and MultiRes Block. Experimental results on the model using CT images of the chest showed a 4.2% improvement in accuracy for Dice and a 4.6% improvement for IoU compared to the original U-Net model.

DOI： 10.23919/ICCAS59377.2023.10316783

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85179183296&origin=inward

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

China  

DOI： https://doi.org/10.1117/12.2663251

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

DOI： https://doi.org/10.1117/12.2655680

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

DOI： https://doi.org/10.1117/12.2655052

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

China  

DOI： https://doi.org/10.1117/12.2655051

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Respiratory diseases are one of the leading causes of death worldwide. Approximately 8 million people die annually from respiratory diseases. Diagnosis is made primarily by auscultation using a stethoscope. The lack of quantitative criteria makes diagnosis difficult in the field where physicians are in short supply. To solve this problem, a computer aided diagnosis (CAD) system that quantitatively analyzes and classifies respiratory sounds and outputs them as a 'second opinion' is needed. In this paper, HPSS (Harmonious/Percussive Sound Separation) is used to separate abnormal respiratory sound features. Images are generated from the spectral envelopes obtained by linear prediction coefficients (LPC) for each of the three types of respiratory sound data before separation. The CNN (convolutional neural networks) framework based on hierarchical structure of the correct labels is introduced. The proposed method was applied to the dataset used in the International Conference on Biomedical and Health Informatics (ICBHI) 2017 Challenge. As a result, we obtained a sensitivity of 63.5%, specificity of 85.1%, average score of 74.3%, harmonic score of 72.7%, area under the curve of 87.8%, and false negative rate of24.5%, respectively.

DOI： 10.23919/ICCAS55662.2022.10003771

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146537757&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Lung cancer is one of the most common cancers worldwide and has become a general medical problem. To lessen the risk of death, early detection and treatment is particularly required. The patients can use molecularly targeted drugs when the driver gene mutations of the cancer are detected, but invasive biopsies are required. So, development of new methods to detect it noninvasively and in a short time are expected. we propose a new machine learning method for identifying the presence or absence of driver gene mutations of lung cancer on Thoracic CT Images that is a non-invasive, in a short time, and low-cost CAD (Computer Aided Diagnosis) system. In the proposed method, radiomics features are given as explanatory variables in addition to Thoracic CT Images, and supervised learning using LightGBM is performed to conduct binary classification with/without driver gene mutations.

DOI： 10.23919/ICCAS55662.2022.10003871

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146575218&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

We are developing a computer-aided diagnosis system for rheumatoid arthritis. X-rays images are widely used to diagnose the rheumatoid arthritis. However, it is difficult for physicians to read minute changes from the images. Therefore, we propose a method to visualize lesions in the phalangeal region by comparing past and current images of the same subject using a temporal subtraction technique. The proposed method consists of three steps: segmentation of phalanges, registration, and generation of subtraction images. First, the phalangeal region is extracted from the hand CR image using DeepLabv3+. Next, the past and current phalangeal region images are aligned by geometric-matching based on a CNN (convolutional neural networks) with instance-specific optimization. Finally, we apply the temporal subtraction technique to those images. We confirmed the effectiveness of the proposed registration method in an experiment using synthetic data. Also, the proposed method was applied to a pair of past and current image sets on same subject to generate a subtraction image. As a result, we confirmed that the proposed method can visualize changes between past and current images.

DOI： 10.23919/ICCAS55662.2022.10003828

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146633083&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Electric wheelchair is an easy-to-operate means of transportation that does not require physical strength. With the number of electric wheelchair users increasing in recent years, the increase in traffic accidents becomes a problem. Therefore, by developing an autonomous electric wheelchair, it is expected that the risk of accidents will be reduced and the convenience of the electric wheelchair will be improved. Environment recognition is indispensable for the development of autonomous electric wheelchairs. We propose a semantic segmentation method for recognizing 16 objects in traffic environment. This paper examines the improvement of problems such as the high price of autonomous electric wheelchairs due to the increase in the number of sensors used, which has been a concern in related research. Therefore, we use panoramic images acquired by a spherical camera as input data, and extern the Multi-Attention Deep Lab algorithms fitting for the recognition of distorted images. A new CNN model is constructed using Deep Lab v3+, scSE Block, Pairwise Self-Attention, and Joint Pyramid Up-sampling. We conducted a recognition experiment using images taken on campus and verified its effectiveness. (Comparing to DeepLab v3+, IoU and Dice showed a 3.5% and 3.6% improvement in accuracy, respectively.)

DOI： 10.23919/ICCAS55662.2022.10003689

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146621100&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Early detection and treatment of the lymph node are important since swelling of the neck is a likely factor in systemic metastasis of cancer. One of the diagnoses of cervical swelling is a CT scan, which has a beneficial influence on the diagnosis of the disease. However, the reading of CT images is burdened by the large number of images, which increases the physician's workload. In addition, since it is based on the subjective judgment of the physician, there may be discrepancies in diagnostic results and undetected cases due to differences in experience. A means of solving these problems requires a CAD system that provides a second opinion to the physician. Therefore, this paper proposes a segmentation method of cervical lymph node region for the purpose of developing a CAD system for the diagnosis of cervical lymphadenopathy from CT images. The proposal method is a CNN model with U-Net++ as the backbone, introducing CBAM (convolution block attention module) and dual-branch multi-scale attention module. The proposed method was applied to CT images of 11 cases, yielding IoU of 62.29, confirming its validity.

DOI： 10.23919/ICCAS55662.2022.10003916

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146553025&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Satellite images can be analyzed and used for a variety of purposes. In the future, satellite image analysis will become more important since the number of satellites launches, and the amount of satellite data increase every year. Under these circumstances, there are some problems to be solved. One is the existence of low-resolution satellite images. To analyze the lower resolution of satellite images there are some technical issues such as reduction of noise, misclassification of object recognition. Therefore, high-resolution images are necessary. However, high-resolution satellite images are expensive, and its images may not be available in the past satellite images. Super-resolution which is introduced in image processing is a method to solve these problems. Convolutional neural network (CNN)-based methods are effective, and there is a need for models that can perform super-resolution with higher accuracy. In this paper, we propose a method for super-resolving satellite images, based on the improved the RCAN (residual channel attention network) model with SRM (style-based recalibration module). The proposed method improves the PSNR (peak signal to noise ratio) by 0.0181 dB compared to the conventional RCAN model.

DOI： 10.23919/ICCAS55662.2022.10003856

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146557504&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Currently, the increase of general waste is a problem in Japan. It is garbage and human waste. One solution to this problem is to reuse the bodies of used plastic bottles to reduce the amount of plastic waste and the increase in general waste. For this purpose, some municipalities provide bags for disposing of plastic bottle bodies. However, in many cases, the bottles are not only discarded in these bags, but also with their caps attached. In such cases, the caps must be found and removed by hand in the sorting and processing facility at the waste treatment plant. This means that the same work is performed by person for a long time, causing problems such as overlooking caps due to fatigue. To solve these problems, we develop a method for automatic identification of plastic bottle bodies and caps using deep learning technique. In this paper, we propose a model that combines multiple Eff-UNets. Specifically, we combine EfficientNetB4 for local segmentation and EfficientNetB5 for global segmentation. By using our method, we conducted an experiment on images of plastic bottles collected from the internet and other sources. We obtained segmentation results of 97.9% for plastic bottle bodies and 86.0% for plastic bottle caps.

DOI： 10.23919/ICCAS55662.2022.10003920

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146540644&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

In recent years, CT examinations have been widely used as a screening method to detect lung cancer. However, reading enormous CT images become a heavy burden to the physician. To avoid this problem, computer-aided diagnosis systems have been introduced on CT screening. In general, physicians consider patient information in addition to image information when they make a diagnosis, new efforts are being made to improve the accuracy of diagnosis by mimicking this information with a machine. In this paper, we propose a method for identifying pulmonary nodules by adding medical record information to images to improve the accuracy of diagnosis. We classify nodules from unknown data by assigning branching information of vascular opacities, straight vascular shadows, and nodular shadows as labeled image, which are a cause of misrecognition based on image features in machine learning. In the experiment, the classification accuracy of the nodule class was improved by adding clinical information to 644 images including 161 nodal images.

DOI： 10.23919/ICCAS55662.2022.10003937

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146569950&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

This paper focuses on separation rubbish, which is one of the causes of global warming. Currently, recyclable waste including garbage is sent to the waste disposal site in an unsorted state, and it is separated by human source. This is difficult work and a cause of manpower shortage. To overcome those problem, it is necessary to introduce automatic sorting algorithms on separation rubbish process. In this study, we focused on separation rubbish process especially plastic bottles based on image recognition technique using a deep learning approach. We implement an improved U-Net as the deep learning scheme to increase segmentation accuracy. As a result, we obtained the segmentation accuracy with 0.789 of cap part and 0.972 of body part on plastic bottle respectively.

DOI： 10.23919/ICCAS55662.2022.10003819

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146598114&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Korea  

DOI： 10.23919/ICCAS55662.2022.10003777

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Computed Tomography (CT) scans are widely used in Japan, and they contribute to public health. On the other hand, there is also a risk of radiation exposure. To solve this problem, attempts are being made to reduce the radiation dose during imaging. However, reducing the radiation dose causes noise and degrades image quality. In this paper, we propose an image analysis method that efficiently removes noise by changing the activation function of Deep Convolutional Neural Network (Deep CNN). Experimental tests using full-body slice CT images of pigs and phantom CT images of lungs with Poisson noise show that the proposed method is helpful by comparing them with normal-dose CT images and evaluating image quality using peak signal-to-noise ratio (PSNR).

DOI： 10.23919/ICCAS55662.2022.10003847

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146610706&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

These days, cancer has been the most primary cause of death in Japan. Cancer often progresses by repeating metastasis, so early detection and early treatment are important. Analysis of Circulating Tumor Cells (CTCs) has come to gather attention as a new biomarker that CTCs can detect primary cancer in human body. However, the number of CTCs in a billion blood cells is only a few, and detecting CTCs is very hard. Accordingly, we propose an automatic detection method of CTCs from fluorescence microscopy images to enable quantitative analysis by computer. This method consists of two parts. The first part, we use some series of filtering to the images and, new dividing method some overlapping nucleus then, from the images cut out the region of interest (ROI). The second part is distinguishing images by using CNN. We applied the proposed method to 5040 images of 6 samples. As a result, we obtained TPR:94.59%, FPR:6.544% by using AlexNet based model.

DOI： 10.23919/ICCAS55662.2022.10003905

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146616673&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Root resorption is a pathological condition which is characterized by the loss of the tooth root. Root resorption is not painful in its early stages. As a result, many people who are potentially affected and the condition are often left untreated until it is detected during regular check-ups. If detected early, good treatment results can be achieved, whereas failure to treat the condition properly can lead to tooth extraction. However, the root resorption is currently difficult to detect on panoramic radiographs and may be treated as caries after it becomes painful. The aim of this paper is to identify root resorption from panoramic X-ray images using a deep metric learning algorithm. As a loss function for distance learning, it is known that the loss function in angle space is consistent. Therefore, a loss function is defined and trained using the cosine value of the angle between the feature and the center position to improve the discrimination performance. We obtained experimental results based on 150 image sets with 0.80 of accuracy, 0.62 of TPR, 0.19 of FPR and 0.78 of AUC, respectively.

DOI： 10.23919/ICCAS55662.2022.10003752

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146551688&origin=inward

担当区分：最終著者,　責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：最終著者,　責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

The temporal subtraction technique is a useful tool for computer aided diagnosis (CAD) in visual screening. The technique subtracts the previous image set from the current one for the same subject to emphasize temporal changes and/or new abnormalities. However, it is difficult to obtain a clear subtraction image without subtraction image artifacts. VoxelMorph in deep learning is a useful method, as preparing large training datasets is difficult for medical image analysis, and the possibilities of incorrect learning, gradient loss, and overlearning are concerns. To overcome this problem, we propose a new method for generating temporal subtraction images of thoracic multi-detector row computed tomography (MDCT) images based on ResidualVoxelMorph, which introduces a residual block to VoxelMorph to enable flexible positioning at a low computational cost. Its high learning efficiency can be expected even with a limited training set by introducing residual blocks to VoxelMorph. We performed our method on 84 clinical images and evaluated based on three-fold cross-validation. The results showed that the proposed method reduced subtraction image artifacts on root mean square error (RMSE) by 11.3% (p < 0.01), and its effectiveness was verified. That is, the proposed temporal subtraction method for thoracic MDCT improves the observer’s performance.

DOI： 10.3390/app12178542

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85137919955&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

Landmark correspondence is one of the key steps in Statistical Shape Model (SSM) building. In this paper, a non-rigid iterative closest point surfaces registration method is introduced to seek proper corresponded landmarks in the multi-organ surface meshes. Surfaces of four abdominal organs are used in the experiment to build SSM from landmarks corresponded by five different registration strategies. The proposed method of individually non-rigid registration of single organs shows the least errors measured by Hausdorff distance and the best model quality of generalization ability, specificity, and compactness.

DOI： 10.18178/joig.10.3.95-101

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85138705827&origin=inward

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：最終著者,　責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

日本  

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

日本  

記述言語：英語   掲載種別：研究論文（学術雑誌）

In computer-aided diagnosis systems for lung cancer, segmentation of lung nodules is important for analyzing image features of lung nodules on computed tomography (CT) images and distinguishing malignant nodules from benign ones. However, it is difficult to accurately and robustly segment lung nodules attached to the chest wall or with ground-glass opacities using conventional image processing methods. Therefore, this study aimed to develop a method for robust and accurate three-dimensional (3D) segmentation of lung nodule regions using deep learning. In this study, a nested 3D fully connected convolutional network with residual unit structures was proposed, and designed a new loss function. Compared with annotated images obtained under the guidance of a radiologist, the Dice similarity coefficient (DS) and intersection over union (IoU) were 0.845 ± 0.008 and 0.738 ± 0.011, respectively, for 332 lung nodules (lung adenocarcinoma) obtained from 332 patients. On the other hand, for 3D U-Net and 3D SegNet, the DS was 0.822 ± 0.009 and 0.786 ± 0.011, respectively, and the IoU was 0.711 ± 0.011 and 0.660 ± 0.012, respectively. These results indicate that the proposed method is significantly superior to well-known deep learning models. Moreover, we compared the results obtained from the proposed method with those obtained from conventional image processing methods, watersheds, and graph cuts. The DS and IoU results for the watershed method were 0.628 ± 0.027 and 0.494 ± 0.025, respectively, and those for the graph cut method were 0.566 ± 0.025 and 0.414 ± 0.021, respectively. These results indicate that the proposed method is significantly superior to conventional image processing methods. The proposed method may be useful for accurate and robust segmentation of lung nodules to assist radiologists in the diagnosis of lung nodules such as lung adenocarcinoma on CT images.

DOI： 10.3389/frai.2022.782225

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85125736621&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

Bone disease including rheumatoid arthritis and osteoporosis is one of se-rious diseases especially elderly people. Diagnostic imaging using computed radiography (CR) imaging is an effective method for diagnosing the bone disease. However, accurate diagnosis should be made within a limited time, increasing the burden on doctors. There-fore, the goal of this paper is to develop a computer aided diagnosis (CAD) system for automatically diagnosing bone disease from CR images. The proposed method based on an improved U-Net and multiscale gradient vector flow snakes (MSGVF Snakes) extracts each phalange region from the CR image and creates an image for each phalange. In the experiment, the proposed method was applied to CR images of 101 cases, and the performance was evaluated. As a result, our proposed model based on improved U-Net for the segmentation obtained intersection over union (IoU) of 0.939, true positive (TP) of 0.984, and false positive (FP) of 0.068. The initial segmentation accuracy was higher than the conventional approach such as U-Net, Residual U-Net. Final segmentation performance of the phalanges based on MSGVF Snakes are TP of 0.945 and FP of 0.030.

DOI： 10.24507/icicelb.13.02.155

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85123095886&origin=inward

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：筆頭著者,　最終著者,　責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

The multiphase scan has enabled an improved detection of liver tumors. However, tumor regions and peripheral tissues are difficult to distinguish and delineate owing to their highly similar image features. Moreover, their characteristics vary significantly in different phases. This is challenging when using segmentation methods that are based on unique training models. Herein, a hybrid framework is proposed for liver tumor segmentation in multiphase images. We first develop a cascade region-based convolutional neural network with refined head to locate the tumors. Meanwhile, phase-sensitive noise filtering is introduced to refine the segmentation conducted by a level-set-based framework. This method is sensitive to the intensity contrast but not to the regions of interest, thereby affording better performance in delineating adjacent tumors. In our experiment, the average precision and recall rates are 76.8% and 84.4%, respectively. The intersection over union, true positive rate, and false positive rate are 72.7%, 76.2%, and 4.75%, respectively.

DOI： 10.1016/j.compeleceng.2021.107626

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85120725147&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

IoT technology has made remarkable progress in recent years, and the world is full of IoT devices that continue to evolve every day. From smartphones, personal computers, and smartwatches to home appliances such as refrigerators and washing machines, and even indoor lights and house keys, IoT devices have become an inseparable part of our lives. In addition to devices used by individuals, IoT technology supports our daily lives from both front and back sides, such as IoT-enabled industrial equipment and satellite positioning systems. Japan has been making a national push to shift to IoT in industries that reduce the burden on workers and have recently been promoting a plan called Smart Agriculture, Forestry, and Fisheries. Among these three types of industries, the agricultural sector is slightly ahead of the others, with the Smart Agriculture Demonstration Project starting in 2019, and 182 districts in Japan are implementing the project by FY2021. The forestry and fisheries industries are also developing daily to become next-generation industries based on the program established in December 2019, although they are behind agriculture. However, the examples mentioned so far are those that have been promoted with the help of companies and the national government, although everyone has benefited from them. In this paper, we propose an IoT device that can be used as an IoT buoy by using a microcomputer, Raspberry Pi, to create a camera device that can distribute underwater images in real-time and track its location.

DOI： 10.1117/12.2655052

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146731793&origin=inward

担当区分：最終著者,　責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

In recent years, the elderly population in Japan has been increasing. Expectations for the utilization of welfare equipment are also increasing. Electric wheelchairs are one of equipment and are widely used as a convenient means of transportation. On the other hand, accidents have also occurred, and dangers have been pointed out when driving the electric wheelchair. Therefore, we believe that the development of an autonomous mobile electric wheelchair can improve the causes of accidents. In addition, it can be expected to reduce accidents and improve the convenience of electric wheelchairs. For the development of an autonomous electric wheelchair, environment recognition such as estimation of the current position, recognition of sidewalks and traffic lights, and prediction of movement of objects is indispensable. To solve these problems, we develop an algorithm to recognize the sidewalks, crosswalks, and traffic lights from video images. In recent years, deep learning has been widely applied in the field of image recognition. Therefore, we improve WideSeg, one of the semantic segmentation algorithms that apply CNN (Convolutional Neural Networks), and develop an object recognition method using a new CNN model. In our approach, we perform adding the sidewalk correction and noise removal processing after performing semantic segmentation with the proposed model.

