A dynamic model of synthetic resting-state brain hemodynamics

Near infrared spectroscopy (NIRS) is an emerging field of brain study. From an engineering perspective, the absence of a ground truth signal or a model for producing synthetic data has hindered understanding of the underlying elements of this signal and validating of existing algorithms. In this paper, a dynamic model of artificial NIRS signal is proposed. The model incorporates arterial pulsations, its possible frequency drifts, Mayer waves, respiratory waves and other very low frequency components. Parameter selection and model fitting has been carried out using measurements from a NIRS database. To be general in the process of parameter selection, our dataset included 4 NIRS devices and 256 channels for each subject, covering all the scalp and therefore providing realistic measures of the varying parameters. Results are compared with the real data in time and frequency domains, both showing high level of resemblance.