On Sat, May 29, 2010 at 4:43 PM, Dominik Bach <[log in to unmask]> wrote:
> Hi Gareth & Vladimir
> perhaps you can give me some advice. Results from my beamformer source
> reconstruction are disappointing on the group level, but the MSP approach
> seems quite exciting. I see induced responses in the hypothesised frequency
> band (40-60Hz) but not in other frequencies, and on the group level the
> significant spots map almost precisely on the left temporal plane; a very
> thin blob that seems to follow the cortical contour (cluster-level
> significant with auditory cortex-based SVC for FWE).
> So I was wondering:
> - does the cortical mesh include the cortical folding pattern, e.g. fissures
> etc. such that this could be an artefact of the method?
The mesh does follow cortical pattern but you should use the smoothed
images (sw...) for your contrasts and in that case the blobs shouldn't
be very thin. Perhaps try smoothing more (you can just smooth the w...
images with a wider kernel) and see what you get then.
> - what could I do to corroborate this finding?
That's a $64,000 question but I'd find it convincing if you also do a
sensor level analysis comparing the corresponding time windows and
conditions and show that there is a pattern on the scalp consistent
with a source in the temporal lobe. I wouldn't necessarily insist that
both sensor and source-level results are significant but at least one
of them should be. You can use the 'Fieldrip multitaper power map'
function in MEEGtools if you want to average over the 1 sec window.
> - how can I extract the timecourse of this gamma response (so far I average
> over a 1000 ms time window).
You can look at the attached function I got from Laurence. I haven't
tested it myself, but I'm planning to include something similar in SPM
in the future.