Update whole-body sCT model

CBCT/CBCT_MODEL_INFO.md

@@ -0,0 +1,157 @@
+# ImpactSynth CBCT
+
+## Summary
+
+- Modality: whole-body CBCT to sCT synthesis
+- Dataset source: SynthRAD 2023 and SynthRAD 2025 Task 2
+- Distributed checkpoints: `CV_0.pt`, `CV_1.pt`, `CV_2.pt`, `CV_3.pt`, `CV_4.pt`
+- Inference-time test-time augmentation: `tta = 2`
+- Monte Carlo dropout: `false`
+- Framework: KonfAI
+- Architecture: 2.5D UNet++ with a ResNet34 encoder
+- Main perceptual prior: SAM 2.1-s through the IMPACT-Synth loss
+- Pair alignment: CT-CBCT pairs are aligned with IMPACTReg
+- Transform files:
+  - `https://huggingface.co/datasets/VBoussot/synthrad2023-impact-registration`
+  - `https://huggingface.co/datasets/VBoussot/synthrad2025-impact-registration`
+
+## Data splits
+
+The repository contains the validation and cross-validation splits used for Task 2.
+
+- Patients in the 2023 Task 2 training cohort: `291`
+- Patients in the 2025 Task 2 training cohort: `638`
+- Total patients across the paired 2023+2025 Task 2 cohorts: `929`
+- Additional out-of-distribution Task 2 patients: `103`
+- Region counts in the paired 2023+2025 Task 2 cohorts:
+  - `AB`: `195`
+  - `HN`: `232`
+  - `TH`: `211`
+  - `brain`: `160`
+  - `pelvis`: `131`
+- `Validation.txt`: `150` patients
+- `CrossValidation_0..4.txt` are validation folds associated with the distributed checkpoints
+
+- `Validation.txt`: 150 patients
+- `CrossValidation_0.txt`: 164 validation patients
+- `CrossValidation_1.txt`: 158 validation patients
+- `CrossValidation_2.txt`: 156 validation patients
+- `CrossValidation_3.txt`: 153 validation patients
+- `CrossValidation_4.txt`: 148 validation patients
+
+## Release CV performance
+
+Overall Release CV metrics:
+
+- Dice is reported with `TS = 3 mm` on the CT body mask.
+- All detailed metrics are reported in the tables below.
+
+- Dice: `0.758 +/- 0.113`
+- MAE: `53.55 +/- 17.79` HU
+- SSIM: `0.939 +/- 0.026`
+- PSNR: `32.09 +/- 2.93` dB
+- SAM: `13.70 +/- 6.02`
+- Reg: `1.04 +/- 0.03`
+- Uncertainty: `784.24 +/- 278.43`
+
+Per-fold overall Release metrics:
+
+| Fold | n | Dice | MAE (HU) | SSIM | PSNR (dB) | SAM | Reg | Uncertainty |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| CV | 150 | 0.758 +/- 0.113 | 53.55 +/- 17.79 | 0.939 +/- 0.026 | 32.09 +/- 2.93 | 13.70 +/- 6.02 | 1.04 +/- 0.03 | 784.24 +/- 278.43 |
+| CV_0 | 150 | 0.744 +/- 0.114 | 59.33 +/- 18.30 | 0.927 +/- 0.028 | 31.31 +/- 2.69 | 13.25 +/- 5.32 | 1.05 +/- 0.03 | n/a |
+| CV_1 | 150 | 0.755 +/- 0.112 | 58.26 +/- 17.45 | 0.931 +/- 0.028 | 31.34 +/- 2.66 | 13.24 +/- 5.48 | 1.04 +/- 0.03 | n/a |
+| CV_2 | 150 | 0.751 +/- 0.114 | 57.70 +/- 17.98 | 0.931 +/- 0.028 | 31.47 +/- 2.75 | 13.25 +/- 5.49 | 1.04 +/- 0.03 | n/a |
+| CV_3 | 150 | 0.748 +/- 0.115 | 57.98 +/- 18.18 | 0.931 +/- 0.028 | 31.42 +/- 2.77 | 13.09 +/- 5.30 | 1.04 +/- 0.03 | n/a |
+| CV_4 | 150 | 0.750 +/- 0.113 | 57.95 +/- 18.16 | 0.931 +/- 0.028 | 31.43 +/- 2.75 | 13.03 +/- 5.32 | 1.04 +/- 0.03 | n/a |
+
+Regional Release CV metrics:
+
+| Region | n | Dice | MAE (HU) | SSIM | PSNR (dB) | SAM | Reg | Uncertainty |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| AB | 32 | 0.673 | 57.18 | 0.926 | 31.12 | 17.71 | 1.05 | 730.95 |
+| HN | 37 | 0.717 | 67.53 | 0.946 | 29.97 | 8.85 | 1.05 | 661.14 |
+| TH | 34 | 0.724 | 57.27 | 0.928 | 32.03 | 16.24 | 1.05 | 793.57 |
+| brain | 26 | 0.905 | 39.42 | 0.970 | 33.96 | 7.22 | 1.01 | 879.35 |
+| pelvis | 21 | 0.835 | 34.84 | 0.924 | 35.12 | 20.07 | 1.02 | 949.44 |
+
+Regional Release metrics for CV_0:
+
+| Region | n | Dice | MAE (HU) | SSIM | PSNR (dB) | SAM | Reg |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| AB | 32 | 0.661 +/- 0.090 | 62.73 +/- 15.73 | 0.916 +/- 0.024 | 30.45 +/- 2.29 | 16.99 +/- 3.72 | 1.06 +/- 0.02 |
+| HN | 37 | 0.704 +/- 0.074 | 73.95 +/- 16.03 | 0.936 +/- 0.024 | 29.38 +/- 2.05 | 8.92 +/- 1.39 | 1.06 +/- 0.04 |
+| TH | 34 | 0.706 +/- 0.077 | 62.31 +/- 15.91 | 0.918 +/- 0.022 | 31.31 +/- 2.53 | 15.72 +/- 3.92 | 1.05 +/- 0.02 |
+| brain | 26 | 0.886 +/- 0.075 | 46.50 +/- 7.67 | 0.958 +/- 0.009 | 32.86 +/- 1.43 | 7.32 +/- 1.14 | 1.02 +/- 0.00 |
+| pelvis | 21 | 0.829 +/- 0.078 | 39.42 +/- 8.34 | 0.906 +/- 0.028 | 34.10 +/- 2.20 | 18.52 +/- 3.94 | 1.03 +/- 0.01 |
+| overall | 150 | 0.744 +/- 0.114 | 59.33 +/- 18.30 | 0.927 +/- 0.028 | 31.31 +/- 2.69 | 13.25 +/- 5.32 | 1.05 +/- 0.03 |
+
+Regional Release metrics for CV_1:
+
+| Region | n | Dice | MAE (HU) | SSIM | PSNR (dB) | SAM | Reg |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| AB | 32 | 0.668 +/- 0.084 | 61.42 +/- 14.23 | 0.919 +/- 0.024 | 30.44 +/- 2.17 | 17.18 +/- 4.16 | 1.06 +/- 0.02 |
+| HN | 37 | 0.719 +/- 0.076 | 71.83 +/- 15.20 | 0.942 +/- 0.022 | 29.43 +/- 1.99 | 8.79 +/- 1.33 | 1.06 +/- 0.04 |
+| TH | 34 | 0.719 +/- 0.070 | 62.74 +/- 15.15 | 0.917 +/- 0.020 | 31.26 +/- 2.47 | 15.74 +/- 3.85 | 1.05 +/- 0.02 |
+| brain | 26 | 0.901 +/- 0.075 | 44.74 +/- 6.76 | 0.963 +/- 0.008 | 33.02 +/- 1.40 | 7.12 +/- 1.09 | 1.02 +/- 0.00 |
+| pelvis | 21 | 0.831 +/- 0.076 | 38.99 +/- 7.73 | 0.911 +/- 0.023 | 34.13 +/- 2.26 | 18.59 +/- 4.00 | 1.03 +/- 0.01 |
+| overall | 150 | 0.755 +/- 0.112 | 58.26 +/- 17.45 | 0.931 +/- 0.028 | 31.34 +/- 2.66 | 13.24 +/- 5.48 | 1.04 +/- 0.03 |
+
+Regional Release metrics for CV_2:
+
+| Region | n | Dice | MAE (HU) | SSIM | PSNR (dB) | SAM | Reg |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| AB | 32 | 0.664 +/- 0.086 | 61.41 +/- 15.08 | 0.918 +/- 0.024 | 30.53 +/- 2.27 | 16.85 +/- 3.66 | 1.06 +/- 0.02 |
+| HN | 37 | 0.710 +/- 0.074 | 71.52 +/- 15.44 | 0.942 +/- 0.021 | 29.53 +/- 2.05 | 8.73 +/- 1.31 | 1.06 +/- 0.04 |
+| TH | 34 | 0.714 +/- 0.073 | 62.22 +/- 15.33 | 0.917 +/- 0.022 | 31.31 +/- 2.50 | 15.93 +/- 3.83 | 1.05 +/- 0.02 |
+| brain | 26 | 0.898 +/- 0.072 | 43.32 +/- 7.04 | 0.965 +/- 0.007 | 33.28 +/- 1.63 | 7.13 +/- 1.06 | 1.01 +/- 0.00 |
+| pelvis | 21 | 0.830 +/- 0.079 | 38.16 +/- 8.44 | 0.914 +/- 0.023 | 34.32 +/- 2.32 | 18.94 +/- 4.50 | 1.03 +/- 0.01 |
+| overall | 150 | 0.751 +/- 0.114 | 57.70 +/- 17.98 | 0.931 +/- 0.028 | 31.47 +/- 2.75 | 13.25 +/- 5.49 | 1.04 +/- 0.03 |
+
+Regional Release metrics for CV_3:
+
+| Region | n | Dice | MAE (HU) | SSIM | PSNR (dB) | SAM | Reg |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| AB | 32 | 0.658 +/- 0.087 | 62.08 +/- 15.29 | 0.917 +/- 0.025 | 30.47 +/- 2.32 | 17.00 +/- 3.63 | 1.06 +/- 0.02 |
+| HN | 37 | 0.706 +/- 0.076 | 72.26 +/- 15.58 | 0.940 +/- 0.022 | 29.44 +/- 2.04 | 8.81 +/- 1.40 | 1.06 +/- 0.04 |
+| TH | 34 | 0.717 +/- 0.075 | 61.98 +/- 15.42 | 0.918 +/- 0.022 | 31.30 +/- 2.50 | 15.61 +/- 3.89 | 1.05 +/- 0.02 |
+| brain | 26 | 0.895 +/- 0.071 | 43.45 +/- 6.54 | 0.964 +/- 0.007 | 33.17 +/- 1.47 | 7.08 +/- 1.23 | 1.02 +/- 0.00 |
+| pelvis | 21 | 0.832 +/- 0.077 | 38.05 +/- 8.24 | 0.913 +/- 0.025 | 34.38 +/- 2.37 | 18.06 +/- 3.95 | 1.03 +/- 0.01 |
+| overall | 150 | 0.748 +/- 0.115 | 57.98 +/- 18.18 | 0.931 +/- 0.028 | 31.42 +/- 2.77 | 13.09 +/- 5.30 | 1.04 +/- 0.03 |
+
+Regional Release metrics for CV_4:
+
+| Region | n | Dice | MAE (HU) | SSIM | PSNR (dB) | SAM | Reg |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| AB | 32 | 0.662 +/- 0.087 | 61.60 +/- 15.14 | 0.918 +/- 0.025 | 30.49 +/- 2.30 | 16.73 +/- 3.58 | 1.06 +/- 0.02 |
+| HN | 37 | 0.711 +/- 0.075 | 71.79 +/- 15.70 | 0.941 +/- 0.022 | 29.48 +/- 2.06 | 8.71 +/- 1.37 | 1.06 +/- 0.04 |
+| TH | 34 | 0.716 +/- 0.073 | 62.64 +/- 15.68 | 0.918 +/- 0.021 | 31.25 +/- 2.44 | 15.62 +/- 3.76 | 1.05 +/- 0.02 |
+| brain | 26 | 0.895 +/- 0.071 | 43.27 +/- 6.67 | 0.964 +/- 0.007 | 33.27 +/- 1.48 | 6.98 +/- 1.13 | 1.01 +/- 0.00 |
+| pelvis | 21 | 0.830 +/- 0.078 | 38.55 +/- 9.06 | 0.914 +/- 0.023 | 34.31 +/- 2.31 | 18.26 +/- 4.22 | 1.03 +/- 0.01 |
+| overall | 150 | 0.750 +/- 0.113 | 57.95 +/- 18.16 | 0.931 +/- 0.028 | 31.43 +/- 2.75 | 13.03 +/- 5.32 | 1.04 +/- 0.03 |
+
+## Split files
+
+### Validation.txt
+
+`2BA047`, `2BA092`, `2BA011`, `2BA083`, `2BA073`, `2BA006`, `2BA001`, `2BA056`, `2BA053`, `2BB062`, `2BB008`, `2BB044`, `2BB095`, `2BB048`, `2BB111`, `2BB026`, `2BB076`, `2BB005`, `2BC075`, `2BC012`, `2BC068`, `2BC010`, `2BC047`, `2BC055`, `2BC085`, `2BC082`, `2PA064`, `2PA016`, `2PA078`, `2PA006`, `2PA055`, `2PA007`, `2PA002`, `2PB110`, `2PB082`, `2PB093`, `2PB021`, `2PB043`, `2PB106`, `2PB004`, `2PC047`, `2PC060`, `2PC013`, `2PC021`, `2PC053`, `2PC096`, `2PC017`, `2ABA082`, `2ABA118`, `2ABA103`, `2ABA135`, `2ABA006`, `2ABA024`, `2ABA077`, `2ABA066`, `2ABB061`, `2ABB059`, `2ABB095`, `2ABB090`, `2ABB117`, `2ABB051`, `2ABB039`, `2ABB018`, `2ABC222`, `2ABC093`, `2ABC049`, `2ABC019`, `2ABC150`, `2ABC063`, `2ABC152`, `2ABC118`, `2ABE158`, `2ABE197`, `2ABE191`, `2ABE169`, `2ABE124`, `2ABE022`, `2ABE184`, `2ABE180`, `2HNA084`, `2HNA005`, `2HNA050`, `2HNA002`, `2HNA058`, `2HNA024`, `2HNA032`, `2HNA110`, `2HNA100`, `2HNA098`, `2HNB105`, `2HNB099`, `2HNB044`, `2HNB067`, `2HNB089`, `2HNB015`, `2HNB047`, `2HNB019`, `2HNB075`, `2HNC028`, `2HNC024`, `2HNC099`, `2HNC078`, `2HNC077`, `2HNC043`, `2HNC021`, `2HNC060`, `2HNC036`, `2HNE054`, `2HNE010`, `2HNE101`, `2HNE079`, `2HNE026`, `2HNE075`, `2HNE007`, `2HNE042`, `2HNE041`, `2THA113`, `2THA043`, `2THA117`, `2THA122`, `2THA005`, `2THA077`, `2THA101`, `2THA082`, `2THA099`, `2THB059`, `2THB038`, `2THB088`, `2THB055`, `2THB015`, `2THB031`, `2THB046`, `2THB014`, `2THC037`, `2THC102`, `2THC107`, `2THC097`, `2THC033`, `2THC068`, `2THC076`, `2THC111`, `2THC056`, `2THE119`, `2THE081`, `2THE069`, `2THE058`, `2THE040`, `2THE046`, `2THE099`, `2THE039`
