CompImg / LST-AI

LST-AI - Deep Learning Ensemble for Accurate MS Lesion Segmentation
https://doi.org/10.1016/j.nicl.2024.103611
MIT License
20 stars 4 forks source link

Group-analysis using LST-AI output #15

Closed MMSchmitgen closed 3 months ago

MMSchmitgen commented 4 months ago

Not an issue, but some questions.

  1. Are the *.nii.gz output-files from LST-AI in MNI space, when using LST-AI with default settings? From reading your preprint on LST-AI (https://doi.org/10.1101/2023.11.23.23298966) this is not clear to me.
  2. I would like to use the seg-lst.nii.gz output images for whole-brain group comparisons. I tried setting up a 2nd-level model in SPM (after decompressing the .nii.gz files), but SPM drops an error using spm_check_orientations: "The orientations etc must be identical for this procedure.". Any suggestions on how I could perform whole-brain group comparisons using LST-AI output?

Cheers and thanks a lot, Mike

twiltgen commented 4 months ago

Hi Mike,

I hope I can address your questions: 1) The default output file of LST-AI is the lesion mask in the FLAIR image space "space-flair_seg-lst.nii.gz". If you use the "--temp path/to/temp/folder" flag, then also the auxiliary files generated during processing will be saved, including the lesion mask in MNI space "sub-X_ses-Y_space-mni_seg-lst.nii.gz". 2) I guess you have used the default (decompressed) "space-flair_seg-lst.nii.gz" output to run the group comparison with SPM? Maybe you can try it with the "space-mni_seg-lst.nii.gz" lesion masks, which will be available after running LST-AI with the "--temp" flag. They should all be in the same image space and have the same orientation.

Cheers, and let me know if this helps. Tun

MMSchmitgen commented 4 months ago

Hi Tun,

thanks a lot for your quick reply. Ah, I see. I will give it a shot and keep you updated (might take some days, though).

Cheers, Mike

MMSchmitgen commented 4 months ago

Hi Tun,

it worked like a charm. Thank you so much for your help. :)

Cheers and have a nice weekend, Mike