An ARX model approach to fNIRS data acquired from migraine and healthy subjects

Abstract

This study is focused on investigating the cerebrovascular dynamics of migraine by analyzing data acquired from healthy and migraine subjects with a noninvasive measuring technique, fNIRS during a breath holding task. Brain hemodynamic responses of subjects are modeled via a parametric identi cation technique, Auto-regressive with Exogenous input (ARX) model. Analysis of modeled signals for healthy and migraine subjects is performed both in frequency and time domains. In frequency domain analysis, frequency intervals in which power spectrum estimates of migraineurs signi cantly di er from healthy ones, are obtained as 0.01-0.03Hz, around 0.13 Hz and higher than 0.2 Hz (p<0.05). The energy of the estimated signals of migraineurs in 0.01-0.03 Hz is approximately ve folds smaller than the healthy ones, whereas in 0.13 Hz and 0.25 Hz this di erence is approximately 1.5 folds. Time domain analysis has shown that the amplitude of peak response of migraineurs is ve folds smaller than the healthy ones during all breath holding procedure (p<0.05). Required model orders to ful ll the dynamics of response are found higher in migraine case. Results obtained show that response of cerebrovascular system of migraine subjects to breath holding task is considerably di erent with respect to normal subjects.|Keywords: Migraine, Cerebrovascular Dynamics, Functional Near Infrared Spectroscopy (fNIRS), Linear Parametric Identi cation, Autoregressive Exogenous Input (ARX) Model.