Misaligned Genu of Corpus Callosum in version 3.2

Submitted by ceierud on

Dear Tortoise Team!

Thanks for a great product and recent improvements, liking Tortoise version 3.2 so much that we want to move from version 3.1.4 to 3.2. However, the DIFFPREP command still seems to have a better aspect in 3.1.4 for our images, seeing that all our DTIs are well registered with 3.1.4 to all our T2s. However, when we use DIFFPREP from Tortoise 3.2, one subject becomes corrupted in the genu of the corpus callosum. Except from using two different DIFFPREP versions (3.1.4 and 3.2), everything else was identical, including registration settings file and images (T2, DTI, bvec, bval). For other reasons we want to move all our scripts to Tortoise 3.2, but we could stretch to use DIFFPREP from Tortoise 3.1.4 and EstimateTensorNLLSRESTORE and other commands from Tortoise 3.2 if possible. However, at web page https://tortoise.nibib.nih.gov/faq I noticed the following quote "We highly recommend not combining processing results from two different versions of the software", wondering if the mix explained above between 3.1.4 and 3.2 could be an okay combination or if you have another suggestion of how we may improve our registration in DIFFPREP of version 3.2 even for our data? Please note that in DIFFPREP 3.2 we have tried other distortion corrections than ANTSSyn, but ANTSSyn seems best.

Best Regards,


Submitted by irfanogl on Tue, 03/03/2020 - 11:23


First of all apologies for the late reply.

For some reason, the system did not send me an email about the question so I was not aware of your question till today.

From your question, I assume you are doing EPI distortion correction in DIFFPREP by registering the b=0 to the structural image?

We changed a couple of things in V3.2 regarding this correction.  It should have made things more robust actually not the other way. However, again, EPI distortion correction with DIFFPREP is not something we focus on as most of our EPI distortion correction is currently done with blip-up blip-down with DRBUDDI.

Coming to your question,  PROCESSING-wise we strongly suggest not to mix versions. For example, DIFFPREP V3.0 + DRBUDDI V3.2 would not work. However, DIFFCALC opreations such as tensor fitting, RESTORE, MAPMRI are not included in this. So you can use the fitting executables from the last version with data processed with previous versions.


Hope this helps.




Submitted by ceierud on Fri, 03/06/2020 - 16:50


Thank you so much for your clarification about what versions not to mix and DRBUDDY. Your distortion correction assumption was correct too. Tortoise rocks!