Dear FSL experts,
I have been undertaking some testing of multiband sequences, and have been using them to produce sequences with fairly short TRs, down to 0.333s for the highest acceleration factor tested (MB6). I have some fairly theoretical questions about FILM and how it might handle such short TRs. Clearly the autocorrelation structure in the data will be somewhat different for TRs this short, compared to 'typical' 2-3s TRs, and an AR(1) model will likely be insufficient to correct for it. My understanding of FILM is that it works quite differently from the AR(1) correction used in SPM, and is a more data-driven approach that is (theoretically) independent of TR (as stated in, Woolrich, M. W., Ripley, B. D., Brady, M., & Smith, S. M. (2001). Temporal autocorrelation in univariate linear modeling of FMRI data. Neuroimage, 14(6), 1370-1386.)
My question is therefore, does FILM provide sufficient correction for autocorrelation, even with such short (<=0.5s) TRs? Can I confidently assert that FILM still performs adequately with such sequences? Is anyone aware of any systematic testing of FILM with such short TRs, or any papers I can cite that have used FILM with such sequences?
Many thanks in advance for your thoughts,
Matt.
|