Santosh B. Katwal1, 2, John C. Gore2, 3, Baxter P. Rogers2, 3
1Electrical Engineering and Computer Science, Vanderbilt University, Nashville, TN, United States; 2Vanderbilt University Institute of Imaging Science, Nashville, TN, United States; 3Biomedical Engineering, Vanderbilt University
We assessed the ability to detect small temporal differences in BOLD responses from fMRI using Granger causality. Task-related-voxels were selected using: i) self-organizing map (SOM), ii) independent component analysis (ICA), iii) statistical parametric mapping (SPM) and iv) localizer scan in conjunction with SPM. Additionally, we fitted curves to average signals with inverse logit functions and estimated the differences in time-to-peak to compare the temporal differences. The combination of SOM and Granger causality detected differences as small as 28 ms. In general, data-driven approach for voxel selection seems to work better than hypothesis-driven approach in detecting small temporal differences from fMRI.