WEKO3
アイテム
{"_buckets": {"deposit": "b1eb82fa-338b-47f5-b086-8c3b3eb5e0cd"}, "_deposit": {"id": "4137", "owners": [], "pid": {"revision_id": 0, "type": "depid", "value": "4137"}, "status": "published"}, "_oai": {"id": "oai:repository.dl.itc.u-tokyo.ac.jp:00004137", "sets": ["280", "329"]}, "item_7_alternative_title_1": {"attribute_name": "その他のタイトル", "attribute_value_mlt": [{"subitem_alternative_title": "脳磁図を用いた運動軌道予測に関する研究"}]}, "item_7_biblio_info_7": {"attribute_name": "書誌情報", "attribute_value_mlt": [{"bibliographicIssueDates": {"bibliographicIssueDate": "2011-03-24", "bibliographicIssueDateType": "Issued"}, "bibliographic_titles": [{}]}]}, "item_7_date_granted_25": {"attribute_name": "学位授与年月日", "attribute_value_mlt": [{"subitem_dategranted": "2011-03-24"}]}, "item_7_degree_grantor_23": {"attribute_name": "学位授与機関", "attribute_value_mlt": [{"subitem_degreegrantor": [{"subitem_degreegrantor_name": "University of Tokyo (東京大学)"}]}]}, "item_7_degree_name_20": {"attribute_name": "学位名", "attribute_value_mlt": [{"subitem_degreename": "博士(科学)"}]}, "item_7_description_5": {"attribute_name": "抄録", "attribute_value_mlt": [{"subitem_description": "Motion is one of the most important ways for humans to communicate with environments, that is, with other people and objects. However, there are patients have a clear consciousness but are not able to move their body because of injuries to their peripheral motor nerves or voluntary muscles. In these cases, Brain Computer Interface (BCI) can bypass the injured parts, directly link brain activity to artificial devices, and enable paralyzed patients to move as healthy people. In recent years, many researchers have focused on BCI with invasive measurements such as implanted microelectrode arrays or electrocorticography (ECoG). These methods directly record brain activities from the brain surface, thus they are accurate but not safe and reliable. On the other hand, non-invasive methods such as electroencephalography (EEG) and magnetoencephalography (MEG) can record brain activity from outside the head and thus are safer and more convenient. However, non-invasive methods are easily contaminated by environmental noise and thus making it difficult to extract motion patterns. Therefore, current non-invasive studies can not provide an efficient prediction and the motion related features in continuous motion are yet to be revealed. The two most important problems in motion trajectory prediction are noise reduction and feature selection because the prediction will not be efficient and robust if the features used have too much noise. Therefore in this thesis, both noise reduction method and feature selection strategy are discussed to perform an efficient continuous motion prediction. To perform accurate prediction of motion trajectory from single trials data, a new effective noise reduction method with almost no brain activity loss is introduced. Firstly, the original spatiotemporal signal space separation (tSSS) method developed by Dr. Samu Taulu in Helsinki University of Technology is applied to our system and the signal loss problem in tSSS method is discussed. Then an innovative improvement on tSSS method, a compensation process which suppresses noise and preserves brain signals simultaneously, is introduced. With the compensation process, our method shows very good noise reduction performance for both simulation of and the application to real MEG data. It should be noted that our method can be applied to all kinds of MEG systems, whereas the original method can be applied only to the MEG system with both gradiometers and magnetometers. A study on 1-D continuous motion using a tool bar is then presented and compensation tSSS method is applied to the recorded MEG data to reduce noise. It was found out that MEG signal spectrums of certain frequency bands closely resemble motion parameters and thus these spectrums are used to predict motion trajectory. Thus, in this study, the correlations averaged across all subjects between brain activity spectrum and motion parameters are investigated, and several motion-related features that have relatively high correlation values are extracted out. Moreover, different channel selection models and time-windows are adopted, and proper features that enable us to improve predictions are determined by multivariate linear regression prediction. With spectral, spatial, and temporal feature