Efficacy of Data-Driven Respiration Compensation Methods in fMRI Data at 1.5T
Kothari A, Talavage T, Hu S
Respiratory noise is a component of interest in fMRI data. This work identifies those circumstances in which respiratory noise compensation techniques are beneficial. A magnitude-only Gaussian band-reject filter and a complex image-space estimation and removal procedure were assessed for improvement (relative to no compensation) in true and false detections of synthetic activations superimposed on human baseline fMRI data. A clear benefit in the specificity of detected activation is obtained using complex image-space filtering algorithms. However, there is no compelling evidence that respiration-induced noise compensation algorithms are effective when the rate of stimulus presentation is not near the rate of respiration.