DOI： 10.1117/12.2655680

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146673279&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

Recently, two-dimensional object detection (2D object detection) has been introduced in numerous applications such as building exterior diagnosis, crime prevention and surveillance, and medical fields. However, the distance (depth) information is not enough for indoor robot navigation, robot grasping, autonomous running, and so on, with conventional object detection. Therefore, in order to improve the accuracy of 3D object detection, this paper proposes an improvement of Point RCNN, which is a segmentation-based method using RPNs and has performed well in 3D detection benchmarks on the KITTI dataset commonly used in recognition tasks for automatic driving. The proposed improvement is to improve the network in the first stage of generating 3D box candidates in order to solve the problem of frequent false positives. Specifically, we added a Squeeze and Excitation (SE) Block to the network of pointnet++ that performs feature extraction in the first stage and changed the activation function from ReLU to Mish. Experiments were conducted on the KITTI dataset, which is commonly used in research aimed at automated driving, and an accurate comparison was conducted using AP. The proposed method outperforms the conventional method by several percent on all three difficulty levels.

DOI： 10.1016/j.cogr.2022.12.001

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85144888388&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

In recent years, robots have been used not only in factories. However, most robots currently used in such places can only perform the actions programmed to perform in a predefined space. For robots to become widespread in the future, not only in factories, distribution warehouses, and other places but also in homes and other environments where robots receive complex commands and their surroundings are constantly being updated, it is necessary to make robots intelligent. Therefore, this study proposed a deep learning grasp position estimation model using depth images to achieve intelligence in pick-and-place. This study used only depth images as the training data to build the deep learning model. Some previous studies have used RGB images and depth images. However, in this study, we used only depth images as training data because we expect the inference to be based on the object's shape, independent of the color information of the object. By performing inference based on the target object's shape, the deep learning model is expected to minimize the need for re-training when the target object package changes in the production line since it is not dependent on the RGB image. In this study, we propose a deep learning model that focuses on the stacked encoder-decoder structure of the Stacked Hourglass Network. We compared the proposed method with the baseline method in the same evaluation metrics and a real robot, which shows higher accuracy than other methods in previous studies.

DOI： 10.1109/COMPSAC54236.2022.00187

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85137004247&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

The robot market in Japan is gradually expanding due to increasing demand. Industrial robots are being actively introduced in the manufacturing industry. The introduction of robots in the industry has three advantages: 1) securing labor, 2) increasing productivity, and 3) improving quality. The robot can run for a long time with constant work efficiency, thus achieving stable production. In addition, by replacing human labor, robots can reduce labor costs and reduce human error. The downside of introducing a robot is that the robot has to be told where to grab, which takes time, and a technician with specialized knowledge. Furthermore, this method cannot perform grasping when the grasping object is not in the specified position. However, the introduction of robot vision may solve these problems. In this study, by using depth images, processed images, and deep learning models, we aim to achieve object color-independent high-accuracy grasp position estimation for thin objects. We sharpen the depth image mainly by applying grayscale transformation and modify the deep learning model. The experimental results show that our design can achieve good results.

DOI： 10.1117/12.2663251

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146708731&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

In the process of transportation, the handling and loading methods of rigid objects are becoming more and more perfect. However, whether in today's transportation system or in daily life, such as packing objects or sorting cables before transportation, the manipulation of deformable objects has been always inevitable and has attracted more and more attention. Due to the super degrees of freedom and the unpredictable physical state of deformed objects. It is difficult for robots to complete tasks under the environment of the deformable object. Therefore, we present a method based on imitation learning. In the generated expert demonstration, the agent is offered to learn the state sequence, and then imitate the expert's trajectory sequence which avoid the above-mentioned difficulties. In addition, compared with the baseline method, our proposed DN-Transporter Networks are more competitive in a simulation environment involving cloth, ropes or bags.

DOI： 10.1109/TITS.2022.3168303

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85129430885&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

Objective: To quantitatively assess three-dimensional (3D) soft tissue facial asymmetry in patients with unilateral cleft lip and palate (UCLP) who have undergone primary lip repair. Design: Clinical, retrospective, comparative, methodological study. Patients/Participants: Twenty patients with UCLP were selected after a review of the records. Inclusion Criteria: Complete UCLP; surgically treated without secondary repair. An age-matched and sex-matched Control group was employed. Main Outcome Measures: A 3D facial symmetry plane (FSP) was obtained by superimposing the point clouds of the original 3D facial image excluding the surgical site and including lip and nose areas and those of a mirrored facial image using the iterative closest point (ICP) adjustment method. The discrepancies in the depth and angle of the normal vector of the facial surface of each point cloud between right and left sides (cleft and non-cleft sides in the UCLP group, respectively) based on FSP were calculated. Results: Facial asymmetry in the UCLP group was significantly greater than in the Control group regarding both the discrepancies in the depth (1.34 ± 0.62, 0.73 ± 0.32 pixels, respectively) (P =.0004) and surface angle (18.0 ± 5.88, 12.8 ± 4.0°, respectively) (P =.0024). Biaxial assessment of the discrepancies in the depth and surface angle allowed us to visually extract UCLP patients with greater facial asymmetry. Conclusions: Facial asymmetry analysis based on 3D FSP effectively facilitates the facial asymmetry quantification and soft tissue surgical outcome evaluation in patients with UCLP.

DOI： 10.1177/10556656221123276

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85138316191&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

In recent years, industrialization and economic development in countries around the world have led to an ever-increasing demand for energy. Renewable energies are attracting attention, but they still often use mineral resources such as coal, petroleum, and natural gas, and onshore resources are depleting day by day. These energy and metal resources, such as copper, support Japan's industries and affluent lifestyle, and if Japan continues to rely on imports for most of these resources, it will become difficult for Japan to secure a stable supply of these energies and resources. Therefore, mining of mineral resources on the seafloor is essential to solve these problems, and research on seafloor resource surveys and mining is underway. Because direct human exploration and mining of seafloor resources are naturally dangerous, underwater robots are used to explore and mine seafloor resources. However, due to light absorption and turbidity in water, the underwater image of an underwater robot is sometimes less visible, making exploration unsatisfactory. Therefore, there is a need for higher-resolution underwater images of underwater robots. In this study, we perform super-resolution of underwater images using an improved SRCNN to support research on underwater images of underwater robots. The conventional SRCNN method uses the ReLU function as the activation function, but the improved SRCNN uses the PReLU function and FReLU function, which are extended activation functions of the ReLU function, to improve accuracy.

DOI： 10.1117/12.2655051

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85146731084&origin=inward

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

記述言語：英語   掲載種別：研究論文（学術雑誌）

The rapid increase of patients with coronavirus disease 2019 (COVID-19) has introduced major challenges to healthcare services worldwide. Therefore, fast and accurate clinical assessment of COVID-19 progression and mortality is vital for the management of COVID-19 patients. We developed an automated image-based survival prediction model, called U-survival, which combines deep learning of chest CT images with the established survival analysis methodology of an elastic-net Cox survival model. In an evaluation of 383 COVID-19 positive patients from two hospitals, the prognostic bootstrap prediction performance of U-survival was significantly higher (P < 0.0001) than those of existing laboratory and image-based reference predictors both for COVID-19 progression (maximum concordance index: 91.6% [95% confidence interval 91.5, 91.7]) and for mortality (88.7% [88.6, 88.9]), and the separation between the Kaplan–Meier survival curves of patients stratified into low- and high-risk groups was largest for U-survival (P < 3 × 10–14). The results indicate that U-survival can be used to provide automated and objective prognostic predictions for the management of COVID-19 patients.

DOI： 10.1038/s41598-021-88591-z

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85105193010&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

In recent years, the rate of deaths from cancer tends to increase in Japan. Especially the number of deaths from lung cancer is increasing. Computed tomography (CT) device is effective screening tool for early detection of lung cancer. Simultaneously, there is concern that an increase in burden on doctors will be caused by high performance of CT improving. Therefore, by presenting the "second opinion"by the computer aided diagnosis (CAD) system, it reduces the burden on the doctor. In this paper, we developed a CAD system for automatic detection of abnormal opacities such as lung nodules or ground glass opacity (GGO) on CT images from temporal subtraction image by incorporating a sparse coding technique to convolutional neural network (CNN). We extend the sparse coding and CNN to 3D and applied them to 32 cases which is obtained different time series on same subject. As a result, we obtained true positive rate (TPR) of 80.556%, false positive rate (FPR) of 22.892% and AUC of 0.768.

DOI： 10.1145/3502871.3502876

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85126727510&origin=inward

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

U-Net frameworks reveal potential on image segmentation tasks of the complex morphologic objects, such as capillaries. To compare the performance of vessels segmentation in fundus images, in this paper, we review U-Net and its three derived architectures: Residual Spatial Attention Network, Generative Adversarial Networks and IterNet. The networks training and testing were completed using the same image datasets under the same configurations. We calculated the accuracy and precision of segmentation results. The results showed that the cascade U-Net architecture provided better results especially on the capillaries parts.

DOI： 10.1145/3502803.3502804

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85124455424&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

DOI： 10.1117/12.2604193

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Korea  

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Korea  

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Korea  

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Korea  

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

記述言語：英語   掲載種別：研究論文（学術雑誌）

Because of the rapid spread and wide range of the clinical manifestations of the coronavirus disease 2019 (COVID-19), fast and accurate estimation of the disease progression and mortality is vital for the management of the patients. Currently available image-based prognostic predictors for patients with COVID-19 are largely limited to semi-automated schemes with manually designed features and supervised learning, and the survival analysis is largely limited to logistic regression. We developed a weakly unsupervised conditional generative adversarial network, called pix2surv, which can be trained to estimate the time-to-event information for survival analysis directly from the chest computed tomography (CT) images of a patient. We show that the performance of pix2surv based on CT images significantly outperforms those of existing laboratory tests and image-based visual and quantitative predictors in estimating the disease progression and mortality of COVID-19 patients. Thus, pix2surv is a promising approach for performing image-based prognostic predictions.

DOI： 10.1016/j.media.2021.102159

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85111029043&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

Currently, auscultation using a stethoscope is performed for the diagnosis of respiratory diseases. Auscultation is a simple and non-invasive diagnostic method; however, the diagnostic results depend on the experience of the doctor, thereby rendering quantitative diagnosis difficult. Therefore, we herein propose a new automatic classification method based on deep learning algorithms for respiratory sounds to support the diagnosis of respiratory diseases. The proposed method comprises two stages. First, a spectrogram is generated by applying a short-time Fourier transform to the respiratory sound data. Subsequently, the obtained spectrogram is classified into normal and abnormal (three classes: crackle, wheeze, and both) respiratory sounds using an improved convolutional recurrent neural network. By classifying the respiratory sounds using the proposed method, the following results are obtained: sensitivity, 0.63; specificity,0.83; average score, 0.73; harmonic score, 0.72. Furthermore, the proposed method yields better accuracy compared with other methods.

DOI： 10.1016/j.compeleceng.2021.107367

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85112012928&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

Industrial Internet of Things (IIoT) provide significant support for observing and controlling industrial machinery. In this article, a novel hierarchical feature enhancement network (HFEN) is proposed by combining signal processing and representation learning. The signal processing block extracts features with definite physical significance. Then, the representability of the physical features is improved by connecting stacked denoising autoencoders and squeeze-and-excitation networks. A novel two-stream architecture is designed for HFEN to fuse two types of features. Consequently, HFEN can extract features that can be analyzed for physical significance and that are also representative in terms of recognizable patterns. The experimental results prove that the performance of HFEN is satisfactory in terms of accuracy and efficiency when compared to other methods. Finally, this article also aims to demonstrate the potential of a new pairing that fuses the model-and data-driven strategies for IIoT.

DOI： 10.1109/TII.2020.3021688

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85104187899&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

Visual servo is a method for robot arm motion control. It is controlled by the end effector velocity that is the result in calculating the internal Jacobi matrix and vector of feature error. In general, automatic robotic task require high quality sensor that can measure a 3-dimential distance, and do calibration in order to suit the sensor frame and robot frame in Euclidean space. In this paper, we only use RGB camera as the data collection, which not requiring the calibration in sensor frame. Thus, our method is simpler than any other automatic motion methods. Meanwhile, the proposed characteristics based visual servo method has varying the hyper parameter, and show the effectiveness for indicating the precision of pose error both simulation and actual environments.

DOI： 10.1016/j.cogr.2021.06.002

DOI： 10.1016/j.cogr.2021.06.002

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85125125810&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

Due to the rising demand for minerals and metals, various deep-sea mining systems have been developed for the detection of mines and mine-like objects on the seabed. However, many of them contain some issues due to the diffusion of dangerous substances and radioactive substances in water. Therefore, efficient and accurate visual monitoring is expected by introducing artificial intelligence. Most recent deep-sea mining machines have little intelligence in visual monitoring systems. Intelligent robotics, e.g., deep learning-based edge computing for deep-sea visual monitoring systems, have not yet been established. In this paper, we propose the concept of a learning-based deep-sea visual monitoring system and use testbeds to show the efficiency of the proposed system. For example, to sense the underwater environment in real time, a large quantity of observation data, including captured images, must be transmitted from the seafloor to the ship, but large-capacity underwater communication is difficult. In this paper, we propose using deep compressed learning for real-time communication. In addition, we propose the gradient generation adversarial network (GGAN) to recover the heavily destroyed underwater images. In the application layer, wavelet-aware superresolution is used to show high-resolution images. Therefore, the development of an intelligent remote control deep-sea mining system with good convenience using deep learning applicable to deep-sea mining is expected.

DOI： 10.1016/j.cogr.2020.12.002

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85109094624&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

To optimize the early diagnosis and detection of lung cancer, computer-aided diagnostic (CAD) systems have been a useful tool for analyzing medical images. The temporal subtraction technique, which is a CAD system, performs the subtraction operation between the current image and the previous image on the same patient, and supports observation by emphasizing the temporal changes. However, the temporal subtraction technique for 3D images, such as thoracic CT images, has not yet been established. There is a need to develop efficient and highly accurate 3D nonrigid registration techniques to reduce subtraction artifacts. This study aims to develop a 3D nonrigid registration technique to establish a 3D temporal subtraction technique. In particular, we focus on the Finite Element Method, which is versatile, applicable to a wide range of fields, and capable of handling any shape. Our new method was examined on 46 clinical cases with multidetector row computed tomography images. As a result, the proposed method improved by 6.93% (p = 3.0 × 10−6) compared to the conventional methods in terms of the rate of reduction of artifacts, and the effectiveness was verified. Therefore, this study contributes to the literature on early detection and treatment.

DOI： 10.1016/j.techfore.2021.120745

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85102645602&origin=inward

記述言語：英語   掲載種別：記事・総説・解説・論説等（学術雑誌）

DOI： 10.21037/qims-20-1367

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85104634131&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

Semantic segmentation is widely applied in autonomous driving, in robotic picking, and for medical purposes. Due to the breakthrough of deep learning in recent years, the fully convolutional network (FCN)-based method has become the de facto standard in semantic segmentation. However, the simple FCN has difficulty in capturing global context information, since the local receptive field is small. Furthermore, there is a problem of low image resolution because of the existence of the pooling layer. In this paper, we address the shortcomings of the FCN by proposing a new architecture called WideSegNeXt, which captures the image context on various spatial scales and is effective in identifying small objects. In addition, there is little loss of position information, since there are no pooling layers in the structure. The proposed method achieves a mean intersection over union (MIoU) of 72.5% and a global accuracy (GA) of 92.4% on the CamVid dataset and achieves higher performance than previous methods without additional input datasets.

DOI： 10.1109/JSEN.2020.3008908

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85104614498&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

DOI： 10.1038/s41598-021-88591-z.

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：筆頭著者   記述言語：日本語   掲載種別：記事・総説・解説・論説等（大学・研究所紀要）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

Multi-organ segmentation is a critical step in Computer-Aided Diagnosis (CAD) system. We proposed a novel method for automatic abdominal multi-organ segmentation by introducing spatial information in the process of supervoxel classification. Supervoxels with boundaries adjacent to anatomical edges are separated from the image by using the Simple Linear Iterative Clustering (SLIC) from the images. Then a random forest classifier is built to predict the labels of the supervoxels according to their spatial and intensity features. Thirty abdominal CT images are used in the experiment of segmentation task for spleen, right kidney, left kidney, and liver region. The experiment result shows that the proposed method achieves a higher accuracy of segmentation compares to our previous model-based method.

DOI： 10.18178/joig.9.1.9-14

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85109040224&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

記述言語：英語   掲載種別：研究論文（学術雑誌）

Purpose: Deep learning can be used for improving the performance of computer-aided detection (CADe) in various medical imaging tasks. However, in computed tomographic (CT) colonography, the performance is limited by the relatively small size and the variety of the available training datasets. Our purpose in this study was to develop and evaluate a flow-based generative model for performing 3D data augmentation of colorectal polyps for effective training of deep learning in CADe for CT colonography. Methods: We developed a 3D-convolutional neural network (3D CNN) based on a flow-based generative model (3D Glow) for generating synthetic volumes of interest (VOIs) that has characteristics similar to those of the VOIs of its training dataset. The 3D Glow was trained to generate synthetic VOIs of polyps by use of our clinical CT colonography case collection. The evaluation was performed by use of a human observer study with three observers and by use of a CADe-based polyp classification study with a 3D DenseNet. Results: The area-under-the-curve values of the receiver operating characteristic analysis of the three observers were not statistically significantly different in distinguishing between real polyps and synthetic polyps. When trained with data augmentation by 3D Glow, the 3D DenseNet yielded a statistically significantly higher polyp classification performance than when it was trained with alternative augmentation methods. Conclusion: The 3D Glow-generated synthetic polyps are visually indistinguishable from real colorectal polyps. Their application to data augmentation can substantially improve the performance of 3D CNNs in CADe for CT colonography. Thus, 3D Glow is a promising method for improving the performance of deep learning in CADe for CT colonography.

DOI： 10.1007/s11548-020-02275-z

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85095125050&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Today, cancer is the number one cause of death in Japan, and cancer accounts for 27.3% of all deaths number. The development of cancer by repeating metastasis, hence it is important to operate early detection and early treatment. The diagnosis of cancer includes various treatments, but it is difficult to judge whether cancer is metastatic or not. Then, analysis of Circulating Tumor Cells (CTCs) in blood has been attracting attention as a new biomarker. However, because the ratio of CTCs in a billion blood cells is only a few, and there is a concern that the burden on doctors will increase. We propose a method for automatic identification of CTCs from fluorescence microscopy images to enable quantitative analysis by computer for the diagnosis of CTCs in blood. First, after detecting the cell candidate regions mainly by filtering, we set the region of interest in the cell candidate regions and reconstruct the region of interest by cutting out the cell nucleus region. In this paper, we applied the proposed method to 5,040 images of 6 samples and conducted experiments on the identification of CTCs. As a result, the number of detections was 148(TPR = 100%) and the number of over-detected non-CTCs was 988.

