fMRI-EEG Fingerprint Regression Model for Motor Cortex

Abstract

The combination of modern machine learning and traditional statistical methods allows the construction of individual regression models for predicting the blood oxygenation level dependent (BOLD) signal of a selected region-of-interest within the brain using EEG signal. Among the many different models for motor cortex, we chose the EEG Fingerprint one-electrode approach, based on rigid regression model with Stockwell EEG signal transformation, used before only for the amygdala. In this study we demonstrate the way of finding suitable model parameters for the cases of BOLD signal reconstruction for five individuals: three of them were healthy, and two were after a hemorrhagic stroke with varying degrees of damage according to Medical Research Council (MRC) Weakness Scale. The principal possibility of BOLD restoring using regressor model was demonstrated for all the cases considered above. The results of direct and indirect comparisons of BOLD signal reconstruction at the motor region for healthy participants and for patients who suffered from a stroke are presented.