Lars Kasper1,
Maximilian Haeberlin1, Benjamin E. Dietrich2, Simon
Gross1, Christoph Barmet2, 3, Christian C.
Ruff4, Klaas E. Stephan1, 5, Klaas P.
Pruessmann1
1Institute
for Biomedical Engineering, University & ETH Zurich, Zurich, Switzerland;
2Institute for Biomedical Engineering, University and ETH Zurich,
Zurich, Switzerland; 3Skope Magnetic Resonance Technologies,
Zurich, Switzerland; 4Department of Economics, University of
Zurich, Zurich, Switzerland; 5Wellcome Trust Centre for
Neuroimaging, University College of London, London, United Kingdom
Matched-filter fMRI utilizes 2D-density weighted EPI to increase BOLD sensitivity. It matches acquisition density in k-space to the desired spatial response typically a Gaussian kernel. Theory predicts significant SNR benefits from such acquisitions, which are maximal when thermal noise dominates and expected to decrease with higher physiological noise contributions. We explore the validity of this argument for different regimes of physiological noise in the brain and report consistent and replicable SNR increases of 20-40 % compared to uniform EPI acquisitions. For task-based fMRI, we observe a consistent increase in BOLD-sensitivity of 30 % (average t-value), and show reproducibility both within and between subjects (N=4).