DOI： 10.23919/ICCAS52745.2021.9650022

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85124240604&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

The number of hospitalized patients and the number of people requiring nursing care are serious social problems in Japan due to the increasing elderly population. The major causes of bedridden patients are bone and joint disorders caused by rheumatoid arthritis and osteoporosis. Early detection and treatment of these bone diseases are important because they significantly interfere with the quality of life (QOL) as the symptoms progress. Visual screening based on CR is used as a diagnosing tool for bone diseases. However, imaging diagnosis is subjective and lacks objectivity, and there is a possibility that lesions may be overlooked. In addition, it is difficult to find out subtle changes from images, increasing the workload for doctors. To solve these problems, there is a need to develop a computer aided diagnosis (CAD) system that can quantitatively diagnose bone diseases. We propose a method for automatic extraction of phalange regions for the CAD system to diagnose these diseases. The proposed method can extract the phalanges with high accuracy by using the improved DeepLabv3+. In this paper, we apply the proposed method to 101 cases of CR images and mIoU of 0.949 was obtained.

DOI： 10.23919/ICCAS52745.2021.9649864

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85124185622&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

The death toll from respiratory illness reached nearly 8 million in 2019. Auscultation is used to diagnose for respiratory illness. Highly accurate diagnosis is required to reduce the number of deaths. However, unlike diagnostic imaging, auscultation of respiratory sounds could not visualize the diagnostic results. In addition, since there is a problem that the experience of a doctor affects the diagnosis results, it is required to develop a diagnostic system for quantitative analysis. In recent years, the development of a diagnostic system using the ICBHI 2017 Challenge Respiratory Sound Database has been carried out in the field of respiratory sound analysis. However, the proposed system still has accuracy problems. Therefore, in this study, we improve the proposed method by classifying the improved CRNN (Convolutional Recurrent Neural Network) by inputting multiple respiratory sound images. As a result, Sensitivity: 0.64, Specificity: 0.83, Average Score: 0.74, Harmonic Score: 0.72 were obtained, and excellent results were achieved compared with other methods.

DOI： 10.23919/ICCAS52745.2021.9649906

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85124181434&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Root resorption is a pathological process characterized by the loss of tooth roots because of inflammation induced by bacterial infection, trauma, physical or chemical irritation. As a result, the development of periodontal disease, increased susceptibility to infection and crooked teeth. In the worst case, it can lead to tooth extraction. Root resorption is often caused by pressure during orthodontic treatment. The presence of root resorption should be checked regularly during orthodontic treatment, as it often occurs. It is necessary to check for root resorption periodically during orthodontic treatment. However, it is difficult to detect the root resorption using a panoramic radiograph. As a result, root resorption is often latent and goes undetected. In this paper, we propose an image analysis method based on deep learning technique for detecting the root resorption on panoramic radiograph. We incorporate the EfficientNet for feature extraction in deep learning to the center loss and triplet loss as the loss function for metric learning. Our proposed method performed to 337 images which is obtained by panoramic radiograph. Accuracy of 71 %, true positive rate of 77%, false positive rate of 30% were obtained.

DOI： 10.23919/ICCAS52745.2021.9649745

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85124211898&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

A power device is a semiconductor device for power control used for power conversion such as converting direct current to alternating current and alternating current to direct current. It is widely used such as refrigerators, air conditioners which is implemented electronic components that are closely related to our daily lives. Therefore, high reliability and safety are required, and power cycle tests are conducted for the purpose of evaluating them. In the conventional test, there is a problem that it is difficult to perform analysis because sparks are generated during the test and the device is severely damaged after the test. To solve this problem, a new technology has been developed that adds ultrasonic that enable internal observation during the test. However, there are remains a problem that the method for analyzing the ultrasonic image obtained in the new technology has not been established. Also, few abnormal images are obtained in the test. In this paper, we propose a method for detection of abnormal devices based on CNN. Especially, we implement a Cycle-GAN to extend the abnormal data and classify the known image based on improved VGG16. As an experimental result, classification accuracy of Precision = 97.06%, Recall = 93.58%, F-measure = 95.17% were obtained.

DOI： 10.23919/ICCAS52745.2021.9650029

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85124191152&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

The number of users of electric wheelchairs has been increasing in recent years because it is easy to operate the electric wheelchair and do not require physical strength. However, the traffic accidents are also increasing because of the large number of wheelchairs. The development of autonomous electric wheelchairs is expected to reduce the risk of accidents and improve the convenience of electric wheelchairs. Environmental recognition is essential for the development of autonomous electric wheelchairs. In this paper, we propose a method for recognizing roads, sidewalks, buildings, electric wheelchair drivers, poles, electric wheelchairs, vegetation, curbs, sky, pedestrians, lanes, cars, steps, and bicycles. For recognizing those objects, we use a panoramic image acquired from a spherical camera. As the machine techniques, we use DeepLab v3+, a semantic segmentation algorithm based on Convolutional Neural Network (CNN). In the proposed method, a new CNN model is constructed by adding deformable convolution, SE-block, and MobileNet v2 to DeepLab v3+ into the original DeepLab v3+. In the experiment, IoU 38.8% and Dice of 46.7% were obtained.

DOI： 10.23919/ICCAS52745.2021.9649846

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85124192140&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

One of the difficulties of cancer is the metastasis of cancer through the lymph nodes. Therefore, early detection and treatment of cancerous is important. Visual screening is one of useful tool for diagnosing of the cancer. However, the number of images obtained at a time is large, and the burden on the reading physician is increasing. Furthermore, reading of image is based on the subjective judgment of the physician, which may lead to different diagnostic results. To solve these problems, computer aided diagnosis (CAD) systems have been attracting attention in recent years. One of the CAD systems is the temporal subtraction image technology. In this paper, we propose an image alignment method for generating temporal subtraction images from images taken before and after contrast agent was administered to the cervical lymph nodes of the same subject. In addition, we propose an image analysis method that suppresses overextraction of lymph node candidate regions on the temporal subtraction image based on the features of lymph nodes. We applied the proposed method to the CT images of three sets and compared the temporal subtraction images with the final lymph node extraction images, and confirmed that the proposed method can suppress the overextraction of lymph node regions.

DOI： 10.23919/ICCAS52745.2021.9649797

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85124183732&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Military camouflage should be various in order to quickly adapt environmental changes of battlefield. The elusiveness is depressed if the constant camouflage patterns of the clothes and vehicles are used, as applying in the most of current military. In this paper, we propose a texture synthesis method from an image using convolutional neural networks, which has an effect on the camouflage generation on a 3D surface. We use the latest advances in style transitions in 2D images and map characteristic textures onto 3D surfaces according to geometric feature. This allows us to make an adaptive texture map of the 3D objects, even in the case of complex topologies.

DOI： 10.1007/978-3-030-46032-7_8

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85091331451&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

It is necessary to work in response to various situations when using robots at disaster areas. So, the complete automation remote control technique of the robot is essential. In recent years, the remote control technique of the robot is mainly focused on recognizing the working environment of the robot from the 2D monitor and performing the operation. In order to operate the robot correctly, a corresponding technique and sufficiently trained engineers are required. To this end, a new remote control technology that can operate more easily is required. Also, depending on the disaster environment that working at night or indoors is assumed, it becomes difficult to recognize the work environment correctly from the image by the sensor such as the camera due to insufficient light source. In this chapter, in order to solve these problems, as a development of new remote control technology, we propose the method of reproducing the working environment and the image improvement method under the low light source environment. In this chapter, we reproduce the working environment of remote areas by using virtual reality. In addition, we improve the image under low light source environment using haze removal algorithm. The work environment of the robot is acquired by RGBD sensor. Image improvement processing is performed on the obtained color image, and the image data is transmitted to a remote place where the operator is located via WebSocket. The experiments show that our method reproduces the working environment of the robot as point group to the virtual reality space and examine its usefulness.

DOI： 10.1007/978-3-030-46032-7_3

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85091329616&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

While diagnosis imaging is an indispensable technology in medical field, it causes an increase in the burden to radiologists. In recent years, computer-aided diagnosis (CAD) system for supporting a radiologist has been developed to solve this problem. Temporal subtraction, which is one of CAD, is a technique to generate images emphasizing temporal changes in lesions and facilitates diagnosis of radiologists. To make a temporal subtraction image, image registration technique is required. In this chapter, we propose an image registration method for image registration of current image and previous image to generate temporal subtraction images in a short time. The proposed method consists of three steps: (i) segmentation of the region of interest (ROI) using position information of the spine based on anatomical information, (ii) using global image matching to select pairs of previous image and current image in which the same portion is depicted, and iii) final image matching based on salient region features (SRF). We performed our registration technique to the synthetic data and confirmed usefulness of the proposed method. The rotated synthesis image gives TP 100.0% and FP 12.16%. The synthesis image obtained by applying a Gaussian filter gives TP 70.40% and FP 0.00%. The synthesis image obtained by adding artificial pseudo-lesion region gives TP 99.45% and FP 17.89%. The synthesis image obtained by adding random noise of 5% gives TP 83.05% and FP 16.95%. Furthermore, radiologist conducted comparative experiments without and with temporal subtraction images created by proposed method. As a result, radiologist showed high reading performance by using temporal subtraction images.

DOI： 10.1007/978-3-030-46032-7_2

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85091327306&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

We developed and evaluated the effect of deep radiomic features, called U-radiomics, on the prediction of the overall survival of patients with the coronavirus disease 2019 (COVID-19). A U-net was trained on chest CT images of patients with interstitial lung diseases to classify lung regions of interest into five characteristic lung tissue patterns. The trained Unet was applied to the chest CT images of patients with COVID-19, and a U-radiomics vector for each patient was identified from the bottleneck layer of the U-net across all the axial CT images of the patient. The U-radiomics vector was subjected to a Cox proportional hazards model with an elastic-net penalty for predicting the survival of the patient. The evaluation was performed by use of bootstrapping, where the concordance index (C-index) was used as the comparative performance metric. Our preliminary comparative evaluation of existing prognostic biomarkers and the proposed U-survival model yielded the C-index values of (a) extent of well-Aerated lung parenchyma: 51%, (b) combination of blood tests of lactic dehydrogenase, lymphocyte, and C-reactive protein: 63%, and (c) U-survival: 87%. Thus, the U-survival significantly outperformed clinical biomarkers in predicting the survival of COVID-19 patients, indicating that the U-radiomics vector of the U-survival model may provide a highly accurate prognostic biomarker for patients with COVID-19.

DOI： 10.1117/12.2581907

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85103199185&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

We developed an image-based unsupervised survival prediction model, called pix2surv, based on a conditional generative adversarial network (cGAN), and evaluated its performance based on chest CT images of patients with the coronavirus disease 2019 (COVID-19). The architecture of the pix2surv model includes a time generator that consists of an encoding convolutional network and a fully connected prediction network, and a discriminator network. The time generator is trained to generate survival-time images from chest CT images of each patient. The discriminator is a patch-based convolutional network that is trained to differentiate between "fake pairs"of a chest CT image and a generated survival-time image from "true pairs"of the chest CT image and the corresponding observed survival-time image of the patient. For evaluation, we retrospectively collected high-resolution chest CT images of COVID-19 patients. The survival predictions of the pix2surv model on these patients were compared with those of existing clinical prognostic biomarkers by use of a two-sided t-test with bootstrapping. Concordance index (C-index) and relative absolute error (RAE) were used as measures of the prediction performance. The bootstrap evaluation yielded C-index and RAE values of 80.4% and 15.6% for the pix2surv model, whereas those for the extent of the well-aerated lung parenchyma were 49.8% and 33.6%, and for a combination of blood tests of lactic dehydrogenase, lymphocyte, and C-reactive protein were 69.8% and 25.5%, respectively. The increase in survival prediction by the pix2surv model was statistically significant (p < 0.0001), indicating high effectiveness of the pix2surv model as a prognostic biomarker for the survival of patients with COVID-19.

DOI： 10.1117/12.2581929

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85103686657&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Cleft lip is one of the most common morphological abnormalities that occur when the fetal lips are torn. Cleft lip is treated with multiple surgical procedures. However, there is a problem that the evaluation of surgical results depends on the subjectivity of doctors and the degree of symmetry is difficult to evaluate quantitatively. Therefore, as a previous study, a method for detecting and evaluating the plane of symmetry of the face with an asymmetric region was proposed, however, there remained a problem with the accuracy of separation of the asymmetry between patients and healthy subjects. In this paper, we propose an asymmetry evaluation method with high separation accuracy using the plane of symmetry of the face. The plane of symmetry used as the evaluation standard is obtained by adjusting the plane of symmetry of the entire face to the plane of symmetry of the surgical region. The asymmetry is evaluated by the evaluation index using the depth information with the plane of symmetry as the reference. By comparing the proposed method with the conventional method, the degree of separation of asymmetry between patients and healthy subjects was improved, and more useful results were obtained as an evaluation index.

DOI： 10.23919/ICCAS52745.2021.9649934

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85124214360&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

In-vehicle cameras and surveillance cameras are used in many situations in our daily lives. Visibility degradation in foggy environments is caused by the scattering of reflected light from real objects by minute water droplets or fog in the medium through which light passes. The degree of degradation depends on the density of suspended microparticles existing between the observed object and the observation point in the medium. In general, the farther the object is from the camera, the more it is affected by the fog. The purpose of image de-fogging is to improve the clarity of an object by removing the effects of fog in the image.

DOI： 10.1117/12.2603771

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85120496362&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Virtual Reality (VR) systems are become popular in recent years, and the capture 3D objects from the real world have been studied. 3D objects have large and complex data. In this paper, we propose a novel method that use the shadow information and Photometric stereo with their surface from one point of view to recover the 3D shapes. The experimental results show that the proposed method performs well accuracy.

DOI： 10.1117/12.2604193

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85120506573&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

In recent years, energy minerals have become more important due to the rapid industrialization worldwide. Due to the rapid industrialization on a global scale, there is a shortage of mineral resources, and there are more opportunities to rely on alternative energy sources. Therefore, the exploration of marine resources, which are abundant in the ocean, is being promoted. However, it is dangerous and impractical for humans to dive and search for marine resources by hand. Therefore, it is possible to proceed with underwater exploration safely by having a robot do the work instead. Robots have been used as a mainstream search tool in the underwater environment due to the existence of various hazardous environmental conditions. However, there are several problems associated with robot control in underwater environments, one of which is poor visibility in the water. One of the problems is the poor visibility in the water. To improve the visibility in the water, we are trying to increase the resolution of underwater images by using super-resolution technology. In this paper, we conduct experiments using SRCNN, which is a basic super-resolution technique for underwater images. In addition, we investigate the effectiveness of "Mish", which has been attracting attention in recent years for its potential to surpass the performance of "ReLU", although "ReLU"is a typical activation function of neural networks, on SRCNN.

DOI： 10.1117/12.2603761

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85120500372&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

With the explosion of amount of low cost satellites, satellite images have been widely used for many non-military applications, such as agriculture, landscape, and recognition of environment. Improving the image resolution to mine useful information becomes one of the immediate problems. Therefore, it is expected to improve the recognition accuracy by increasing the resolution of satellite images. Recently, deep learning technique has been proposed to increase the resolution of images. However it requires a large number of learning parameters, which results in huge computational cost. To overcome this problem, we develop a new deep learning model based on ghost module to reduce the parameters while maintaining the quality of results. We utilized Google Earth Pro satellite imagery for the network training and testing. Comparing to the classical convolutional neural network module based methods, the number of parameters used in our model was reduced 49.31 % but keeping the same level of Peak Signal-to-Noise Ratio (24.1578) and Structural Similarity (0.7174).

DOI： 10.23919/ICCAS52745.2021.9649982

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85124232834&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Magnetic Particle Testing (MPT) is a method for determining the presence or absence of a defect by magnetizing the object to be inspected and sprinkling magnetic powder, which is absorbed by the defective part such as a crack and appears as a magnetic powder pattern, which is then evaluated by a specialist. By using the MTP, inspection can be performed without breaking the object to be inspected. However, there are some problems such as the possibility of overlooking defects. In this paper, to solve the problems we develop a classification method of defect images by deep learning for the automation of MPT. The proposed method is based on the structure of U-Net, which has excellent segmentation capability in image processing, and performs segmentation using an improved model that adds convolutional layers to U-Net. Then, an algorithm that combines the result with the last part of the encoder of U-Net is used to discriminate the presence or absence of defects. Using this method, defects were classified from the images obtained during MPT. The results showed that Accuracy of 85.8%, TPR of 65.2%, and FPR of 13.8%.

DOI： 10.23919/ICCAS52745.2021.9650052

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85124232188&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

記述言語：日本語   掲載種別：研究論文（学術雑誌）

DOI： 10.11409/mit.38.217

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2020 Institute of Control, Robotics, and Systems - ICROS. In recent years, the computer-aided diagnosis (CAD) systems for supporting to the physician is attracting attention in medical research field. One of them is temporal subtraction technique. It is a technique to generate images emphasizing temporal changes in lesions by performing a differential operation between current and previous image of the same subject. However, there is a problem of mistakenly selecting current and previous slice of a spine region with similar geometry as the same slice because many of the spine regions in axial plane have similar geometries. In this paper, we propose an image registration method for the detection of bone metastases from the spine region by creating a temporal subtraction images from CT images. Especially, we develop an image registration system for spine region considering sagittal plane to accurately select the same slice for the current image and the previous image. In the experiment, we applied the proposed method to CT images of 27 cases with bone metastases and the results were compared with the markings of the lesions.

DOI： 10.23919/ICCAS50221.2020.9268246

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85098087691&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2020 Institute of Control, Robotics, and Systems - ICROS. Cleft lip occurs in about 1 in 500 people and is a congenital anomaly that occurs when the lips are not completely formed in the process of fetal face fusion. The treatment of the cleft lip is performed for the purpose of forming bilaterally symmetrical lips and external nose. However, it is based on a criterion that depends on subjectivity, quantitatively determining the degree of symmetry of the surgical site is required. In this paper, we propose a method for detecting the face that serves as the symmetry basis of the face to analyze the symmetry of the surgical site, and an index that indicates the symmetry. In the proposed method, after detecting the facial organs as points in the 3D point cloud data of the subject's face, we align the mirror image inversion excluding the region from upper lip to the tip of the nose where the effect of shape change due to cleft lip is remarkable with the original point cloud. Next, the symmetric reference plane is set by finding a plane that bisects vertically between one point in the original point cloud and the corresponding point in the mirror image inversion. As a result, the target reference plane is detected with high accuracy. Furthermore, as a result of applying the symmetry analysis method based on the target reference plane, it was possible to visually represent the left-right asymmetric region.

DOI： 10.23919/ICCAS50221.2020.9268262

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85098071444&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2020 Institute of Control, Robotics, and Systems - ICROS. The aging of Japan is remarkable, and attention has been focused on the use and utilization of assistive devices. One of them is electric wheelchair, which enables physical disability people to easily operate it using a handle or a joystick. However, accidents are occurring frequently with increasing demand by using electric wheelchair. Therefore, developing an autonomous electric wheelchair is required to reduce accidents such as maneuvering mistakes, reduce the accident rate, improve convenience, and reduce the burden on caregivers. In this paper, we focus on the recognition of obstacles and use panoramic images obtained from a spherical camera that can easily handle information from all directions at low cost. A spherical camera is attached to an electric wheelchair, and images are cut out from the sequential images obtained by running. For image analysis, YOLOv3, which has been successful in the field of image recognition in recent years, is used. In the proposed method, considering the distortion of the image caused by using the spherical camera, the improvement of the model of YOLOv3 is examined, and the validity with the actual data is verified.

