Meeting Banner
Abstract #2076

A Novel Temporal Filtering Strategy for Functional MRI Using UNFOLD

Sebastian Domsch1, Andreas Lemke1, Sebastian Weingrtner1, Lothar R. Schad1

1Computer Assisted Clinical Medicine, Heidelberg University, Mannheim, Germany

UNFOLD to increase sampling rates in fMRI to detect transient BOLD signal modulations requires temporal filtering before statistical mapping in the time domain is carried out. Low-pass filtering and zero-filling has been proposed. Both strategies are non-optimal since low-pass filtering degrades temporal resolution and zero-filling leads to serial correlations (non-white noise) known to bias linear modeling of neuronal responses. We propose a more sophisticated temporal filtering strategy intrinsically avoiding non-white noise to increase statistical inference.

Keywords

accelerated activated activation actual alias aliasing anterior applied approx artifacts assisted attenuated besides bilateral body bold carried channel characteristics clearly clinical cluster coil components consistently cortex counted curves dataset datasets decrease degrades depended detectability detected deviation distribution domain effective eightfold encoding equipped estimation event excluded exhibits experiment false filtering fine finger foot frame frequencies frequency fundamental ground head healthy important improves inference interleaved known linear location maps material matrix measured medicine minutes modeling modified modulations motor narrowed nearly noise novel null objective operator optimal particularly pass peaks picture positive positives post power process processed propose proposed pulse receiver related remaining removed replacing representative resolution responses resting robust sampling saturation scale sensitivity sensory short significance significantly since slab slice slight smith smoothing space spatial spectrum statistical statistically stimulus strategies strategy subject subjects successfully supplementary system table tapping temporal thresholds transient trend truth twice typical typically unbiased unfold versus white width years zero