Improving the accuracy of the method for removing motion artifacts from fNIRS data using ICA and an accelerometer

Independent component analysis (ICA) is one of the most preferred methods for removing motion artifacts from functional near-infrared spectroscopy (fNIRS) data. In this method, the fNIRS signal is separated into components by ICA and the component that shows high correlation between the fNIRS signal and motion artifact is determined. This component is removed, and the fNIRS signal without motion artifacts is derived. However, fNIRS data are often delayed temporally compared with accelerometer data because the blood flow changes slowly after the subject's head moves. It is necessary to consider the temporal delay in fNIRS data in order to remove motion artifacts when we use ICA method. In this method, the correlation coefficient is used to identify the motion artifact component. However, the cerebral blood flow has a small change because the biological signal fluctuates minutely. Hence, the correlation is reduced, and it is difficult to determine whether the component has been derived from the motion artifact. We propose a method that uses t-tests and the correlation coefficient to identify the motion artifact. In this proposed method, we used t-tests for comparing accelerometer data and signals separated by ICA. The separated signal with no significant difference from accelerometer data were identified as motion artifacts and removed. To examine the validity of this method, we used data sets including motion artifacts caused by sleepiness. Results obtained using only the correlation coefficient were compared with those obtained using the correlation coefficient and t-tests. We found that the proposed method improved that accuracy of removing motion artifacts. In addition, the signs of the accelerometer data were inverted, and t-tests were performed. Consequently, the accuracy of removing the motion artifact was improved.