DOI： 10.23919/ICCAS50221.2020.9268308

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85098088521&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2020 Institute of Control, Robotics, and Systems - ICROS. In recent years, the number of lung cancer deaths has been increasing. In Japan, CT (Computed Tomography) equipment is used for its visual screening. However, there is a problem that seeing huge number of images taken by CT is a burden on the doctor. To overcome this problem, the CAD (Computer Aided Diagnosis) system is introduced on medical fields. In CT screening, LDCT (Low Dose Computed Tomography) screening is desirable considering radiation exposure. However, the image quality which is caused the lower the dose is another problem on the screening. A CAD system that enables accurate diagnosis even at low doses is needed. Therefore, in this paper, we propose a registration method for generating temporal subtraction images that can be applied to low-quality chest LDCT images. Our approach consists of two major components. Firstly, global matching based on the center of gravity is performed on the preprocessed images, and the region of interest (ROI) is set. Secondly, local matching by free-form deformation (FFD) based on B-Spline is performed on the ROI as final registration. In this paper, we apply our proposed method to LDCT images of 6 cases, and reduce 57.29% in the calculation time, 26.1% in the half value width, and 29.6% in the sum of histogram of temporal subtraction images comparing with the conventional method.

DOI： 10.23919/ICCAS50221.2020.9268267

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85098056092&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2020 Institute of Control, Robotics, and Systems - ICROS. Power devices are semiconductor devices that handle high voltages and large currents, which are used in electric vehicles, televisions, and trains. Therefore, high reliability and safety are required, and to ensure this, power cycle tests are performed to analyze the breakdown process. Conventional tests are often difficult to analyze due to the influence of sparks generated during the test. Therefore, new tests are being developed by adding ultrasound to conventional methods. The new technology is capable of continuously recording structural changes inside the device during testing, which is expected to make testing much easier than conventional testing. However, the new technology still has some challenges. The main problems are the lack of a method for analyzing large amounts of image data and the extraction of small changes in image features that are difficult to distinguish with the human eye, and the establishment of such a system is required. In this paper, we use deep learning for image classification of the obtained ultrasound images. We propose a new network model with the addition of Batch normalization and Global average pooling to VGG16, which is a pre-trained model. In the experiment, accuracy=98.29%, TPR=98.96% and FPR=7.43% classification accuracy was obtained.

DOI： 10.23919/ICCAS50221.2020.9268275

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85098091061&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2020 Institute of Control, Robotics, and Systems - ICROS. According to the 2016 World Health Organization (WHO) survey, respiratory diseases are serious diseases that account for four of the top ten causes of death in the world, accounting for more than 8 million deaths worldwide. Currently, the diagnosis of respiratory disease is made by auscultation, but in order to make an accurate diagnosis, a number of abnormal patterns of respiratory sounds need to be memorized, and the results of the diagnosis are dependent on the proficiency of the physician. Therefore, a computer aided diagnosis (CAD) system is needed to quantitatively classify the respiratory sounds and output the results as a "second opinion". In this paper, a short-time Fourier transformed spectrogram, a Constant-Q transformed logarithmic frequency spectrogram, and a continuous wavelet transformed scalogram are simultaneously input to VGG16 which is one of the network models of CNN(Convolutional Neural Network) and classified by LSTM (Long short-term memory). The proposed method is applied to 26 respiratory sounds, and the 0.90 of accuracy, sensitivity of 0.97, and specificity of 0.90 is obtained.

DOI： 10.23919/ICCAS50221.2020.9268428

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85098088561&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

DOI： 10.1145/3436349.3436365

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85100769772&origin=inward

記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2020 IEEE. Big data-driven deep learning methods have been widely used in image or video segmentation. The main challenge is that a large amount of labeled data is required in training deep learning models, which is important in real-world applications. To the best of our knowledge, there exist few researches in the deep learning-based visual segmentation. To this end, this paper summarizes the algorithms and current situation of image or video segmentation technologies based on deep learning and point out the future trends. The characteristics of segmentation that based on semi-supervised or unsupervised learning, all of the recent novel methods are summarized in this paper. The principle, advantages and disadvantages of each algorithms are also compared and analyzed.

DOI： 10.1109/COMPSAC48688.2020.00-84

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85094163880&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2018, Springer Science+Business Media, LLC, part of Springer Nature. Circulating tumor cells (CTCs) is a clinical biomarker for cancer metastasis. CTCs are cells circulating in the body of patients by being separated from primary cancer and entering into blood vessel. CTCs spread every positions in the body, and this is one of the cause of cancer metastasis. To analyze them, pathologists get information about metastasis without invasive test. CTCs test is conducted by analyzing the blood sample from patient. The fluorescence microscope generates a large number of images per each sample, and images contain a lot of cells. There are only a few CTCs in images and cells often have blurry boundaries. So CTCs identification is not an easy work for pathologists. In this paper, we develop an automatic CTCs identification method in fluorescence microscopy images. This proposed method has three section. In the first approach, we conduct the cell segmentation in images by using filtering methods. Next, we compute feature values from each CTC candidate region. Finally, we identify CTCs using artificial neural network algorithm. We apply the proposed method to 5895 microscopy images (7 samplesas), and evaluate the effectiveness of our proposed method by using leave-one-out cross validation. We achieve the result of performance tests, a true positive rate is 92.57% and false positive rate is 9.156%.

DOI： 10.1007/s11036-018-1121-0

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85053481889&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2018, Springer Science+Business Media, LLC, part of Springer Nature. Deep-sea organism automatic tracking has rarely been studied because of a lack of training data. However, it is extremely important for underwater robots to recognize and to predict the behavior of organisms. In this paper, we first develop a method for underwater real-time recognition and tracking of multi-objects, which we call “You Only Look Once: YOLO”. This method provides us with a very fast and accurate tracker. At first, we remove the haze, which is caused by the turbidity of the water from a captured image. After that, we apply YOLO to allow recognition and tracking of marine organisms, which include shrimp, squid, crab and shark. The experiments demonstrate that our developed system shows satisfactory performance.

DOI： 10.1007/s11036-018-1117-9

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85055349873&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2018, The Natural Computing Applications Forum. Designing the loudspeaker could be concerned as an optimization problem. Like most electromagnetic device design issues, it demonstrates multimodal, multidimensional and constrained. The traditional design method cannot achieve a satisfactory model within a certain period of time. In this paper, a high-quality-guided artificial bee colony algorithm is proposed to increase its convergence speed and search accuracy by gradually changing the number of updating dimensions and searching closer to better locations. The algorithm is first tested on some representative basic benchmark functions and then is applied to a loudspeaker design problem. By comparing with some classical algorithms, the performance of our algorithm is verified.

DOI： 10.1007/s00521-018-3568-0

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85048041015&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

In order to get with the environmental changes of battlefield quickly, the military camouflage should be changeable. If the camouflage patterns of the clothes and vehicles like tanks are different from the environment, it's very easy for cameras of enemies to find. As we all know that the same patterns is used in the most of current military all over the world. In this paper, we propose a novel feature-extraction method from an image using convolutional neural networks. Then the pattern will be combined with the environmental style pattern. The composite image is mapped onto the surface of the actual 3D clothes and vehicles finally. In this paper, the Eye-Movement equipment is applied to evaluate the results for better comparison. We can produce the proper pattern even the different and complicated environment.

DOI： 10.1145/3404555.3404637

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85092212124&origin=inward

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：筆頭著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2020 IEEE. Temporal subtraction (TS) technique is one of computer aided diagnosis (CAD) systems. A TS image is obtained by subtracting a previous image, which are warped to match between the structures of the previous image and one of a current image, from the current image. TS technique removes normal structures and enhances interval changes. However, many subtraction artifacts that can be detected as false positives still remain on a TS image. In this paper, we propose 3D-CNNs with residual sub-blocks based on 3D-VGG16-like architecture for detection of nodules accurately from TS images.

DOI： 10.1109/ICAIIC48513.2020.9065197

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85084033837&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2012 IEEE. In this article, a method to reduce artifacts on temporal difference images is introduced. The proposed method uses a nonrigid registration method for ground glass opacification (GGO), which is light in concentration and difficult to detect early. In this method, global matching, local matching, and three-dimensional (3D) elastic matching are performed on the current and previous images, and an initial temporal subtraction image is generated. After that, we use an Iris filter, which is the gradient vector concentration degree filter, to determine the initial GGO candidate regions and use supervoxel and graph cuts to segment region of interest in the 3D images. For each extracted region, a support vector machine is used to reduce the oversegmentation. The voxel matching is applied to generate the final temporal difference image, emphasizing the GGO regions while reducing the artifact.

DOI： 10.1109/MCE.2019.2941468

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85076374234&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2020, Springer Nature Switzerland AG. In this paper, we design a novel underwater hyperspectral imaging technique for deep-sea mining detection. The spectral sensitivity peaks are in the region of the visible spectrum, 580, 650, 720, 800 nm. In addition, to the underwater objects recognition, because of the physical properties of the medium, the captured images are distorted seriously by scattering, absorption and noise effect. Scattering is caused by large suspended particles, such as in turbid water, which contains abundant particles, algae, and dissolved organic compounds. In order to resolve these problems of recognizing mineral accurately, fast and effectively, an identifying and classifying algorithm is proposed in this paper. We take the following steps: firstly, through image preprocessing, hyperspectral images are gained by using denoising, smoothness, image erosion. After that, we segment the cells by the method of the modified active contour method. These methods are designed for real-time execution on limited-memory platforms, and are suitable for detecting underwater objects in practice. The Initial results are presented and experiments demonstrate the effectiveness of the proposed imaging system.

DOI： 10.1007/978-3-030-04946-1_50

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85062907061&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2020, Springer Nature Switzerland AG. The deep-sea tripod systems are designed and built at the U.S. Geological Survey (USGS) Pacific Coastal and Marine Science Center (PCMSC) in Santa Cruz, California. They are recovered in late September 2014 after spending about half a year collecting data on the floor of the South China Sea. The deep-sea tripod systems are named as Free-Ascending Tripod (FAT), are deployed at 2,100 m water depth—roughly 10 times as deep as most tripods dedicated to measuring currents and sediment movement at the seafloor. Deployment at this unusual depth was made possible by the tripod’s ability to rise by itself to the surface rather than being pulled up by a line. Instruments mounted on the tripod took bottom photographs and measured such variables as water temperature, current velocity, and suspended-sediment concentration. FAT is used to better understand how and where deep-seafloor sediment moves and accumulates. Besides of this, we also use them to study the deep-sea biology. The obtained the images from the camera, the biology animals are hardly to be distinguished. In this project, we are concerned to use novel underwater imaging technologies for recovering the deep-sea scene.

DOI： 10.1007/978-3-030-04946-1_49

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85062904328&origin=inward

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

<p>呼吸器疾患の診断方法としては，聴診器を用いた呼吸音の聴診が長年用いられてきた．これは簡便で安全な診断方法である一方，聴診音の診断には定量的な評価基準がないため，医師の診断支援を行うシステムの開発が必要である．そこで本論文では，畳み込みニューラルネットワーク（CNN： convolutional neural network）を用いた呼吸音の自動分類手法の提案を行う．おもな手法の流れとしては，呼吸音データに対して短時間フーリエ変換と連続ウェーブレット変換を適用し，スペクトログラム画像およびスカログラム画像を生成する．その後，生成した画像を用いてCNN による正常呼吸音，連続性ラ音，断続性ラ音の識別を行う．提案手法を呼吸音データ22 症例に適用した結果，分類性能として，全体正解率は 79.44［％］，ROC（receiver operating characteristic）曲線に基づくAUC（area under the curve）は0.942 を得た．</p>

DOI： 10.11409/mit.38.40

CiNii Article

CiNii Research

その他リンク： https://ci.nii.ac.jp/naid/130007796400

担当区分：責任著者   掲載種別：研究論文（学術雑誌）

© 2020, Springer Science+Business Media, LLC, part of Springer Nature. The original version of this article unfortunately contained a mistake in the Affiliation section.

DOI： 10.1007/s11036-020-01600-9

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85087566169&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2020 SPIE. Colorectal cancer is the second leading cause of cancer deaths worldwide. Computed tomographic colonography (CTC) can detect large colorectal polyps and cancers at a high sensitivity, whereas it can miss some of the smaller but still clinically significant 6 - 9 mm polyps. Dual-energy CTC (DE-CTC) can be used to provide more detailed information about scanned materials than does conventional single-energy CTC. We compared the classification performance of a 3D convolutional neural network (DenseNet) with those of four traditional 3D machine-learning models (AdaBoost, support vector machine, random forest, Bayesian neural network) and their cascade and ensemble classifier variants in the detection of small polyps in DE-CTC. Twenty patients with colonoscopy-confirmed polyps were examined by DE-CTC with a reduced one-day bowel preparation. The traditional machine-learning models were designed to identify polyps based on native radiomic dual-energy features of the DE-CTC image volumes. The performance of the machine-learning models was evaluated by use of the leave-one-patient-out method. The DenseNet was trained with a large independent external dataset of single-energy CTC cases and tested on blended image volumes of the DE-CTC cases. Although the DenseNet yielded the highest detection accuracy for typical polyps, AdaBoost and its cascade classifier variant yielded the highest overall polyp detection performance.

DOI： 10.1117/12.2549793

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85085516392&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2020 SPIE. Three-dimensional (3D) convolutional neural networks (CNNs) can process volumetric medical imaging data in their native volumetric input form. However, there is little information about the comparative performance of such models in medical imaging in general and in CT colonography (CTC) in particular. We compared the performance of a 3D densely connected CNN (3D-DenseNet) with those of the popular 3D residual CNN (3D-ResNet) and 3D Visual Geometry Group CNN (3D-VGG) in the reduction of false-positive detections (FPs) in computer-aided detection (CADe) of polyps in CTC. VGG is the earliest CNN design of these three models. ResNet has been used widely as a de-facto standard model for constructing deep CNNs for image classification in medical imaging. DenseNet is the most recent of these models and improves the flow of information and reduces the number of network parameters as compared to those of ResNet and VGG. For the evaluation, we used 403 CTC datasets from 203 patients. The classification performance of the CNNs was evaluated by use of 5-fold cross-validation, where the area under the receiver operating characteristic curve (AUC) was used as the figure of merit. Each training fold was balanced by use of data augmentation of the samples of real polyps. Our preliminary results showed that the AUC value of the 3D-DenseNet (0.951) was statistically significantly higher than those of the reference models (P < 0.005), indicating that the 3D-DenseNet has the potential of substantially outperforming the other models in reducing FPs in CADe for CTC. This improvement was highest for the smallest polyps.

DOI： 10.1117/12.2549103

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85085513046&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© COPYRIGHT SPIE. Downloading of the abstract is permitted for personal use only. We developed a novel survival prediction model for images, called pix2surv, based on a conditional generative adversarial network (cGAN), and evaluated its performance based on chest CT images of patients with idiopathic pulmonary fibrosis (IPF). The architecture of the pix2surv model has a time-generator network that consists of an encoding convolutional network, a fully connected prediction network, and a discriminator network. The fully connected prediction network is trained to generate survival-time images from the chest CT images of each patient. The discriminator network is a patchbased convolutional network that is trained to differentiate the "fake pair" of a chest CT image and a generated survivaltime image from the "true pair" of an input CT image and the observed survival-time image of a patient. For evaluation, we retrospectively collected 75 IPF patients with high-resolution chest CT and pulmonary function tests. The survival predictions of the pix2surv model on these patients were compared with those of an established clinical prognostic biomarker known as the gender, age, and physiology (GAP) index by use of a two-sided t-test with bootstrapping. Concordance index (C-index) and relative absolute error (RAE) were used as measures of the prediction performance. Preliminary results showed that the survival prediction by the pix2surv model yielded more than 15% higher C-index value and more than 10% lower RAE values than those of the GAP index. The improvement in survival prediction by the pix2surv model was statistically significant (P < 0.0001). Also, the separation between the survival curves for the low- and high-risk groups was larger with pix2surv than that of the GAP index. These results show that the pix2surv model outperforms the GAP index in the prediction of the survival time and risk stratification of patients with IPF, indicating that the pix2surv model can be an effective predictor of the overall survival of patients with IPF.

DOI： 10.1117/12.2551369

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85082610220&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2020 SPIE. We developed and evaluated the effect of U-Net-based radiomic features, called U-radiomics, on the prediction of the overall survival of patients with idiopathic pulmonary fibrosis (IPF). To generate the U-radiomics, we retrospectively collected lung CT images of 72 patients with interstitial lung diseases. An experienced observer delineated regions of interest (ROIs) from the lung regions on the CT images, and labeled them into one of four interstitial lung disease patterns (ground-glass opacity, reticulation, consolidation, and honeycombing) or a normal pattern. A U-Net was trained on these images for classifying the ROIs into one of the above five lung tissue patterns. The trained U-Net was applied to the lung CT images of an independent test set of 75 patients with IPF, and a U-radiomics vector for each patient was identified as the average of the bottleneck layer of the U-Net across all the CT images of the patient. The U-radiomics vector was subjected to a Cox proportional hazards model with elastic-net penalty for predicting the survival of the patient. The evaluation was performed by using bootstrapping with 500 replications, where concordance index (C-index) was used as the comparative performance metric. The preliminary results showed the following C-index values for two clinical biomarkers and the U-radiomics: (a) composite physiologic index (CPI): 64.6%, (b) gender, age, and physiology (GAP) index: 65.5%, and (c) U-radiomics: 86.0%. The U-radiomics significantly outperformed the clinical biomarkers in predicting the survival of IPF patients, indicating that the U-radiomics provides a highly accurate prognostic biomarker for patients with IPF.

DOI： 10.1117/12.2551273

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85085531092&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

In recent years, there has been an increasing demand for automation of cargo handling operations in harbors. However, its automation has not been realized in Japan since there are many fundamental technologies to be solved until now. In this study, we propose a tracking method of the container gripping area based on Re3 for the purpose of automation of cargo handling work. Re3, an object tracking method used in the conventional method, has a problem that global features cannot be extracted well. In order to solve this problem, our method incorporates a model called Stochastic Depth.

DOI： 10.5954/ICAROB.2020.GS5-4

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85108782511&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

In recent years, the rate of death from cancer has tended to increase in Japan. Also, there is a concern that increasing the performance of CT will increase the burden on doctors. Therefore, by presenting a "second opinion" in the CAD system, the burden on doctors can be reduced. We develop a CAD system for automatic detection of lung cancer. In this paper we propose a method to detect abnormalities based on temporal subtraction technique, sparse coding and 3D-CNN. We obtain the result that sparse level contributed most to the score.

