コロキアムB発表

日時: 9月26日(木)2限(11:00~12:30)


会場: L2

司会: 丁明
時末 尚悟 M, 2回目発表 生体医用画像 佐藤 嘉伸, 小笠原 司, 大竹 義人, スーフィー マーゼン
title: Musculotendinous modeling by using fiber tractography of high-resolution cryosectioned images
abstract: Musculotendinous junctions and ligaments play an essential role in supporting human body by connecting between the bones and skeletal muscles. However, current musculoskeletal models do not take into account the detailed structure and complex geometry of those structures. Therefore, in this research, we aimed at developing a technique for modeling the target structures based on high-resolution cryosectioned three-dimensional images. First, we automatically extracted the region including the tendons or ligaments from the image using a convolutional neural network (CNN). Next, the target structures, i.e. the merged region consisting of the muscle with the tendons or ligaments, were mathematically modeled by estimating the constituent fiber orientations based on structure tensors and the application of a muscle fiber tractography approach. The technique was applied to the region including gluteus maximus muscle in a high-resolution image (voxel size: 0.1 mm) selected from the Visible Korean Human dataset. We additionally clustered the analyzed fibers with respect to the individual anatomical structures. In the future, we will validate the proposed technique on the muscles at different anatomical regions.
language of the presentation: Japanese
 
土井畑 禅 M, 2回目発表 数理情報学 池田 和司☆, 小笠原 司, 浦岡 行治 (MS), 川鍋 一晃(客員), 森本 淳(客員)
title:Neural control strategies underlying human-human joint action
abstract:Social and physical joint actions are characterized by the emergence of naturally occurring mutual cooperative relationships and the role of this cooperative relationship varies depending on various factors, but the mechanism of its appearance is not understood. In previous studies, subjects assigned roles to two subjects and conducted experiments to lift objects in cooperation. As a result, it was confirmed that the subjects designated as leaders made the object more stable, but what determines how to cooperate when no role is given is not understood. Once this mechanism is understood, it can be applied to support human-to-human motor learning and improve human-robot interaction. In this research, in order to understand the generation mechanism of the cooperative relationship that appears by the joint task, we investigated the conditions under which the role occurs using the learnable virtual joint task that forms the cooperation method by the TVINS system.
language of the presentation:Japanese
 
伊藤 佑起 M, 2回目発表 数理情報学 池田 和司☆, 佐藤 嘉伸, 川鍋 一晃(客員), 森本 淳(客員)
title:Resting-state current source estimation using spectral graph theory
abstract:Magnetoencephalography (MEG) is non-invasive imaging techniques which can record brain activity with high temporal resolution. However, the spatial resolution isn't better than functional magnetic resonance imaging (fMRI). Moreover, the problem of current source estimation using MEG generally cannot be solved because it fails in ill-posed problem. It is required some constrains to solve the problem but methods suitable for resting-state have not been proposed yet. In our research, we aim to a good method for spatio-temporally by using the graph Laplacian spectrum of brain structure connectivity. We applied graph Laplacian eigen-mode as our model constraint because oscillations of the brain network at resting-state match synchronic wave patterns of certain frequencies. In this result, our model show a better estimation for kuramoto simulation data than minimum norm estimation model , and our model is also more correlated for the functional connectivity for empirical resting fmri data than MNE. In conclusion, we could confirm that proposed methods worked as we aimed.
language of the presentation: Japanese
 
鈴木 啓大 D, 中間発表 数理情報学 池田 和司☆, 佐藤 嘉伸, 川鍋 一晃(客員), 森本 淳(客員)
title: Relevant meta-analysis data selection for prior information of MEG source estimation
abstract: MEG source estimation is an ill-posed inverse problem. We need constraints to solve this problem. One of the possible constraint is fMRI data. Because of the high spatial resolution of fMRI, we can obtain accurate estimated sources. FMRI data should be measured from the same subject, but if it is not possible, we can synthesize meta-analysis data from large fMRI studies. However, the result of fMRI meta-analysis extremely depends on the interests of analysis (e.g. target studies, tasks, terms). The estimated sources are also considerably biased by meta-analysis data selection. This is the biggest problem to combine MEG and meta-analysis fMRI for scientific means. Therefore, we developed novel meta-analysis data selection method for MEG source estimation. It can take in the distance between different meta-analysis data and represent local sparsity. We demonstrated that our method can construct plausible constraints using simulation and empirical data. Furthermore, the estimated constraints is interpretable due to labels of meta-analysis.
language of the presentation: Japanese