Classification of fNIRS based brain hemodynamic response to mental arithmetic tasks

Abstract

Specific characteristics of the functional near infrared spectroscopy (fNIRS) of the hemodynamic response may represent the brain cortical activity levels during mental arithmetic tasks. In this paper, we use hemodynamic response signals of the prefrontal cortex, acquired by a 4-channel fNIRS system to identify the difficulty level of an arithmetic task. To this end, twelve temporal features and several classification methods are used. In addition, most discriminating features are identified by principle component analysis (PCA) method. Experimental results show that the highest accuracy rate of 92.2% is achieved by a linear Support Vector Machine (SVM) classifier. They also show that skewness and total area of the signal from the 3 cm channel on the left prefrontal lobe are the most discriminating features.