Dear Mike,
A SVC does not alter the T map / T values in the table, only the p values. This can indeed lead to two different critical T (or F) values for different regions. In the end it's up to you how to present results in a meaningful way then, either with a single bar but some markers to indicate the two critical values, or separate bars.
> I still don't know how to correctly illustrate small volume corrections (SVCs) results and hope some experts could help
I would suggest to either go with thresholded maps based on the thresholds reported in the text * (i.e. small-volume corrected .05 FWE on peak level) or try to display the entire activation pattern of the small volumes, not just the sig. voxels (e.g. like suggested in Fig. 3b of Allen et al., 2012, http://dx.doi.org/10.1016/j.neuron.2012.05.001 , which combines beta estimates and T values via hue and transparency settings). In both instances, it would be very useful to also show the extent of the a-priori region.
In my opinion figures should provide an easy-to-interpret overview of the results (which you would get with the former, e.g. so that one doesn't has to enter coordinates to get an impression of the exact location), or additional information (which you would get with the latter). This is why I find it very confusing to read papers in which a SVC was conducted with one threshold, while the figure shows results obtained with a different threshold. I mean, what do the authors want to show? Why is it useful to see the data thresholded at e.g. an uncorrected .005 at all, and why this particular value? I came across this frequently in the context of SVC, while it seems to be much less common to switch initial thresholds for whole-brain maps. I have a vague idea why people might do so, one reason might be that they think the initial threshold is relevant, another reason might be that they do not feel well with single-voxel findings or very small clusters even if the peak is sig., and that they turn to an uncorrected threshold that results in a lower critical T/F value instead, with clusters being larger and looking "more appropriate".
Best
Helmut
* This shouldn't be too complicated to achieve with viewing tools like MRIcroN/MRIcroGL that allow for different layers, you basically need a mask image that represents the SV, and you need thresholded T maps for each of the SVs. For the latter you would probably load the contrast in SPM, specify a mask (to hide the non-SV voxels), enter the critical T value, enter an extent threshold if you want to do so, save the thresholded T map. Same for the other SV. Then turn to your viewing tool, load the two T maps and the two mask images as separate overlays, play around with colours. In theory you could combine this with a third T map for the non-SV voxels e.g. thresholded at .001 uncorrected combined with cluster correction.
|