open-discussion > Scaling of overlaid values
Showing 1-2 of 2 posts
Display:
Results per page:
Aug 24, 2017  01:08 AM | Katharina Kunzelmann
Scaling of overlaid values
Hi there,

first of all thanks for that great tool! The 3D rendering is great!

I have a question concerning the visualization of an overlay.
I performed data analysis on fMRI data in SPM and used SnPM (non-parametrical toolbox for SPM) to do 2nd level analysis. With that, I also receive t-contrasts similar to those in SPM and I can display them as an overlay in MRIcroGL. I just have not found out, how they are scaled. The values don't make sense for any values I would imagine :) Mine are between .002 and 1. But what do they mean? My actual t-values would be around 3 and 4 and my actual p-values <.001. Or are these transformed values? The visualization is nice but I now need to know what it actually means :)

I would be grateful for an answer :)

Katy
Attachment: scaleMeaning.jpg
Aug 24, 2017  04:08 AM | Chris Rorden
RE: Scaling of overlaid values
Katy-
  Please look at section 7 of the PDF manual included with MRIcroGL. You can adjust the minimum and maximum ranges of your overlay by adjusting the 'min' and 'max' values in the Overlay panel, so in Figure 7.1 the red overlay hides values darker than 2 and scales the intensity of red from 2 to 4. In contrast, the green overlay hides values with less extreme values than -2 and linearly scales the green intenstiy from -2 to -4. The intensity of your overlay will depend on the software that generated the statistical map, for example NiiStat saves maps as Z-scores, whereas SPM typically saves T-scores. The software that generated your map will generally suggest a threshold for your map. If you open an overlay in MRIcroGL and choose Display/Multiplanar you will see crosshairs on top of your image, as you click to select different locations of the image, the intensity of the overlay is shown in the title bar. 
 In sum, the "meaning" of your map depends on the software that created the overlay. It might be a T-score, a Z-score, a binary mask, etc. So in your case I would look at the SnPM tools for more details.