統計的識別による聴覚誘発電位の特徴分析

摘要

聴覚刺激に対して惹起する脳波（聴覚誘発電位）の波形形態分析の手法として，統計的手法による脳波からの刺激の識別が可能であることを報告する．6種類の異なる刺激持続時間の聴覚刺激に対する誘発電位を7箇所の電極で測定した（健常成人25名）のち，各電極の誘発電位に（1）識別実験に使用するデータ区間を設定し，（2）主成分分析により無相関化と次元圧縮を行い，（3）ナイーブベイズ識別器を作成し，刺激に対する識別率を調べた．その結果から，識別対象刺激・電極・データ区間と識別率の関係について議論する．%To estimate the input stimulus from observer's evoked brain potentials (ERPs), we have developed a classification technique using Principal Component Analysis (PCA) and Naive Bayes method. Our method does not require a priori knowledge about stimuli, and it is easy to apply because of its simplicity and stability. In this study, we applied the technique to classify auditory ERPs that were elicited by temporal gaps embedded in a continuous pure tone. The classifier that was constructed from ERP data elicited by long gap showed higher correct rate than that of short gap. Accuracy was significantly higher than the chance level, indicating that statistical discrimination of temporal gap stimuli was successful and would be available for quantitative feature analysis method of ERPs.