+
+### CrossValidation_0.txt
+
+`2BA040`, `2BA058`, `2BA023`, `2BA027`, `2BA041`, `2BA064`, `2BA061`, `2BA009`, `2BA071`, `2BA055`, `2BB034`, `2BB003`, `2BB189`, `2BB033`, `2BB030`, `2BB090`, `2BB173`, `2BB145`, `2BB200`, `2BC048`, `2BC027`, `2BC042`, `2BC015`, `2BC089`, `2BC046`, `2BC001`, `2BC044`, `2BC090`, `2PA045`, `2PA074`, `2PA066`, `2PA017`, `2PA053`, `2PA069`, `2PA024`, `2PA070`, `2PB081`, `2PB087`, `2PB109`, `2PB070`, `2PB104`, `2PB003`, `2PB033`, `2PC049`, `2PC007`, `2PC029`, `2PC050`, `2PC063`, `2PC019`, `2PC008`, `2PC010`, `2ABA068`, `2ABA122`, `2ABA041`, `2ABA044`, `2ABA101`, `2ABA051`, `2ABA027`, `2ABA110`, `2ABA138`, `2ABB078`, `2ABB009`, `2ABB002`, `2ABB105`, `2ABB036`, `2ABB040`, `2ABB115`, `2ABB056`, `2ABB006`, `2ABC212`, `2ABC036`, `2ABC188`, `2ABC102`, `2ABC079`, `2ABC176`, `2ABC040`, `2ABC114`, `2ABE200`, `2ABE174`, `2ABE131`, `2ABE126`, `2ABE165`, `2ABE162`, `2ABE038`, `2ABE136`, `2HNA092`, `2HNA033`, `2HNA044`, `2HNA021`, `2HNA025`, `2HNA095`, `2HNA011`, `2HNA107`, `2HNA046`, `2HNA088`, `2HNA040`, `2HNB006`, `2HNB010`, `2HNB058`, `2HNB039`, `2HNB007`, `2HNB024`, `2HNB086`, `2HNB023`, `2HNB102`, `2HNB078`, `2HNC100`, `2HNC049`, `2HNC091`, `2HNC088`, `2HNC113`, `2HNC062`, `2HNC034`, `2HNC082`, `2HNC112`, `2HNC119`, `2HNE092`, `2HNE109`, `2HNE105`, `2HNE037`, `2HNE086`, `2HNE087`, `2HNE043`, `2HNE067`, `2HNE070`, `2HNE089`, `2THA056`, `2THA084`, `2THA132`, `2THA045`, `2THA081`, `2THA123`, `2THA052`, `2THA066`, `2THA092`, `2THA017`, `2THB106`, `2THB058`, `2THB068`, `2THB009`, `2THB064`, `2THB052`, `2THB098`, `2THB109`, `2THB026`, `2THC032`, `2THC099`, `2THC002`, `2THC067`, `2THC112`, `2THC091`, `2THC095`, `2THC117`, `2THC092`, `2THC084`, `2THE049`, `2THE080`, `2THE014`, `2THE015`, `2THE020`, `2THE117`, `2THE001`, `2THE006`, `2THE071`
+
+### CrossValidation_1.txt
+
+`2BA014`, `2BA039`, `2BA013`, `2BA049`, `2BA077`, `2BA066`, `2BA044`, `2BA051`, `2BA012`, `2BB082`, `2BB028`, `2BB099`, `2BB071`, `2BB066`, `2BB109`, `2BB051`, `2BB002`, `2BB100`, `2BC011`, `2BC021`, `2BC024`, `2BC043`, `2BC030`, `2BC033`, `2BC051`, `2BC061`, `2BC038`, `2PA088`, `2PA035`, `2PA012`, `2PA061`, `2PA037`, `2PA062`, `2PA063`, `2PA036`, `2PB057`, `2PB013`, `2PB117`, `2PB060`, `2PB024`, `2PB096`, `2PB084`, `2PC066`, `2PC097`, `2PC051`, `2PC041`, `2PC082`, `2PC069`, `2PC036`, `2PC042`, `2ABA075`, `2ABA032`, `2ABA038`, `2ABA064`, `2ABA005`, `2ABA048`, `2ABA037`, `2ABA136`, `2ABB043`, `2ABB113`, `2ABB015`, `2ABB005`, `2ABB033`, `2ABB087`, `2ABB024`, `2ABB029`, `2ABB100`, `2ABC018`, `2ABC094`, `2ABC146`, `2ABC089`, `2ABC066`, `2ABC151`, `2ABC230`, `2ABC109`, `2ABE187`, `2ABE120`, `2ABE107`, `2ABE010`, `2ABE170`, `2ABE206`, `2ABE073`, `2ABE177`, `2HNA083`, `2HNA113`, `2HNA097`, `2HNA014`, `2HNA045`, `2HNA068`, `2HNA039`, `2HNA089`, `2HNA041`, `2HNA082`, `2HNB002`, `2HNB065`, `2HNB107`, `2HNB085`, `2HNB054`, `2HNB081`, `2HNB091`, `2HNB093`, `2HNB045`, `2HNB071`, `2HNC097`, `2HNC087`, `2HNC042`, `2HNC101`, `2HNC108`, `2HNC057`, `2HNC085`, `2HNC030`, `2HNC055`, `2HNC123`, `2HNE103`, `2HNE083`, `2HNE006`, `2HNE050`, `2HNE018`, `2HNE093`, `2HNE049`, `2HNE028`, `2HNE095`, `2HNE084`, `2THA075`, `2THA107`, `2THA106`, `2THA070`, `2THA141`, `2THA069`, `2THA074`, `2THA022`, `2THA144`, `2THB004`, `2THB117`, `2THB063`, `2THB082`, `2THB101`, `2THB125`, `2THB120`, `2THB019`, `2THB115`, `2THC060`, `2THC029`, `2THC022`, `2THC045`, `2THC119`, `2THC017`, `2THC016`, `2THC116`, `2THC005`, `2THE068`, `2THE028`, `2THE109`, `2THE089`, `2THE035`, `2THE022`, `2THE021`, `2THE107`
+
+### CrossValidation_2.txt
+
+`2BA069`, `2BA037`, `2BA008`, `2BA063`, `2BA067`, `2BA028`, `2BA090`, `2BA094`, `2BA004`, `2BB179`, `2BB041`, `2BB075`, `2BB049`, `2BB177`, `2BB205`, `2BB052`, `2BB011`, `2BB017`, `2BC073`, `2BC022`, `2BC056`, `2BC007`, `2BC059`, `2BC026`, `2BC023`, `2BC050`, `2BC064`, `2PA081`, `2PA030`, `2PA058`, `2PA095`, `2PA050`, `2PA005`, `2PA038`, `2PB055`, `2PB080`, `2PB092`, `2PB030`, `2PB056`, `2PB048`, `2PB007`, `2PC034`, `2PC091`, `2PC002`, `2PC026`, `2PC003`, `2PC087`, `2PC012`, `2PC037`, `2ABA039`, `2ABA133`, `2ABA109`, `2ABA084`, `2ABA132`, `2ABA030`, `2ABA043`, `2ABA040`, `2ABB121`, `2ABB065`, `2ABB096`, `2ABB086`, `2ABB085`, `2ABB120`, `2ABB047`, `2ABB093`, `2ABB034`, `2ABC035`, `2ABC097`, `2ABC132`, `2ABC059`, `2ABC142`, `2ABC129`, `2ABC177`, `2ABC137`, `2ABE088`, `2ABE114`, `2ABE152`, `2ABE150`, `2ABE172`, `2ABE201`, `2ABE189`, `2ABE141`, `2HNA086`, `2HNA108`, `2HNA047`, `2HNA111`, `2HNA104`, `2HNA035`, `2HNA077`, `2HNA036`, `2HNA034`, `2HNA087`, `2HNB063`, `2HNB098`, `2HNB055`, `2HNB082`, `2HNB053`, `2HNB020`, `2HNB037`, `2HNB050`, `2HNB011`, `2HNB109`, `2HNC066`, `2HNC052`, `2HNC086`, `2HNC048`, `2HNC081`, `2HNC029`, `2HNC058`, `2HNC121`, `2HNC118`, `2HNE107`, `2HNE072`, `2HNE030`, `2HNE003`, `2HNE036`, `2HNE068`, `2HNE108`, `2HNE104`, `2HNE051`, `2HNE090`, `2THA140`, `2THA105`, `2THA058`, `2THA076`, `2THA047`, `2THA136`, `2THA030`, `2THA065`, `2THA059`, `2THB036`, `2THB030`, `2THB035`, `2THB070`, `2THB040`, `2THB007`, `2THB017`, `2THB094`, `2THB104`, `2THC030`, `2THC108`, `2THC073`, `2THC080`, `2THC085`, `2THC042`, `2THC094`, `2THC090`, `2THC001`, `2THE105`, `2THE072`, `2THE066`, `2THE010`, `2THE083`, `2THE101`, `2THE019`, `2THE023`
+
+### CrossValidation_3.txt
+
+`2BA076`, `2BA002`, `2BA035`, `2BA025`, `2BA085`, `2BA030`, `2BA029`, `2BA052`, `2BA080`, `2BB091`, `2BB182`, `2BB073`, `2BB059`, `2BB007`, `2BB152`, `2BB079`, `2BB098`, `2BB043`, `2BC070`, `2BC017`, `2BC019`, `2BC076`, `2BC081`, `2BC045`, `2BC074`, `2BC016`, `2PA019`, `2PA092`, `2PA040`, `2PA049`, `2PA026`, `2PA003`, `2PA059`, `2PB121`, `2PB119`, `2PB126`, `2PB068`, `2PB026`, `2PB083`, `2PB022`, `2PC080`, `2PC005`, `2PC058`, `2PC052`, `2PC039`, `2PC070`, `2PC015`, `2PC098`, `2ABA128`, `2ABA098`, `2ABA115`, `2ABA126`, `2ABA047`, `2ABA002`, `2ABA046`, `2ABA130`, `2ABB119`, `2ABB022`, `2ABB019`, `2ABB106`, `2ABB011`, `2ABB032`, `2ABB072`, `2ABB108`, `2ABC011`, `2ABC038`, `2ABC065`, `2ABC199`, `2ABC085`, `2ABC135`, `2ABC043`, `2ABC168`, `2ABE144`, `2ABE128`, `2ABE173`, `2ABE121`, `2ABE145`, `2ABE209`, `2ABE140`, `2ABE125`, `2HNA019`, `2HNA027`, `2HNA099`, `2HNA105`, `2HNA031`, `2HNA106`, `2HNA102`, `2HNA073`, `2HNA026`, `2HNA091`, `2HNB070`, `2HNB051`, `2HNB064`, `2HNB059`, `2HNB108`, `2HNB009`, `2HNB043`, `2HNB106`, `2HNB101`, `2HNB031`, `2HNC051`, `2HNC061`, `2HNC094`, `2HNC084`, `2HNC072`, `2HNC018`, `2HNC125`, `2HNC080`, `2HNC053`, `2HNE031`, `2HNE012`, `2HNE076`, `2HNE011`, `2HNE024`, `2HNE001`, `2HNE097`, `2HNE046`, `2HNE064`, `2THA097`, `2THA131`, `2THA108`, `2THA040`, `2THA019`, `2THA139`, `2THA126`, `2THA048`, `2THA096`, `2THB039`, `2THB053`, `2THB023`, `2THB072`, `2THB057`, `2THB079`, `2THB087`, `2THB011`, `2THB043`, `2THC025`, `2THC063`, `2THC015`, `2THC104`, `2THC020`, `2THC072`, `2THC040`, `2THC077`, `2THC011`, `2THE114`, `2THE104`, `2THE003`, `2THE042`, `2THE025`, `2THE004`, `2THE013`, `2THE097`
+
+### CrossValidation_4.txt
+
+`2BA003`, `2BA032`, `2BA010`, `2BA024`, `2BA031`, `2BA068`, `2BA036`, `2BA007`, `2BA045`, `2BB083`, `2BB072`, `2BB050`, `2BB198`, `2BB006`, `2BB151`, `2BB039`, `2BB096`, `2BB171`, `2BC072`, `2BC071`, `2BC040`, `2BC057`, `2BC086`, `2BC013`, `2BC088`, `2BC018`, `2PA056`, `2PA075`, `2PA067`, `2PA015`, `2PA047`, `2PA048`, `2PA073`, `2PB050`, `2PB125`, `2PB098`, `2PB036`, `2PB114`, `2PB018`, `2PC092`, `2PC024`, `2PC065`, `2PC011`, `2PC032`, `2PC073`, `2PC075`, `2ABA102`, `2ABA094`, `2ABA086`, `2ABA134`, `2ABA071`, `2ABA143`, `2ABA031`, `2ABA008`, `2ABB042`, `2ABB054`, `2ABB053`, `2ABB092`, `2ABB118`, `2ABB004`, `2ABB025`, `2ABB037`, `2ABC051`, `2ABC122`, `2ABC229`, `2ABC017`, `2ABC010`, `2ABC171`, `2ABC130`, `2ABC156`, `2ABE130`, `2ABE207`, `2ABE119`, `2ABE135`, `2ABE159`, `2ABE208`, `2ABE185`, `2HNA029`, `2HNA016`, `2HNA056`, `2HNA028`, `2HNA093`, `2HNA070`, `2HNA090`, `2HNA094`, `2HNA010`, `2HNA076`, `2HNB001`, `2HNB104`, `2HNB072`, `2HNB014`, `2HNB033`, `2HNB103`, `2HNB030`, `2HNB016`, `2HNB111`, `2HNC017`, `2HNC063`, `2HNC025`, `2HNC075`, `2HNC020`, `2HNC079`, `2HNC037`, `2HNC040`, `2HNC117`, `2HNE023`, `2HNE100`, `2HNE059`, `2HNE106`, `2HNE022`, `2HNE098`, `2HNE102`, `2HNE034`, `2HNE060`, `2THA010`, `2THA124`, `2THA115`, `2THA062`, `2THA078`, `2THA039`, `2THA034`, `2THA103`, `2THA146`, `2THB133`, `2THB003`, `2THB021`, `2THB051`, `2THB129`, `2THB074`, `2THB111`, `2THB002`, `2THC061`, `2THC019`, `2THC109`, `2THC105`, `2THC126`, `2THC114`, `2THC101`, `2THC039`, `2THC089`, `2THE106`, `2THE060`, `2THE077`, `2THE115`, `2THE100`, `2THE031`, `2THE045`, `2THE112`