selection, an effective motion trajectory prediction with relatively high correlation coefficients (average value across all subjects is 0.59, p \u003c 0.001) is achieved on the 5 epoch averaged data preprocessed by compensation tSSS method. To further improve the prediction performance and provide an acceptable single trial motion prediction, differences in brain mechanism between subjects are considered and a more robust frequency feature selection model is proposed. In this study, motion related frequency features ranged from μ (8-16Hz) and β rhythm (18-24Hz) to low frequency δ rhythm (5-7Hz) and some part of high frequency γ bands (30-50Hz, 60-70Hz) are determined for each subject. By combining these selected subject-dependent frequency features, the prediction performance is further improved and this improvement is significant for most subjects. From the prediction performance of all subjects, it is concluded that using the correlation based feature selection method, single-trial MEG data could also predict continuous motion well (R = 0.46) with few features (less than 100). Moreover, two other tasks are performed and the robustness of our feature selection method on different motion cycles and external devices is proven. In task 1, similar motion using a different device (trackball) is performed and the prediction results confirm that our feature selection method worked equally well on different devices which indicates a robustness in different devices. In task 2, a different motion cycle without visual guidance is considered, and the efficiency of our feature selection method on different motion cycle, which showed a robustness in different motion, is confirmed. As there is no visual guidance, the selected features are verified to be from motion brain activities. From further contour map and source estimation studies, it is also confirmed that the sources of frequency features selected by our method is really located in the contralateral motor cortex and sensorimotor cortex, and thus motion related features. Our study reveals detailed characteristics of motion related activities which are consistent with ECoG and EEG studies. It also provides a guidance to select features and achieves a successful single trial motion trajectory prediction. The high quality prediction demonstrates that non-invasive measurement can predict motion comparably well as invasive measurement such as ECoG. Also, the prediction of continuous motion trajectory in our study provides a possibility of controlling external prosthetic devices.", "subitem_description_type": "Abstract"}]}, "item_7_dissertation_number_26": {"attribute_name": "学位授与番号", "attribute_value_mlt": [{"subitem_dissertationnumber": "甲第27217号"}]}, "item_7_full_name_3": {"attribute_name": "著者別名", "attribute_value_mlt": [{"nameIdentifiers": [{"nameIdentifier": "9508", "nameIdentifierScheme": "WEKO"}], "names": [{"name": "斉, 亮"}]}]}, "item_7_identifier_registration": {"attribute_name": "ID登録", "attribute_value_mlt": [{"subitem_identifier_reg_text": "10.15083/00004128", "subitem_identifier_reg_type": "JaLC"}]}, "item_7_select_21": {"attribute_name": "学位", "attribute_value_mlt": [{"subitem_select_item": "doctoral"}]}, "item_7_subject_13": {"attribute_name": "日本十進分類法", "attribute_value_mlt": [{"subitem_subject": "007", "subitem_subject_scheme": "NDC"}]}, "item_7_text_22": {"attribute_name": "学位分野", "attribute_value_mlt": [{"subitem_text_value": "Science (Kagaku)(科学)"}]}, "item_7_text_24": {"attribute_name": "研究科・専攻", "attribute_value_mlt": [{"subitem_text_value": "Department of Complexity Science and Engineering, Graduate School of Frontier Sciences (新領域創成科学研究科複雑理工学専攻)"}]}, "item_7_text_27": {"attribute_name": "学位記番号", "attribute_value_mlt": [{"subitem_text_value": "博創域第664号"}]}, "item_7_text_36": {"attribute_name": "資源タイプ", "attribute_value_mlt": [{"subitem_text_value": "Thesis"}]}, "item_7_text_4": {"attribute_name": "著者所属", "attribute_value_mlt": [{"subitem_text_value": "東京大学大学院新領域創成科学研究科基盤科学研究系複雑理工学専攻"}, {"subitem_text_value": "Department of Complexity Science and Engineering, Graduate School of Frontier Sciences, The University of Tokyo"}]}, "item_creator": {"attribute_name": "著者", "attribute_type": "creator", "attribute_value_mlt": [{"creatorNames": [{"creatorName": "Qi, Liang"}], "nameIdentifiers": [{"nameIdentifier": "9507", "nameIdentifierScheme": "WEKO"}]}]}, "item_files": {"attribute_name": "ファイル情報", "attribute_type": "file", "attribute_value_mlt": [{"accessrole": "open_date", "date": [{"dateType": "Available", "dateValue": "2017-06-01"}], "displaytype": "detail", "download_preview_message": "", "file_order": 