Data organization/Group Preprocessing/Statistics

1

I’m going to perform the seed-based tractography analyses with a prior defined mask.

I have a couple of questions regarding the MRtrix.

  1. What would be the most efficient way to prepare the data sets for each of subjects within the different clinical group before preprocessing?

  2. Would it possible to incorporate batch processing with “foreach” command to handle the multiple tasks at the same time?

  3. How MRtrix handle the group comparison between control and disease group (For example, two sample-t, ANCOVA, and non-parametric TFCE permutation)?

  4. How to export the results from the FBA for further external statistical analysis?

Thanks for your help and considering my request.

Cheers

Larry

2

Hi Larry,

  1. It’s not clear exactly what the intent of the question is here. Since you stated “before preprocessing”, this sounds like a question about data management. If that’s the case, then certainly converting from raw DICOM into some kind of standardised filesystem convention would be a good start, and a necessity if you want to automate processing across subjects. Personally at this point I convert everything to BIDS, and automate my analyses to operate on such, but that requires either an existing pipeline designed to operate on such, or the ability to create such.

  2. Yes, it is precisely the situation of performing the same operation across a large number of subjects that foreach was built for.

  3. There are multiple commands in MRtrix3 for performing non-parametric permutation testing using a general linear model. The detail missing here is what quantities you want to compare between groups; “seed-based tractography analysis” is not adequately specific to communicate this. You mention FBA in point 4, but even if using tratography to define a fixel mask to constrain fixel-based statistical inference, I wouldn’t refer to such a pipeline as “seed-based tractography analysis”.

  4. It depends on exactly what information you want, and what you expect to be able to do with it. The outputs from fixel-based statistical inference are fixel data files; while the raw underlying intensities can be accessed fairly trivially, other software packages are simply not designed to handle data of this nature.

Rob