A Dynamic Model of Brain Hemodynamics in Near-Infrared Spectroscopy.

Abstract

Near-infrared spectroscopy (NiRS) is a noninvasive technology used in measuring oxy- and deoxy-hemoglobin changes, neural activation, functional connectivity, and vascular health assessment. In this paper, we propose a dynamic model of the NiRS signal to facilitate a better understanding of the underlying elements of this signal and as a means of validation for existing and new NiRS signal processing algorithms. The model incorporates arterial pulsations, its possible frequency drifts and the reflected waves, the hemodynamic response function (HRF), Mayer waves, respiratory waves and other very low-frequency components of the NiRS signal. Parameter selection and model fitting have been carried out using measurements from a NiRS database. Our database includes 25 participants each with 64 channels, covering all the scalp and therefore providing realistic measures of the varying parameters. We compared synthetic resting-state and HRF-included model outputs with in vivo resting and task-included measurements. The results showed a significant equivalence of the in vivo and synthetic signals. The proposed signal model generates realistic NiRS signals. The model accepts simple physiological and physical parameters to produce realistic NiRS signals and will accelerate the growth of optical signal processing algorithms.