0, "filename": "K-03003.pdf", "filesize": [{"value": "3.8 MB"}], "format": "application/pdf", "future_date_message": "", "is_thumbnail": false, "licensetype": "license_free", "mimetype": "application/pdf", "size": 3800000.0, "url": {"label": "本文(fulltext)", "url": "https://repository.dl.itc.u-tokyo.ac.jp/record/4137/files/K-03003.pdf"}, "version_id": "2e9a2e17-d597-4775-ba69-bf8f20d69c95"}]}, "item_language": {"attribute_name": "言語", "attribute_value_mlt": [{"subitem_language": "eng"}]}, "item_resource_type": {"attribute_name": "資源タイプ", "attribute_value_mlt": [{"resourcetype": "thesis", "resourceuri": "http://purl.org/coar/resource_type/c_46ec"}]}, "item_title": "MEG study on the prediction of motion trajectory", "item_titles": {"attribute_name": "タイトル", "attribute_value_mlt": [{"subitem_title": "MEG study on the prediction of motion trajectory"}]}, "item_type_id": "7", "owner": "1", "path": ["280", "329"], "permalink_uri": "https://doi.org/10.15083/00004128", "pubdate": {"attribute_name": "公開日", "attribute_value": "2012-11-12"}, "publish_date": "2012-11-12", "publish_status": "0", "recid": "4137", "relation": {}, "relation_version_is_last": true, "title": ["MEG study on the prediction of motion trajectory"], "weko_shared_id": null}
MEG study on the prediction of motion trajectory
https://doi.org/10.15083/00004128
https://doi.org/10.15083/000041285369fcff-fbe2-4cf6-bd15-7525f0b4091e
名前 / ファイル | ライセンス | アクション |
---|---|---|
本文(fulltext) (3.8 MB)
|
|
Item type | 学位論文 / Thesis or Dissertation(1) | |||||
---|---|---|---|---|---|---|
公開日 | 2012-11-12 | |||||
タイトル | ||||||
タイトル | MEG study on the prediction of motion trajectory | |||||
言語 | ||||||
言語 | eng | |||||
資源タイプ | ||||||
資源 | http://purl.org/coar/resource_type/c_46ec | |||||
タイプ | thesis | |||||
ID登録 | ||||||
ID登録 | 10.15083/00004128 | |||||
ID登録タイプ | JaLC | |||||
その他のタイトル | ||||||
その他のタイトル | 脳磁図を用いた運動軌道予測に関する研究 | |||||
著者 |
Qi, Liang
× Qi, Liang |
|||||
著者別名 | ||||||
識別子 | 9508 | |||||
識別子Scheme | WEKO | |||||
姓名 | 斉, 亮 | |||||
著者所属 | ||||||
著者所属 | 東京大学大学院新領域創成科学研究科基盤科学研究系複雑理工学専攻 | |||||
著者所属 | ||||||
著者所属 | Department of Complexity Science and Engineering, Graduate School of Frontier Sciences, The University of Tokyo | |||||
Abstract | ||||||
内容記述タイプ | Abstract | |||||
内容記述 | Motion is one of the most important ways for humans to communicate with environments, that is, with other people and objects. However, there are patients have a clear consciousness but are not able to move their body because of injuries to their peripheral motor nerves or voluntary muscles. In these cases, Brain Computer Interface (BCI) can bypass the injured parts, directly link brain activity to artificial devices, and enable paralyzed patients to move as healthy people. In recent years, many researchers have focused on BCI with invasive measurements such as implanted microelectrode arrays or electrocorticography (ECoG). These methods directly record brain activities from the brain surface, thus they are accurate but not safe and reliable. On the other hand, non-invasive methods such as electroencephalography (EEG) and magnetoencephalography (MEG) can record brain activity from outside the head and thus are safer and more convenient. However, non-invasive methods are easily contaminated by environmental noise and thus making it difficult to extract motion patterns. Therefore, current non-invasive studies can not provide an efficient prediction and the motion related features in continuous motion are yet to be revealed. The two most important problems in motion trajectory prediction are noise reduction and feature selection because the prediction will not be efficient and robust if the features used have too much noise. Therefore in this thesis, both noise reduction method and feature selection strategy are discussed to perform an efficient continuous motion prediction. To perform accurate prediction of motion trajectory from single trials data, a new effective noise reduction method with almost no brain activity loss is introduced. Firstly, the original spatiotemporal signal space separation (tSSS) method developed by Dr. Samu Taulu in Helsinki University of Technology is applied to our