DOI： 10.5954/ICAROB.2020.GS3-3

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85108848194&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

A system that investigates deep sea automatically has never developed. A purpose of this study is developing such a system. We employed a technique of recognition and tracking of multi-objects, called “You Only Look Once: YOLO.” This method provides us very fast and accurate tracker. In our system, we remove the haze, which is caused by turbidity of water, from image. After its process, we apply “YOLO” to tracking and recognizing the marine organisms, which includes shrimp, squid, crab, and shark. Our developed system shows generally satisfactory performance.

DOI： 10.1007/978-3-030-17763-8_6

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85097568303&origin=inward

記述言語：日本語   掲載種別：記事・総説・解説・論説等（商業誌、新聞、ウェブメディア）

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2019 Elsevier B.V. With the global popularity of Internet of Things (IoT) technology, increasingly numbers of digital mobile products have been developed, and they have increased the productivity of people's daily lives. These electronic products are used in all aspects of life, such as medical care, office life, home services, and sports. However, most of these products are designed for healthy people with high literacy rates. For disabled people, these products cannot be widely used. In this paper, new, differently shaped touch sensors are proposed for body sensor network-based devices. This touch sensor can be formed into any shape because of the use of conductive fabric adhesive tape as a switch. That property is why the sensor can change positions in the body sensor network in which the human body is used as a trigger to safely activate the touch switch. The number of switch sensors can easily be increased or decreased without changing the wiring of the central controller. The number of sensors in a switch sensor system is greater than that in other touch switch systems, and the accuracy is higher.

DOI： 10.1016/j.comcom.2019.07.019

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85069818135&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2019 Institute of Control, Robotics and Systems - ICROS. In recent years, liver cancer has become the fourth-largest number of deaths in the world. Surgery is a typical treatment for liver cancer. Therefore, advance information about the number and size of cancer is important for surgery. Multi-phase CT images are well known diagnostic method. By extracting the region of the liver and the region of cancer from the obtained CT image, the shape can be finally restored in 3D. In this paper, as a preliminary step to construct an image analysis method for efficiently extracting cancerous regions in multi-phase CT, we propose a method of obtaining a rectangular region as a rough cancerous region of interest. As a method, after preprocessing the input image, using Faster R-CNN, the region of interest including the cancer region is extracted as a rectangle. As a result of applying this method to 11 cases of arterial phase of multi-phase CT, the detection performance was different depending on the network model adopted for backbone part.

DOI： 10.23919/ICCAS47443.2019.8971627

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85079102007&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2019 Institute of Control, Robotics and Systems - ICROS. Auscultation of respiratory sounds is very important for discovering the respiratory disease. However, there is no quantitative evaluation method for the diagnosis of respiratory sounds until now. It is necessary to develop a system to support the diagnosis of respiratory sounds. In addition, there are few studies using dataset suitable for generating realistic classification models that can be used in clinical sites in algorithm development for automatic analysis of respiratory sounds. We describe the development of an algorithm for the automatic classification of the large-scale respiratory sound dataset used in ICBHI 2017 Challenge as containing crackles, containing wheeze, containing both, and normal. Our approach consists of two major components. Firstly, transformation of one-dimensional signals into two-dimensional time-frequency representation images using short-time Fourier transform and continuous wavelet transform. Secondly, classification of transferred images using convolutional neural networks. In this paper, we apply our proposed method to 920 respiratory sound data, and achieve score of 28[%], harmonic score of 81[%], sensitivity of 54[%] and specificity of 42[%].

DOI： 10.23919/ICCAS47443.2019.8971689

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85079096950&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2019 Institute of Control, Robotics and Systems - ICROS. Bone diseases include rheumatoid arthritis and osteoporosis. Although visual screening using computed radiography (CR) images is an effective method for diagnosing osteoporosis, there are some similar diseases that exhibit low bone mass status. To this end, we aim to develop a computer-aided diagnostic (CAD) system to support the automatic diagnosis of osteoporosis from CR images. In this paper, we use convolutional neural network (CNN) and multiscale gradient vector flow snakes (MSGVF Snakes) algorithms to segment each finger bone regions from the CR image. The proposed method is applied to 15 cases, 92.95 [%] of the true positive rates, 2.21 [%] of the false positive rates, 7.05 [%] of the false negative rates are obtained respectively.

DOI： 10.23919/ICCAS47443.2019.8971740

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85079102296&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2019 Institute of Control, Robotics and Systems - ICROS. Circulating Tumor Cells (CTC) is expected as a useful biomarker test that can evaluate cancer metastasis. CTC exists in the blood of cancer patients and is considered to be an incentive of cancer metastasis. Pathologists analyze the blood to find these metastasis cancers from three colors of fluorescence microscopy images, but the manual analysis is time-consuming. In this paper, we develop an automatic CTC classification method in fluorescence microscopy images to reduce the burden of pathologists. In the proposed method, we detect cell regions by the bacterial foraging-based edge detection (BFED) algorithm and classify CTC by SqueezeNet, which is the kind of convolutional neural network (CNN). We apply the proposed method to 5040 microscopy images (6 samples) and evaluate the effectiveness. The experimental results demonstrate that the proposed method has a true positive rate is 89.86% and a false positive rate is 3.27%.

DOI： 10.23919/ICCAS47443.2019.8971646

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85079092431&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2019 Institute of Control, Robotics and Systems - ICROS. Cleft lip is a birth defect that occurs when the lips are not completely formed during healing of the face of the fetus. In Japan, it occurs in about 1 in 500 people. Although treatment is performed to form a symmetrical outer nose, it is necessary to evaluate the degree of symmetry of the surgical site quantitatively because it is based on the judgment criteria that depend on the doctor's subjectivity. In this paper, we propose a method to detect the plane which is the symmetry basis of the face to analyze the symmetry of the operation site. In the proposed method, the face organ is detected as points from 3D point cloud of the face. Then, the mirror image inversion excluding the area affected by shape change due to cleft lip is aligned with the original point cloud. Next, a symmetric plane is set by finding a plane that bisects vertically between one point in the original point cloud and the corresponding point in the mirror image. As a result of applying the proposed method to real 3D point cloud, we could detect the symmetric plane with good accuracy.

DOI： 10.23919/ICCAS47443.2019.8971537

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85079096298&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2019 Institute of Control, Robotics and Systems - ICROS. In recent years, the number of death due to lung cancer is increasing year by year worldwide. Early detection and early treatment of lung cancer are important. Especially, early detection of the abnormalities on thoracic MDCT images detection of small nodules is required in visual screening. Although a CT apparatus is used for the examination, the burden on the image interpretation doctor is large due to the high performance of the CT, so the diagnostic accuracy may be reduced. In this paper, we propose an image analysis method to detect abnormal shadows from chest CT images automatically. The initial lesion candidate areas are extracted by using temporal subtraction technique that emphasizes temporal change by subtracting from a current image to previous one which is obtained same subject. The image of the area is given as input and classification is performed by CNN (Convolutional Neural Network). In the discrimination experiment based on our proposed method, 90.26 [%] of true positive rates and 13.58 [%] of false positive rates are obtained from the 49 clinical cases.

DOI： 10.23919/ICCAS47443.2019.8971533

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85079099157&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2019 Institute of Control, Robotics and Systems - ICROS. At present, the aging population is growing in Japan. Along with that, the expectation for the utilization of welfare equipment is increasing. Electric wheelchair, a convenient transportation tool, is popularized rapidly. On the other hand, accidents have occurred, and the dangers for driving are pointed out. Therefore, it needs to improve accident factors, reduce accidents and improve the convenience of electric wheelchair by automation. Environmental recognition is necessary for the development of autonomous electric wheelchair. Environmental recognition includes self-position estimation, recognition of sidewalks, crosswalks and traffic lights, moving object prediction, etc. In order to solve these various problems, this paper examines the segmentation of sidewalks, crosswalks and traffic lights. We develop the WideSeg that is one of semantic segmentation algorithms applying convolutional neural networks (CNN).

DOI： 10.23919/ICCAS47443.2019.8971608

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85079103064&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2019 Institute of Control, Robotics and Systems - ICROS. Segmentation of the bone metastasis area in medical images can reduce the workload for diagnosis and treatment. However, there are various shapes and sizes of bone metastasis also affected by noise. As a result, it is difficult to segment using classical segmentation methods. In this paper, we propose a convolutional neural network model-based segmentation method. The proposed method easily predicts the contour and location of the lesion area using side connection and modified network. In this study, we modified again the modified HED network to match the characteristics of bone metastasis. The experimental results using the proposed method for segmenting bone metastasis in the lesion area has 79.8[%] of TP rate and 69.2[%] of IOU rate.

DOI： 10.23919/ICCAS47443.2019.8971539

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85079084833&origin=inward

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2019 Elsevier Ltd An image diagnosis by deep learning was applied to failure analysis of power devices. A series of images during a process to failure by power cycling test was used for this method. The images were obtained by a scanning acoustic microscopy of our real-time monitoring system. An image classifier was designed based on a convolutional neural network (CNNs). A developed classifier successfully diagnosed input image into a normal device and an abnormal device. The accuracy of classification was improved by introducing a pre-training and an overlapping pooling into the system. A technique to extract a feature related a failure is essential for the failure analysis based on the real-time monitoring and the deep learning is one likely candidate for it.

DOI： 10.1016/j.microrel.2019.113399

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85074484253&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

中国   成都   2019年08月13日  -  2019年08月15日

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2019 Association for Computing Machinery. An automatic multi-organ segmentation method from upper abdominal CT image is proposed in this paper. A group of statistical shape models for multiple organs are generated by learning the statistical distribution of organs' shapes and intensity profiles. Then, a random forest regression model is trained to find the candidate position to initialize the statistical shape model. The proposed method is evaluated at segmentation of four abdomen organs (spleen, right kidney, left kidney and liver) from training set of 26 cases of upper abdominal CT images. The accuracy shows that the initialization improves the accuracy for statistical shape model-based segmentation.

DOI： 10.1145/3354031.3354042

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85073475971&origin=inward

担当区分：筆頭著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85073475807&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

DOI： 10.1007/978-3-030-46032-7

記述言語：英語   掲載種別：研究論文（学術雑誌）

OBJECTIVE: To visualize and quantitatively analyze facial surface asymmetry following primary cleft lip repair in patients with unilateral cleft lip and palate (UCLP) and to compare this with noncleft controls. DESIGN: Retrospective comparative study. PATIENTS: Twenty-two patients with complete UCLP who underwent primary lip repair from 2009 to 2013 were enrolled in this study. The preserved 3-dimensional (3D) data of 23 healthy Japanese participants with the same age were used as controls. INTERVENTIONS: All patients had received primary labioplasty in accordance with Cronin triangular flap method with orbicular oris muscle reconstruction. MAIN OUTCOME MEASURES: Shadow and zebra images established from moiré images, which were reconstructed from 3D facial data using stereophotogrammetry, were bisected and reversed by the symmetry axes (the middle line of the face). The discrepancies of the gravity and density between cleft and noncleft sides in 2 regions of interest, facial and lip areas, were then calculated and compared with those of healthy participants. RESULTS: In the UCLP group, the mean discrepancies of gravity on shadow and zebra images were 1.76 ± 0.70 and 2.63 ± 1.72 pixels, respectively, in the facial area and 1.31 ± 0.36 and 3.83 ± 2.08 pixels, respectively, in the lip area. There was a significant difference in the mean discrepancies of gravity and density on zebra images in the lip area between the UCLP and control groups. CONCLUSIONS: Our image analysis of digital facial surface asymmetry in patients with UCLP provides visual and quantitative information, and it may contribute to improvements in muscle reconstruction on cleft lip repair.

DOI： 10.1177/1055665618819645

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85069287671&origin=inward

記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

<p>A temporal subtraction image is obtained by subtracting a previous image, which are warped to match between the structures of the previous image and one of a current image, from the current image. The temporal subtraction technique removes normal structures and enhances interval changes such as new lesions and changes of existing abnormalities from a medical image. However, many artifacts remain on a temporal subtraction image and these can be detected as false positives on the subtraction images. In this paper, we propose a 3D-CNN after initial nodule candidates are detected using temporal subtraction technique. To compare the proposed 3D-CNN, we used 7 model architectures, which are 3D ShallowNet, 3D-AlexNet, 3D-VGG11, 3D-VGG13, 3D-ResNet8, 3D-ResNet20, 3D-ResNet32, with these performance on 28 thoracic MDCT cases including 28 small-sized lung nodules. The higher performance is showed on 3D-AlexNet.</p>

DOI： 10.11318/mii.36.77

CiNii Article

CiNii Research

その他リンク： https://search.jamas.or.jp/link/ui/2019298826

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2019 IEEE. The scientific and technological revolution of the Internet of Things has begun in the area of oceanography. Historically, humans have observed the ocean from an external viewpoint in order to study it. In recent years, however, changes have occurred in the ocean, and laboratories have been built on the sea floor. Approximately 70.8 percent of the Earth's surface is covered by oceans and rivers. The Ocean of Things is expected to be important for disaster prevention, ocean resource exploration, and underwater environmental monitoring. Unlike traditional wireless sensor networks, the Ocean Network has its own unique features, such as low reliability and narrow bandwidth. These features will be great challenges for the Ocean Network. Furthermore, the integration of the ocean network with artificial intelligence has become a topic of increasing interest for oceanology researchers. The cognitive ocean network (CONet) will become the mainstream of future ocean science and engineering developments. In this article, we define the CONet. The contributions of the article are as follows: a CONet architecture is proposed and described in detail; important and useful demonstration applications of the CONet are proposed; and future trends in CONet research are presented.

DOI： 10.1109/MWC.2019.1800325

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85061026040&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

France  

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

© 1986-2012 IEEE. As it combines AI and IoT, autonomous driving has attracted a great deal of attention from both academia and industry because of its benefits to the economy and society. However, ultra-low delay and ultra-high reliability cannot be guaranteed by individual autonomous vehicles with limited intelligence and the existing architectures of the Internet of Vehicles. In this article, based on a cloud/fog-computing pattern and the IoT AI service framework, we propose a cross-domain solution for auto-driving. In contrast to existing studies, which mainly focus on communication technologies, our solution achieves intelligent and flexible autonomous driving task processing and enhances transportation performance with the help of the Cognitive Internet of Vehicles. We first present an overview of the enabling technology and the architecture of the Cognitive Internet of Vehicles for autonomous driving. Then we discuss the autonomous driving Cognitive Internet of Vehicles specifically from the perspectives of what to compute, where to compute, and how to compute. Simulations are then conducted to prove the effect of the Cognitive Internet of Vehicles for autonomous driving. Our study explores the research value and opportunities of the Cognitive Internet of Vehicles in autonomous driving.

DOI： 10.1109/MNET.2019.1800339

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85068858755&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）

記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2018 Elsevier B.V. Currently, the aging population is growing in Japan, and the needs for the utilization of welfare equipment are consequently increasing. The electric wheelchair, which is a convenient transportation tool, has rapidly become popular. However, many accidents have occurred when using electric wheelchairs, and the dangers of driving have been noted. Therefore, there is a need to improve accident factors, reduce accidents and improve the convenience of electric wheelchairs by using automation. Environmental recognition is the key technology for developing autonomous electric wheelchairs. Environmental recognition includes self-position estimation, the recognition of sidewalks, crosswalks and traffic lights, and moving object predictions. To solve these problems, this paper develops a system for detecting sidewalks, crosswalks and traffic lights. We develop the object recognition methods using a modified YOLOv2, which is an object detection algorithm that applies convolutional neural networks (CNNs). We detect the object through YOLOv2 and perform processing steps, such as unnecessary bounding box deletion and interpolation. The experimental results demonstrate that the average AUC of the detection rate is 0.587.

DOI： 10.1016/j.future.2018.09.068

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85054855462&origin=inward

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2017 Elsevier Ltd Automatic segmentation of organs from medical images is indispensable for the computer-assisted medical applications. Statistical Shape Models (SSMs) based scheme has been developed as an accurate and robust approach for extraction of anatomical structures, in which a crucial step is the need to place the sampled points (landmarks) with well corresponding across the whole training set. On the one hand, the correspondence of landmarks is related the quality of shape model. On the other hand, in clinical application some key positions of landmarks should be specified by physicians referring to the anatomic structure. In this paper, we develop an interactive method to build SSM that the landmark distribution can be modified manually without influencing the model quality. We extend an existing remeshing method to produce a model prototype in advance and surface features driven registration to insure the universal optimization of correspondence. The key landmarks are fixed during the prototype generation. We experimented and evaluated the proposed SSM method for lung regions, the deformations of which are considerable large.

DOI： 10.1016/j.optlastec.2017.09.018

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85030029527&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2017 Elsevier Ltd In recent years, underwater image processing has been a focus of many studies. Most underwater image processing technologies are focused on descattering, absorption correction and reflection correction. For a deep sea mining machine, dust seriously affects visual acuity. To correct for this, a two-part dust removal approach is proposed. Underwater red-green minimum channel prior descattering is used to remove fine dust in the first stage. However, the impact of dust streaks on images is always undesirable. Consequently, we propose a further deep convolutional neural-network-based dust removal method. The experimental results show that the proposed method performs better in removing haze-like scatters and dust-like scatters.

DOI： 10.1016/j.optlastec.2017.09.017

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85029658908&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2013 IEEE. To monitor the sedimentary process and morphological evolution in the South China Sea, free-ascending deep-sea tripod (FDT) has been developed. This FDT was equipped with a deep-sea camera and landed on the sea floor at a depth of 2100 m. Although the FDT was equipped with an artificial light, the battery capacity limited the duration and intensity of light. Therefore, enhancing such low-illumination images to obtain clear visual effects is an important advancement for analyzing the geological evolution process. In this paper, an adaptive bright-color channel-based low-light underwater image-enhancement method and a denoising method are proposed to enhance such images and remove noise and artifacts. The experimental results demonstrated that the proposed method outperformed state-of-the-art methods.

DOI： 10.1109/ACCESS.2019.2897691

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85064689140&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2019 SPIE. We developed a novel ensemble three-dimensional residual network (E3D-ResNet) for the reduction of false positives (FPs) in computer-aided detection (CADe) of polyps on CT colonography (CTC). To capture the volumetric multiscale information of CTC images, each polyp candidate was represented with three different sizes of volumes of interest (VOIs), which were enlarged to a common size and were individually subjected to three 3D-ResNets. These 3D-ResNets were trained to calculate three polyp-likelihood probabilities, p1, p2 and p3, corresponding to each input VOI. The final polyp likelihood, p, was obtained as the maximum of p1, p2 and p3. We compared the classification performance of the E3D-ResNet with that of a non-ensemble 3D-ResNet, ensemble 2D-ResNet, and ensemble of 2D-and 3D-convolutional neural network (CNN) models. All models were trained and evaluated with 21,021 VOIs of polyps and 19,557 VOIs of FPs that were sampled with data augmentation from the CADe detections on the CTC data of 20 patients. We evaluated the classification performance of the models with receiver operating characteristics (ROC) analysis using cross-validation, where the area under the ROC curve (AUC) was used as the figure of merit. Preliminary results showed that AUC value (0.98) of the E3D-ResNet was significantly higher than that of the reference models (P < 0.001), indicating that the E3D-ResNet has the potential of substantially reducing the FPs in CADe of polyps on CTC.

