Motion artifact removal algorithm for portable fNIRS system using artificial neural network

Abstract

Functional near-infrared spectroscopy (fNIRS), a technique for measuring changes in oxygen saturation in cerebral blood resulting from activation of neurons, has been widely used in cognitive research thanks to its portability and high resolution. However, during brain fNIRS measurement, drift of the blood flow by movement occurs due to the absence of the valves in brain vein. This movement appears to be a very important artifact in fNIRS measurements which can seriously contaminate original signal. This artifact is very difficult to remove due to the complexity, individuality, and autonomic regulation of the cerebral vasculature. In this study, proposed artificial neural network model learn and predict the motion artifact based on the angular information from the motion sensor attached to portable fNIRS system. This model effectively predicts and removes motion artifact from head movement.