Faculty Profiles - TOKUNAGA Terumasa

Papers - TOKUNAGA Terumasa

Division display1 - 20 of about 50／ All the affair displays >>

Global-Local Consistency for Cost-Effective Anomaly Localization with Point Annotations Reviewed International journal

Nakanishi Keiichi, Katafuchi Ryoya, Tokunaga Terumasa

2024 16th IIAI International Congress on Advanced Applied Informatics (IIAI-AAI) 2024.10

Authorship：Last author,　Corresponding author Language：English Publishing type：Research paper (international conference proceedings)

Creating pixel-level ground-truth (GT) masks is quite costly for deep learning-based image segmentation. Specialists in areas such as anomaly detection and medical diagnostics face difficulties in producing many GT masks due to limited resources. To reduce this burden, we propose a cost-effective image segmentation framework with point annotations (CoSPA) that performs image segmentation with only point annotations. Point annotations refer to the partial and sparse labeled coordinates in an image. The key idea is to ensure consistency between the predictions for the same coordinates from the two different networks of CoSPA. This new consistency enables the CoSPA to improve segmentation performance and extend to semi-supervised learning. For the MVTec AD dataset, we verified the cost-effectiveness of CoSPA through an anomaly detection task. We demonstrated that the point annotating costs were reduced by 80% compared to creating GT masks. Subsequently, the CoSPA realized by 87% of the mean Intersection over Union (mIoU) achieved using the fully supervised method, DeepLabV3+. Moreover, the mIoU of CoSPA using only 30% of all point annotations defeated that of the unsupervised method, PaDiM. This study offers a new direction for economic anomaly localization.

CiNii Research

Other Link： https://kyutech.repo.nii.ac.jp/records/2000992

Bidirectional 2D Reservoir Computing for Image Anomaly Detection without any Training Reviewed International journal

Keiichi Nakanishi, Terumasa Tokunaga

Nonlinear Theory and Its Applications ( The Institute of Electronics, Information and Communication Engineers ) 15 ( 4 ) 838 - 850 2024.10

Mechanism of sensory perception unveiled by simultaneous measurement of membrane voltage and intracellular calcium Reviewed International journal

Terumasa Tokunaga, Noriko Sato, Mary Arai, Takumi Nakamura, Takeshi Ishihara

Communications Biology, ( Nature Portfolio (Springer Nature) ) 7 ( 1150 ) 2024.09

Authorship：Lead author,　Corresponding author Language：English Publishing type：Research paper (scientific journal)

Measuring neuronal activity is important for understanding neuronal function. Ca2+ imaging by genetically encoded calcium indicators (GECIs) is a powerful way to measure neuronal activity. Although it revealed important aspects of neuronal function, measuring the neuronal membrane voltage is important to understand neuronal function as it triggers neuronal activation. Recent progress of genetically encoded voltage indicators (GEVIs) enabled us fast and precise measurement of neuronal membrane voltage. To clarify the relation of the membrane voltage and intracellular Ca2+, we analyzed neuronal activities of olfactory neuron AWA in Caenorhabditis elegans by GCaMP6f (GECI) and paQuasAr3 (GEVI) responding to odorants. We found that the membrane voltage encodes the stimuli change by the timing and the duration by the weak semi-stable depolarization. However, the change of the intracellular Ca2+ encodes the strength of the stimuli. Furthermore, ODR- 3, a G-protein alpha subunit, was shown to be important for stabilizing the membrane voltage. These results suggest that the combination of calcium and voltage imaging provides a deeper understanding of the information in neural circuits.

DOI： 10.1038/s42003-024-06778-2

Scopus

Other Link： https://link.springer.com/article/10.1038/s42003-024-06778-2

Global-Local Consistency for Cost-Effective Anomaly Localization with Point Annotations Reviewed International journal

Keiichi Nakanishi, Ryoya Katafuchi, Terumasa Tokunaga

The proceedings of 16th IIAI International Congress on Advanced Applied Informatics ( IEEE Computer Society ) 53 - 57 2024.07

Authorship：Last author,　Corresponding author Language：English Publishing type：Research paper (international conference proceedings)

日本高松 2024.07.06 - 2024.07.12

DOI： 10.1109/IIAI-AAI63651.2024.00019

Kyutacar

Scopus

Other Link： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85208092849&origin=inward

Layer-wise External Attention by well-localized attention map for efficient deep anomaly detection Invited Reviewed International journal

Keiichi Nakanishi, Ryo Shiroma, Tokihisa Hayakawa, Ryoya Katafuchi, Terumasa Tokunaga

SN Computer Science ( Springer Nature ) 592 ( 5 ) 2024.05

Authorship：Last author,　Corresponding author Language：English Publishing type：Research paper (scientific journal)

