Correcting physiological noise in whole-head functional near-infrared spectroscopy

Systemic physiology augmented functional near-infrared spectroscopy: a powerful approach to study the embodied human brain

“…Concerning the question how the systemic physiological signals should be processed and analyzed within the SPA-fNIRS approach, there is currently no standard procedure available, and the research how to analyze fNIRS data along systemic physiological data just started. Different methods have been employed so far, including the calculation of block-averages of stimulus-evoked changes in the signals, 95 , 96 block-averaging with subsequent correlation analysis to investigate the neurosystemic functional connectivity, 87 the use of a general linear model (GLM) that treats the systemic physiological signals as additional regressors, 89 , 91 , 92 , 97 – 99 wavelet coherence analysis, 97 the use coupling functions derived from the phases of the signals via the continuous wavelet transform, 24 oblique subspace projections signal decomposition, 100 or the recent approach using a GLM and regularized temporally embedded canonical correlation analysis (tCCA). 27 , 101 The use of tCCA allows one to create optimal nuisance regressors by considering non-instantaneous and non-constant coupling between the recorded signals and by intelligently combining any available auxiliary signals (e.g., systemic physiology and short-channel fNIRS recordings).…”

“… 118 , 119 Various methods how to use short-channels have been developed, 118 , 120 – 126 relying all on regressing out the extracerebral information from the long-channels (short-channel regression). The methods improve the ability to measure changes related to NVC 98 , 118 , 126 – 129 and increase the reproducibility of fNIRS measurements on the single-subject level. 130 …”

“…That the combination of short-channels and systemic physiological signals help to remove physiological noise in a GLM framework with fNIRS data from whole-head recordings has been shown recently. 98 We expect that the number of such investigations will increase in the near future enabling one to determine when each method works best.…”

“…118,119 Various methods how to use short-channels have been developed, 118,[120][121][122][123][124][125][126] relying all on regressing out the extracerebral information from the long-channels (short-channel regression). The methods improve the ability to measure changes related to NVC 98,118,[126][127][128][129] and increase the reproducibility of fNIRS measurements on the single-subject level. 130 A comparison between short-channel regression, and SPA-fNIRS-based regression has been recently published by Abdalmalak et al 99 investigating the impact of both methods on the detection and quantification of functional resting-state brain networks.…”

Identifying Individuals by fNIRS-Based Brain Functional Network Fingerprints

“…First, a band-pass filtering is implemented by the third-order Butterworth filter with a cut-off frequency of 0.01-0.1 Hz to eliminate the physiological noise (approximately 0.1 Hz for Mayer wave, 0.25 Hz for respiration, and 1 Hz for a heartbeat). Then, the baseline correction is used to subtract the global signal (i.e., the average signals of all channels) from all signals (Nguyen et al, 2018;Zhang et al, 2021). All the preprocessing mentioned above is done through the BBCI toolkit (Blankertz et al, 2016).…”