CBCT/CV_0.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:085b0eceff7ade696b7f82df61956445f1a458e9a80300989fc1754d3541a574
-size 104526414
+oid sha256:4d838bd6ac56d6fc70360e6da9f0a06fdf184c497c8177e33a140544d0fa4551
+size 208936878

CBCT/CV_1.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4ca50c6be923c37b1f2cc2a803eb60626501dfdbf17af6ee94bd6e86575616b7
-size 104526414
+oid sha256:fac5aaa8879ab7ccd60aff670943ca394923e034de6544f7592a4baeb7fd4da6
+size 208936878

CBCT/CV_2.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:29e50e6d555ea9738615619bdf0d16a56120c1e6c2362937e6ae158a5d89db1c
-size 104526414
+oid sha256:0817dd34dd01fb89e0eb95fdd7c9fee760bb8d4356c6c19540d8e305d086c279
+size 208936878

CBCT/CV_3.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:927d9fa5bbee134bc6cb67bbf981a0bebf2dbab64685031924495d94b09f1f38
-size 104526414
+oid sha256:b6399bd93a6858848075febabc2805dc375cd2dbff08ece33e67700acdc9c30d
+size 208936878

CBCT/CV_4.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a1c4d4eaf66cd6818917fc6f851c1773176fea40c542b444a453b40cc52ad622
-size 104526414
+oid sha256:270543067874d1ddd4a59cdaf1d129d0b2f1250fbae461f273fd886f5119b30f
+size 208936878

CBCT/Config.yml

@@ -2,13 +2,6 @@ Trainer:
   Model:
     classpath: Model:UNetpp
     UNetpp:
-      schedulers:
-        StepLR:
-          step_size: 10
-          gamma: 0.75
-          last_epoch: -1
-          verbose: deprecated
-          nb_step: 0
       outputs_criterions:
         Head:Tanh:
           targets_criterions:
@@ -25,82 +18,120 @@ Trainer:
                   stop: None
                   accumulation: false
                   reduction: mean
-                IMPACTSynth:
+                IMPACTReg/1:
+                  name: SAM
+                  model_name: SAM2.1/SAM2.1_Small.pt
+                  shape:
+                  - 0
+                  - 0
+                  in_channels: 3
+                  loss: torch:nn:L1Loss
+                  weights:
+                  - 0
+                  - 1
+                  - 1
+                  size_average: None
+                  reduce: None
+                  reduction: mean
+                  schedulers:
+                    Constant:
+                      nb_step: 0
+                      value: 0.5
                   is_loss: true
                   group: 0
                   start: 0
                   stop: None
                   accumulation: false
-                  model_name: SAM2.1/Tiny_3_Layers.pt
-                  shape:
-                  - 0
-                  - 0
-                  in_channels: 3
-                  losses:
-                    torch:nn:L1Loss:
-                      weights:
-                      - 1
-                      - 1
-                      - 1
-                      size_average: None
-                      reduce: None
-                      reduction: mean
+            CT;MASK:
+              criterions_loader:
+                MAE:
                   schedulers:
                     Constant:
                       nb_step: 0
                       value: 1
-      pretrained: false
-      Optimizer:
-        name: AdamW
-        lr: 0.001
-        betas:
-        - 0.9
-        - 0.999
-        eps: 1e-08
-        weight_decay: 0.001
-        amsgrad: false
+                  is_loss: false
+                  group: 0
+                  start: 0
+                  stop: None
+                  accumulation: false
+                  reduction: mean
+      schedulers:
+        PolyLRScheduler:
+          initial_lr: 0.01
+          max_steps: 500
+          exponent: 0.9
+          current_step: 0
+          nb_step: 0
+      Patch: None
+      optimizer:
+        name: SGD
+        lr: 0.01
+        momentum: 0.99
+        dampening: 0
+        weight_decay: 3e-05
+        nesterov: true
         maximize: false
         foreach: None
-        capturable: false
         differentiable: false
         fused: None
+      nb_channel: 5
   Dataset:
     groups_src:
+      MASK:
+        groups_dest:
+          MASK:
+            transforms: None
+            patch_transforms: None
+            is_input: false
       CT:
         groups_dest:
           CT:
-            transforms: 
+            transforms:
               Clip:
                 min_value: -1024
                 max_value: 3071
                 save_clip_min: true
                 save_clip_max: true
                 mask: None
+              Statistics: {}
+              Normalize:
+                lazy: true
+                channels: None
+                min_value: -1
+                max_value: 1
+                inverse: false
+            patch_transforms:
               Normalize:
                 lazy: false
                 channels: None
                 min_value: -1
                 max_value: 1
-                inverse: true
-            patch_transforms: None
+                inverse: false
             is_input: false
-      CBCT:
+      CBCT_IMPACT:
         groups_dest:
           CBCT:
-            transforms: 
+            transforms:
               Clip:
                 min_value: min
                 max_value: percentile:99.5
                 save_clip_min: false
                 save_clip_max: false
                 mask: None
+              Statistics: {}
+              Normalize:
+                lazy: true
+                channels: None
+                min_value: -1
+                max_value: 1
+                inverse: false
+            patch_transforms:
               Normalize:
                 lazy: false
                 channels: None
                 min_value: -1
                 max_value: 1
-                inverse: true
-            patch_transforms: None
+                inverse: false
             is_input: true
     augmentations:
       DataAugmentation_0:
@@ -120,7 +151,7 @@ Trainer:
       overlap: None
       mask: None
       pad_value: -1
-      extend_slice: 2
+      extend_slice: 4
     subset: None
     shuffle: true
     filter: None
@@ -128,7 +159,7 @@ Trainer:
     - ./Dataset/:a:mha
     inline_augmentations: true
     use_cache: true
-    batch_size: 1
+    batch_size: 32
     validation: None
   train_name: FT_0
   manual_seed: 32

CBCT/Evaluation.yml

@@ -12,6 +12,22 @@ Evaluator:
               dynamic_range: None
             SSIM:
               dynamic_range: None
+            SAM_Perceptual: {}
+            IMPACTReg:
+              name: Reg
+              model_name: TS/M291.pt
+              shape:
+              - 0
+              - 0
+              - 0
+              in_channels: 1
+              loss: torch:nn:MSELoss
+              weights:
+              - 0
+              - 1
+              size_average: None
+              reduce: None
+              reduction: mean
     Output_seg:
       targets_criterions:
         Reference_seg;Mask:
@@ -25,19 +41,27 @@ Evaluator:
       Mask_0:
         groups_dest:
           Mask:
-            transforms: None
+            transforms:
+              TensorCast:
+                dtype: uint8
       Volume_0:
         groups_dest:
           Output:
-            transforms: None
+            transforms: 
+              Statistics: {}
+              TensorCast:
+                dtype: float32
+            patch_transforms: None
+            is_input: false
           Output_seg:
             transforms:
               KonfAIInference:
-                repo_id: VBoussot/MRSegmentator-KonfAI
-                model_name: MRSegmentator
-                number_of_ensemble: 1
+                repo_id: VBoussot/TotalSegmentator-KonfAI
+                model_name: total-3mm
+                checkpoints_name:
+                - M297.pt
                 number_of_tta: 0
-                number_of_mc_dropout: 0
+                number_of_mc: 0
                 per_channel: false
               Save:
                 dataset: ./Evaluations/ImpactSynth/Output:mha
@@ -45,15 +69,21 @@ Evaluator:
       Reference_0:
         groups_dest:
           Reference:
-            transforms: None
+            transforms:               
+              Statistics: {}
+              TensorCast:
+                dtype: float32
+            patch_transforms: None
+            is_input: false
           Reference_seg:
             transforms:
               KonfAIInference:
-                repo_id: VBoussot/MRSegmentator-KonfAI
-                model_name: MRSegmentator
-                number_of_ensemble: 1
+                repo_id: VBoussot/TotalSegmentator-KonfAI
+                model_name: total-3mm
+                checkpoints_name:
+                - M297.pt
                 number_of_tta: 0
-                number_of_mc_dropout: 0
+                number_of_mc: 0
                 per_channel: false
               Save:
                 dataset: ./Evaluations/ImpactSynth/Output:mha

CBCT/Model.py

@@ -16,12 +16,12 @@ class UNetpp(network.Network):
                     "default:ReduceLROnPlateau": network.LRSchedulersLoader(0)
                 },
                 outputs_criterions: dict[str, network.TargetCriterionsLoader] = {"default" : network.TargetCriterionsLoader()},
-                pretrained: bool = False):
-        super().__init__(in_channels = 3, optimizer = optimizer, schedulers = schedulers, outputs_criterions = outputs_criterions, dim = 2)
+                nb_channel : int = 5):
+        super().__init__(in_channels = nb_channel, optimizer = optimizer, schedulers = schedulers, outputs_criterions = outputs_criterions, dim = 2)
         self.add_module("model", smp.UnetPlusPlus(
             encoder_name="resnet34",
-            encoder_weights=None if not pretrained else "imagenet", 
-            in_channels=3,
+            encoder_weights=None, 
+            in_channels=nb_channel,
             classes=1,  
             activation=None  
         ))

CBCT/Prediction.yml

@@ -3,11 +3,25 @@ Predictor:
     classpath: Model:UNetpp
     UNetpp:
       outputs_criterions: None
-      pretrained: false
+      nb_channel: 5
   Dataset:
     groups_src:
       Volume_0:
         groups_dest:
+          MASK:
+            transforms:
+              KonfAIInference:
+                repo_id: VBoussot/TotalSegmentator-KonfAI
+                model_name: body
+                checkpoints_name:
+                - M598.pt
+                number_of_tta: 0
+                number_of_mc_dropout: 0
+                per_channel: false
+              Save:
+                dataset: ./Dataset:mha
+            patch_transforms: None
+            is_input: false
           Volume:
             transforms:
               Clip/0:
@@ -57,7 +71,7 @@ Predictor:
       overlap: None
       mask: None
       pad_value: -1
-      extend_slice: 2
+      extend_slice: 4
     subset: None
     filter: None
     dataset_filenames:
@@ -75,6 +89,9 @@ Predictor:
           TensorCast:
             dtype: int16
             inverse: false
+          Mask:
+            path: MASK
+            value_outside: -1024
         after_reduction_transforms: 
           InferenceStack:
             dataset: Predictions/ImpactSynth/Output:mha
@@ -82,10 +99,10 @@ Predictor:
             mode: mean
         final_transforms: None
         dataset_filename: Output:mha
+        inverse_transform: true
         group: sCT
         same_as_group: Volume_0:Volume
         patch_combine: None
-        inverse_transform: true
         reduction: Concat
         Concat: {}
   train_name: ImpactSynth

