I had multi-shell data processed with the ACT pipeline in July (right after upgrading and building MRtrix). The results of dwi2fod are in the image below.
With the latest version of MRtrix upgraded last week and with the same data, I started the dwiintensitynorm pipeline. I did bias correction, dwiintensitynorm and then I used the average response functions. Before dwi2fod, I aligned the images with the anatomicals, as in the previous case. I obtained the following results:
First of all, it seems that the model is not selecting different tissues.
Below old and new images for csf:
Do you have any recommendation on were to start looking to solve this problem? I think I repeated the same steps as before (aside from the initial group normalization to the data).
PS This is the command I issued in both cases:
dwi2fod msmt_csd -mask brainmask_file Norm_dwi_file wmAvgResponse_file wmCSD_file gmAvgResponse_file gmCSD_file csfAvgResponse_file csfCSD_file -grad grad_file
I’m not sure exactly what the problem is, but maybe the first thing to look at are the response functions - can you post the output of:
I’m also struggling to make sense of the images you’re showing. When you say:
Is it the ODFs that we should be looking at…? The background image seems identical, and appears to correspond to the WM image. But if the ODFs shown do correspond to CSF, then this is definitely wrong, the ODFs should be isotropic - i.e. it should be a simple scalar density, with no orientation structure… Looking at the response files should help clarify this.
Thank you very much @jdtournier , there was a problem with the average response files indeed.
I am sending the corrected ones just in case.
And yes, all images are ODF images over anatomicals.
853.03442 0.00000 0.00000 0.00000 0.00000
442.25191 -142.03415 31.49591 -4.34229 0.00000
284.38205 -125.43598 58.77800 -19.86846 4.75597