Resting State ICA Enhanced with Multi-Echo fMRI

Introduction Independent Components Analysis (ICA) uses mutual information to reduce an fMRI dataset to a small group of source timecourses and corresponding localization maps. Resting state network (RSN) activity can then be distinguished from noise sources based on gray matter localization and temporal smoothness. In practice, components frequently have equivocal localization and mixed frequency temporal character. Providing ICA more temporal data acquired from longer resting fMRI scans alleviates this problem, but this is uneconomical. It is proposed that multi-echo (ME) fMRI can do the same while maintaining short scan times. ME fMRI samples signal at several echo-times (TE) during T2* relaxation, providing multiple fMRI timecourses of different BOLD contrasts for the same period. Since robust hemodynamic activity should be expressed across all contrasts within the TE range for BOLD, providing fMRI data of several echo times should improve ICA decomposition by increasing the representation of a true hemodynamic source and decreasing the relative ratios of TE-specific RF noise and weighting contribution of non-hemodynamic physiological signal to some TEs over others.