Improved short-channel regression for mapping resting-state functional connectivity networks using functional near-infrared spectroscopy

Resting-state functional connectivity (rsFC) is an attractive biomarker of brain function that can vary with brain injury. The simplicity of resting-state protocols coupled with the main features of functional near-infrared spectroscopy (fNIRS), such as portability and versatility, can facilitate the monitoring of unresponsive patients in acute settings at the bedside. However, accurately mapping rsFC networks is challenging due to signal contamination from non-neural components, such as scalp hemodynamics and systemic physiology. Physiological noise may be mitigated through the use of short channels which may be able to provide sufficient information to eliminate the need for additional measurement devices, decreasing the complexity of the experimental setup. To this end, we examined the extent to which systemic physiology is embedded in the short-channel data and improved short-channel regression to account for temporal heterogeneity in the scalp hemodynamics. Our findings indicate that using temporal shifts in the short-channel data increases the agreement, by 70% on average, between short-channel regression and regression that includes short channels and physiological recordings. Overall, this method decreases the need for additional physiological recordings when mapping rsFC networks, providing a viable alternative when such measurements are not available or feasible.