CBCT/Uncertainty.yml

@@ -19,22 +19,25 @@ Evaluator:
           Uncertainty:
             transforms:
               Variance: {}
+              Percentage: 
+                baseline: 784.24
               Save:
                 dataset: ./Uncertainties/ImpactSynth/Output:mha
                 group: None
           Comformity:
             transforms:
               KonfAIInference:
-                repo_id: VBoussot/MRSegmentator-KonfAI
-                model_name: MRSegmentator
-                number_of_ensemble: 1
+                repo_id: VBoussot/TotalSegmentator-KonfAI
+                model_name: total-3mm
+                checkpoints_name:
+                - M297.pt
                 number_of_tta: 0
-                number_of_mc_dropout: 0
+                number_of_mc: 0
                 per_channel: true
               Save/1:
                 dataset: ./Uncertainties/ImpactSynth/Output:mha
                 group: None
-              Variance: {}
+              SegmentationDisagreement: {}
               Save/2:
                 dataset: ./Uncertainties/ImpactSynth/Output:mha
                 group: Comformity_var

CBCT/app.json

@@ -1,7 +1,7 @@
 {
     "display_name": "Synthesis: CBCT",
-    "short_description": "<b>Description:</b><br>Supervised CBCT synthesis model developed as part of the <a href=\"https://synthrad2025.grand-challenge.org/\">SynthRAD 2025 Challenge (Task 2)</a>.<br><b>⚠️ Warning:</b> Models were trained with an anatomical mask, but no mask is used at inference. Artifacts may appear outside the anatomy. Future models will be trained without masks.<br><br><b>How to cite:</b><br><cite>V. Boussot et al., <i>Why Registration Quality Matters: Enhancing sCT Synthesis with IMPACT-Based Registration</i>, arXiv preprint arXiv:2510.21358, <a href=\"https://doi.org/10.48550/arXiv.2510.21358\">https://doi.org/10.48550/arXiv.2510.21358</a>.</cite>",
-    "description": "<b>Description:</b><br>Supervised CBCT synthesis model developed as part of the <a href=\"https://synthrad2025.grand-challenge.org/\">SynthRAD 2025 Challenge (Task 2)</a>.<br><br><b>Architecture:</b><br>Based on a 2.5D UNet++ with a ResNet34 encoder, the model was optimized using the <b>IMPACT-Synth loss</b>, a perceptual loss leveraging semantic priors from <b>SAM 2.1-s</b>. Training was conducted with the <b>KonfAI</b> deep learning framework.<br><br><b>Training data:</b><br>Paired CT–CBCT volumes from the <a href=\"https://synthrad2025.grand-challenge.org/\">SynthRAD 2025 Challenge (Task 2)</a>, <b>aligned using IMPACT-based registration</b>. Corresponding B-spline deformation fields are available in the <a href=\"https://huggingface.co/datasets/VBoussot/synthrad2025-impact-registration\">SynthRAD2025-IMPACT (aligned)</a> dataset repository.<br><br><b>How to cite:</b><br><cite>V. Boussot et al., <i>Why Registration Quality Matters: Enhancing sCT Synthesis with IMPACT-Based Registration</i>, arXiv preprint arXiv:2510.21358, <a href=\"https://doi.org/10.48550/arXiv.2510.21358\">https://doi.org/10.48550/arXiv.2510.21358</a>.</cite>",
+    "short_description": "Supervised whole-body CBCT-to-sCT model for SynthRAD Task 2.<br><br><b>Training data:</b><br>929 paired CT-CBCT cases from SynthRAD 2023+2025.<br><br><b>How to cite:</b><br><cite>V. Boussot et al., <i>Why Registration Quality Matters: Enhancing sCT Synthesis with IMPACT-Based Registration</i>, arXiv preprint arXiv:2510.21358, <a href=\"https://doi.org/10.48550/arXiv.2510.21358\">https://doi.org/10.48550/arXiv.2510.21358</a>.</cite>",
+    "description": "<b>Description:</b><br>Supervised whole-body CBCT-to-sCT model trained using SynthRAD 2023 and SynthRAD 2025 Task 2 datasets, distributed as five validation-fold checkpoints (<code>CV_0.pt</code> to <code>CV_4.pt</code>). Training pairs are aligned with IMPACTReg. The available Task 2 data include 291 patients from SynthRAD 2023 and 638 patients from SynthRAD 2025, for 929 patients in total.<br><br><b>Architecture and loss:</b><br>2.5D UNet++ with a ResNet34 encoder, trained with KonfAI and optimized using IMPACT-Synth, a perceptual loss leveraging semantic priors from <b>SAM 2.1-s</b>.<br><br><b>Training data:</b><br>Whole-body paired CT-CBCT volumes come from SynthRAD 2023 Task 2 (<code>brain</code>, <code>pelvis</code>) and SynthRAD 2025 Task 2 (<code>AB</code>, <code>HN</code>, <code>TH</code>). Additional OOD evaluation cohorts are available for <code>AB_ood</code>, <code>HN_ood</code>, and <code>TH_ood</code>. Pairs are aligned with IMPACTReg, and the associated transform files are available at <a href=\"https://huggingface.co/datasets/VBoussot/synthrad2023-impact-registration\">synthrad2023-impact-registration</a> and <a href=\"https://huggingface.co/datasets/VBoussot/synthrad2025-impact-registration\">synthrad2025-impact-registration</a>. The files <code>CrossValidation_0.txt</code> to <code>CrossValidation_4.txt</code> correspond to validation folds for the distributed checkpoints, and <code>Validation.txt</code> contains 150 release-validation patients.<br><br><b>Release performance (CV overall, n=150):</b><br>Dice 0.758, MAE 53.55 HU, SSIM 0.939, PSNR 32.09 dB, SAM 13.70, Reg 1.04, Uncertainty 784.24.<br><br><b>Inference note:</b><br>The prediction pipeline supports test-time augmentation and writes an averaged <code>InferenceStack</code> output that can also be used for uncertainty estimation.<br><br><b>How to cite:</b><br><cite>V. Boussot et al., <i>Why Registration Quality Matters: Enhancing sCT Synthesis with IMPACT-Based Registration</i>, arXiv preprint arXiv:2510.21358, <a href=\"https://doi.org/10.48550/arXiv.2510.21358\">https://doi.org/10.48550/arXiv.2510.21358</a>.</cite>",
     "tta": 2,
     "mc_dropout": false,
     "models": ["CV_0.pt", "CV_1.pt", "CV_2.pt", "CV_3.pt", "CV_4.pt"],
@@ -37,4 +37,4 @@
         }
       }
     }
-}
+}

CBCT/requirements.txt

@@ -1,2 +1 @@
-segmentation_models_pytorch
-scikit-image
+segmentation_models_pytorch

MR/CV_0.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1e200b775b0970feb194578723b05476cb6b8c8e08e7847863f16525c79917a4
-size 104526414
+oid sha256:ddd3e9250ea81a9f3003ec6b01aa95f200c1102c529f97b7a2ad2068cf28f57c
+size 208936686

MR/CV_1.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f3715933e9402889a1fa59555c140da06ef8968753acd762c2b8fe6aa76e26ef
-size 104526414
+oid sha256:6c396d218f32947f8e0af07fa8c62f4ba9e1a084ee40799369b488957b359eb4
+size 208936686

MR/CV_2.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ed09ba33bd8e86c6e3af4ad7d6c26137b697a7b6f670addccdb1af3247c461fd
-size 104526414
+oid sha256:b2933e1a8bd2a10c5a7d242c15ab71d323fab242995a28503aa42488087d73c0
+size 208936686

MR/CV_3.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:94ae1de52438b3e17c185bd33241b8821faaaae3a0d5c861ee1b00242a9abaa3
-size 104526414
+oid sha256:b963f174fd2f2d8324a37aa88666a458944a26f7cf3712593af5f49dd43e5ac5
+size 208936686

MR/CV_4.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9b261788d918bf12691e34625d5aa5c69715481d8ae91292ee8de1c081e8a63e
-size 104526414
+oid sha256:4e0a3bbb8513252b68dfa3fc84e16ebcb740289b96e9acdaac7ecf1e76a4062a
+size 208936686

MR/Config.yml

@@ -0,0 +1,181 @@
+Trainer:
+  Model:
+    classpath: Model:UNetpp
+    UNetpp:
+      outputs_criterions:
+        Head:Tanh:
+          targets_criterions:
+            CT:
+              criterions_loader:
+                MAE:
+                  schedulers:
+                    Constant:
+                      nb_step: 0
+                      value: 1
+                  is_loss: true
+                  group: 0
+                  start: 0
+                  stop: None
+                  accumulation: false
+                  reduction: mean
+                IMPACTReg/1:
+                  name: SAM
+                  model_name: SAM2.1/SAM2.1_Small.pt
+                  shape:
+                  - 0
+                  - 0
+                  in_channels: 3
+                  loss: torch:nn:L1Loss
+                  weights:
+                  - 0
+                  - 1
+                  - 1
+                  size_average: None
+                  reduce: None
+                  reduction: mean
+                  schedulers:
+                    Constant:
+                      nb_step: 0
+                      value: 0.5
+                  is_loss: true
+                  group: 0
+                  start: 0
+                  stop: None
+                  accumulation: false
+            CT;MASK:
+              criterions_loader:
+                MAE:
+                  schedulers:
+                    Constant:
+                      nb_step: 0
+                      value: 1
+                  is_loss: false
+                  group: 0
+                  start: 0
+                  stop: None
+                  accumulation: false
+                  reduction: mean
+      schedulers:
+        PolyLRScheduler:
+          initial_lr: 0.01
+          max_steps: 500
+          exponent: 0.9
+          current_step: 0
+          nb_step: 0
+      Patch: None
+      optimizer:
+        name: SGD
+        lr: 0.01
+        momentum: 0.99
+        dampening: 0
+        weight_decay: 3e-05
+        nesterov: true
+        maximize: false
+        foreach: None
+        differentiable: false
+        fused: None
+      nb_channel: 5
+  Dataset:
+    groups_src:
+      MASK:
+        groups_dest:
+          MASK:
+            transforms: None
+            patch_transforms: None
+            is_input: false
+      CT:
+        groups_dest:
+          CT:
+            transforms:
+              Clip:
+                min_value: -1024
+                max_value: 3071
+                save_clip_min: true
+                save_clip_max: true
+                mask: None
+              Statistics: {}
+              Normalize:
+                lazy: true
+                channels: None
+                min_value: -1
+                max_value: 1
+                inverse: false
+            patch_transforms:
+              Normalize:
+                lazy: false
+                channels: None
+                min_value: -1
+                max_value: 1
+                inverse: false
+            is_input: false
+      MR_IMPACT:
+        groups_dest:
+          MR:
+            transforms:
+              Clip:
+                min_value: min
+                max_value: percentile:99.5
+                save_clip_min: false
+                save_clip_max: false
+                mask: None
+              Statistics: {}
+              Normalize:
+                lazy: true
+                channels: None
+                min_value: -1
+                max_value: 1
+                inverse: false
+            patch_transforms:
+              Normalize:
+                lazy: false
+                channels: None
+                min_value: -1
+                max_value: 1
+                inverse: false
+            is_input: true
+    augmentations:
+      DataAugmentation_0:
+        data_augmentations:
+          Flip:
+            f_prob:
+            - 0
+            - 0.5
+            - 0.5
+            prob: 1
+        nb: 1
+    Patch:
+      patch_size:
+      - 1
+      - 320
+      - 320
+      overlap: None
+      mask: None
+      pad_value: -1
+      extend_slice: 4
+    subset: None
+    shuffle: true
+    filter: None
+    dataset_filenames:
+    - ./Dataset/:a:mha
+    inline_augmentations: true
+    use_cache: true
+    batch_size: 32
+    validation: None
+  train_name: FT_0
+  manual_seed: 32
+  epochs: 100
+  it_validation: 2500
+  autocast: false
+  gradient_checkpoints: None
+  gpu_checkpoints: None
+  ema_decay: 0
+  data_log:
+  - CT/IMAGES/5
+  - MR/IMAGES/5
+  - Head:Tanh/IMAGES/5
+  save_checkpoint_mode: ALL
+  EarlyStopping:
+    monitor: []
+    patience: 30
+    min_delta: 0.0
+    mode: min

MR/Evaluation.yml

@@ -12,6 +12,27 @@ Evaluator:
               dynamic_range: None
             SSIM:
               dynamic_range: None
+            SAM_Perceptual: {}
+            IMPACTReg:
+              name: Reg
+              model_name: MRSeg/MRSeg.pt
+              shape:
+              - 128
+              - 128
+              - 128
+              in_channels: 1
+              loss: torch:nn:MSELoss
+              weights:
+              - 0
+              - 0
+              - 0
+              - 0
+              - 0
+              - 0
+              - 1
+              size_average: None
+              reduce: None
+              reduction: mean
     Output_seg:
       targets_criterions:
         Reference_seg;Mask:
@@ -25,19 +46,27 @@ Evaluator:
       Mask_0:
         groups_dest:
           Mask:
-            transforms: None
+            transforms:
+              TensorCast:
+                dtype: uint8
       Volume_0:
         groups_dest:
           Output:
-            transforms: None
+            transforms: 
+              Statistics: {}
+              TensorCast:
+                dtype: float32
+            patch_transforms: None
+            is_input: false
           Output_seg:
             transforms:
               KonfAIInference:
-                repo_id: VBoussot/MRSegmentator-KonfAI
-                model_name: MRSegmentator
-                number_of_ensemble: 1
+                repo_id: VBoussot/TotalSegmentator-KonfAI
+                model_name: total-3mm
+                checkpoints_name:
+                - M297.pt
                 number_of_tta: 0
-                number_of_mc_dropout: 0
+                number_of_mc: 0
                 per_channel: false
               Save:
                 dataset: ./Evaluations/ImpactSynth/Output:mha
@@ -45,15 +74,21 @@ Evaluator:
       Reference_0:
         groups_dest:
           Reference:
-            transforms: None
+            transforms:               
+              Statistics: {}
+              TensorCast:
+                dtype: float32
+            patch_transforms: None
+            is_input: false
           Reference_seg:
             transforms:
               KonfAIInference:
-                repo_id: VBoussot/MRSegmentator-KonfAI
-                model_name: MRSegmentator
-                number_of_ensemble: 1
+                repo_id: VBoussot/TotalSegmentator-KonfAI
+                model_name: total-3mm
+                checkpoints_name:
+                - M297.pt
                 number_of_tta: 0
-                number_of_mc_dropout: 0
+                number_of_mc: 0
                 per_channel: false
               Save:
                 dataset: ./Evaluations/ImpactSynth/Output:mha

