help
help > RE: CONN with harmonized data
Sep 16, 2021 05:09 AM | Yann Quide
RE: CONN with harmonized data
Hi all,
Bumping this one up as I am also interested.
I have few questions though to confirm my approach is correct.
Seed to voxel analyses
I have been using the neuroHarmonize tools to harmonize the seed to voxels maps, replacing the original maps by the harmonized maps in the firstlevel folder (I kept a copy of the original maps somewhere just in case...), and re-ran the second-level analyses, over-writing the previous analyses.
Does that sound appropriate?
Does CONN need to access the resultsROI_Subjects0XX_Condition001.mat for these analyses?
What would be the best way to deal with this?
I am not sure I see how to first harmonize these matrices.
ROI-to-ROI analyses
For the ROI-to-ROI analyses, I believe CONN is reading the correlation in resultsROI_Subjects0XX_Condition001.mat, right?
Or would it be more appropriate to harmonize other files following preprocessing?
Again, I am not sure I see how to first harmonize these matrices.
se_*.nii files
How are the se_Subject0XX_Condition001.nii files used? Harmonization of these files failed (mean of empty slice?).
Thanks for your time.
Yann
Originally posted by Alexandra Muratore:
Bumping this one up as I am also interested.
I have few questions though to confirm my approach is correct.
Seed to voxel analyses
I have been using the neuroHarmonize tools to harmonize the seed to voxels maps, replacing the original maps by the harmonized maps in the firstlevel folder (I kept a copy of the original maps somewhere just in case...), and re-ran the second-level analyses, over-writing the previous analyses.
Does that sound appropriate?
Does CONN need to access the resultsROI_Subjects0XX_Condition001.mat for these analyses?
What would be the best way to deal with this?
I am not sure I see how to first harmonize these matrices.
ROI-to-ROI analyses
For the ROI-to-ROI analyses, I believe CONN is reading the correlation in resultsROI_Subjects0XX_Condition001.mat, right?
Or would it be more appropriate to harmonize other files following preprocessing?
Again, I am not sure I see how to first harmonize these matrices.
se_*.nii files
How are the se_Subject0XX_Condition001.nii files used? Harmonization of these files failed (mean of empty slice?).
Thanks for your time.
Yann
Originally posted by Alexandra Muratore:
Hi all,
I recently harmonized data collected from different scanners using ComBat. To do so, I parcellated the T1-weighed images into 136 ROIs (using the default CONN atlas), extracted the connectivity matrix for each subject, and put it through the ComBat pipeline to get adjusted values that correct for site specific effects. Now, I want to import these data back into CONN and complete my seed-to-ROI and seed-to-voxel analyses. To do this, I labeled the harmonized matrix for each subject as "x" and loaded it into the Matlab environment, which I believe CONN uses when conducting second level analysis.
My questions are:
1. If I want to run a seed-to-voxel analysis, will CONN use the connectivity matrix of 136 ROIs to analyze these data, or will it look at all voxels in the brain (unharmonized)?
2. If it does indeed look at the connectivity between the seed and all the voxels in the brain, how can I tell what values the analysis is based on? I tried loading in an SPM.mat file from the second analysis and looking at the "x" variable within this, and it shows a matrix of 136x136x201 (my number of subjects). Does this mean it is relying on the harmonized matrix?
3. Once I close out of Matlab, is "x" reset to its original data, or is it permanently replaced by the harmonized data that I loaded in?
Thanks so much,
Alex
I recently harmonized data collected from different scanners using ComBat. To do so, I parcellated the T1-weighed images into 136 ROIs (using the default CONN atlas), extracted the connectivity matrix for each subject, and put it through the ComBat pipeline to get adjusted values that correct for site specific effects. Now, I want to import these data back into CONN and complete my seed-to-ROI and seed-to-voxel analyses. To do this, I labeled the harmonized matrix for each subject as "x" and loaded it into the Matlab environment, which I believe CONN uses when conducting second level analysis.
My questions are:
1. If I want to run a seed-to-voxel analysis, will CONN use the connectivity matrix of 136 ROIs to analyze these data, or will it look at all voxels in the brain (unharmonized)?
2. If it does indeed look at the connectivity between the seed and all the voxels in the brain, how can I tell what values the analysis is based on? I tried loading in an SPM.mat file from the second analysis and looking at the "x" variable within this, and it shows a matrix of 136x136x201 (my number of subjects). Does this mean it is relying on the harmonized matrix?
3. Once I close out of Matlab, is "x" reset to its original data, or is it permanently replaced by the harmonized data that I loaded in?
Thanks so much,
Alex
Threaded View
| Title | Author | Date |
|---|---|---|
| Alex G | Sep 2, 2021 | |
| Yann Quide | Sep 16, 2021 | |
| Yann Quide | Sep 28, 2021 | |