I wasn’t sure what this line was supposed to mean:
#Unclear if the line below is needed; it seems to make the coregistration worse. Maybe it is only needed for atlases aside from the default FreeSurfer atlases
mrtransform {SUBJ}_parcels.mif -interp nearest -linear diff2struct_mrtrix.txt -inverse -datatype uint32 {SUBJ}_parcels_coreg.mif
It says that this line of commands can make coregisteration ‘worse’.
I used the Desikan-Killiany Atlas via freesurfer for parcellation, so I supposed that it may be better for me to not use the line.
But i still did try and got these two different images:
this image was created without running mrtransform,
and this image was made with running mrtransform.
Would you recommend running the line or not?
Thank you in advance!
there are many people more experienced than me to talk about this, but I will try . For me the first thing would be to visualy check it - anytime you are trying to evaluate what is better for registration, first step should be to look at your data before and after the registration step. Can you load your preprocessed DWI image (you used to generate wmfod_norm image for generating tractography) and look at the ${SUBJ}_parcels.mif and ${SUBJ}_parcels_coreg.mif as an overlay? Something like this:
Well, you can clearly see that it was beneficial here
About the actual connectivity matrix. Maybe it is just a misunderstanding, please provide a colorbar, but it looks like it is going from blue (zero on diagonal), through green to yellow (highest connectivity). So what is, for example, second image saying to me is that connectivity between homologous regions in different hemispheres is higher than a structural connectivity within hemispheres. That is not anatomically correct. Maybe somebody in a connectomics can answer this better. I assume you are using Matlab’s imagesc, what is a vector you are using to plot this image?
I checked the coregistered and non-coregistered images, and the coregistered ones were clearly better. I would definitely see the line as beneficial, thank you.
For the mean FA matrix, it is true that inter-hemispheric connections are showing stronger than the intra-hemispheric connections. I’ve seen explanations saying that it could happen (like Weighted mean FA connectomes ?)… but it still seems problematic to me too. I’ve been searching for a good thresholding value, and also been thinking of going for another weighting measure instead of mean FA😂
these are the lines I used for making the connectome, just to add:
Just to add - I’ve been doing some research and figured that the stronger connection among inter-hemispheric regions could be due to high FA value in the corpus callosum. Probably will be taking that into account.
Hi Kelly,
you can still try CSD. There is a single shell version and even it is not solving partial volume problem it is way better in voxels with crossing-fiber problem
Yes, can be. I am not experienced in FA connectivity matrix, but this does make sense.
. For me the first thing would be to visualy check it - anytime you are trying to evaluate what is better for registration, first step should be to look at your data before and after the registration step. Can you load your preprocessed DWI image (you used to generate wmfod_norm image for generating tractography) and look at the 
