I would add that this open-close hand task is a very robust localizer, and unless something goes wrong, you should be able to see the data clearly in the timecourse with the naked eye. Sadly, I don't have a ready example to share, but it would be comparable to a primary visual response to a flashing checkerboard.
If you have someone who can help you with this, find a voxel near the central sulcus at the "hand notch." This landmark is easy to find in a decent anatomical image. You should see a robust block pattern in that timecourse.
Just one more thing to check!
Dav
On Jan 21, 2011, at 10:12 AM, Mark Jenkinson wrote:
> Dear Nicolo,
>
> It is possible that your activation does not quite pass this
> threshold but that the console analysis showed a much
> more liberally thresholded version of the results. To see
> if you also have similar activation, just below significance,
> load the zstat1.nii.gz file from the stats subdirectory (inside
> the feat directory) into FSLView to look at the unthresholded
> statistics. This should give you a feeling for whether the
> analysis is showing you similar things, but the statistics are
> just a little weak, or whether something is very different and
> you can't detect things that were shown on the console, which
> is probably indicative of a bigger error in setting up the
> analysis.
>
> It is not worth turning to FLOBS at this stage, and I would
> only consider it if you activations are close to statistical
> significance and you have an expectation that the HRF
> in your group might be reasonably different from the standard
> population. With a 10s on and 10s off design (which is quite
> quick), the details of the HRF are largely washed out.
>
> It may also be worthwhile running ICA to see if there are
> any major artefacts in your data that you could clean up
> with ICA-based denoising. But first look at your
> unthresholded stats.
>
> Hope this helps.
> All the best,
> Mark
>
>
> On 21 Jan 2011, at 15:48, Nicoḷ Cardobi wrote:
>
>> Thanks for the very quick reply.
>>
>> I change the convolution to gamma (I think that's the default setting of convolution with square basic shape wave), and I set the other parameters as you kindly advised, however I still not obtain a decent activation map.
>>
>> I post here a zmap thresholded with cluster correction (z threshold = 2.3 and p < 0.05)
>>
>> [URL=http://img200.imageshack.us/i/renderedthreshzstat1.png/][IMG]http://img200.imageshack.us/img200/8932/renderedthreshzstat1.png[/IMG][/URL]
>>
>> I add also a time series plot, witch may reveal the problem:
>>
>> [URL=http://img806.imageshack.us/i/tsplotzstat1.png/][IMG]http://img806.imageshack.us/img806/2495/tsplotzstat1.png[/IMG][/URL]
>>
>> The full mode fit and the cope one seems to be very different by the data. Is this the problem? Do I have to use the FLOBS in order to create an optimal basis set that fits better to the data?
>>
>> Best regards.
>>
>> P.s.: the z-threshold set in Misc tab has any relationship between the z-threshold set in the cluster thresholding? The default setting is 5.3 and changing it produces a modification in the design efficiency.
>>
|