MR/MR_MODEL_INFO.md

@@ -0,0 +1,175 @@
+# ImpactSynth MR
+
+## Summary
+
+- Modality: whole-body MR to sCT synthesis
+- Dataset source: SynthRAD 2023 and SynthRAD 2025 Task 1
+- Distributed checkpoints: `CV_0.pt`, `CV_1.pt`, `CV_2.pt`, `CV_3.pt`, `CV_4.pt`
+- Inference-time test-time augmentation: `tta = 2`
+- Monte Carlo dropout: `false`
+- Framework: KonfAI
+- Architecture: 2.5D UNet++ with a ResNet34 encoder
+- Main perceptual prior: SAM 2.1-s through the IMPACT-Synth loss
+- Pair alignment: CT-MR pairs are aligned with IMPACTReg
+- Transform files:
+  - `https://huggingface.co/datasets/VBoussot/synthrad2023-impact-registration`
+  - `https://huggingface.co/datasets/VBoussot/synthrad2025-impact-registration`
+
+## Data splits
+
+The repository contains the validation and cross-validation splits used for Task 1.
+
+- Patients in the 2023 Task 1 training cohort: `317`
+- Patients in the 2025 Task 1 training cohort: `410`
+- Total patients across the paired 2023+2025 Task 1 cohorts: `727`
+- Additional out-of-distribution Task 1 patients: `78`
+- Region counts in the paired 2023+2025 Task 1 cohorts:
+  - `AB`: `138`
+  - `HN`: `129`
+  - `TH`: `143`
+  - `brain`: `172`
+  - `pelvis`: `145`
+- OOD cohort counts:
+  - `CSIRO`: `24`
+  - `GA`: `17`
+  - `MR_Linac`: `25`
+  - `MR_Rennes`: `12`
+- `Validation.txt`: `140` patients
+- `CrossValidation_0..4.txt` are validation folds associated with the distributed checkpoints
+
+- `Validation.txt`: 140 patients
+- `CrossValidation_0.txt`: 126 validation patients
+- `CrossValidation_1.txt`: 125 validation patients
+- `CrossValidation_2.txt`: 123 validation patients
+- `CrossValidation_3.txt`: 121 validation patients
+- `CrossValidation_4.txt`: 117 validation patients
+
+## Release CV performance
+
+Overall Release CV metrics:
+
+- Dice is reported with `TS = 3 mm` on the CT body mask.
+- All detailed metrics are reported in the tables below.
+
+- Dice: `0.698 +/- 0.106`
+- MAE: `67.49 +/- 15.97` HU
+- SSIM: `0.959 +/- 0.056`
+- PSNR: `29.28 +/- 2.10` dB
+- SAM: `21.98 +/- 7.69`
+- Reg: `0.622 +/- 0.303`
+- Uncertainty: `1822.49 +/- 1035.84`
+
+Per-fold overall Release metrics:
+
+| Fold | n | Dice | MAE (HU) | SSIM | PSNR (dB) | SAM | Reg | Uncertainty |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| CV | 140 | 0.698 +/- 0.106 | 67.49 +/- 15.97 | 0.959 +/- 0.056 | 29.28 +/- 2.10 | 21.98 +/- 7.69 | 0.622 +/- 0.303 | 1822.49 +/- 1035.84 |
+| CV_0 | 140 | 0.682 +/- 0.112 | 76.04 +/- 18.29 | 0.956 +/- 0.060 | 28.28 +/- 1.99 | 20.69 +/- 6.65 | 0.632 +/- 0.303 | n/a |
+| CV_1 | 140 | 0.688 +/- 0.110 | 75.40 +/- 17.93 | 0.957 +/- 0.059 | 28.37 +/- 2.00 | 20.73 +/- 6.81 | 0.628 +/- 0.297 | n/a |
+| CV_2 | 140 | 0.696 +/- 0.108 | 70.98 +/- 15.90 | 0.958 +/- 0.057 | 28.69 +/- 1.99 | 20.29 +/- 7.17 | 0.634 +/- 0.302 | n/a |
+| CV_3 | 140 | 0.695 +/- 0.110 | 71.23 +/- 16.07 | 0.959 +/- 0.056 | 28.67 +/- 1.99 | 20.12 +/- 6.97 | 0.633 +/- 0.300 | n/a |
+| CV_4 | 140 | 0.696 +/- 0.106 | 71.53 +/- 16.25 | 0.957 +/- 0.057 | 28.64 +/- 2.04 | 20.22 +/- 7.13 | 0.634 +/- 0.305 | n/a |
+
+Regional Release CV metrics:
+
+| Region | n | Dice | MAE (HU) | SSIM | PSNR (dB) | SAM | Reg | Uncertainty |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| AB | 22 | 0.598 | 60.10 | 0.975 | 30.01 | 27.41 | 0.511 | 1543.81 |
+| HN | 21 | 0.645 | 80.89 | 0.949 | 28.44 | 13.41 | 0.435 | 1308.83 |
+| TH | 23 | 0.645 | 58.45 | 1.000 | 31.21 | 21.64 | 0.466 | 751.61 |
+| brain | 27 | 0.751 | 79.71 | 0.973 | 27.44 | 13.21 | 1.199 | 3402.08 |
+| pelvis | 23 | 0.756 | 48.16 | 0.940 | 31.28 | 25.95 | 0.574 | 2348.75 |
+| CSIRO | 24 | 0.774 | 75.99 | 0.920 | 27.67 | 30.90 | 0.434 | 1272.28 |
+
+Regional Release metrics for CV_0:
+
+| Region | n | Dice | MAE (HU) | SSIM | PSNR (dB) | SAM | Reg |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| AB | 22 | 0.580 +/- 0.073 | 65.69 +/- 9.96 | 0.972 +/- 0.049 | 29.20 +/- 1.35 | 25.49 +/- 3.71 | 0.542 +/- 0.118 |
+| HN | 21 | 0.620 +/- 0.109 | 92.70 +/- 15.86 | 0.943 +/- 0.092 | 27.55 +/- 1.34 | 13.13 +/- 1.84 | 0.427 +/- 0.066 |
+| TH | 23 | 0.624 +/- 0.059 | 64.64 +/- 11.69 | 1.000 +/- 0.000 | 30.38 +/- 1.76 | 19.99 +/- 3.63 | 0.481 +/- 0.098 |
+| brain | 27 | 0.731 +/- 0.096 | 93.73 +/- 9.48 | 0.970 +/- 0.066 | 26.23 +/- 0.82 | 13.50 +/- 2.76 | 1.208 +/- 0.136 |
+| pelvis | 23 | 0.746 +/- 0.113 | 56.77 +/- 9.75 | 0.932 +/- 0.042 | 29.68 +/- 1.26 | 25.11 +/- 3.92 | 0.581 +/- 0.091 |
+| CSIRO | 24 | 0.770 +/- 0.032 | 80.45 +/- 11.20 | 0.917 +/- 0.024 | 27.04 +/- 1.03 | 27.44 +/- 3.56 | 0.442 +/- 0.059 |
+| overall | 140 | 0.682 +/- 0.112 | 76.04 +/- 18.29 | 0.956 +/- 0.060 | 28.28 +/- 1.99 | 20.69 +/- 6.65 | 0.632 +/- 0.303 |
+
+Regional Release metrics for CV_1:
+
+| Region | n | Dice | MAE (HU) | SSIM | PSNR (dB) | SAM | Reg |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| AB | 22 | 0.590 +/- 0.063 | 65.69 +/- 10.19 | 0.972 +/- 0.047 | 29.19 +/- 1.37 | 25.32 +/- 3.88 | 0.537 +/- 0.115 |
+| HN | 21 | 0.627 +/- 0.112 | 90.39 +/- 14.10 | 0.944 +/- 0.090 | 27.66 +/- 1.25 | 13.07 +/- 1.86 | 0.431 +/- 0.058 |
+| TH | 23 | 0.631 +/- 0.055 | 64.10 +/- 11.72 | 1.000 +/- 0.000 | 30.42 +/- 1.77 | 19.88 +/- 3.54 | 0.483 +/- 0.103 |
+| brain | 27 | 0.739 +/- 0.097 | 92.55 +/- 9.65 | 0.970 +/- 0.065 | 26.40 +/- 0.86 | 13.43 +/- 2.77 | 1.191 +/- 0.139 |
+| pelvis | 23 | 0.750 +/- 0.112 | 55.27 +/- 9.81 | 0.935 +/- 0.039 | 29.99 +/- 1.24 | 25.03 +/- 3.99 | 0.579 +/- 0.092 |
+| CSIRO | 24 | 0.772 +/- 0.034 | 82.03 +/- 10.49 | 0.917 +/- 0.022 | 26.92 +/- 0.90 | 28.13 +/- 3.80 | 0.437 +/- 0.061 |
+| overall | 140 | 0.688 +/- 0.110 | 75.40 +/- 17.93 | 0.957 +/- 0.059 | 28.37 +/- 2.00 | 20.73 +/- 6.81 | 0.628 +/- 0.297 |
+
+Regional Release metrics for CV_2:
+
+| Region | n | Dice | MAE (HU) | SSIM | PSNR (dB) | SAM | Reg |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| AB | 22 | 0.593 +/- 0.069 | 65.10 +/- 10.26 | 0.972 +/- 0.048 | 29.19 +/- 1.42 | 25.59 +/- 3.90 | 0.540 +/- 0.117 |
+| HN | 21 | 0.646 +/- 0.104 | 84.06 +/- 12.54 | 0.948 +/- 0.085 | 27.95 +/- 1.29 | 12.72 +/- 1.86 | 0.437 +/- 0.067 |
+| TH | 23 | 0.638 +/- 0.054 | 61.92 +/- 11.88 | 1.000 +/- 0.000 | 30.67 +/- 1.82 | 20.23 +/- 3.62 | 0.481 +/- 0.097 |
+| brain | 27 | 0.753 +/- 0.098 | 82.15 +/- 7.73 | 0.972 +/- 0.063 | 27.06 +/- 0.78 | 11.80 +/- 2.32 | 1.207 +/- 0.132 |
+| pelvis | 23 | 0.757 +/- 0.109 | 51.15 +/- 8.23 | 0.935 +/- 0.036 | 30.49 +/- 1.35 | 23.99 +/- 4.21 | 0.585 +/- 0.096 |
+| CSIRO | 24 | 0.769 +/- 0.033 | 80.05 +/- 10.62 | 0.919 +/- 0.021 | 27.10 +/- 0.98 | 28.14 +/- 3.98 | 0.439 +/- 0.058 |
+| overall | 140 | 0.696 +/- 0.108 | 70.98 +/- 15.90 | 0.958 +/- 0.057 | 28.69 +/- 1.99 | 20.29 +/- 7.17 | 0.634 +/- 0.302 |
+
+Regional Release metrics for CV_3:
+
+| Region | n | Dice | MAE (HU) | SSIM | PSNR (dB) | SAM | Reg |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| AB | 22 | 0.586 +/- 0.072 | 64.82 +/- 10.07 | 0.974 +/- 0.044 | 29.22 +/- 1.41 | 25.37 +/- 3.94 | 0.536 +/- 0.119 |
+| HN | 21 | 0.649 +/- 0.107 | 84.18 +/- 12.60 | 0.948 +/- 0.085 | 27.96 +/- 1.28 | 12.66 +/- 1.99 | 0.435 +/- 0.063 |
+| TH | 23 | 0.639 +/- 0.056 | 61.98 +/- 12.31 | 1.000 +/- 0.000 | 30.59 +/- 1.88 | 20.09 +/- 3.62 | 0.478 +/- 0.093 |
+| brain | 27 | 0.751 +/- 0.102 | 82.98 +/- 8.67 | 0.972 +/- 0.062 | 26.96 +/- 0.88 | 11.86 +/- 2.45 | 1.203 +/- 0.135 |
+| pelvis | 23 | 0.752 +/- 0.112 | 52.13 +/- 9.38 | 0.938 +/- 0.035 | 30.44 +/- 1.40 | 23.98 +/- 4.07 | 0.584 +/- 0.093 |
+| CSIRO | 24 | 0.772 +/- 0.035 | 79.75 +/- 10.80 | 0.918 +/- 0.022 | 27.15 +/- 0.95 | 27.48 +/- 3.31 | 0.448 +/- 0.056 |
+| overall | 140 | 0.695 +/- 0.110 | 71.23 +/- 16.07 | 0.959 +/- 0.056 | 28.67 +/- 1.99 | 20.12 +/- 6.97 | 0.633 +/- 0.300 |
+
+Regional Release metrics for CV_4:
+
+| Region | n | Dice | MAE (HU) | SSIM | PSNR (dB) | SAM | Reg |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| AB | 22 | 0.593 +/- 0.067 | 65.18 +/- 10.15 | 0.972 +/- 0.047 | 29.19 +/- 1.41 | 25.76 +/- 3.90 | 0.532 +/- 0.123 |
+| HN | 21 | 0.643 +/- 0.103 | 84.34 +/- 12.78 | 0.948 +/- 0.085 | 27.93 +/- 1.28 | 12.76 +/- 1.98 | 0.435 +/- 0.069 |
+| TH | 23 | 0.640 +/- 0.057 | 62.20 +/- 11.92 | 1.000 +/- 0.000 | 30.56 +/- 1.85 | 20.04 +/- 3.65 | 0.484 +/- 0.108 |
+| brain | 27 | 0.751 +/- 0.092 | 83.10 +/- 8.84 | 0.972 +/- 0.062 | 26.94 +/- 0.90 | 11.72 +/- 2.62 | 1.211 +/- 0.134 |
+| pelvis | 23 | 0.755 +/- 0.103 | 51.79 +/- 9.44 | 0.934 +/- 0.039 | 30.55 +/- 1.48 | 24.09 +/- 4.09 | 0.584 +/- 0.096 |
+| CSIRO | 24 | 0.771 +/- 0.036 | 80.96 +/- 10.79 | 0.917 +/- 0.022 | 26.98 +/- 0.98 | 27.68 +/- 3.85 | 0.442 +/- 0.061 |