DOI： 10.1117/12.2512173

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85068173052&origin=inward

担当区分：最終著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

<p>骨のおもな疾患として，骨粗しょう症が挙げられる．骨粗しょう症に対する画像診断は有効であるが，医師の負担増加や経験差による診断結果のばらつき，病変部の未検出等が懸念されている．そこで本稿では，指骨computed radiography（CR）画像から骨粗しょう症の識別を行い，医師に提示するための診断支援手法を提案する．提案手法では，畳み込みニューラルネットワークの一種である，Residual Network （ResNet）を用いた識別器を構築し，骨粗しょう症有無の識別を行う．ResNetへの入力画像には，CR画像から生成した画像を用いる．本稿では，3種類の入力画像を提案し，各画像で学習および，識別の評価を行う．実験では，101症例に対し提案手法を適用し，receiver operating characteristics（ROC）曲線上のarea under the curve（AUC）値を用いて評価したところ，最大で0.931という結果を得た．</p>

DOI： 10.11409/mit.37.107

CiNii Article

CiNii Research

その他リンク： https://ci.nii.ac.jp/naid/130007633096

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

<p>Osteoporosis is known as one of the main diseases of bone. Although image diagnosis for osteoporosis is effective, there are concerns about increased burden of radiologists associated with diagnostic imaging, uneven diagnostic results due to experience difference, and undetected lesions. Therefore, in this study, we propose a diagnosis supporting method for classifying osteoporosis from phalanges computed radiography images and presenting classification results to physicians. In the proposed method, we construct classifiers using convolution neural network and classify normal cases and abnormal cases about osteoporosis. In our experiments, two kinds of CNN models were constructed using input images generated from 101 cases of CR images and evaluated using Area Under the Curve（AUC）value on Receiver Operating Characteristics（ROC）curve. Finaly, AUC of 0.995 was obtained.</p>

DOI： 10.11318/mii.36.72

CiNii Article

CiNii Research

その他リンク： https://ci.nii.ac.jp/naid/130007668726

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© ICROS. In the task of object recognition, convolutional neural networks (CNNs) have achieved high performance. In addition, these CNNs are also applied to the field of semantic image segmentation. However, applying the classification models to semantic segmentation tasks has a problem, lack of global context and reduction in resolution. In this work, we propose global context module and high resolution path in order to solve above problems. By simply combining them with an existing classification model (wide residual networks), our methods yield high-accuracy segmentation models. Our proposed approaches produce competitive results, the mean intersection over union (IoU) 67.6% and global accuracy 91.1%, on CamVid test set.

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85060445374&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© ICROS. In recent years, the proportion of deaths from cancer tends to increase in Japan, especially the number of deaths from lung cancer is increasing. CT device is effective for early detection of lung cancer. However, there is concern that an increase in burden on doctors will be caused by high performance of CT improving. Therefore, by presenting the “second opinion” by the CAD system, it reduces the burden on the doctor. In this paper, we develop a CAD system for automatic detection of lesion candidate regions such as lung nodules or ground glass opacity (GGO) from 3D CT images. Our proposed method consists of three steps. In the first step, lesion candidate regions are extracted using temporal subtraction technique. In the second step, the image is reconstructed by sparse coding for the extracted region. In the final step, 3D Convolutional Neural Network (3D-CNN) identification using reconstructed images is performed. We applied our method to 51 cases and True Positive rate (TP) of 79.81 % and False Positive rate (FP) of 37.65 % are obtained.

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85060466308&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© ICROS. Recently, visual screening based on CT images become useful tools in the medical fields. However, due to the large number of images and the complexity of the image processing algorithms, image processing technique for the high screening quality is still required. To overcome this problem, some computer aided diagnosis (CAD) algorithms are proposed. Cancer is a leading cause of death both in Japan and worldwide. Detection of cancer region in CT images is the most important task to early detection and early treatment. We have designed and developed a framework combining machine learning based on multi-phase convolutional neural networks (CNN) and temporal subtraction techniques based on non-rigid image registration algorithm. Our main classification method can be built into three main steps; i) preprocessing for image segmentation, ii) image matching for registration, and iii) classification of abnormal regions based on machine learning algorithms. We performed our proposed technique to 25 thoracic MDCT sets and obtained true positive rates of 93.55%, false positive rates of 10.93 /case.

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85060434756&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© ICROS. Recently, consistent container transportation of roads and ships is mainstream of international freight transport. Because of various factors, automation of cargo handling work is required at the container terminal. Various causes are decrement of future labor force population by an increasing trend of container moving amount and declining birthrate and aging population. Therefore, this study presents the relative position of hanger and container measurement technology using Single Shot Multibox Detector (SSD) for the purpose of improvement of cargo handling work efficiency and unmanned container terminal. In the case of undetected by SSD, it will be detected using AKAZE feature. The proposed method is applied to 368 images of container gripping taken by a camera installed in a container crane. As a result, Interaction of Union (IoU) targeted for container gripping is 87.79%, and a detection rate is 94.57%.

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85060451358&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© ICROS. Osteoporosis is one of the famous bone diseases. It is a major cause of deteriorating the quality of life, and early detection and early treatment are becoming socially important. Visual screening using Computed Radiography (CR) images is effective for diagnosis of osteoporosis, but there are problems of increasing the burden on doctors, variation in diagnostic results due to differences in experiences of doctors, and undetected lesions. In order to solve this problem, we are working on a computer-aided diagnosis (CAD) system for osteoporosis. In this paper, we propose segmentation methods of the phalange region from the phalangeal CR images as a preprocessing of classification of osteoporosis. In the proposed method, we construct a segmentation model using U-Net, which is a type of deep convolution neural network (DCNN). The proposed method was applied to input images generated from CR images of 101 patients with both hands, and evaluated using the Intersection over Union (IoU) values. The result was 0.914 in IoU.

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85060468143&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© ICROS. In recent years, the development of the computer-aided diagnosis (CAD) systems to support radiologist is attracting attention in medical research field. One of them is temporal subtraction technique. It is a technique to generate images emphasizing temporal changes in lesions by performing a differential operation between current and previous image of the same subject. In this paper, we propose an image registration method for image registration of current and previous image, to generate temporal subtraction images from CT images and enhanced bone metastasis region. The proposed registration method is composed into three main steps: i) automatic segmentation of the region of interest (ROI) using position information of the spine based on biology, ii) use global image matching to select pairs from previous and current image, and iii) final image matching based on salient region feature. We perform registration technique on synthetic data and confirm usefulness of the proposed method. Furthermore, radiologist conduct comparative experiments without and with temporal subtraction images created by proposed method. As a result, they show high reading performance by using temporal subtraction images.

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85060475609&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© ICROS. Cleft lip is a kind of congenital facial morphological abnormality. In the clinical field of cleft lip, it is necessary to analyze symmetric shape. However, there is no method to analyze the cleft lip technique based on symmetrical viewpoints. On the other hand, in our previous method to find a symmetric axis using a 2D image, since the middle line is extracted only from the front view of the face moire image. There was a problem that low accuracy was obtained by slight rotation of the face and it was not possible to consider 3D information. In this paper, we propose a method to extract the median plane of the face by analyzing based on bilateral symmetry by using 3D point cloud on the face of front. By extracting the median plane, we believe that not only surgical assistance of doctor be possible but also become a clue to development of simulation software which is the end goal.

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85060498535&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© ICROS. In Japan, the number of patients with osteoporosis and rheumatoid arthritis is increasing. Image diagnosis using CR images is effective for osteoporosis and rheumatoid arthritis. Development of a CAD system is important for reducing burdens on doctors. In this paper, we propose an automatic registration algorithm in the CAD system. In the proposed method, the genetic algorithm is used to register bone regions between identical parts of the same subject with different time series. In the experiment, the proposed method is applied to 176 bone area, and 98.14 % of TPR, 1.85 % of FPR are obtained respectively. Even when the area difference is used as the fitness of the genetic algorithm, it has cross-correlation and positioning accuracy equivalent to mutual information.

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85060451078&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© ICROS. In this paper, we propose a registration method for fully automated liver tumor detection. Multiple phases CT is used for the detection of the liver tumor because multiple phase CT can give different characteristic features of lesions for each time phases. Registration accuracy is important when obtaining image features from multiple time phases. However, since each time phases have different image density characteristics, therefore registration of multi-phase CT is a challenging task. In this paper, we propose a robust initial alignment method independent of changing image density features in each time phase, and deformable registration method with region of interests (ROI) as liver region extracted by U-Net. Our proposed method is evaluated on 15 patient image sets. This method is applied to the early arterial phase and the equilibrium phase to registries. Experimental results show that segmentation of early arterial phase is 83% and registration is 93% accuracy.

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85060464031&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© ICROS. In recent years, dysphagia is problem among elderly people. Therefore, it is necessary to accurately evaluate swallowing function in order to prevent swallowing disorder beforehand or to detect it early. And it is considered that evaluation of swallowing function using Magnetic Resonance Imaging (MRI) is useful. In order to accurately analyzing of the swallowing motion using a computer aided diagnosis (CAD) system on MR imaging, automatic extraction of the esophagus region, which is a region of interest by the image analysis method, is required. Extraction of the spinal region is required as a preliminary step of the esophagus region extraction. Therefore, in this paper, we develop an analysis method of swallowing movement by three steps of extraction of spinal region, extraction of esophageal region, and analysis of swallowing movement. As an analytical method of swallowing movement, we emphasize the liquid part at the time of swallowing movement using an emphasis map, then follow the liquid tracing by using the AKAZE feature quantity and the particle filter algorithm, and analyze the swallowing motion.

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85060441189&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2018, Springer Science+Business Media, LLC, part of Springer Nature. In this paper, we propose a method to reduce artifacts on temporal difference images by improving the conventional method using a non-rigid registration method for ground glass opacification (GGO), which is light in concentration and difficult to detect early. In this method, global matching, local matching, and 3D elastic matching are performed on the current image and past image, and an initial temporal difference image is generated. After that, we use an Iris filter, which is the gradient vector concentration degree filter, to determine the initial GGO candidate regions and perform segmentation using SuperVoxel and Graph Cuts in which a superpixel is extended to three dimensions for each region of interest. For each extracted region, a support vector machine (SVM) is used to reduce the over-segmentation. Finally, in the method that greatly reduces artifacts other than the remaining GGO candidate regions, Voxel Matching is applied to generate the final temporal difference image, emphasizing the GGO regions while reducing the artifact. The resulting ratio of artifacts to lung volume is 0.101 with an FWHM of 28.3, which is an improvement over the conventional method and shows the proposed method’s effectiveness.

DOI： 10.1007/s11036-018-1111-2

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85053445997&origin=inward

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Since the declining birthrate and the aging of society, there is concern about the labor shortage in Japan. There is a movement to compensate for the labor shortage by automation of factories by robots. Automation technique is wildly promoted in logistics industry, while there is few studies in objects picking. To solve this issue, we develop an image detection scheme for robotics picking from a video image. It is difficult to recognize and grasp different types of objects in robot vision field. Therefore, in the proposed method, object detection and object recognition method are proposed using Region-based Fully Convolutional Networks that is a type of object detection using deep learning. After detecting the object individually, final target object can select by applying the GrowCut algorithm. As a result, we achieve 0.6773 of the average precision and 0.6395 of Intersection over Union as the segmentation result respecively.

DOI： 10.1109/ICT-ROBOT.2018.8549879

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85060010049&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2018 IEEE. Tumor of spinal cord (Spinal Cord Neoplasms (SCN)) is considered as one of the life threatening diseases that causes death. Early detection of the SCN plays an important role in the management of the lesions. To analyze the treatment, it is necessary to segment the spinal canal based on accurate three-dimensional image processing technique. This paper presents a segmentation algorithm based on 3D region growing for extracting spinal canal from CT images with high accuracy. Intersection over union (IoU) is used to compare the results of segmentation with the manual segmentation results. In the experiment, the proposed method was tested on 3373 CT slices of 10 patients. The proposed method has an average accuracy of 0.7732 and a variance of 0.0061. Satisfactory results have been achieved rapidly, which demonstrates the effectiveness and superiority of the proposed method.

DOI： 10.1109/ICT-ROBOT.2018.8549913

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85060055563&origin=inward

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2018 Elsevier B.V. Purpose: To assess the effectiveness of a CT temporal subtraction (TS) method on radiologists’ performance in sclerotic metastasis detection in the thoracolumbar spine. Materials and methods: 20 pairs (current and previous CTs) of standard-dose CT and their TS images in patients with sclerotic bone metastasis and 20 pairs (current and previous CTs) of those in patients without bone metastasis were used for an observer performance study. A total of 135 lesions were identified as the reference standard of actionable lesions (sclerotic metastasis newly appeared or increased in size or in attenuation). 4 attending radiologists and 4 radiology residents participated in this observer study. Ratings and locations of “lesions” determined by the observers were utilized for assessing the statistical significance of differences between radiologists’ performances without and with the CT-TS images in JAFROC analysis. The statistical significance of differences in the reviewing time was determined by a two-tailed paired t-test. Results: The average figure-of-merit (FOM) values for all but one radiologist increased to a statistically significant degree, from 0.856 without the CT-TS images to 0.884 with the images (P =.037). The average sensitivity for detecting the actionable lesions was improved from 60.7 % to 72.5% at a false-positive rate of 0.15 per case by use of the CT-TS images. The average reading time with CT-TS images was significantly shorter than that without (150.6 s vs. 166.5 s, P =.004). Conclusion: The use of CT-TS would improve the observer performance for the detection of the sclerotic bone metastasis in the thoracolumbar spine.

DOI： 10.1016/j.ejrad.2018.07.017

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85051670298&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2018 Association for Computing Machinery. Statistical shape models (SSMs) is a robust and efficient method in medical image segmentation. In this paper, a novel landmark corresponding method based on k-means clustering and demons registration is proposed to train a 3-D statistical shape model with higher quality. The k-means clustering method is performed on the original geometric surface to obtain a simplified surface as standard set of landmarks to find correspondent landmarks on each mapped spherical surface obtained from demon registration in the training set. Twenty cases of left lung and right lung regions in thoracic MDCT images are used in the experiment to build two SSMs. Performance evaluation results show that SSMs generated by the proposed method achieve better generalization ability and specificity while maintaining the same compactness and accuracy of segmentation as those reported by state-of-the-art methods.

DOI： 10.1145/3277453.3277467

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85060643606&origin=inward

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Korea   Busan  

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2018, © 2018 Informa UK Limited, trading as Taylor & Francis Group. This paper presents a method to recognise actions, which are overlapping and multi-dimensionalities. A spatio-temporal representation is illustrated on local interest points to compute global features. Motion history image (MHI) is computed and motion overwriting the motion overwriting problem of the MHI. The main contribution of this paper is that it demonstrates a higher discriminative ability of various complex actions when compared to the other MHI-based approaches. It selects local interest feature points to capture motion information using Speeded-Up Robust Features (SURF). These key interest points are exploited to compute gradient-based optical flow into four channels. RANSAC is exploited to remove outliers. It incorporates frame-subtracted accumulated image so that we can mask out points that are not required. Afterwards, feature vectors are computed based on moments. Actions are recognised by employing a nearest neighbour classifier and leave-one-out cross-validation partitioning scheme. The proposed method provides satisfactory recognition rates over several other approaches for some challenging actions in outdoor scenes.

DOI： 10.1080/21681163.2017.1298472

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85016247850&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2017, Springer Science+Business Media New York. Convolutional networks are currently the most popular computer vision methods for a wide variety of applications in multimedia research fields. Most recent methods have focused on solving problems with natural images and usually use a training database, such as Imagenet or Openimage, to detect the characteristics of the objects. However, in practical applications, training samples are difficult to acquire. In this study, we develop a powerful approach that can accurately learn marine organisms. The proposed filtering deep convolutional network (FDCNet) classifies deep-sea objects better than state-of-the-art classification methods, such as AlexNet, GoogLeNet, ResNet50, and ResNet101. The classification accuracy of the proposed FDCNet method is 1.8%, 2.9%, 2.0%, and 1.0% better than AlexNet, GooLeNet, ResNet50, and ResNet101, respectively. In addition, we have built the first marine organism database, Kyutech10K, with seven categories (i.e., shrimp, squid, crab, shark, sea urchin, manganese, and sand).

DOI： 10.1007/s11042-017-4585-1

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85015699675&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2018 IEEE. Human wearable helmet is a useful tool for monitoring the status of miners in the mining industry. However, there is little research regarding human emotion recognition in an extreme environment. In this paper, an emotional state evoked paradigm is designed to identify the brain area where the emotion feature is most evident. Next, the correct electrode position is determined for the collection of the negative emotion by the electroencephalograph (EEG) based on the international 10-20 system of electrode placement. And then, a fusion algorithm of the anxiety level is proposed to evaluate the person's mental state using the θ, α, and β rhythms of an EEG. Experiments demonstrate that the position Fp2 is the best electrode position for obtaining the anxiety level parameter. The most visible EEG changes appear within the first two seconds following stimulation. The amplitudes of the θ rhythm increase most significantly in the negative emotional state.

DOI： 10.1109/IWCMC.2018.8450382

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85053937610&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2018 Association for Computing Machinery. Circulating tumor cells (CTC) attract attention as a biomarker that can evaluate cancer metastasis and therapeutic effects. The CTC exists in the blood of cancer patients, so pathologists analyze blood by using a fluorescence microscope. However, manual analysis by pathologists is hard-work since the number of CTC to substances contained in the blood is very few and the cell regions are often unclear depending on shooting environments. In addition, there are few studies on automatic identification of CTC. In this paper, we develop an automatic detection method of cell regions in microscope images based on bacterial foraging-based edge detection (BFED) algorithm to analyze CTC. In the first step, we detect the initial cell regions by BFED algorithm. Second, we identify whether the region is a single cell or multiple cells come in connect with other cell(s) by SVM. Third, when a cell is connected with other one, we separate the connecting cells by branch and bound algorithm and obtain the final cell regions. We applied our proposed method to 1680 microscopy images (6 cases). The experimental results demonstrate that the proposed method has a true positive rate of 93.9% and a false positive 1.29 /case.

DOI： 10.1145/3278229.3278232

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85058561021&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2018 Association for Computing Machinery. At present, the aging population is growing in Japan. Along with that, the need for the utilization of welfare equipment is increasing. Electric wheelchair, a convenient transportation tool, is popularized rapidly. However, many accidents have occurred by using electric wheelchair, and the dangers for driving are pointed out. Therefore, it needs to improve accident factors, reduce accidents and improve the convenience of electric wheelchair by automation. Environmental recognition is the key technology for developing autonomous electric wheelchair. Environmental recognition includes self-position estimation, recognition of sidewalks, crosswalks, traffic lights, and moving object prediction, etc. In order to solve these problems, this paper describes a system for the detection of sidewalks, crosswalks and traffic lights. We develop the object recognition methods using a modified YOLOv2 that is one of object detection algorithms applying convolutional neural networks (CNN). We detect the object through YOLOv2 and perform processing such as unnecessary bounding box deletion and interpolation. The experimental results demonstrate that the area under the curve (AUC) of the detection rate is 0.620.

