Comparing different motion correction approaches for resting-state functional connectivity analysis with functional near-infrared spectroscopy data.

Abstract

Motion artifacts are a notorious challenge in the functional near-infrared spectroscopy (fNIRS) field. However, little is known about how to deal with them in resting-state data. We assessed the impact of motion artifact correction approaches on assessing functional connectivity, using semi-simulated datasets with different percentages and types of motion artifact contamination. Thirty-five healthy adults underwent a 15-min resting-state acquisition. Semi-simulated datasets were generated by adding spike-like and/or baseline-shift motion artifacts to the real dataset. Fifteen pipelines, employing various correction approaches, were applied to each dataset, and the group correlation matrix was computed. Three metrics were used to test the performance of each approach. When motion artifact contamination was low, various correction approaches were effective. However, with increased contamination, only a few pipelines were reliable. For datasets mostly free of baseline-shift artifacts, discarding contaminated frames after pre-processing was optimal. Conversely, when both spike and baseline-shift artifacts were present, discarding contaminated frames before pre-processing yielded the best results. This study emphasizes the need for customized motion correction approaches as the effectiveness varies with the specific type and amount of motion artifacts present.