Mathijs & Raphael, See inline comments below.
Donna On 08/19/2010 08:53 AM, Raemaekers-2, M. wrote: > Dear Rafael, > > Not sure of this, but the results that you describe may be the result of a > misalignment between your statistical map and your surface. > > It somehow seems that SPM and CARET are not using the niftii positioning > information exactly the same, Any details you can provide would be appreciated. > but for me it works when I coregister&reslice a statistical map to the > T1-image (as put in Talairach orientation before segmentation), and then map > the resliced statistical image to the surface. Mathijs' comments here suggest, to me, that he is generating his own surfaces and mapping individual maps to the individual's surface. this differs from what Raphael is doing, which is mapping group results to the PALS atlas, at least as I understand it. But in both cases, your statistical maps do need to align with the target surface. In Mathijs individual case, the map must align with the T1 used to generate the segmentation and resulting surface. In Raphael's case, it must be in MNI space (i.e., spatially normalized to the standard templates in the SPM distribution). Now on to Raphael's comments below... > Hope this helps. > > Regards, > Mathijs > > ________________________________ > > Van: [email protected] namens Raphael Hilgenstock > Verzonden: do 19-8-2010 13:56 > Aan: [email protected] > Onderwerp: [caret-users] SPM8 - Mapping Functionals to Volume problem > > > > Dear List, > > I`ve got a problem mapping SPM8 t- and F-contrasts to Caret surfaces. > Unfortunately I even don`t know if it is possible to map SPM8 images to > Caret as haven`t found any informationen on the compatibility of both > programs (even on the list). Is there a difference between SPM5- and > SPM8-space that has to be accounted for when projecting activations?! > We haven't yet generated SPM8 surfaces for the PALS atlas, but we should do so. Odds are, John Ashburner made significant improvements to the spatial normalization between the SPM5 and SPM8 revisions. Still, using SPM5 is an acceptable substitute. I can tell you that when I compared the SPM2 and SPM5 surfaces, differences were modest. I would just specify which PALS_B12 target you used in your methods. > I`m asking because I ran across the following problem: > > I processed data to map activations using the Tutorial_Sept_06 (provided > on the Sumsdb site) click by click with the > Human.PALS_B12.BOTH.TEMPLATE-for-fMRI-MAPPING.73730.spec-file (is this > approach still the standard of mapping functional files to volumes?). > Yes. > When I mapped an F-contrast (spm_F.hdr and .nii) on an inflated brain > there where more spots (and some spots in strange locations) on my > MFM-image than on my SPM. This is not unusual with MFM, which maps your volume to each of the 12 PALS subjects' surfaces, and then averages the resulting 12 maps. The tutorial talks about the pros and cons of the MFM and AFM methods. David feels the MFM method is more representative, but in practice many WUSTL researchers use the AFM method, because it appears more conservative. There is a way to constrain the display to an extent equal to the average supra-threshold area in the contributing subjects. The tutorial talks about that, too. > When I mapped a t-contrast (spm_T.hdr and > .nii) all activations were gone! When lowering the threshold also this > (T-)map is showing activations that can be expected from my SPM but also > "new" acitavtions. I am confused about this. If all activity disappeared (using the same contrast that gave positive F results) on your t-map, then I start doubting the volume -- either it really is all zeroes, or the origin is way off and/or not specified -- something like that. Upload your spm_T.hdr and .img and/or .nii here: http://pulvinar.wustl.edu/cgi-bin/upload.cgi > The MFMs were thresholded with my SPM t/F-treshold > and the treshold type in Caret set to "column" (does this acutally mean > that just the information (t/F-values) form the the spm_T/F image is > read without any changes to the threshold by Caret? I infered that from > posts I read but I´m not yet definitively sure). This gets confusing, but relates to what I mentioned earlier about the MFM threshold that keeps the supra-threshold area equal to the mean supra-threshold area of the 12 PALS subjects. If you select this option and enter a threshold when you map, then Caret computes the threshold needed to hold the area at this level, and it stores it in the user threshold, I believe. You over-ride this with your own column threshold. > As suggested in another > post I also used different algorithms (e.g. Metric-Gaussian). However > the result is still the same. I´m most concerned about strange > activation foci that I won`t find in my SPMs. > This is a normal consequence of MFM mapping -- a reflection of normal anatomical variability. It might go away with proper thresholding. But first, try AFM mapping and see if your concerns disappear. > Is there anything I did wrong?! Don't know. The disappearing t activations puzzle me, which is why I want to see your volume. > Any hint or help would be greatly > appreciated. > > Thanks & best regards > Raphael > _______________________________________________ caret-users mailing list [email protected] http://brainvis.wustl.edu/mailman/listinfo/caret-users
