Poster F104, Tuesday, March 28, 8:00 – 10:00 am, Pacific Concourse
An evaluation of fNIRS preprocessing techniques using concurrent fNIRS-fMRI measurements
Aaron M. Piccirilli1, S.M. Hadi Hosseini1, Joseph M. Baker1, Jennifer L. Bruno1, Andrew Gundran1, Zachary Stuart1, Lene K. Harbott1, J. Christian Gerdes1, Allan L. Reiss1; 1Stanford University
Functional near-infrared spectroscopy (fNIRS) is a noninvasive optical brain imaging technique that measures relative changes in oxygenated and deoxygenated hemoglobin (HbO, HbR) as a proxy for brain activity. It is considered a promising alternative to functional MRI (fMRI) because of its relative portability, cost-effectiveness, and motion-tolerance. However, fNIRS preprocessing methods are still developing. Here, we investigate different preprocessing methods by comparing the correlation between simultaneously-recorded fNIRS and fMRI timecourses. We measured brain activity in nineteen healthy adults during a visuomotor task using fMRI and twenty-one fNIRS channels spanning the prefrontal cortex. From these data, we generated spherical regions of interest (ROIs) centered on the cortical projections of each fNIRS channel and calculated a mean BOLD timecourse for each ROI. Finally, we correlated the resultant timecourses while varying motion-correction and channel removal methods. The results indicated that HbO was more highly correlated with the BOLD signal than HbR, regardless of preprocessing methods. Motion artifact removal using wavelet filtering improved correlations for HbO, but decreased correlations for HbR. Compared to not removing channels, all of the methods that we used to remove noisy channels improved correlations. The highest correlations came from rejecting channels based on qualitative examination of the data in the frequency domain, as well as thresholding on instrument gain and signal-to-noise ratio. Our findings suggest that wavelet filtering, a widely-used technique, may not be optimal for low-motion paradigms. Further, they emphasize the value of bad channel removal and an approach informed by frequency-domain analysis and instrumentation parameters.
Topic Area: METHODS: Neuroimaging