+| overall | 140 | 0.696 +/- 0.106 | 71.53 +/- 16.25 | 0.957 +/- 0.057 | 28.64 +/- 2.04 | 20.22 +/- 7.13 | 0.634 +/- 0.305 |
+
+## Split files
+
+### Validation.txt
+
+`1BA131`, `1BA336`, `1BA253`, `1BA143`, `1BA175`, `1BA022`, `1BA189`, `1BA091`, `1BA328`, `1BB007`, `1BB076`, `1BB008`, `1BB072`, `1BB048`, `1BB095`, `1BB171`, `1BB109`, `1BB179`, `1BC014`, `1BC090`, `1BC027`, `1BC020`, `1BC019`, `1BC073`, `1BC083`, `1BC039`, `1BC009`, `1PA127`, `1PA174`, `1PA147`, `1PA163`, `1PA030`, `1PA138`, `1PA117`, `1PA164`, `1PA133`, `1PA111`, `1PA169`, `1PA113`, `1PA156`, `1PA098`, `1PA019`, `1PC006`, `1PC001`, `1PC036`, `1PC023`, `1PC015`, `1PC070`, `1PC010`, `1PC073`, `1ABA080`, `1ABA044`, `1ABA098`, `1ABA065`, `1ABA085`, `1ABA099`, `1ABA062`, `1ABB151`, `1ABB150`, `1ABB011`, `1ABB109`, `1ABB139`, `1ABB037`, `1ABB161`, `1ABB073`, `1ABB113`, `1ABB069`, `1ABB061`, `1ABB170`, `1ABC121`, `1ABC002`, `1ABC007`, `1HNA116`, `1HNA025`, `1HNA121`, `1HNA061`, `1HNA049`, `1HNA120`, `1HNA136`, `1HNA133`, `1HNA056`, `1HNA102`, `1HNA109`, `1HNA089`, `1HNA086`, `1HNA033`, `1HNC088`, `1HNC003`, `1HNC109`, `1HNC103`, `1HNC112`, `1HNC046`, `1HNC083`, `1THA230`, `1THA249`, `1THA015`, `1THA224`, `1THA262`, `1THA002`, `1THA017`, `1THA272`, `1THA291`, `1THA243`, `1THB103`, `1THB017`, `1THB225`, `1THB214`, `1THB121`, `1THB048`, `1THB224`, `1THB216`, `1THB063`, `1THB057`, `1THB212`, `1THB150`, `1THB215`, `B006`, `B037`, `B038`, `B040`, `D031`, `H007`, `H017`, `J026`, `K008`, `K019`, `K042`, `M004`, `M015`, `M020`, `M023`, `M030`, `M036`, `R024`, `R034`, `R039`, `S028`, `S035`, `W012`, `W029`
+
+### CrossValidation_0.txt
+
+`1BA325`, `1BA185`, `1BA105`, `1BA368`, `1BA300`, `1BA085`, `1BA220`, `1BA294`, `1BA307`, `1BA288`, `1BB005`, `1BB073`, `1BB182`, `1BB200`, `1BB091`, `1BB039`, `1BB151`, `1BB145`, `1BB100`, `1BB062`, `1BC081`, `1BC004`, `1BC008`, `1BC086`, `1BC017`, `1BC085`, `1BC034`, `1BC028`, `1BC023`, `1BC036`, `1PA054`, `1PA012`, `1PA185`, `1PA180`, `1PA165`, `1PA063`, `1PA150`, `1PA059`, `1PA177`, `1PA025`, `1PA048`, `1PA151`, `1PA049`, `1PA081`, `1PA134`, `1PA118`, `1PA010`, `1PC004`, `1PC022`, `1PC027`, `1PC071`, `1PC080`, `1PC017`, `1PC069`, `1PC052`, `1ABA061`, `1ABA081`, `1ABA086`, `1ABA030`, `1ABA104`, `1ABA103`, `1ABA012`, `1ABA068`, `1ABB070`, `1ABB053`, `1ABB044`, `1ABB029`, `1ABB056`, `1ABB009`, `1ABB025`, `1ABB030`, `1ABB155`, `1ABB039`, `1ABB112`, `1ABB120`, `1ABB118`, `1ABC016`, `1ABC018`, `1ABC010`, `1HNA115`, `1HNA029`, `1HNA077`, `1HNA069`, `1HNA104`, `1HNA084`, `1HNA093`, `1HNA132`, `1HNA072`, `1HNA019`, `1HNA067`, `1HNA015`, `1HNA129`, `1HNA135`, `1HNA071`, `1HNC125`, `1HNC124`, `1HNC076`, `1HNC130`, `1HNC094`, `1HNC061`, `1HNC084`, `1THA026`, `1THA286`, `1THA270`, `1THA207`, `1THA028`, `1THA283`, `1THA212`, `1THA213`, `1THA251`, `1THA261`, `1THA217`, `1THB074`, `1THB021`, `1THB043`, `1THB015`, `1THB058`, `1THB011`, `1THB135`, `1THB143`, `1THB003`, `1THB149`, `1THB072`, `1THB052`, `1THB035`, `1THB124`
+
+### CrossValidation_1.txt
+
+`1BA001`, `1BA040`, `1BA292`, `1BA260`, `1BA159`, `1BA012`, `1BA359`, `1BA082`, `1BA032`, `1BA222`, `1BB205`, `1BB030`, `1BB102`, `1BB011`, `1BB090`, `1BB026`, `1BB152`, `1BB006`, `1BB085`, `1BB017`, `1BC056`, `1BC065`, `1BC049`, `1BC051`, `1BC082`, `1BC080`, `1BC062`, `1BC063`, `1BC052`, `1BC038`, `1PA065`, `1PA142`, `1PA076`, `1PA064`, `1PA182`, `1PA094`, `1PA157`, `1PA100`, `1PA095`, `1PA101`, `1PA056`, `1PA114`, `1PA088`, `1PA044`, `1PA187`, `1PA168`, `1PA084`, `1PC029`, `1PC032`, `1PC041`, `1PC066`, `1PC037`, `1PC035`, `1PC048`, `1PC000`, `1ABA051`, `1ABA110`, `1ABA054`, `1ABA092`, `1ABA018`, `1ABA112`, `1ABA071`, `1ABA115`, `1ABB169`, `1ABB041`, `1ABB116`, `1ABB168`, `1ABB034`, `1ABB040`, `1ABB036`, `1ABB103`, `1ABB114`, `1ABB078`, `1ABB135`, `1ABB045`, `1ABB042`, `1ABC012`, `1ABC122`, `1ABC116`, `1HNA023`, `1HNA008`, `1HNA032`, `1HNA103`, `1HNA110`, `1HNA040`, `1HNA105`, `1HNA048`, `1HNA039`, `1HNA066`, `1HNA012`, `1HNA124`, `1HNA106`, `1HNA035`, `1HNA097`, `1HNC128`, `1HNC029`, `1HNC110`, `1HNC004`, `1HNC127`, `1HNC037`, `1HNC002`, `1THA019`, `1THA279`, `1THA250`, `1THA293`, `1THA237`, `1THA267`, `1THA273`, `1THA055`, `1THA260`, `1THA052`, `1THB022`, `1THB220`, `1THB062`, `1THB151`, `1THB100`, `1THB016`, `1THB047`, `1THB120`, `1THB200`, `1THB031`, `1THB222`, `1THB045`, `1THB217`, `1THB027`
+
+### CrossValidation_2.txt
+
+`1BA014`, `1BA278`, `1BA151`, `1BA206`, `1BA076`, `1BA116`, `1BA201`, `1BA058`, `1BA005`, `1BB096`, `1BB016`, `1BB034`, `1BB083`, `1BB071`, `1BB002`, `1BB051`, `1BB066`, `1BB111`, `1BB052`, `1BC084`, `1BC006`, `1BC077`, `1BC022`, `1BC046`, `1BC041`, `1BC010`, `1BC094`, `1BC074`, `1BC088`, `1PA073`, `1PA093`, `1PA152`, `1PA080`, `1PA026`, `1PA022`, `1PA105`, `1PA047`, `1PA178`, `1PA108`, `1PA159`, `1PA140`, `1PA188`, `1PA079`, `1PA005`, `1PA173`, `1PC063`, `1PC098`, `1PC007`, `1PC078`, `1PC019`, `1PC065`, `1PC042`, `1PC092`, `1ABA031`, `1ABA060`, `1ABA097`, `1ABA014`, `1ABA005`, `1ABA019`, `1ABA084`, `1ABA067`, `1ABB084`, `1ABB062`, `1ABB117`, `1ABB057`, `1ABB164`, `1ABB143`, `1ABB001`, `1ABB123`, `1ABB031`, `1ABB058`, `1ABB046`, `1ABB035`, `1ABB119`, `1ABC019`, `1ABC100`, `1ABC009`, `1HNA107`, `1HNA014`, `1HNA096`, `1HNA006`, `1HNA034`, `1HNA053`, `1HNA099`, `1HNA108`, `1HNA041`, `1HNA082`, `1HNA031`, `1HNA004`, `1HNA068`, `1HNA142`, `1HNA085`, `1HNC001`, `1HNC012`, `1HNC043`, `1HNC104`, `1HNC101`, `1HNC117`, `1HNC031`, `1THA282`, `1THA039`, `1THA205`, `1THA001`, `1THA018`, `1THA005`, `1THA011`, `1THA244`, `1THA004`, `1THA288`, `1THB006`, `1THB095`, `1THB211`, `1THB076`, `1THB053`, `1THB134`, `1THB106`, `1THB028`, `1THB141`, `1THB068`, `1THB078`, `1THB179`, `1THB195`, `1THB226`
+
+### CrossValidation_3.txt
+
+`1BA256`, `1BA184`, `1BA379`, `1BA239`, `1BA054`, `1BA358`, `1BA158`, `1BA172`, `1BA100`, `1BB082`, `1BB003`, `1BB049`, `1BB098`, `1BB173`, `1BB175`, `1BB177`, `1BB041`, `1BB079`, `1BB189`, `1BC066`, `1BC035`, `1BC050`, `1BC070`, `1BC048`, `1BC001`, `1BC025`, `1BC064`, `1BC087`, `1BC021`, `1PA167`, `1PA053`, `1PA110`, `1PA058`, `1PA020`, `1PA107`, `1PA119`, `1PA146`, `1PA112`, `1PA074`, `1PA040`, `1PA062`, `1PA070`, `1PA004`, `1PA035`, `1PA031`, `1PC046`, `1PC088`, `1PC033`, `1PC096`, `1PC011`, `1PC038`, `1PC061`, `1PC084`, `1ABA113`, `1ABA072`, `1ABA057`, `1ABA070`, `1ABA102`, `1ABA094`, `1ABA011`, `1ABA114`, `1ABB153`, `1ABB128`, `1ABB137`, `1ABB004`, `1ABB173`, `1ABB021`, `1ABB020`, `1ABB066`, `1ABB102`, `1ABB024`, `1ABB130`, `1ABB124`, `1ABB054`, `1ABC014`, `1ABC004`, `1HNA126`, `1HNA143`, `1HNA021`, `1HNA018`, `1HNA060`, `1HNA091`, `1HNA100`, `1HNA013`, `1HNA037`, `1HNA010`, `1HNA059`, `1HNA042`, `1HNA119`, `1HNA130`, `1HNC067`, `1HNC038`, `1HNC040`, `1HNC068`, `1HNC087`, `1HNC120`, `1HNC035`, `1THA203`, `1THA271`, `1THA265`, `1THA253`, `1THA248`, `1THA228`, `1THA289`, `1THA269`, `1THA252`, `1THA280`, `1THB002`, `1THB033`, `1THB060`, `1THB054`, `1THB046`, `1THB114`, `1THB191`, `1THB221`, `1THB008`, `1THB034`, `1THB199`, `1THB201`, `1THB207`, `1THB019`
+
+### CrossValidation_4.txt
+
+`1BA103`, `1BA234`, `1BA125`, `1BA075`, `1BA164`, `1BA345`, `1BA227`, `1BA266`, `1BA247`, `1BB059`, `1BB198`, `1BB050`, `1BB099`, `1BB075`, `1BB043`, `1BB028`, `1BB033`, `1BB184`, `1BC047`, `1BC076`, `1BC031`, `1BC075`, `1BC037`, `1BC058`, `1BC067`, `1BC054`, `1BC007`, `1PA176`, `1PA009`, `1PA086`, `1PA141`, `1PA041`, `1PA136`, `1PA060`, `1PA170`, `1PA024`, `1PA154`, `1PA083`, `1PA171`, `1PA148`, `1PA001`, `1PA011`, `1PA137`, `1PC054`, `1PC044`, `1PC039`, `1PC085`, `1PC049`, `1PC018`, `1PC059`, `1PC095`, `1ABA049`, `1ABA082`, `1ABA119`, `1ABA033`, `1ABA117`, `1ABA087`, `1ABA109`, `1ABB166`, `1ABB115`, `1ABB067`, `1ABB083`, `1ABB008`, `1ABB098`, `1ABB002`, `1ABB145`, `1ABB047`, `1ABB127`, `1ABB059`, `1ABB006`, `1ABB138`, `1ABC127`, `1ABC008`, `1HNA051`, `1HNA026`, `1HNA139`, `1HNA047`, `1HNA030`, `1HNA001`, `1HNA098`, `1HNA113`, `1HNA043`, `1HNA090`, `1HNA028`, `1HNA141`, `1HNA138`, `1HNA036`, `1HNC111`, `1HNC107`, `1HNC102`, `1HNC071`, `1HNC020`, `1HNC121`, `1HNC072`, `1THA256`, `1THA257`, `1THA010`, `1THA255`, `1THA258`, `1THA276`, `1THA281`, `1THA013`, `1THA003`, `1THA284`, `1THB029`, `1THB038`, `1THB119`, `1THB205`, `1THB218`, `1THB067`, `1THB202`, `1THB126`, `1THB073`, `1THB050`, `1THB196`, `1THB004`, `1THB122`
+
+## Notes
+
+- Inference uses the pipeline defined in `Prediction.yml`, including automatic body-mask estimation with `VBoussot/TotalSegmentator-KonfAI` (`body`, checkpoint `M598.pt`).
+- Evaluation uses the pipeline defined in `Evaluation.yml`, including segmentation consistency with `TotalSegmentator-KonfAI` (`total-3mm`, checkpoint `M297.pt`) and two perceptual/registration metrics:
+  - `IMPACTReg/1`: `SAM2.1/SAM2.1_Small.pt`
+  - `IMPACTReg/2`: `MRSeg/MRSeg.pt`

