Datasets to add for the next version of the model

[naga-karthik]

Datasets included in v2.4 model:

• "basel-mp2rage"
• "canproco"
• "data-multi-subject"
• "dcm-zurich"
• "lumbar-epfl"
• "lumbar-vanderbilt"
• "sct-testing-large"

Datasets to be included next:

• [ ] Beijing MS (has several different contrasts)
• [ ] Karolinska (mainly T1w, T2w, T2star)
• [ ] lumbar-marseille
• [x] dcm-brno
• [ ] philadelphia-paediatric (at the moment, T1w and T2w)
• [x] sci-colorado
• [x] sci-paris
• [x] sci-zurich
• [x] nih-ms-mp2rage

Optional (NOTE - this could destroy the model):

• [ ] stanford-epi (open-neuro epi spinalcord; good to use as testing data)

EDIT: updated with correct list of datasets in v2.4 (cross-ref'd from datasplits.md in the v2.4 release)

[naga-karthik]

Update: Here is the current list of datasets used for training the model (potentially v2.5). This function outputs the following info in a txt file gathering all the datasets statistics.

datasets stats

``` DATASETS USED FOR MODEL TRAINING (n=14): - datasplit_dcm-zurich-lesions-20231115_seed50.json - datasplit_sci-paris_seed50.json - datasplit_lumbar-epfl_seed50.json - datasplit_data-multi-subject_seed50.json - datasplit_dcm-brno_seed50.json - datasplit_basel-mp2rage_seed50.json - datasplit_sct-testing-large_seed50.json - datasplit_canproco_seed50.json - datasplit_sci-colorado_seed50.json - datasplit_dcm-zurich_seed50.json - datasplit_nih-ms-mp2rage_seed50.json - datasplit_lumbar-vanderbilt_seed50.json - datasplit_dcm-zurich-lesions_seed50.json - datasplit_sci-zurich_seed50.json SPLITS ACROSS DIFFERENT CONTRASTS (n=11): | Contrast | #train_images | #validation_images | #test_images | |:-----------|----------------:|---------------------:|---------------:| | dwi | 176 | 37 | 50 | | mp2rage | 142 | 16 | 18 | | mt-off | 158 | 36 | 49 | | mt-on | 167 | 37 | 50 | | psir | 217 | 49 | 67 | | stir | 56 | 15 | 18 | | t1map | 71 | 8 | 9 | | t1w | 248 | 45 | 63 | | t2star | 653 | 102 | 111 | | t2w | 1434 | 226 | 254 | | unit1 | 152 | 19 | 20 | | TOTAL | 3474 | 590 | 709 | PATHOLOGY-WISE SPLIT: | Pathology | Number of Subjects | |:--------------------|---------------------:| | MS | 894 | | HC | 407 | | MildCompression | 57 | | DCM | 239 | | MildCompression/DCM | 88 | | SCI | 19 | | ALS | 32 | | NMO | 19 | | TOTAL | 1755 | ```

[naga-karthik]

Datasets that could be used for evaluation:

• whole-spine
• lumbar-vanderbilt sagittal T2w images
• dcm-oklahoma (maybe)
• EPI datasets