心算和想象运动二分类的近红外光谱脑机接口范式重测信度

In this paper, two classical paradigms, mental arithmetic and right-hand motor imagery are utilized in the fNIRS-BCI binary classification and the test-retest reliability is investigated in terms of classification accuracy, classification feature and brain activation response strength. Three repeated experiments sessions are conducted on 10 male subjects over a month span and use the mean value of oxygenated hemoglobin (HBO) as classification feature and SVM as classification method. Results show that in terms of classification accuracy, intra-session and inter-session classification accuracies have good test-retest reliability. Then the intraclass correlations coefficients, the quantity of significant activated channels and the same location of significant activated channels are used to evaluate the test-retest reliability of mean features and brain activation response strength. The results show that the test-retest reliability of the classification accuracy is highly consistent with the reliability of mean features and brain activation response strength for all subjects.