system and the signal loss problem in tSSS method is discussed. Then an innovative improvement on tSSS method, a compensation process which suppresses noise and preserves brain signals simultaneously, is introduced. With the compensation process, our method shows very good noise reduction performance for both simulation of and the application to real MEG data. It should be noted that our method can be applied to all kinds of MEG systems, whereas the original method can be applied only to the MEG system with both gradiometers and magnetometers. A study on 1-D continuous motion using a tool bar is then presented and compensation tSSS method is applied to the recorded MEG data to reduce noise. It was found out that MEG signal spectrums of certain frequency bands closely resemble motion parameters and thus these spectrums are used to predict motion trajectory. Thus, in this study, the correlations averaged across all subjects between brain activity spectrum and motion parameters are investigated, and several motion-related features that have relatively high correlation values are extracted out. Moreover, different channel selection models and time-windows are adopted, and proper features that enable us to improve predictions are determined by multivariate linear regression prediction. With spectral, spatial, and temporal feature selection, an effective motion trajectory prediction with relatively high correlation coefficients (average value across all subjects is 0.59, p < 0.001) is achieved on the 5 epoch averaged data preprocessed by compensation tSSS method. To further improve the prediction performance and provide an acceptable single trial motion prediction, differences in brain mechanism between subjects are considered and a more robust frequency feature selection model is proposed. In this study, motion related frequency features ranged from μ (8-16Hz) and β rhythm (18-24Hz) to low frequency δ rhythm (5-7Hz) and some part of high frequency γ bands (30-50Hz, 60-70Hz) are determined for each subject. By combining these selected subject-dependent frequency features, the prediction performance is further improved and this improvement is significant for most subjects. From the prediction performance of all subjects, it is concluded that using the correlation based feature selection method, single-trial MEG data could also predict continuous motion well (R = 0.46) with few features (less than 100). Moreover, two other tasks are performed and the robustness of our feature selection method on different motion cycles and external devices is proven. In task 1, similar motion using a different device (trackball) is performed and the prediction results confirm that our feature selection method worked equally well on different devices which indicates a robustness in different devices. In task 2, a different motion cycle without visual guidance is considered, and the efficiency of our feature selection method on different motion cycle, which showed a robustness in different motion, is confirmed. As there is no visual guidance, the selected features are verified to be from motion brain activities. From further contour map and source estimation studies, it is also confirmed that the sources of frequency features selected by our method is really located in the contralateral motor cortex and sensorimotor cortex, and thus motion related features. Our study reveals detailed characteristics of motion related activities which are consistent with ECoG and EEG studies. It also provides a guidance to select features and achieves a successful single trial motion trajectory prediction. The high quality prediction demonstrates that non-invasive measurement can predict motion comparably well as invasive measurement such as ECoG. Also, the prediction of continuous motion trajectory in our study provides a possibility of controlling external prosthetic devices. | |||||
書誌情報 | 発行日 2011-03-24 | |||||
日本十進分類法 | ||||||
主題 | 007 | |||||
主題Scheme | NDC | |||||
学位名 | ||||||
学位名 | 博士(科学) | |||||
学位 | ||||||
値 | doctoral | |||||
学位分野 | ||||||
Science (Kagaku)(科学) | ||||||
学位授与機関 | ||||||
学位授与機関名 | University of Tokyo (東京大学) | |||||
研究科・専攻 | ||||||
Department of Complexity Science and Engineering, Graduate School of Frontier Sciences (新領域創成科学研究科複雑理工学専攻) | ||||||
学位授与年月日 | ||||||
学位授与年月日 | 2011-03-24 | |||||
学位授与番号 | ||||||
学位授与番号 | 甲第27217号 | |||||
学位記番号 | ||||||
博創域第664号 |