DOI： 10.1145/3278229.3278231

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85058546828&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85058508395&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2017 IEEE. Unmanned aerial vehicles (UAVs) are used in many fields including weather observation, farming, infrastructure inspection, and monitoring of disaster areas. However, the currently available UAVs are prone to crashing. The goal of this paper is the development of an anomaly detection system to prevent the motor of the drone from operating at abnormal temperatures. In this anomaly detection system, the temperature of the motor is recorded using DS18B20 sensors. Then, using reinforcement learning, the motor is judged to be operating abnormally by a Raspberry Pi processing unit. A specially built user interface allows the activity of the Raspberry Pi to be tracked on a Tablet for observation purposes. The proposed system provides the ability to land a drone when the motor temperature exceeds an automatically generated threshold. The experimental results confirm that the proposed system can safely control the drone using information obtained from temperature sensors attached to the motor.

DOI： 10.1109/JIOT.2017.2737479

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85028968216&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2018 Elsevier B.V. In recent years, unmanned aerial vehicle (UAV) technologies have rapidly developed. Drones, which are one type of UAV, are used in many industrial fields, such as photography, delivery and agriculture. However, a commercial drone can fly for only approximately 20 min on one charge. Furthermore, drones are prohibited from flying in some areas, and cannot be operated in bad weather. Due to the development of drone technologies, we must reduce energy consumption and achieve long-range movement. To overcome these limitations, we develop a new air–land amphibious car drone that can fly and requires less power consumption in land mode; this extends the range of mobility of the drone. Moreover, land mode can be used to pass through restricted areas or bad weather conditions by sliding. Furthermore, we develop a Convolutional Neural Network (CNN)-based algorithm for detecting the road in a captured scene. To more accurately segment the road region based on images from the equipped camera of the drone, we propose atrous spatial pyramid pooling (ASPP) ResNet blocks, instead of Resblocks, which were proposed by DeepLab. The experimental results demonstrate that the proposed method improves the pixel accuracy (PA) to 85.6% and achieves a mean Intersection over Union (mIoU) of 55.8%.

DOI： 10.1016/j.future.2018.02.036

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85043254725&origin=inward

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）

担当区分：筆頭著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2018 Elsevier B.V. Underwater optical images are usually influenced by low lighting, high turbidity scattering and wavelength absorption. To solve these issues, a great deal of work has been performed to improve the quality of underwater images. Most of them use the high-intensity LEDs for lighting to obtain the high contrast images. However, in high turbidity water, high-intensity LEDs cause strong scattering and absorption. In this paper, we propose a light field imaging approach for solving underwater imaging problems in a low-intensity light environment. As a solution, we tackle the problem of de-scattering from light field images by using deep convolutional neural networks with depth estimation. Furthermore, a spectral characteristic-based color correction method is used for recovering the color reduction. Experimental results show the effectiveness of the proposed method by challenging real-world underwater imaging.

DOI： 10.1016/j.future.2018.01.001

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85040510117&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2017, Springer Science+Business Media, LLC. Artificial intelligence (AI) is an important technology that supports daily social life and economic activities. It contributes greatly to the sustainable growth of Japan’s economy and solves various social problems. In recent years, AI has attracted attention as a key for growth in developed countries such as Europe and the United States and developing countries such as China and India. The attention has been focused mainly on developing new artificial intelligence information communication technology (ICT) and robot technology (RT). Although recently developed AI technology certainly excels in extracting certain patterns, there are many limitations. Most ICT models are overly dependent on big data, lack a self-idea function, and are complicated. In this paper, rather than merely developing next-generation artificial intelligence technology, we aim to develop a new concept of general-purpose intelligence cognition technology called “Beyond AI”. Specifically, we plan to develop an intelligent learning model called “Brain Intelligence (BI)” that generates new ideas about events without having experienced them by using artificial life with an imagine function. We will also conduct demonstrations of the developed BI intelligence learning model on automatic driving, precision medical care, and industrial robots.

DOI： 10.1007/s11036-017-0932-8

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85029688332&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

2018年04月01日

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2017, Springer Science+Business Media, LLC. Optical underwater images often demonstrate low contrast, heavy scatter, and color distortion. Contrast enhancement methods have been proposed to solve these issues. However, such methods typically do not consider high-level inhomogeneous scatter removal and do not focus on real-scene color restoration. We proposed a hierarchical transmission fusion method and a color-line ambient light estimation method for image de-scattering from a single input image. Our proposed method can be summarized into three steps. Firstly, we take the dark channel as prior information to estimating the preliminary transmission and ambient light. In the second step, we then use color lines to estimate the refined ambient light in selected patches. The refined transmission is obtained by hierarchical transmission maps using maximum local energy-based fusion at different turbidity levels. We then use a joint normalized filter to obtain the final transmission. Finally, a chromatic color correction method and de-blurring algorithm are used to recover the scene color. Experimental results demonstrate that the accurate estimation of the depth map and ambient light by the proposed method can recover visually appealing images with sharp details.

DOI： 10.1007/s11036-017-0933-7

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85029692278&origin=inward

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2017 IEEE. In this paper, a method is proposed for detecting and tracking a specific moving object (e.g., a bus) on the road from images of a camera attached to the head of a user, aiming at developing a system to support daily lives of visually impaired people. The proposed method traces feature points on the images, extracts a moving object region, and detects a bus by applying Haar-like feature and random trees to the region. The effectiveness of the proposed method is shown experimentally.

DOI： 10.1109/SII.2017.8279323

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85050871522&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

2018年01月18日

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

2018年01月18日

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© Springer International Publishing AG 2018. Underwater optical images are usually influenced by low lighting, high turbidity scattering and wavelength absorption. In order to solve these issues, a great deal of work has bee n used to improve the quality of underwater images. Most of them used the high-intensity LED for lighting to obtain the high contrast images. However, in high turbidity water, high-intensity LED causes strong scattering and absorption. In this paper, we firstly propose a light field imaging approach for solving underwater depth map estimation problems in low-intensity lighting environment. As a solution, we tackle the problem of de-scattering from light field images by using deep convolutional neural fields in depth estimation. Experimental results show the effectiveness of the proposed method through challenging real world underwater imaging.

DOI： 10.1007/978-3-319-69877-9_33

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85033692294&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

© Springer Science+Business Media, LLC 2017. Although radiographic assessment of joint damage is essential in characterizing disease progression and prognosis in patients with rheumatoid arthritis (RA), it is often difficult even for trained radiologists to find radiographic changes on hand and foot radiographs because lesion changes are often subtle. This paper proposes a novel quantitative method for automatically detecting bone erosion on hand radiographs to assist radiologists. First, the proposed method performs with the crude segmentation of phalanges regions from hand radiograph and extracts the detailed phalanges regions by the multiscale gradient vector flow (MSGVF) Snakes method. Subsequently, the region of interest (ROI; 40 × 40 pixels) is automatically set on the contour line of the segmented phalanges by the MSVGF algorithm. Finally, these selected ROIs are identified by the presence or absence of bone erosion using a deep convolutional neural network classifier. This proposed method is applied to the hand radiographs of 30 cases with RA. The true-positive rate and the false-positive rate of the proposed method are 80.5% and 0.84%, respectively. The number of false-positive ROIs is 3.3 per case. We believe that the proposed method is useful for supporting radiologists in imaging diagnosis of RA.

DOI： 10.1007/s11042-017-5449-4

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85056120147&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© Springer International Publishing AG 2018. Automatic segmentation of organs from medical images is indispensable for the applications of computer-aided diagnosis (CAD) and computer-assisted surgery (CAS). Statistical Shape Models (SSMs) based scheme have been proved as the accurate and robust methods for extraction of anatomical structures. A key step of this approach is the need to place the sampled points(landmarks) with correspondence across the training set. On the one hand, the correspondence of landmarks is related the quality of SSMs. On the other hand, in many cases the location of key landmarks should be manipulated by physicians, since an unattended system is hard to use in most clinical applications. In this paper, we establish a dense correspondence across the whole training set automatically by surface features, which are registered using diffeomorphic demons approach. And the optimization is executed on spherical domain. We establish the SSM for lung regions, the deformation of where is greatly. Finally, we derive quantitative measures of model quality and comparison of segmentation results using the model with non-optimized correspondence.

DOI： 10.1007/978-3-319-69877-9_14

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85033680194&origin=inward

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

骨のおもな疾患として，骨粗しょう症が挙げられる．骨粗しょう症に対する画像診断は有効であるが，類似した低骨量を呈する画像も多く，画像診断における客観性や再現性の問題がある．そこで本稿では，指骨computed radiography（CR）画像から骨粗しょう症の自動識別手法を提案する．提案手法では，深層畳み込みニューラルネットワーク（DCNN）を用いた識別器を構築し，骨粗しょう症有無の識別を行う．DCNNの学習および識別には，CR画像から3種類の画像を作成し，各指骨領域内部からROIを抽出後，この3種類のROIをR，G，Bチャンネルに割り当て生成した疑似カラー画像を用いる．実験では，101症例に対し提案手法を適用し，真陽性率（TPR）：75.5［％］，偽陽性率（FPR）：13.9［％］という結果を得た．

DOI： 10.11409/mit.36.90

CiNii Article

CiNii Research

その他リンク： https://ci.nii.ac.jp/naid/130006588793

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Segmentation of liver from Multi-phase CT images is one of the essential technology for computer aided diagnosis. Contrast medium gives multi-phase CT images different intensity feature which enables to detect tumor. It is a challenging problem to segment liver region from multi-phase CT images. There are many approaches for solving this problem, however, these methods depend on other phases or registration. In order to solve this problem, we propose anatomy feature-based method which is mostly independent for each phase in this paper. This method uses level set method for final segmentation. The accuracy of segmentation result by level set methods relay on initial contour, so we preprocess initial region of liver by anatomical feature. Then we introduced contour constrain by using ribs information to improve segmentaion accuracy. Our segmentation was evaluated on 5 multi-phase CT images which have 4 phases. Experimental results show that the proposed method is good accuracy for each phase.

DOI： 10.23919/ICCAS.2017.8204240

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85044443256&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Circulating tumor cells (CTCs) is a useful biomarker for cancer metastasis. The blood from a cancer patient is analyzed by a fluorescence microscope. Each case takes a large number of images, which usually have a lot of cell regions. Thus, analyzing the images is hard work for pathologists, and misdiagnosis may happen. In this paper, we develop an automatic CTCs identification method for fluorescence microscopy images. The proposed method consists of three steps. First, we extract cell regions in images using filtering methods. Second, we compute features of each CTC candidate regions. Finally, we identify the CTCs using AdaBoost algorithm. And we analyze the features to know which ones are effective for characterizing CTCs and normal cells. We apply the proposed method to 5040 microscopy images, and evaluate the effectiveness of our method by using leave-one-out cross validation. We achieve a true positive rate of 97.30 [%] and a false positive rate of 12.82 [%].

DOI： 10.23919/ICCAS.2017.8204219

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85044436784&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2017 Institute of Control, Robotics and Systems - ICROS. In recent years, the development of the computer-aided diagnosis (CAD) systems to support radiologist is attracting attention in medical research field. Temporal subtraction technique, which is one of CAD, is a technique to generate images emphasizing temporal changes in lesions by performing a differential operation between current and previous image of the same subject. In this paper, we propose an image registration method for image registration of current and previous image, to generate temporal subtraction images from CT images and enhanced bone metastasis region. The proposed registration method is composed into three main steps: i) segmentation of the region of interest (ROI) using graph cut, ii) use global image matching to select pairs from previous and current image, and iii) final image matching based on salient region feature. We perform our proposed method to synthesis and satisfactory registration experiments. The rotated synthesis image give TP 100.0[%] and FP 12.16[%] . The synthesis image obtained by applying a Gaussian filter give TP 70.40[%] and FP 0.00[%] . The synthesis image obtained by adding artificial pseudo lesion region give TP 99.45[%] and FP 17.89[%] . The synthesis image obtained by adding random noise of 5[%], which gave TP 83.05[%] and FP 16.95[%].

DOI： 10.23919/ICCAS.2017.8204242

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85044465807&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2017 Institute of Control, Robotics and Systems - ICROS. Cancer is a leading cause of death both in Japan and worldwide. Detection of cancer region in CT images is the most important task to early detection. Recently, visual screening based on CT images become useful tools for cancer detection. However, due to the large number of images and the complexity of the image processing algorithms, image processing technique is still required a high screening quality. To overcome this problem, some computer aided diagnosis (CAD) algorithms are proposed. In this paper, we have designed and developed a framework combining machine learning based on deep convolutional neural networks (DCNN) and temporal subtraction techniques based on non-rigid image registration algorithm. Our main classification method can be built into three main steps; i) pre-processing for image segmentation, ii) image matching for registration, and iii) classification of abnormal regions based on machine learning algorithms. We performed our proposed technique to 25 thoracic MDCT sets and obtained true positive rates of 92.31 [%], false positive rates of 6.32 [/case] were obtained.

DOI： 10.23919/ICCAS.2017.8204218

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85044482182&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2017 Institute of Control, Robotics and Systems - ICROS. The Japanese population has rapidly aged and the number of aged persons who have lower physical ability has increased recently. Thus the development of medical and healthcare devices is expected. Wheelchair requires care support in most cases. Therefore the development of autonomous wheelchair is meaningful since we can expect to improve convenience and to reduce burden of caregivers. The autonomous wheelchair requires several techniques. Our research is to develop a navigation system based on image processing techniques. However, we assume that the system instructs an appropriate direction to head towards the destination when a wheelchair user comes to a crossing. Incidentally, deep learning, a kind of artificial neural network, has attracted attention in the field of machine learning in recent years. This paper proposes methodology for supporting autonomous driving by use of a classifier trained on a video images with deep learning. Also, we apply visual odometry to generate training data.

DOI： 10.23919/ICCAS.2017.8204217

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85044473264&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2017 Institute of Control, Robotics and Systems - ICROS. Lung cancer is the leading cause of death which accounts for the number of deaths in cancer in the world. Early detection and early treatment are regarded as an important. Especially, the ground glass opacity (GGO) is a shadow called pre-cancerous lesion, but it is a shadow which is difficult to detect by a radiologist because of haze and complicated shape. Therefore, in recent years, a computer aided diagnosis (CAD) system has been developed for the purpose of improving the detection accuracy for early detection and reducing the burden to radiologists. In this paper, we extract the GGO using Deep Convolutional Neural Network (DCNN) based on emphasized images. Before detect a GGO region, we apply preprocessing such as isotropic voxel to the original images, and extraction of the lung area. Next, we remove the vessel and bronchial region by 3D line filter based on Hessian matrix, and extract the initial candidate regions using density gradient, volume and sphericity. Subsequently, we segment the candidate regions, extraction of features, and reducing false positive shadows. Finally we create emphasize images and identify with DCNN using those images. As a result of applying the proposed method to 31 cases on Lung Image Database Consortium (LIDC), we obtained a true positive rate (TP) of 86.05 [%] and false positive number (FP) of 4.81 [/case].

DOI： 10.23919/ICCAS.2017.8204464

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85044442253&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2017 Institute of Control, Robotics and Systems - ICROS. Osteoporosis is known as a disease of bone. Visual screening using Computed Radiography (CR) images is an effective method for osteoporosis, however, there are many similar diseases that exhibit state of low bone mass. In this paper, we propose an automatic identification method of osteoporosis from phalanges CR images. In the proposed method, we implement a classifier based on Deep Convolutional Neural Network (DCNN), and identify unknown CR images as normal or abnormal. For training and evaluating of CNN, we use pseudo color images. In the experiment, we apply our proposal method to 101 cases and TPR of 64.7 [%] and FPR of 6.51 [%] were obtained.

DOI： 10.23919/ICCAS.2017.8204241

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85044473373&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2017 Institute of Control, Robotics and Systems - ICROS. In recent years, the death rate caused by lung cancer is increasing. To detect the lung cancer, multi detector-row computed tomography (MDCT) images are used in visual screening. Lung cancer can be easily detected by using the chest MDCT images, however, it has enormous images and burden to radiologists. Research and development of the computer aided diagnosis (CAD) system have been assisted the diagnosis. As one of the CAD technologies, temporal subtraction technique is possible to emphasize the changing interval on the CT images. It uses subtraction operation between previous and current CT images of the same patient. On the other hand, pattern recognition using image reconstruction by sparse coding method has attracted attention. This technique is mathematically modeling the information processing by the primary visual cortex of human. It is the technique for representing images by the linear combination of a small number of basis. In this paper, candidate nodules under 20[mm] were segmented from temporal subtraction images based on the 3D sparse coding technique.3D sparse coding is three dimensional expansion of the sparse coding. Also, we classified the final candidate nodules using support vector machine (SVM) method based on coefficient matrix which are obtained by the 3D sparse coding. We applied proposed method to 31 cases of chest MDCT images in which the number of nodules was more than one. We achieved experimental result with true positive rates (TPR) of 70.2[[%] , and false positive rates (FP) of 34.7[/scan], respectively.

DOI： 10.23919/ICCAS.2017.8204220

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85044482344&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2017 Institute of Control, Robotics and Systems - ICROS. In recent years, dysphagia is problem among elderly people. Therefore, it is necessary to accurately evaluate swallowing function in order to prevent swallowing disorder beforehand or to detect it early, and it is considered that evaluation of swallowing function using MRI is useful, and the demand for development of CAD system using MRI is increasing. In order to accurately analyze the swallowing motion, automatic extraction of the region of interest by the image analysis method is necessary. In this paper, as a pretreatment of a method for automatic extraction of esophageal region on dental MR image, we develop automatic extraction method of spinal candidate region. We first extract the posterior region of the spine and then extract the intervertebral disc region. Then we perform the detection of the anterior region of the spine and finally we extract the region between the anterior region of the spine and the posterior region of the spine as a spinal region. The proposed method was applied to 5 cases of dental MR images, and the results of TP 91.2 [%] and FP 8.8 [%] are obtained.

DOI： 10.23919/ICCAS.2017.8204243

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85044477204&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2017 Institute of Control, Robotics and Systems - ICROS. About 4% of babies in Japan suffer from congenital anomalies. Cleft lip is the most common disorder among external malformations occurring at a rate of 1 in 500 people. The goal of treatment is to recover symmetric and functional lips and nose forms. However, in the case of unilateral cleft lip where the lips and nose themselves are shifted from the midline of the face, it is difficult to set the midline which is the symmetry axis for evaluating the degree of symmetry of the face. In this paper, we propose an image processing method for extracting the midline using zebra image and shadow image which are two dimensional image created based on three dimensional analysis. Also, we evaluate the degree of asymmetry of the face by using the midline as the axis. As a result of applying the proposed method to 25 cases, which indicated that the proposed method is useful.