The external attention mechanism offers a promising approach to enhance image anomaly detection (Hayakawa et al., in: IMPROVE, pp. 100-–110, 2023). Nevertheless, the effectiveness of this method is contingent upon the judicious selection of an intermediate layer with external attention. In this study, we performed a comprehensive series of experiments to clarify the mechanisms through which external attention improves detection performance. We assessed the performance of the LEA-Net (Hayakawa et al., in: IMPROVE, pp. 100–110, 2023), which implements layer-wise external attention, using MVTec AD and Plant Village datasets. The detection performances of the LEA-Net were compared with that of the baseline model under different anomaly maps generated by three unsupervised approaches. In addition, we investigated the relationship between the detection performance of LEA-Net and the selection of an attention point, which means an intermediate layer where external attention is applied. The findings reveal that the synergy between the dataset and the generated anomaly map influenced the effectiveness of the LEA-Net. For poorly localized anomaly maps, the selection of the attention point becomes a pivotal factor in determining detection efficiency. At shallow attention points, a well-localized attention map successfully notably improves the detection performance. For deeper attention points, the overall intensity of the attention map is essential; this intensity can be substantially amplified by layer-wise external attention, even for a low-intensity anomaly map. Overall, the results suggest that for layer-wise external attention, the positional attributes of anomalies hold greater significance than the overall intensity or visual appearance of the anomaly map.

DOI： 10.1007/s42979-024-02912-3

Kyutacar

Other Link： https://www.scopus.com/record/display.uri?eid=2-s2.0-85194873765&origin=resultslist

Layer-Wise External Attention by Well-Localized Attention Map for Efficient Deep Anomaly Detection Invited Reviewed International journal

Nakanishi Keiichi, Shiroma Ryo, Hayakawa Tokihisa, Katafuchi Ryoya, Tokunaga Terumasa

SN Computer Science 5 ( 5 ) 592 2024.05

Authorship：Last author,　Corresponding author Language：English Publishing type：Research paper (scientific journal)

CiNii Research

Other Link： https://kyutech.repo.nii.ac.jp/records/2000766

Bidirectional 2D reservoir computing for image anomaly detection without any training Reviewed International journal

Nakanishi K., Tokunaga T.

Nonlinear Theory and its Applications, IEICE 15 ( 4 ) 838 - 850 2024.01

Layer-wise External Attention for Efficient Deep Anomaly Detection Reviewed

Tokihisa Hayakawa, Keiichi Nakanishi, Ryoya Katafuchi, Terumasa Tokunaga

Proceedings of the 3rd International Conference on Image Processing and Vision Engineering 100 - 110 2023.01

Behavioral forgetting of olfactory learning is mediated by interneuron-regulated network plasticity in Caenorhabditis elegans Reviewed International journal

Jamine Teo, Itsuki Kurokawa, Yuuki Onishi, Noriko Sato, Tomohiro Kitazono, Terumasa Tokunaga, Manabi Fujiwara, Takeshi Ishihara

eNeuro ( the Society for Neuroscience ) 9 ( 4 ) 2022.08

Language：English Publishing type：Research paper (scientific journal)

Forgetting is important for animals to manage acquired memories to enable adaptation to changing environments; however, the neural network in mechanisms of forgetting is not fully understood. To understand the mechanisms underlying forgetting, we examined olfactory adaptation, a form of associative learning, in Caenorhabditis elegans (C. elegans). The forgetting of diacetyl olfactory adaptation in C. elegans is regulated by secreted signals from AWC sensory neurons via the TIR-1/JNK-1 pathway. These signals cause a decline of the sensory memory trace in AWA neurons where diacetyl is mainly sensed. To further understand the neural network that regulates this forgetting, we investigated the function of interneurons downstream of AWA and AWC neurons. We found that a pair of interneurons, AIA, is indispensable for the proper regulation of behavioral forgetting of diacetyl olfactory adaptation. Loss of or inactivation of AIA caused the impairment of the chemotaxis recovery after adaptation without causing severe chemotaxis defects in naïve animal. AWA Ca2+ imaging analyses suggested that loss or inactivation of AIA interneurons did not affect the decline of the sensory memory trace after the recovery. Furthermore, AIA responses to diacetyl were observed in naïve and after the recovery, but not just after the conditioning, suggesting that AIA responses after the recovery are required for the chemotaxis to diacetyl. We propose that the functional neuronal circuit for attractive chemotaxis to diacetyl is changed temporally at the recovery phase so that AIA interneurons are required for chemotaxis, although AIAs are dispensable for attractive chemotaxis to diacetyl in naïve animals.

