Validate the entire CVR pipeline.

[servoz]

Basically, there are two parts: the "pipeline" itself and the "reporting" part (which is necessarily linked to the results obtained in the "pipeline" part).

There are already some points to consider:

• Reporting part:

  • 1 We observe undefined values in the Mia report compared with the Amigo report. A quick preliminary examination indicates that the necessary tags are not in the database (see with mri_conv?).
  • 2 Check that all data is correct (are we taking the right parameters from the database?).
  • 3 Try to have the same slices in the mosaic display (the slices represented in the Amigo and Mia reports are not exactly the same).
  • 4 Absence of Greek alphabet (depending on the station, consider adding it to the code?).
  • 5 In python (Matplotlib), there is no color map like the one used in Matlab to generate parametric maps (as in the Amigo report). Do we need to create our own to keep the same color codes?
  • 6 Some territories are not visible, page 11 - 12. They seem to be the territories in "chocolate" and "moccasin" colors => posterior inferior cerebellar artery, insular lobe, striatum).
  • 7 The beta map is not visible in the report.
  • 8 The territories and ROIs at the end of the report seem corrupted (numerous artefacts).

• Pipeline part:

  • 9 Different laterality index values (there's already an open ticket for this exact point #41 ).
  • 10 Vascular territories and lobes do not appear identical (probably linked to the already open ticket #41 ).

There are most certainly other points to add! We either have to understand and validate, or modify the codes so that we no longer observe any differences.

[servoz]

• 9: done
• 10: done

[servoz]

• 8: done
• 7: done

[servoz]

• 3: done. For the anat it's OK. for the func it's not exactly the same thing because in Amigo there's a resampling to the size of the anat. We don't do that in Mia at the moment. So there are far fewer cuts and it's impossible in these conditions to get the same thing as with Amigo. It's just a cosmetic issue and we might as well wait and see if users want it.

[servoz]

• 4: done.

[servoz]

• 6: Done

[servoz]

Our results are now in line with those of Amigo (there are still some very small differences which do not affect the result of the GLM analysis). The report is similar to that of Amigo.

I think the CVR pipeline can now be run into production.

There are still 3 points, in this ticket:

• 5: I really don't think this is a priority. We've moved from matlab/SPM to python and the default cmaps aren't exactly the same. Maybe it's just a habit to change. If users want to ... we can always work on it later!
• 1 and 2: to do this correctly, you'd have to see with mri_conv. I'm going to open a ticket with mri_conv ASAP.

I think this ticket can now be closed.