open-discussion > File/Overlay (Remove Small Clusters) missing?
Showing 1-5 of 5 posts
Sep 5, 2019 05:09 PM | mullinsr
File/Overlay (Remove Small Clusters) missing?
I am using MRIcroGL 1.2.20190902 Cocoa on Mac OS Sierra 10.12.6
On prior versions, under File/Add Overlay, there was an option to remove small clusters which no longer appears.
Is this option going to return? I need to remove small clusters to match my extent thresholded figures in SPM.
Thanks,
Jake
On prior versions, under File/Add Overlay, there was an option to remove small clusters which no longer appears.
Is this option going to return? I need to remove small clusters to match my extent thresholded figures in SPM.
Thanks,
Jake
Sep 5, 2019 05:09 PM | Chris Rorden
RE: File/Overlay (Remove Small Clusters) missing?
I have submitted an enhancement request for you. Feel free to submit any others you have. While
this is not hard, I have a lot of other things on my plate, so
don't expect a fast response. The project is open source, so you
can make a fork and add features - once you are happy with them you
can issue a pull request to share your solution.
Since you note you are using SPM, this suggests you have Matlab. Therefore, you can use my Matlab script to create a cluster threshold.
Since you note you are using SPM, this suggests you have Matlab. Therefore, you can use my Matlab script to create a cluster threshold.
Sep 5, 2019 06:09 PM | mullinsr
RE: File/Overlay (Remove Small Clusters) missing?
Originally posted by Chris Rorden:
Thanks, I will give this a shot, using SPM to save all clusters (n-ary) style the Display/Save in the lower Results window seems to work as well.
I have submitted an enhancement request for you. Feel free to submit any
others you have. While this is not hard, I have a lot of other
things on my plate, so don't expect a fast response. The project is
open source, so you can make a fork and add features - once you are
happy with them you can issue a pull request to share your
solution.
Since you note you are using SPM, this suggests you have Matlab. Therefore, you can use my Matlab script to create a cluster threshold.
Since you note you are using SPM, this suggests you have Matlab. Therefore, you can use my Matlab script to create a cluster threshold.
Thanks, I will give this a shot, using SPM to save all clusters (n-ary) style the Display/Save in the lower Results window seems to work as well.
Sep 5, 2019 07:09 PM | Chris Rorden
RE: File/Overlay (Remove Small Clusters) missing?
Assuming your statistical map is lower resolution than the
background image, I would strongly encourage you to threshold the
data AFTER it has been upsampled, rather than BEFORE. Due to
spatial smoothing, the non-significant neighbors of significant
voxels are very likely to be ALMOST significant. However, this
information is lost when you threshold the data. If you plan to
upsample data that has been already thresholded, I would male sure
you have "Load Smooth Overlays" unchecked in the "Options" pulldown
of the "Layers" panel. The overlays may appear jagged if you do not
smooth them, but smoothing data that has been masked can be
misleading. You want the interpolation to work with all the
available data.
I mention this briefly in the cluster threshold scripting page.
I mention this briefly in the cluster threshold scripting page.
Sep 9, 2019 07:09 PM | mullinsr
RE: File/Overlay (Remove Small Clusters) missing?
Originally posted by Chris Rorden:
Assuming your statistical map is lower
resolution than the background image, I would strongly encourage
you to threshold the data AFTER it has been upsampled, rather than
BEFORE. Due to spatial smoothing, the non-significant neighbors of
significant voxels are very likely to be ALMOST significant.
However, this information is lost when you threshold the data. If
you plan to upsample data that has been already thresholded, I
would male sure you have "Load Smooth Overlays" unchecked in the
"Options" pulldown of the "Layers" panel. The overlays may appear
jagged if you do not smooth them, but smoothing data that has been
masked can be misleading. You want the interpolation to work with
all the available data.
I mention this briefly in the cluster threshold scripting page.
Good point, I had not considered
this.I mention this briefly in the cluster threshold scripting page.