DOI： 10.23919/ICCAS.2017.8204239

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85044463365&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2017, Springer Science+Business Media New York. Underwater cameras are widely used to observe the sea floor. They are usually included in autonomous underwater vehicles (AUVs), unmanned underwater vehicles (UUVs), and in situ ocean sensor networks. Despite being an important sensor for monitoring underwater scenes, there exist many issues with recent underwater camera sensors. Because of light’s transportation characteristics in water and the biological activity at the sea floor, the acquired underwater images often suffer from scatters and large amounts of noise. Over the last five years, many methods have been proposed to overcome traditional underwater imaging problems. This paper aims to review the state-of-the-art techniques in underwater image processing by highlighting the contributions and challenges presented in over 40 papers. We present an overview of various underwater image-processing approaches, such as underwater image de-scattering, underwater image color restoration, and underwater image quality assessments. Finally, we summarize the future trends and challenges in designing and processing underwater imaging sensors.

DOI： 10.1007/s11036-017-0863-4

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85018366927&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Japan   Kitakyushu   2017年11月25日

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

2017年11月25日

記述言語：英語   掲載種別：研究論文（学術雑誌）

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

2017年11月25日

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）

2017年11月20日

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2016 Institute of Control, Robotics and Systems - ICROS.Circulating tumor cells (CTCs) is an informative biomarker which assists pathologists in early diagnosis and evaluating therapeutic effects of patients with malignant tumors. The blood from a cancer patient is analyzed by a microscope and a large number of pictures including many cells are generated for each case. Thus, analyzing them is time-consuming work for pathologists, and misdiagnosis may happen since the diagnosis of CTCs tends to depend on the individual skill of pathologist. In this paper, we propose a method which detects cell candidate regions in microscopy images automatically to make quantitative analysis possible by computer. Our proposed method consists of three steps. In the first step, we extract initial cell candidate regions in microscopy images based on the saliency map. In the second step, we choose non-single cell regions from the initial candidates based on the SVM algorithm. In the third step, we separate connected regions into single cell regions based on the branch and bound algorithm. We demonstrated the effectiveness of our proposed method using 540 microscopy images and we achieved a true positive rate of 99.04[%] and a false positive rate of 3.95[%].

DOI： 10.1109/ICCAS.2016.7832397

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85014007441&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2016 Institute of Control, Robotics and Systems - ICROS.In recent years, development of the computer-aided diagnosis (CAD) systems for the purpose of reducing the false positive on visual screening and improving accuracy of lesion detection has been advanced. Lung cancer is the leading cause of cancer death in the world. Among them, GGO (Ground Glass Opacity) that exhibited early in the before cancer lesion and carcinoma in situ shows a pale concentration, have been concerned about the possibility of undetected on the screening. In this paper, we propose an automatic extraction method of GGO candidate regions from the chest CT image. Our proposed image processing algorithms is consist of four main steps; 1) segmentation of volume of interest from the chest CT image and removing the blood vessel regions, bronchus regions based on 3D line filter, 2) first detection of GGO regions based on density and gradient which is selected the initial GGO candidate regions, 3) identification of the final GGO candidate regions based on DCNN (Deep Convolutional Neural Network) algorithms. Finally, we calculates the statistical features for reducing the false-positive (FP) shadow by the rule-based method, performs identification of the final GGO candidate regions by SVM (Support Vector Machine). Our proposed method performed on to the 31 cases of the LIDC (Lung Image Database Consortium) database, and final identification performance of TP: 93.02[%], FP: 128.52[/case] are obtained respectively.

DOI： 10.1109/ICCAS.2016.7832398

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85014043688&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

韓国   2017年10月18日

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

韓国  

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

韓国  

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Korea  

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Korea  

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Korea  

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Korea  

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Korea  

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2017, CARS. Purpose: A temporal subtraction (TS) image is obtained by subtracting a previous image, which is warped to match the structures of the previous image and the related current image. The TS technique removes normal structures and enhances interval changes such as new lesions and substitutes in existing abnormalities from a medical image. However, many artifacts remaining on the TS image can be detected as false positives. Method: This paper presents a novel automatic segmentation of lung nodules using the Watershed method, multiscale gradient vector flow snakes and a detection method using the extracted features and classifiers for small lung nodules (20 mm or less). Result: Using the proposed method, we conduct an experiment on 30 thoracic multiple-detector computed tomography cases including 31 small lung nodules. Conclusion: The experimental results indicate the efficiency of our segmentation method.

DOI： 10.1007/s11548-017-1598-1

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85019139416&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

Japan   Kitakyshu   2017年08月23日

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2017 Association for Computing Machinery. Computer-aided diagnosis (CAD) system can assistant radiologists to diagnose bone metastasis, which not only reduces burden on workload but also improves diagnostic accuracy. As key step in CAD system, vertebral segmentation can directly affect diagnostic results. In order to obtain high accurate segmentation results, we propose a connected component Labeled Graph Cuts (LGC) algorithm. The proposed method is tested on 100 computed tomography (CT) slices. The assessed quantitatively of experimental results is compared with those by radiologist. The proposed method has a 96.72[%] of True Positive Rate (TPR), and 1.84[%] of False Positive Rate (FPR), which have better performance than conventional Graph Cuts algorithm, 90.07[%] of TPR and 2.32[%] of FPR.

DOI： 10.1145/3133793.3133795

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85052716817&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2017 Association for Computing Machinery. Statistical shape modelling is an efficient and robust method for medical image segmentation in computer-aided diagnosis. The key step in building a statistical shape model is to find corresponding landmarks in each instance of a training set. In this paper, a novel landmark correspondence estimation method that uses edge collapse surface simplification and the sphere registration is proposed. All the landmarks are selected and transformed by spherical conformal mapping from the instances of the training set and the associated correspondence are automatically found on the spheres. We applied our method on 21 cases of 3-D right lung shapes. The results of image segmentation experiment indicate that our method has a positive influence on the accuracy of segmentation result.

DOI： 10.1145/3133793.3133796

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85052679703&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2017 Association for Computing Machinery. Circulating tumor cells (CTCs) are a useful biomarker since they may have some information about cancer metastasis. The blood from cancer patient is analyzed by a fluorescence microscope. It takes a large number of photos for each case, and many cells are contained in the microscopy images. Thus, analyzing them is hard work for pathologists. This work tends to depend on the individual skill of pathologist so misdiagnosis may be happen. In this paper, we develop an automatic CTCs identification method in fluorescence microscopy images based on artificial neural network. We applied our proposed method to 5040 microscopy images (6 cases), and evaluated the effectiveness of our method by using leave-one-out cross validation. We achieve a true positive rate of 98.65 [%] and a false positive rate of 18.24 [%].

DOI： 10.1145/3133793.3133798

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85052677091&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

日本   北九州   2017年08月23日

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

日本   北九州   2017年08月23日

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

日本   北九州   2017年08月23日

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）

日本   北九州   2017年07月02日

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2017, ISAROB.Detection of objects from a video is one of the basic issues in computer vision study. It is obvious that moving objects detection is particularly important, since they are those to which one should pay attention in walking, running, or driving a car. This paper proposes a method of detecting moving objects from a video as foreground objects by inferring backgrounds frame by frame. The proposed method can cope with various changes of a scene including large dynamical change of a scene in a video taken by a stationary/moving camera. Experimental results show satisfactory performance of the proposed method.

DOI： 10.1007/s10015-016-0347-9

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85009833761&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2016 IEEE.The construction of statistical shape model (SSM) is an important research topic in medical imaging benefited from its robust and nature represent of anatomical structures. Place-march of corresponding landmarks is one of the major factors influencing 3D SSM quality. In this paper, we present a supervised correspondence method for fast building SSM, which includes two main steps, i.e., surface data alignment and landmarks specified based on surface parameterization. The framework is validated with statistical models of the liver constructed from contrast CT images. The experiment results demonstrate that the generated model is statistical and anatomically meaningful.

DOI： 10.1109/SIPROCESS.2016.7888219

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85018716962&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

© 2016 IEEE.This paper describes a portable system for finger-spelling recognition, employing a PC and a video taken by an ego-camera mounted on the body of a person who does finger-spelling. The system is intended to be a useful tool for an orally impaired person to communicate to anyone at any place by carrying it with him/her. The images of the finger-spelling hand of a user who is carrying the system is captured by an ego-camera. The hand is extracted from arbitrary backgrounds by use of a Gaussian mixture model and skin color evaluation: The trimmed and normalized image of the extracted hand is recognized employing the feature space defined by applying the principal component analysis to the learning data containing 45 finger-spelled Japanese Hiragana letters each with 50 samples and the nearest neighbor method. The on-line performance of the proposed system is experimentally shown.

DOI： 10.1109/TENCON.2016.7848021

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85015432805&origin=inward

担当区分：筆頭著者   記述言語：日本語   掲載種別：研究論文（研究会，シンポジウム資料等）

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2013 IEEE. Underwater images are degraded due to scatters and absorption, resulting in low contrast and color distortion. In this paper, a novel self-similarity-based method for descattering and super resolution (SR) of underwater images is proposed. The traditional approach of preprocessing the image using a descattering algorithm, followed by application of an SR method, has the limitation that most of the high-frequency information is lost during descattering. Consequently, we propose a novel high turbidity underwater image SR algorithm. We first obtain a high resolution (HR) image of scattered and descattered images by using a self-similarity-based SR algorithm. Next, we apply a convex fusion rule for recovering the final HR image. The super-resolved images have a reasonable noise level after descattering and demonstrate visually more pleasing results than conventional approaches. Furthermore, numerical metrics demonstrate that the proposed algorithm shows a consistent improvement and that edges are significantly enhanced.

DOI： 10.1109/ACCESS.2017.2648845

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85018527963&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2017 IEEE. Underwater optical cameras are widely used for security monitoring in ocean, such as earthquake prediction and tsunami alarming. Optical cameras recognize objects for autonomous underwater vehicles and provide security protection for sea-floor networks. However, there are many issues for underwater optical imaging, such as forward and backward scattering, light absorption, and sea snow. Many underwater image processing techniques have been proposed to overcome these issues. Among these techniques, the depth map gives important information for many applications of the post-processing. In this paper, we propose a Kinect-based underwater depth map estimation method that uses a captured coarse depth map by Kinect with the loss of depth information. To overcome the drawbacks of low accuracy of coarse depth maps, we propose a corresponding reconstruction architecture that uses the underwater dual channels prior dehazing model, weighted enhanced image mode filtering, and inpainting. Our proposed method considers the influence of mud sediments in water and performs better than the traditional methods. The experimental results demonstrated that, after inpainting, dehazing, and interpolation, our proposed method can create high-accuracy depth maps.

DOI： 10.1109/ACCESS.2017.2690455

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85028360504&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

© MYU K.K. Smart touch sensors are used in many applications, such as the iPhone and some smart home systems. Recent touch sensors perform well in consumer electrical devices; however, there are some drawbacks. For example, most touch sensors have low accuracy for detecting human movements. Many require significant power and have a fixed shape. We propose a flexible, highly accurate, and energy-conserving touch sensor. Our primary contributions are as follows: (1) an energy-conserving touch sensor is developed and tested experimentally; (2) a flexible and arbitrarily shaped touch sensor is designed; and (3) the manufacturing cost is very low and up to 200 touch sensors can be connected to the system. As a result, this energy-conserving touch sensor can be fabricated using common manufacturing processes for consumer electronic devices.

DOI： 10.18494/SAM.2017.1458

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85020809645&origin=inward

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

We developed a screening equipment for ceramic substrate level power module of IGBT. The equipment realizes a new screening test with current distribution. The equipment acquires magnetic field signals over bonding wires and finally classifies to normal/abnormal module automatically. We established statistics based normal/abnormal classification with image processing. It is expected to be applied for screening in a production line and analysis to prevent the failure of power modules.

DOI： 10.23919/ISPSD.2017.7988970

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85028510316&origin=inward

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

<p>Computed tomographic colonography（CTC）, also known as virtual colonoscopy, provides a minimally invasive screening method for early detection of colorectal lesions. It can be used to solve the problems of accuracy, capacity, cost,and safety that have been associated with conventional colorectal screening methods. Computer-aided detection（CADe）has been shown to increase radiologists' sensitivity and to reduce inter-observer variance in detecting colonic polyps in CTC. However, although CADe systems can prompt locations of abnormalities at a higher sensitivity than that of radiologists,they also prompt relatively large numbers of false positives（FPs）. In this study, we developed and evaluated the effect of a transfer-learning deep convolutional neural network（TL-DCNN）on the classification of polyp candidates detected by a CADe system from dual-energy CTC images. A deep convolution neural network（DCNN）that had been pre-trained with millions of natural non-medical images was fine-tuned to identify polyps by use of pseudo-colored images that were generated by assigning axial, coronal, and sagittal images of the polyp candidates to the red, green, and blue channels of the images, respectively. The classification performances of the TL-DCNN and the corresponding non-transfer-learning DCNN were evaluated by use of 5-fold cross validation on 20 clinical CTC cases. The TL-DCNN yielded true- and falsepositive rates of 73.6［％］and 1.79［％］, respectively, which were significantly higher than those of the non-transferlearning DCNN. This preliminary result demonstrates the effectiveness of the TL-DCNN in the classification of polyp candidates from CTC images.</p>

DOI： 10.11318/mii.34.80

CiNii Article

CiNii Research

その他リンク： https://ci.nii.ac.jp/naid/130006846731

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

<p>Lung cancer is one of the most important cancer in the world. Among them, Ground Glass Opacity（GGO）has a hazy area of increased attenuation in the lung image. In recent years, development of a Computer Aided Diagnosis （CAD）system for reducing the burden on work load and improving the detection rate of lesions has been advanced. In this paper, we propose a CAD system to extract GGO from CT images. Firstly, we extract the lung region from the input CT images and remove the vessel, and bronchial region based on 3 D line filter algorithm. After that, we extract initial GGO regions using concentration and gradient information. Next, we calculate the statistical features on the segmented regions. After that, we classify GGO regions using support vector machine（SVM）. Finally, we detect the final GGO regions using deep convolutional neural network（DCNN）. The proposed method is tested on 31 cases of CT images from the Lung Image Database Consortium（LIDC）. The results demonstrate that the proposed method has 86.05[%] of true positive rate and 39.03[/case] of false positive number.</p>

DOI： 10.11318/mii.34.70

CiNii Article

CiNii Research

その他リンク： https://ci.nii.ac.jp/naid/130006846732

担当区分：責任著者   記述言語：日本語   掲載種別：研究論文（学術雑誌）

<p>Circulating tumor cells（CTCs）can be a useful biomarker. They may have some information about the malignant disease, since they are one of causes of the cancer metastasis. The blood sample from cancer patient is analyzed by fluorescence microscope. This microscope takes enlarged images with three types of lights（red, green and blue）,and specific materials are reacted respectively. The blood contains a lot of cells, but there are few CTCs. Therefore analyzing them is not easy work for pathologists. In this study, we develop a method which detects circulating tumor cells in fluorescence microscopy images automatically. Our proposed method has three steps. First, we extract cell regions in microscopy images by using filtering processing. Second, we separate the connecting cell regions into single cell regions,based on the branch and bound algorithm. Finally, we identify CTCs by using logical conjunction method. We demonstrated the effectiveness of our proposed method using 6 cases（5040 microscopy images）, and we evaluated the performance of CTCs identification. Our proposed method achieved, a true positive rate of 95.27 [%] and a false positive rate of 6.172 [%] respectively. And we confirmed the effectiveness of the logical conjinction for CTCs identicication.</p>

DOI： 10.11318/mii.34.151

CiNii Article

CiNii Research

その他リンク： https://ci.nii.ac.jp/naid/130006267703

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（学術雑誌）

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2016 ICIC International.Rheumatoid arthritis and osteoporosis are two of the major diseases related to the phalanges. Diagnostic imaging is often used to diagnosis them. Especially, observing the temporal changes of the shape or internal structure of the phalanges in phalange CR images is an effective way to detect these diseases. However, there are various problems in image diagnosis, such as the evaluation of diagnosis is generally empiric, and the burden of reading images is heavy. In order to solve these problems, in this paper, we develop a computer aided diagnosis (CAD) system for automatically diagnosing, which includes segmentation of knuckles, registration of temporal images and features analysis of phalange regions. In the segmentation part, we perform the multi scale gradient vector flow (MSGVF) snakes to improve the precision of extracted results. Also we have developed an image registration technique based on salient region features (SRF) method. Two image features are used to train the artificial neural network (ANN) to identify the abnormal knuckle regions.

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84991355197&origin=inward

担当区分：責任著者   記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）

中国  

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2016.This paper proposes a method of detecting prominent regions from an image using a global and local saliency measure. Detection of prominent regions from an image is important, because it is expected to improve the precision and the processing speed of object recognition through image processing. The proposed method employs two spatial redundancies, global saliency and local saliency. Global saliency is calculated by comparing the value of an interested pixel with the mode of the values of all pixels. On the other hand, local saliency is calculated using the relation among local pixels. In calculating the local saliency, the proposed method employs human visual characteristics, i.e., complementary color harmony to detect salient regions. Saliency map is made by integrating the global saliency and the local saliency considering some weights. Experimental results show the effectiveness of the proposed method.

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84975702036&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2016 ICIC International. In recent years, the equipment on walk support for visually impaired persons has spread to a certain extent. A studded paving block and a sound-type signal are set on the sidewalk or on the road in many places. However, there are still problems such that the studded paving block is still installed at limited places, or the sound of the signal is obscure because of the environmental noise such as roaring traffic or heavy rain. Therefore, a system that can support visually impaired persons in more effective way is necessary, such as a system which finds a signal or a traffic sign automatically and provides its information to the visually impaired. In order to realize such a system, this paper proposes a method of detecting pedestrian signals and crosswalk signs by a camera and a computer which we call MY VISION system. In the proposed method, color information is used in the first step to restrict the search for traffic signal and sign candidates, and then the HOG feature is introduced to describe the candidates by feature vectors. Recognition of the signals and the signs is performed by applying randomized trees to the candidates. The color sign of the pedestrian signal is also discriminated by using color information. Experimental results are shown and the method is evaluated.

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84959357672&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）

© 2016 ICIC International.There are not a few visually impaired people in our society. When they want to go for a walk, it is difficult for them to perceive moving objects around them. When they walk on the sidewalk, they must pay strong attention to those coming closer to them or to moving objects. Therefore, it is indispensable for them to obtain the information on moving objects around them. This paper proposes a method of extracting moving objects, in particular, pedestrians, from self-wearable camera images. In the proposed system, a user is supposed to wear a camera and a PC and, by performing image analysis of the video taken from the camera, it acquires surrounding pedestrians’ information including their peculiar characteristics. The information is fed back to the user to realize his/her safe walk. Experimental results are shown and the performance of the proposed method is evaluated. The proposed system is primarily intended to be used by visually impaired people, but it may also be used by any pedestrian who is not paying much attention to his/her surroundings.

Scopus

その他リンク： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84959361584&origin=inward

記述言語：英語   掲載種別：研究論文（学術雑誌）