Denoising fNIRS Signals to Enhance Brain Imaging Diagnosis

Abstract

Functional Near Infrared Spectroscopy (fNIRS) signals have the potential to permit accurate analysis of intracortical brain physiologic disease. The HomER graphical user interface is used to display the NIRS data, FastICA is updated to reduce data dimension and combined Wavelet & Pca method is developed to denoise NIRS signals. These signals include several types of noise spread from low to high frequencies such as respiratory interference frequency band of 0.1-0.3Hz, NIRS Mayer wave which is about 0.1Hz, cardiac interference frequency band which is 0.8-2.0Hz, artifacts from head and facial motions, and high frequency noise generated from electronic components. Wavelet & Pca is an efficient method to reduce biological noise, motion artifact and high-frequency noise. The applied processing technique consists of adaptively modifying the wavelet coefficients based on the degree of noise contaminating the processed NIRS signal. This is done subsequently to signal pre-processing by reducing data dimension using the FastICA method. The feasibility of the method was demonstrated by testing it on experimental fNIRS data collected from 47 subjects. Preliminary results, through signal-to-noise ratio and correlation indicators show that the technique reduces noise and improves the quality of the acquired NIRS signals and its corresponding analysis.