propagate_and_score

Self‑contained inference pipeline to propagate tumor masks and/or derive future masks from a Time‑to‑Event (T2E) map using trained models.

TODO

• Add code for training
• Make scripts easier to understand/follow

1) Run only prediction (produces a T2E map)

python src/run_pipeline.py predict
--bravo data/cas_1/t1_bravo_bet.nii.gz
--flair data/cas_1/t2_flair_bet.nii.gz
--mask data/cas_1/tumor_1.nii.gz
--out_dir data/cas_1/output_1
--t2e_ckpt saved_models/v17/propagation_unet_best.pth
--mni_ref data/mni_masked.nii.gz

2) Run only scoring (uses an already-produced T2E map)

python src/run_pipeline.py score
--R data/cas_1/tumor_1_dil.nii.gz
--T1 data/cas_1/tumor_1.nii.gz
--G data/cas_1/output_1/t2e_map.nii.gz
--out_dir data/cas_1/output_1 --save_maps
--mni_ref data/mni_masked.nii.gz

3) Full pipeline (predict → score); by default score uses the predicted G

python src/run_pipeline.py both
--bravo data/cas_1/t1_bravo_bet.nii.gz
--flair data/cas_1/t2_flair_bet.nii.gz
--mask data/cas_1/tumor_1.nii.gz
--pred_out_dir data/cas_1/output_1
--t2e_ckpt saved_models/v17/propagation_unet_best.pth
--model unet3d_larger_skip
--R data/cas_1/tumor_1_dil.nii.gz
--save_maps
--mni_ref data/mni_masked.nii.gz

python src/run_pipeline.py both
--bravo data/cas_1/t1_bravo_bet.nii.gz
--flair data/cas_1/t2_flair_bet.nii.gz
--mask data/cas_1/tumor_2.nii.gz
--pred_out_dir data/cas_1/output_2
--t2e_ckpt saved_models/v19/propagation_unet_best.pth
--model unet3d
--R data/cas_1/tumor_2_dil.nii.gz
--save_maps
--mni_ref data/mni_masked.nii.gz