MR/Model.py

@@ -16,12 +16,12 @@ class UNetpp(network.Network):
                     "default:ReduceLROnPlateau": network.LRSchedulersLoader(0)
                 },
                 outputs_criterions: dict[str, network.TargetCriterionsLoader] = {"default" : network.TargetCriterionsLoader()},
-                pretrained: bool = False):
-        super().__init__(in_channels = 3, optimizer = optimizer, schedulers = schedulers, outputs_criterions = outputs_criterions, dim = 2)
+                nb_channel : int = 5):
+        super().__init__(in_channels = nb_channel, optimizer = optimizer, schedulers = schedulers, outputs_criterions = outputs_criterions, dim = 2)
         self.add_module("model", smp.UnetPlusPlus(
             encoder_name="resnet34",
-            encoder_weights=None if not pretrained else "imagenet", 
-            in_channels=3,
+            encoder_weights=None, 
+            in_channels=nb_channel,
             classes=1,  
             activation=None  
         ))

MR/Prediction.yml

@@ -3,16 +3,30 @@ Predictor:
     classpath: Model:UNetpp
     UNetpp:
       outputs_criterions: None
-      pretrained: false
+      nb_channel: 5
   Dataset:
     groups_src:
       Volume_0:
         groups_dest:
+          MASK:
+            transforms:
+              KonfAIInference:
+                repo_id: VBoussot/TotalSegmentator-KonfAI
+                model_name: body
+                checkpoints_name:
+                - M598.pt
+                number_of_tta: 0
+                number_of_mc_dropout: 0
+                per_channel: false
+              Save:
+                dataset: ./Dataset:mha
+            patch_transforms: None
+            is_input: false
           Volume:
             transforms:
               Clip/0:
-                min_value: -32000
-                max_value: 32000
+                min_value: -1024
+                max_value: 5000
                 save_clip_min: false
                 save_clip_max: false
                 mask: None
@@ -57,7 +71,7 @@ Predictor:
       overlap: None
       mask: None
       pad_value: -1
-      extend_slice: 2
+      extend_slice: 4
     subset: None
     filter: None
     dataset_filenames:
@@ -75,6 +89,9 @@ Predictor:
           TensorCast:
             dtype: int16
             inverse: false
+          Mask:
+            path: MASK
+            value_outside: -1024
         after_reduction_transforms: 
           InferenceStack:
             dataset: Predictions/ImpactSynth/Output:mha

MR/Uncertainty.yml

@@ -19,22 +19,25 @@ Evaluator:
           Uncertainty:
             transforms:
               Variance: {}
+              Percentage: 
+                baseline: 1822.49
               Save:
                 dataset: ./Uncertainties/ImpactSynth/Output:mha
                 group: None
           Comformity:
             transforms:
               KonfAIInference:
-                repo_id: VBoussot/MRSegmentator-KonfAI
-                model_name: MRSegmentator
-                number_of_ensemble: 1
+                repo_id: VBoussot/TotalSegmentator-KonfAI
+                model_name: total-3mm
+                checkpoints_name:
+                - M297.pt
                 number_of_tta: 0
-                number_of_mc_dropout: 0
+                number_of_mc: 0
                 per_channel: true
               Save/1:
                 dataset: ./Uncertainties/ImpactSynth/Output:mha
                 group: None
-              Variance: {}
+              SegmentationDisagreement: {}
               Save/2:
                 dataset: ./Uncertainties/ImpactSynth/Output:mha
                 group: Comformity_var
@@ -42,4 +45,4 @@ Evaluator:
     dataset_filenames:
     - ./Dataset:mha
     validation: None
-  train_name: ImpactSynth
+  train_name: ImpactSynth

MR/app.json

@@ -1,38 +1,38 @@
 {
-    "display_name": "Synthesis: MR",
-    "short_description": "<b>Description:</b><br>Supervised MRI (T1-weighted) synthesis model developed as part of the <a href=\"https://synthrad2025.grand-challenge.org/\">SynthRAD 2025 Challenge (Task 1)</a>.<br><b>⚠️ Warning:</b> Models were trained with an anatomical mask, but no mask is used at inference. Artifacts may appear outside the anatomy. Future models will be trained without masks.<br><br><b>How to cite:</b><br><cite>V. Boussot et al., <i>Why Registration Quality Matters: Enhancing sCT Synthesis with IMPACT-Based Registration</i>, arXiv preprint arXiv:2510.21358, <a href=\"https://doi.org/10.48550/arXiv.2510.21358\">https://doi.org/10.48550/arXiv.2510.21358</a>.</cite>",
-    "description": "<b>Description:</b><br>Supervised MRI (T1-weighted) synthesis model developed as part of the <a href=\"https://synthrad2025.grand-challenge.org/\">SynthRAD 2025 Challenge (Task 1)</a>.<br><br><b>Architecture:</b><br>Based on a 2.5D UNet++ with a ResNet34 encoder, the model was optimized using the <b>IMPACT-Synth loss</b>, a perceptual loss leveraging semantic priors from <b>SAM 2.1-s</b>. Training was conducted with the <b>KonfAI</b> deep learning framework.<br><br><b>Training data:</b><br>Paired CT–MRI (T1-weighted) volumes from the <a href=\"https://synthrad2025.grand-challenge.org/\">SynthRAD 2025 Challenge (Task 1)</a>, <b>aligned using IMPACT-based registration</b>. Corresponding B-spline deformation fields are available in the <a href=\"https://huggingface.co/datasets/VBoussot/synthrad2025-impact-registration\">SynthRAD2025-IMPACT (aligned)</a> dataset repository.<br><br><b>How to cite:</b><br><cite>V. Boussot et al., <i>Why Registration Quality Matters: Enhancing sCT Synthesis with IMPACT-Based Registration</i>, arXiv preprint arXiv:2510.21358, <a href=\"https://doi.org/10.48550/arXiv.2510.21358\">https://doi.org/10.48550/arXiv.2510.21358</a>.</cite>",
-    "tta": 2,
-    "mc_dropout": false,
-    "models": ["CV_0.pt", "CV_1.pt", "CV_2.pt", "CV_3.pt", "CV_4.pt"],
-    "inputs": {
-      "MR": {
-        "display_name": "MR",
-        "volume_type": "VOLUME",
-        "required": true
-      }
-    },
-    "outputs": {
-      "sCT": {
-        "display_name": "sCT",
-        "volume_type": "VOLUME",
-        "required": true
-      }
-    },
-    "inputs_evaluations": {
-      "Image": {
-        "Evaluation.yml": {
-          "sCT": {
-            "display_name": "sCT",
-            "volume_type": "VOLUME",
-            "required": true
-          },
-          "CT": {
-            "display_name": "CT",
-            "volume_type": "VOLUME",
-            "required": true
-          }
+  "display_name": "Synthesis: MR",
+  "short_description": "Supervised whole-body MR-to-sCT model for SynthRAD Task 1.<br><br><b>Training data:</b><br>805 paired CT-MR cases including OOD cohorts.<br><br><b>How to cite:</b><br><cite>V. Boussot et al., <i>Why Registration Quality Matters: Enhancing sCT Synthesis with IMPACT-Based Registration</i>, arXiv preprint arXiv:2510.21358, <a href=\"https://doi.org/10.48550/arXiv.2510.21358\">https://doi.org/10.48550/arXiv.2510.21358</a>.</cite>",
+  "description": "<b>Description:</b><br>Supervised whole-body MR-to-sCT model trained using SynthRAD 2023 and SynthRAD 2025 Task 1 datasets, distributed as five validation-fold checkpoints (<code>CV_0.pt</code> to <code>CV_4.pt</code>). Training pairs are aligned with IMPACTReg. The available Task 1 data include 317 patients from SynthRAD 2023, 410 patients from SynthRAD 2025, and 78 additional OOD paired CT-MR cases, for 805 training patients in total.<br><br><b>Architecture and loss:</b><br>2.5D UNet++ with a ResNet34 encoder, trained with KonfAI and optimized using IMPACT-Synth, a perceptual loss leveraging semantic priors from <b>SAM 2.1-s</b>.<br><br><b>Training data:</b><br>Whole-body paired CT-MR volumes come from SynthRAD 2023 Task 1 (<code>brain</code>, <code>pelvis</code>), SynthRAD 2025 Task 1 (<code>AB</code>, <code>HN</code>, <code>TH</code>), and additional OOD cohorts including <code>CSIRO</code>, <code>GA</code>, <code>MR_Linac</code>, and <code>MR_Rennes</code>. Pairs are aligned with IMPACTReg, and the associated transform files are available at <a href=\"https://huggingface.co/datasets/VBoussot/synthrad2023-impact-registration\">synthrad2023-impact-registration</a> and <a href=\"https://huggingface.co/datasets/VBoussot/synthrad2025-impact-registration\">synthrad2025-impact-registration</a>. The files <code>CrossValidation_0.txt</code> to <code>CrossValidation_4.txt</code> correspond to validation folds for the distributed checkpoints, and <code>Validation.txt</code> contains 140 release-validation patients.<br><br><b>Release performance (CV overall, n=140):</b><br>Dice 0.698, MAE 67.49 HU, SSIM 0.959, PSNR 29.28 dB, SAM 21.98, Reg 0.622.<br><br><b>Inference note:</b><br>A body mask is estimated automatically at inference time and used to crop valid anatomy support before writing the output volume.<br><br><b>How to cite:</b><br><cite>V. Boussot et al., <i>Why Registration Quality Matters: Enhancing sCT Synthesis with IMPACT-Based Registration</i>, arXiv preprint arXiv:2510.21358, <a href=\"https://doi.org/10.48550/arXiv.2510.21358\">https://doi.org/10.48550/arXiv.2510.21358</a>.</cite>",
+  "tta": 2,
+  "mc_dropout": false,
+  "models": ["CV_0.pt", "CV_1.pt", "CV_2.pt", "CV_3.pt", "CV_4.pt"],
+  "inputs": {
+    "MR": {
+      "display_name": "MR",
+      "volume_type": "VOLUME",
+      "required": true
+    }
+  },
+  "outputs": {
+    "sCT": {
+      "display_name": "sCT",
+      "volume_type": "VOLUME",
+      "required": true
+    }
+  },
+  "inputs_evaluations": {
+    "Image": {
+      "Evaluation.yml": {
+        "sCT": {
+          "display_name": "sCT",
+          "volume_type": "VOLUME",
+          "required": true
+        },
+        "CT": {
+          "display_name": "CT",
+          "volume_type": "VOLUME",
+          "required": true
         }
       }
     }
-}
+  }
+}

MR/requirements.txt

@@ -1,2 +1 @@
-segmentation_models_pytorch
-scikit-image
+segmentation_models_pytorch

MR_CBCT/CV_0.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ceb14b248f8876544210acc8d94a0bd6eeb26eacdb7d959c3a30057a7694c3fc
+size 208936878