DOI： https://doi.org/10.1523/ENEURO.0084-22.2022

Other Link： https://www.eneuro.org/content/early/2022/08/17/ENEURO.0084-22.2022

半教師あり二値分類のためのクラス事前確率を用いたコスト関数の提案 Reviewed

中西慶一, 徳永旭将

第25回画像の認識・理解シンポジウム ( 情報処理学会コンピュータビジョンとイメージメディア(CVIM)研究会 ) 2022.07

Authorship：Last author,　Corresponding author Language：Japanese Publishing type：Research paper (conference, symposium, etc.)

日本姫路 2022.07.25 - 2022.07.28

本稿は半教師あり二値分類のための新しいコスト関数を提案する. 提案するコスト関数は, Focal Loss を半教師あり分類のために拡張した損失関数とクラス事前確率罰則項から構成される. クラス事前確率罰則項を導入することで, ラベルなしデータにおける正例の割合がわかる場合, その事前情報を直接的に推論に活用することが可能である. 提案するコスト関数の有用性を評価するため, CIFAR-10 を用いた画像二値分類試験を行った. その結果, 少量のラベルありデータセットや不均衡なデータセットにおいて, 提案するコスト関数は既存のコスト関数と比べて優れた性能を示すことを確認した.

Other Link： https://sites.google.com/view/miru2022/program#h.44rbhnwbmvlj

Behavioral Forgetting of Olfactory Learning Is Mediated by Interneuron-Regulated Network Plasticity in Caenorhabditis elegans Reviewed

Teo J.H.M., Kurokawa I., Onishi Y., Sato N., Kitazono T., Tokunaga T., Fujiwara M., Ishihara T.

eNeuro 9 ( 4 ) 2022.07

Signal and Noise Separation from Satellite Magnetic Field Data through Independent Component Analysis: Prospect of Magnetic Measurements without Boom and Noise Source Information Reviewed International journal

Shun Imajo, Masahito Nosé, Mari Aida, Haruhisa Matsumoto, Nana Higashio, Terumasa Tokunaga , Ayako Matsuoka

Journal of Geophysical Research: Space Physics ( American Geophysical Union ) 126 ( 5 ) 2021.04

Language：English Publishing type：Research paper (scientific journal)

We propose an application of the independent component analysis (ICA) to separate satellite-induced time-varying stray fields from magnetic field data obtained using onboard multiple magnetometers. The ICA is a method for estimating source signals at multiple sites so that the estimated source signals can become statistically independent of each other. Since stray field variations are statistically independent of external natural field variations, the ICA method is expected to separate the natural variations from stray fields. Thus, we applied the ICA to magnetic field data from the first Quasi-Zenith Satellite, which has two triaxial fluxgate magnetometers, without using an extendable boom. First, we removed the long-period trend from the original data to create detrended data. Then, we applied the FastICA algorithm to the detrended data and obtained six independent components (ICs). The stray fields were successfully separated into three ICs (noise ICs), and the natural signals were represented by the other three ICs (signal ICs). Finally, we restored the observed signals from the signal ICs, and confirmed that the natural phenomena variations were not altered by the processing step. We also proposed a selection method of the noise ICs using the C coefficient, which is the coefficient of the variance of the mixing vectors. There was a large difference in C between the ICs whose C coefficients are the largest 3rd and 4th ones. Overall, these results demonstrate the possibility that the ICA method can support for boom-less magnetic observations in future satellite missions.

DOI： 10.1029/2020JA028790

Scopus

Other Link： https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2020JA028790

Image-based Plant Disease Diagnosis with Unsupervised Anomaly Detection based on Reconstructability of Colors Reviewed International journal

Katafuchi, R. and Tokunaga, T.

In Proceedings of the International Conference on Image Processing and Vision Engineering - IMPROVE ( SciTePress ) 112 - 120 2021.04

Authorship：Last author,　Corresponding author Language：English Publishing type：Research paper (international conference proceedings)

Virtual Virtual 2021.04.28 - 2021.04.30

This paper proposes an unsupervised anomaly detection technique for image-based plant disease diagnosis. The construction of large and publicly available datasets containing labeled images of healthy and diseased crop plants led to growing interest in computer vision techniques for automatic plant disease diagnosis. Although supervised image classifiers based on deep learning can be a powerful tool for plant disease diagnosis, they require a huge amount of labeled data. The data mining technique of anomaly detection includes unsupervised approaches that do not require rare samples for training classifiers. We propose an unsupervised anomaly detection technique for image-based plant disease diagnosis that is based on the reconstructability of colors; a deep encoder-decoder network trained to reconstruct the colors of healthy plant images should fail to reconstruct colors of symptomatic regions. Our proposed method includes a new image-based framework for plant disease detection that u tilizes a conditional adversarial network called pix2pix and a new anomaly score based on CIEDE2000 color difference. Experiments with PlantVillage dataset demonstrated the superiority of our proposed method compared to an existing anomaly detector at identifying diseased crop images in terms of accuracy, interpretability and computational efficiency.

