Per voxel hemodynamic response estimation by sinusoid tracking

Motivation: Sinusoid tracking has several important advantages over DFT to detect hemodynamic response (HRF). We propose only per voxel HRF estimation to detect activation. GLM globally assumes identical HRF over all voxels to detect activation. While the general contour of HRF is widely accepted, little is known about its variability in different brain regions. Goal(s): Determine activation and HRF across whole brain simultaneously. Approach: A periodic impulsive stimulus generates BOLD response. Sinusoid tracking returns seven-harmonic parameters required from BOLD signal. Results: HRF is reconstructed from these harmonics, and ratios of harmonic amplitudes define activation. Impact: A custom sinusoid tracking method, which (unlike windowed DFT) imposes no restriction on sample rate, stimulus task frequency, and record length, can detect hemodynamic response function and fMRI activation simultaneously per voxel across the whole brain.