A Software Platform for High-Performance Real-Time Resting-State fMRI Analysis

Motivation: Real-time resting-state fMRI (rsfMRI) is a novel modality with considerable clinical potential. However, denoising and statistics approaches lack the performance of widely used offline rsfMRI analysis approaches. Goal(s): To enhance sensitivity and specificity of real-time rsfMRI analysis by integrating advanced signal processing approaches. Approach: Dual sliding window partial correlation with PCA-based confound regression and spectral segmentation of regression vectors and correction for nonstationary autocorrelations enabled simultaneous mapping of static and dynamic connectivity. Results: Frequency segmented regression substantially reduced false-positive connectivity in motion corrupted data. Ten networks with whole-band regression of motion parameters, white matter and CSF signals were mapped with 400 ms TR. Impact: This study demonstrates advances in seed-based real-time resting-state fMRI analysis for high-speed data acquisition that approach the performance and utility of conventional offline resting-state fMRI analysis toolboxes.