search

spinal-cord-7t / coil-qc-code

7T Spinal Cord Coil QC Analysis Code
0 stars 0 forks source link

Combining flip-angle maps from different DREAM voltages #12

Closed Kyota-exe closed 5 months ago

Kyota-exe commented 7 months ago

In PR #9, it was discovered that the scanner already outputs both the FA and RefV maps for DREAM scans in this comment, and that we can simply use those for the dataset. However, it was stated earlier in the discussion in this comment that the FA maps from different voltages should be combined to minimize bias.

At the moment, the bids_conversion.py script simply takes the DREAM scans (FA map and RefV) from the optimal reference voltage (x1.0), instead of combining different voltages into a single FA map. Is this an issue? Should we modify the script so that it combines the DREAM scans?

evaalonsoortiz commented 7 months ago

No, it is better if the notebook retrieves all 3 scans (0.66*refV, refV, 1.5refV) and combines them.

Kyota-exe commented 7 months ago

I apologize if this is obvious, but how would we go about combining these scans? This is something that is not clear to me, and I would appreciate some clarification; my understanding was that we would combine all 3 FAmap NIFTI scans into a single NIFTI file, and put this in the dataset. I am beginning to understand now though that you may have meant that we should simply take the average of the data we get in all 3 voltages when we process the data in the analysis code? So we don't actually combine the scans themselves, but the results acquired from the scans...?

Assuming you do mean that we should combine the 3 NIFTI scans from different voltages into a single NIFTI scan: As you recently stated in this comment from the original bids_conversion.py PR, the 0.66x and 1.5x scans are not currently included in the dataset. Should I include them all in the dataset and do the combining in the analysis code, or handle this "combining" in the bids_conversion.py script and have it already be combined in the OpenNeuro dataset? For the same reason @jcohenadad recommended we put the processing code to generate the FAmap DREAM scans in the analysis code instead of in a pre-processing script here, my vote would be to put it in the analysis script.

If we put the FAmap DREAM scans for all 3 voltages in the OpenNeuro dataset however, how should we name these files to be BIDS compliant? The 1.0x FAmap scan is currently named sub-SUBJECT_acq-famp_TB1DREAM, so maybe sub-SUBJECT_acq-famp-0.66_TB1DREAM and sub-SUBJECT_acq-famp-1.5_TB1DREAM? Maybe @jcohenadad can pitch in on this?

evaalonsoortiz commented 7 months ago

The 3 DREAM scans should be part of the OpenNeuro dataset. We will combine them into one scan using the approach described here https://github.com/spinal-cord-7t/coil-qc-code/issues/7#issuecomment-1939582828 in the analysis script.

I'm not sure about the labeling, and what is and isn't accepted. I'll try to see if I can find examples.

evaalonsoortiz commented 7 months ago

Based on the acq-

Kyota-exe commented 7 months ago

Alright, thank you. I've added the 0.66x and 1.5x DREAM FAmap scans to the database with the filenames I proposed. Before I update the database on the Google Drive, should I also include the REFVOLTMAP scans for all 3 voltages, or is there no use for this?

evaalonsoortiz commented 7 months ago

I'm not sure yet if we need them or not. It might be that we have all the info we need from the flip angle maps, I'll consult with others. In the meantime add them, and we can remove them later if we don't need them.

Kyota-exe commented 5 months ago

This has been handled by 8522cc5. Can we close this issue?