DOI： 10.5220/0010463201120120

Scopus

Other Link： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85115692116&origin=inward

Cohesive and anisotropic vascular endothelial cell motility driving angiogenic morphogenesis Reviewed International journal

Takubo Naoko, Yura Fumitaka, Naemura Kazuaki, Yoshida Ryo, Tokunaga Terumasa, Tokihiro Tetsuji, Kurihara Hiroki

Scientific Reports 9 ( 1 ) 9304-1 - 9304-9 2019.12

Language：English Publishing type：Research paper (scientific journal)

Vascular endothelial cells (ECs) in angiogenesis exhibit inhomogeneous collective migration called “cell mixing”, in which cells change their relative positions by overtaking each other. However, how such complex EC dynamics lead to the formation of highly ordered branching structures remains largely unknown. To uncover hidden laws of integration driving angiogenic morphogenesis, we analyzed EC behaviors in an in vitro angiogenic sprouting assay using mouse aortic explants in combination with mathematical modeling. Time-lapse imaging of sprouts extended from EC sheets around tissue explants showed directional cohesive EC movements with frequent U-turns, which often coupled with tip cell overtaking. Imaging of isolated branches deprived of basal cell sheets revealed a requirement of a constant supply of immigrating cells for ECs to branch forward. Anisotropic attractive forces between neighboring cells passing each other were likely to underlie these EC motility patterns, as evidenced by an experimentally validated mathematical model. These results suggest that cohesive movements with anisotropic cell-to-cell interactions characterize the EC motility, which may drive branch elongation depending on a constant cell supply. The present findings provide novel insights into a cell motility-based understanding of angiogenic morphogenesis.

DOI： 10.1038/s41598-019-45666-2

Other Link： https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85068037379&origin=inward

SPF-CellTracker: Tracking Multiple Cells with Strongly-Correlated Moves Using a Spatial Particle Filter Reviewed International journal

Hirose O., Kawaguchi S., Tokunaga T., Toyoshima Y., Teramoto T., Kuge S., Ishihara T., Iino Y., Yoshida R.

IEEE/ACM Transactions on Computational Biology and Bioinformatics ( Institute of Electrical and Electronics Engineers Inc. ) 15 ( 6 ) 1822 - 1831 2018.11

Phase analysis of whole-brain activity data in C. elegans

Iwasaki Y., Teramoto T., Oe S., Tokunaga T., Hirose O., Wu S., Toyoshima Y., Jang M. S., Yoshida R., Iino Y., Ishihara T.

Meeting Abstracts of the Physical Society of Japan ( The Physical Society of Japan ) 72.1 ( 0 ) 2817 - 2817 2017.01

Accurate Automatic Detection of Densely Distributed Cell Nuclei in 3D Space Reviewed International journal

Toyoshima Y., Tokunaga T., Hirose O., Kanamori M., Teramoto T., Jang M.S., Kuge S., Ishihara T., Yoshida R., Iino Y.

PLoS Computational Biology ( Public Library of Science ) 12 ( 6 ) 2016.06

Automated detection and tracking of many cells by using 4D live-cell imaging data Reviewed

Tokunaga T., Hirose O., Kawaguchi S., Toyoshima Y., Teramoto T., Ikebata H., Kuge S., Ishihara T., Iino Y., Yoshida R.

Bioinformatics ( Oxford University Press ) 30 ( 12 ) 2014.06

3次元動画像内の非常に多数の細胞領域を自動追跡するための粒子フィルタ手法の開発

広瀬修, 川口翔太郎, 徳永旭将, 豊島有, 寺本孝行, 佐藤賢二, 池端久貴, 佐藤博文, 久下小百合, 石原健, 飯野雄一, 吉田亮

人工知能学会全国大会論文集 ( 一般社団法人人工知能学会 ) JSAI2014 ( 0 ) 2C32in - 2C32in 2014.01

Language：Japanese Publishing type：Research paper (conference, symposium, etc.)

<p>細胞動画像の典型的な特徴として多数の追跡対象物が密に存在し視覚的に類似していることが挙げられる．このような動画像に対し標準的な粒子フィルタを利用した場合，本来の追跡対象を別の対象物と誤認し追跡に容易に失敗する．この問題に対し多数の細胞の共変動性に着目し追跡のための補助情報としてこれを利用することで追跡精度の向上を目指す．</p>

DOI： 10.11517/pjsai.jsai2014.0_2c32in

CiNii Article

CiNii Research

Data Assimilation for Reconstructing a Whole Neuronal System of C. Elegans : The Current State and Issue(<Special Feature>Big Data and Data Assimilation) Invited Reviewed

32 ( 4 ) 287 - 294 2013.12

　PREV- NEXT 　

<TOKUNAGA Terumasa>

Papers - TOKUNAGA Terumasa