help > Overlay underestimation issue
Showing 1-2 of 2 posts
Feb 4, 2020 02:02 PM | Damien Marie - Brain and Language Lab
Overlay underestimation issue
Hi,
I'm running into troubles with the multi-planar overlay of a functional map (macOS Catalina 10.15.2).
I manually selected a threshold of t values, however the clusters are way smaller than when I display the map in the SPM gui or MRIcro at the same threshold (see screen attached capture). I did not select anything in the overlay options.
Any ideas?
Thank you and best,
Damien
I'm running into troubles with the multi-planar overlay of a functional map (macOS Catalina 10.15.2).
I manually selected a threshold of t values, however the clusters are way smaller than when I display the map in the SPM gui or MRIcro at the same threshold (see screen attached capture). I did not select anything in the overlay options.
Any ideas?
Thank you and best,
Damien
Feb 4, 2020 03:02 PM | Chris Rorden
RE: Overlay underestimation issue
The issue is that you have thresholded an image and now you are
resizing it. This will either result in jagged artifacts (if you
use nearest neighbor, e.g. SPM) or shrinking of the volume (if you
use any other interpolation method). The original data is
inherently smooth, both due to partial volume effects and the fact
we Gaussian blur data. Therefore, sub-threshold voxels near
super-threshold voxels are typically very close to the threshold.
However, the thresholding artificially sets all sub-threshold
voxels to have an intensity of zero. You have two options:
1.) Upsample your data to match the resolution of your anatomical scan before applying the threshold
2.) Switch to nearest neighbor interpolation and live with the results.
This is described in the wiki.
I would suggest you upgrade to the current version of MRIcroGL 1.2 (1.2.20190902). You can choose whether overlays are loaded smooth or not using the Layers panel's "Options" pull down menu and toggling "Load Smooth Overlays". Likewise, the Python function "gl.overlayloadsmooth(1)" and "gl.overlayloadsmooth(0)" turn interpolation On/Off. The Scripting/Templates/Jagged provides a Python script that demonstrates this.
1.) Upsample your data to match the resolution of your anatomical scan before applying the threshold
2.) Switch to nearest neighbor interpolation and live with the results.
This is described in the wiki.
I would suggest you upgrade to the current version of MRIcroGL 1.2 (1.2.20190902). You can choose whether overlays are loaded smooth or not using the Layers panel's "Options" pull down menu and toggling "Load Smooth Overlays". Likewise, the Python function "gl.overlayloadsmooth(1)" and "gl.overlayloadsmooth(0)" turn interpolation On/Off. The Scripting/Templates/Jagged provides a Python script that demonstrates this.