HRfunc: a tool for modeling hemodynamic response variability in fNIRS

Neural activation in functional near-infrared spectroscopy (fNIRS) signals is inherently convolved with, and temporally blurred by, a hemodynamic response function (HRF). Accurately modeling HRF variability during deconvolution improves neural activity recovery.

We present the Python-based HRfunc tool for estimating local HRF distributions and neural activity from fNIRS through deconvolution. HRFs are stored within a tree and a hash table hybrid data structure for efficient spatial and contextual identification of relevant HRFs.

To test the HRfunc tool, we conducted two analyses with hemoglobin and estimated neural activity, a general linear model (GLM) analysis on a single subject, child executive function task ( $n=79$ ), and a neural synchrony analysis assessing wavelet coherence between child-parent dyads (92 dyads).

Estimated HRFs contained a generally canonical shape. Within estimated neural activity, kurtosis increased, skew remained stable, and signal-to-noise ratio decreased. Neural synchrony lateralization effects emerged, and consistent GLM outcomes were observed.

These results support the use of the HRfunc tool for estimating event-based HRFs and neural activity in fNIRS studies. Through collective sharing of HRFs, an HRF database will be established to provide access to estimated HRFs across brain regions, subject ages, and experimental contexts.