Quite a lot of people have asked whether we have published the GLM method
at the heart of FEAT, the model-based FMRI analysis tool in FSL. A paper
on our GLM method (FILM - FMRIB's Improved Linear Model) has now been
accepted for publication in NeuroImage, and should appear later in the
year. A closely related technical report can be found at
http://www.fmrib.ox.ac.uk/analysis/research/feat
*** Brief summary of what FILM does, for the technically interested ***
FILM, the work of Mark Woolrich in the FMRIB analysis group, applies the
General Linear Model, with maximum possible estimation efficiency whilst
remaining valid and robust. It achieves this by "prewhitening" the time
series at each voxel separately - this is the most efficient possible
(univariate GLM-based) estimator of activation, resulting in increased
sensitivity of up to a factor of 2 compared with methods which precolour
(smooth) the data instead (this factor being greatest for single-event
designs). It achieves this robustly by regularising the autocorrelation
estimates first by applying a Tukey taper to the estimates, and then
applying a small amount of within-tissue-type spatial regularisation to
the estimates. The estimates are then used (in a non-parametric fashion)
to provide a prewhitening filter.
Stephen M. Smith
Head of Image Analysis, FMRIB
Oxford University Centre for Functional MRI of the Brain
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
|