Voxel And Vicinity-based algOrithm for nOise reductiOn of fMri data sets (VAVOOOM).

Motivation: Improve the low SNR in high-resolution fMRI and show the benefit of denoising strategies. Goal(s): Use a combined Hankel-PCA denoising method to remove thermal noise at the voxel as well as vicinity level, in contrast to existing methods exploiting vicinity-based PCA algorithms. Approach: High resolution task fMRI data were obtained using two protocols: 65 slices/2s TR and 5 slices/0.5s TR. Data were denoised with 7 different settings (voxel and vicinity-based combinations) and tSNR, task-related activation and temporal dynamics were evaluated. Results: tSNR improved with all denoising strategies. VAVOOOM, the proposed denoising method, shows promise for high-resolution applications and laminar fMRI. Impact: The presented denoising method (VAVOOOM) can be implemented in pre-processing pipelines to improve the tSNR and facilitate accurate characterization of fMRI metrics in high-resolution data such as laminar resolution fMRI at ultra-high fields.