個人差を考慮した聴性定常反応の検出
個人差を考慮した聴性定常反応の検出
カテゴリ: 研究会(論文単位)
論文No: MBE22020
グループ名: 【C】電子・情報・システム部門 医用・生体工学研究会
発行日: 2022/03/13
タイトル(英語): Auditory Steady-State Response with Individual Characteristics
著者名: 高橋 斗威(東京大学),可部 泰生(東京大学),森本 隆司(リオン),高橋 宏知(東京大学)
著者名(英語): Toi Takahashi(The University of Tokyo),Yasuo Kabe(The University of Tokyo),Takashi Morimoto(RION),Hirokazu Takahashi(The University of Tokyo)
キーワード: 個人差|聴性定常反応|ASSR|他覚的聴力検査|機械学習|LSTM|individual characteristics|auditory steady-state response|ASSR|objective hearing test|machine learning|LSTM
要約(日本語): LSTMを用いた機械学習モデルで聴性定常反応(ASSR)の検出を試みた.訓練用データの個人特性の類似度とASSRの検出精度に,相関のある被験者とない被験者が存在した.機械学習モデルの中間出力を用いて,個人特性の類似した被験者のデータを訓練用に選択すると,ASSRの検出精度が向上した.したがって,脳波の個人特性を考慮することで,ASSRの検出精度は向上されることが示唆される.
要約(英語): We hypothesized that steady-state responses in electroencephalogram (EEG) exhibits individual differences, which cause deterioration of EEG characterization. We attempted to characterize individual differences in auditory steady-state response (ASSR) and improve the detection by taking the individual differences into account. We constructed a deep learning model with long-short term memory (LSTM) for ASSR detection. The intermediate output from the LSTM layer in the model suggested that ASSR exhibited individual characteristics depending on subjects. In addition, the area under curve of the receiver operating characteristic indicated that, as compared to a conventional method, ASSR detection improved by 6% when EEG data with similar individual characteristics were selected as training data. Thus, deep learning with LSTM that takes individual characteristics into account is able to improve EEG characterization.
本誌掲載ページ: 21-26 p
原稿種別: 日本語
PDFファイルサイズ: 2,101 Kバイト
受取状況を読み込めませんでした
