Abstract #3638

Fully Automated fMRI Denoising using Multi-Echo fMRI & TE-Dependent Properties

Prantik Kundu¹, Souheil J. Inati¹, Jennifer W. Evans¹, Ziad S. Saad², Peter A. Bandettini¹

¹Section on Functional Imaging Methods, National Institute of Mental Health, Bethesda, MD, United States; ²Scientific & Statistical Computing Core, National Institute of Mental Health, Bethesda, MD, United States

A novel configuration of spatial ICA was applied to multi-echo fMRI data. The combination allowed robust, prior-free, and fully automated differentiation of physiological artifact components from BOLD components using an amplitude model for TE-dependence. No anatomical templates, bandpass filtering, physiology regressors, or motion parameters were required. The technique was applied to both task and rest data. Demonstrated is the improvement of including artifact component timecourses as baseline regressors in both activation mapping and seed-voxel functional connectivity.