MR_CBCT/Config.yml

@@ -0,0 +1,181 @@
+Trainer:
+  Model:
+    classpath: Model:UNetpp
+    UNetpp:
+      outputs_criterions:
+        Head:Tanh:
+          targets_criterions:
+            CT:
+              criterions_loader:
+                MAE:
+                  schedulers:
+                    Constant:
+                      nb_step: 0
+                      value: 1
+                  is_loss: true
+                  group: 0
+                  start: 0
+                  stop: None
+                  accumulation: false
+                  reduction: mean
+                IMPACTReg/1:
+                  name: SAM
+                  model_name: SAM2.1/SAM2.1_Small.pt
+                  shape:
+                  - 0
+                  - 0
+                  in_channels: 3
+                  loss: torch:nn:L1Loss
+                  weights:
+                  - 0
+                  - 1
+                  - 1
+                  size_average: None
+                  reduce: None
+                  reduction: mean
+                  schedulers:
+                    Constant:
+                      nb_step: 0
+                      value: 0.5
+                  is_loss: true
+                  group: 0
+                  start: 0
+                  stop: None
+                  accumulation: false
+            CT;MASK:
+              criterions_loader:
+                MAE:
+                  schedulers:
+                    Constant:
+                      nb_step: 0
+                      value: 1
+                  is_loss: false
+                  group: 0
+                  start: 0
+                  stop: None
+                  accumulation: false
+                  reduction: mean
+      schedulers:
+        PolyLRScheduler:
+          initial_lr: 0.01
+          max_steps: 500
+          exponent: 0.9
+          current_step: 0
+          nb_step: 0
+      Patch: None
+      optimizer:
+        name: SGD
+        lr: 0.01
+        momentum: 0.99
+        dampening: 0
+        weight_decay: 3e-05
+        nesterov: true
+        maximize: false
+        foreach: None
+        differentiable: false
+        fused: None
+      nb_channel: 5
+  Dataset:
+    groups_src:
+      MASK:
+        groups_dest:
+          MASK:
+            transforms: None
+            patch_transforms: None
+            is_input: false
+      CT:
+        groups_dest:
+          CT:
+            transforms:
+              Clip:
+                min_value: -1024
+                max_value: 3071
+                save_clip_min: true
+                save_clip_max: true
+                mask: None
+              Statistics: {}
+              Normalize:
+                lazy: true
+                channels: None
+                min_value: -1
+                max_value: 1
+                inverse: false
+            patch_transforms:
+              Normalize:
+                lazy: false
+                channels: None
+                min_value: -1
+                max_value: 1
+                inverse: false
+            is_input: false
+      IMAGE_IMPACT:
+        groups_dest:
+          IMAGE:
+            transforms:
+              Clip:
+                min_value: min
+                max_value: percentile:99.5
+                save_clip_min: false
+                save_clip_max: false
+                mask: None
+              Statistics: {}
+              Normalize:
+                lazy: true
+                channels: None
+                min_value: -1
+                max_value: 1
+                inverse: false
+            patch_transforms:
+              Normalize:
+                lazy: false
+                channels: None
+                min_value: -1
+                max_value: 1
+                inverse: false
+            is_input: true
+    augmentations:
+      DataAugmentation_0:
+        data_augmentations:
+          Flip:
+            f_prob:
+            - 0
+            - 0.5
+            - 0.5
+            prob: 1
+        nb: 1
+    Patch:
+      patch_size:
+      - 1
+      - 320
+      - 320
+      overlap: None
+      mask: None
+      pad_value: -1
+      extend_slice: 4
+    subset: None
+    shuffle: true
+    filter: None
+    dataset_filenames:
+    - ./Dataset/:a:mha
+    inline_augmentations: true
+    use_cache: true
+    batch_size: 32
+    validation: None
+  train_name: FT_0
+  manual_seed: 32
+  epochs: 100
+  it_validation: 2500
+  autocast: false
+  gradient_checkpoints: None
+  gpu_checkpoints: None
+  ema_decay: 0
+  data_log:
+  - CT/IMAGES/5
+  - IMAGE/IMAGES/5
+  - Head:Tanh/IMAGES/5
+  save_checkpoint_mode: ALL
+  EarlyStopping:
+    monitor: []
+    patience: 30
+    min_delta: 0.0
+    mode: min

MR_CBCT/Evaluation.yml

@@ -0,0 +1,95 @@
+Evaluator:
+  metrics:
+    Output:
+      targets_criterions:
+        Reference;Mask:
+          criterions_loader:
+            MAESaveMap:
+              reduction: mean
+              dataset: ./Evaluations/ImpactSynth/Output:mha
+              group: MAE_map
+            PSNR:
+              dynamic_range: None
+            SSIM:
+              dynamic_range: None
+            SAM_Perceptual: {}
+            IMPACTReg:
+              name: Reg
+              model_name: TS/M291.pt
+              shape:
+              - 0
+              - 0
+              - 0
+              in_channels: 1
+              loss: torch:nn:MSELoss
+              weights:
+              - 0
+              - 1
+              size_average: None
+              reduce: None
+              reduction: mean
+    Output_seg:
+      targets_criterions:
+        Reference_seg;Mask:
+          criterions_loader:
+            DiceSaveMap:
+              labels: None
+              dataset: ./Evaluations/ImpactSynth/Output:mha
+              group: Seg_MAE_map
+  Dataset:
+    groups_src:
+      Mask_0:
+        groups_dest:
+          Mask:
+            transforms:
+              TensorCast:
+                dtype: uint8
+      Volume_0:
+        groups_dest:
+          Output:
+            transforms: 
+              Statistics: {}
+              TensorCast:
+                dtype: float32
+            patch_transforms: None
+            is_input: false
+          Output_seg:
+            transforms:
+              KonfAIInference:
+                repo_id: VBoussot/TotalSegmentator-KonfAI
+                model_name: total-3mm
+                checkpoints_name:
+                - M297.pt
+                number_of_tta: 0
+                number_of_mc: 0
+                per_channel: false
+              Save:
+                dataset: ./Evaluations/ImpactSynth/Output:mha
+                group: None
+      Reference_0:
+        groups_dest:
+          Reference:
+            transforms:               
+              Statistics: {}
+              TensorCast:
+                dtype: float32
+            patch_transforms: None
+            is_input: false
+          Reference_seg:
+            transforms:
+              KonfAIInference:
+                repo_id: VBoussot/TotalSegmentator-KonfAI
+                model_name: total-3mm
+                checkpoints_name:
+                - M297.pt
+                number_of_tta: 0
+                number_of_mc: 0
+                per_channel: false
+              Save:
+                dataset: ./Evaluations/ImpactSynth/Output:mha
+                group: None
+    subset: None
+    dataset_filenames:
+    - ./Dataset:a:mha
+    validation: None
+  train_name: ImpactSynth

MR_CBCT/Model.py

@@ -0,0 +1,28 @@
+from konfai.network import network
+import segmentation_models_pytorch as smp
+import torch
+
+class Head(network.ModuleArgsDict):
+
+    def __init__(self):
+        super().__init__()
+        self.add_module("Tanh", torch.nn.Tanh())
+
+class UNetpp(network.Network):
+    
+    def __init__(self,
+                optimizer : network.OptimizerLoader = network.OptimizerLoader(),
+                schedulers: dict[str, network.LRSchedulersLoader] = {
+                    "default:ReduceLROnPlateau": network.LRSchedulersLoader(0)
+                },
+                outputs_criterions: dict[str, network.TargetCriterionsLoader] = {"default" : network.TargetCriterionsLoader()},
+                nb_channel : int = 5):
+        super().__init__(in_channels = nb_channel, optimizer = optimizer, schedulers = schedulers, outputs_criterions = outputs_criterions, dim = 2)
+        self.add_module("model", smp.UnetPlusPlus(
+            encoder_name="resnet34",
+            encoder_weights=None, 
+            in_channels=nb_channel,
+            classes=1,  
+            activation=None  
+        ))
+        self.add_module("Head", Head())

MR_CBCT/Prediction.yml

@@ -0,0 +1,115 @@
+Predictor:
+  Model:
+    classpath: Model:UNetpp
+    UNetpp:
+      outputs_criterions: None
+      nb_channel: 5
+  Dataset:
+    groups_src:
+      Volume_0:
+        groups_dest:
+          MASK:
+            transforms:
+              KonfAIInference:
+                repo_id: VBoussot/TotalSegmentator-KonfAI
+                model_name: body
+                checkpoints_name:
+                - M598.pt
+                number_of_tta: 0
+                number_of_mc_dropout: 0
+                per_channel: false
+              Save:
+                dataset: ./Dataset:mha
+            patch_transforms: None
+            is_input: false
+          Volume:
+            transforms:
+              Clip/0:
+                min_value: -1024
+                max_value: 5000
+                save_clip_min: false
+                save_clip_max: false
+                mask: None
+              TensorCast:
+                dtype: int16
+                inverse: false
+              ResampleToResolution:
+                spacing:
+                - 1
+                - 1
+                - 3
+                inverse: true
+              Clip/1:
+                min_value: min
+                max_value: percentile:99.5
+                save_clip_min: false
+                save_clip_max: false
+                mask: None
+              Normalize:
+                lazy: false
+                channels: None
+                min_value: -1
+                max_value: 1
+                inverse: false
+            patch_transforms: None
+            is_input: true
+    augmentations:
+      DataAugmentation_0:
+        data_augmentations:
+          Flip:
+            f_prob:
+            - 0
+            - 0.5
+            - 0.5
+            prob: 1
+        nb: 2
+    Patch:
+      patch_size:
+      - 1
+      - 512
+      - 512
+      overlap: None
+      mask: None
+      pad_value: -1
+      extend_slice: 4
+    subset: None
+    filter: None
+    dataset_filenames:
+    - ./Dataset:mha
+    use_cache: false
+    batch_size: 8
+  outputs_dataset:
+    Head:Tanh:
+      OutputDataset:
+        name_class: OutSameAsGroupDataset
+        before_reduction_transforms: 
+          UnNormalize:
+            min_value: -1024
+            max_value: 3071
+          TensorCast:
+            dtype: int16
+            inverse: false
+          Mask:
+            path: MASK
+            value_outside: -1024
+        after_reduction_transforms: 
+          InferenceStack:
+            dataset: Predictions/ImpactSynth/Output:mha
+            name: InferenceStack
+            mode: mean
+        final_transforms: None
+        dataset_filename: Output:mha
+        inverse_transform: true
+        group: sCT
+        same_as_group: Volume_0:Volume
+        patch_combine: None
+        reduction: Concat
+        Concat: {}
+  train_name: ImpactSynth
+  manual_seed: 32
+  gpu_checkpoints: None
+  images_log: None
+  combine: Concat
+  autocast: false
+  data_log: None
+  Concat: {}

MR_CBCT/Uncertainty.yml

@@ -0,0 +1,48 @@
+Evaluator:
+  metrics:
+    Uncertainty:
+      targets_criterions:
+        None:
+          criterions_loader:
+            Mean:
+              name: Uncertainty
+    Comformity:
+      targets_criterions:
+        None:
+          criterions_loader:
+            Mean:
+              name: Comformity
+  Dataset:
+    groups_src:
+      Volume_0:
+        groups_dest:
+          Uncertainty:
+            transforms:
+              Variance: {}
+              Percentage: 
+                baseline: 1822.49
+              Save:
+                dataset: ./Uncertainties/ImpactSynth/Output:mha
+                group: None
+          Comformity:
+            transforms:
+              KonfAIInference:
+                repo_id: VBoussot/TotalSegmentator-KonfAI
+                model_name: total-3mm
+                checkpoints_name:
+                - M297.pt
+                number_of_tta: 0
+                number_of_mc: 0
+                per_channel: true
+              Save/1:
+                dataset: ./Uncertainties/ImpactSynth/Output:mha
+                group: None
+              SegmentationDisagreement: {}
+              Save/2:
+                dataset: ./Uncertainties/ImpactSynth/Output:mha
+                group: Comformity_var
+    subset: None
+    dataset_filenames:
+    - ./Dataset:mha
+    validation: None
+  train_name: ImpactSynth

MR_CBCT/app.json

@@ -0,0 +1,38 @@
+{
+  "display_name": "Synthesis: MR/CBCT",
+  "short_description": "Supervised whole-body MR/CBCT-to-sCT model.<br><br><b>Training data:</b><br>1734 paired cases from Task 1 + Task 2.<br><br><b>How to cite:</b><br><cite>V. Boussot et al., <i>Why Registration Quality Matters: Enhancing sCT Synthesis with IMPACT-Based Registration</i>, arXiv preprint arXiv:2510.21358, <a href=\"https://doi.org/10.48550/arXiv.2510.21358\">https://doi.org/10.48550/arXiv.2510.21358</a>.</cite>",
+  "description": "<b>Description:</b><br>Supervised whole-body synthesis model distributed as a single checkpoint (<code>CV_0.pt</code>) for MR/CBCT-style inputs. The model follows the same 2.5D synthesis setup used in the ImpactSynth family and produces an sCT volume from a single input volume after automatic body-mask estimation and intensity normalization.<br><br><b>Architecture and loss:</b><br>2.5D UNet++ with a ResNet34 encoder, trained with KonfAI and optimized using IMPACT-Synth, a perceptual loss leveraging semantic priors from <b>SAM 2.1-s</b>.<br><br><b>Training data:</b><br>The model is trained on 1734 IMPACTReg-aligned paired synthesis cases spanning both Task 1 and Task 2 cohorts, combining MR and CBCT supervision in a unified synthesis setup.<br><br><b>Input handling:</b><br>The prediction pipeline accepts a single input volume exposed here as <code>MR or CBCT</code>, estimates a body mask automatically, resamples to the working resolution, applies normalization, and writes an averaged sCT output using test-time augmentation.<br><br><b>How to cite:</b><br><cite>V. Boussot et al., <i>Why Registration Quality Matters: Enhancing sCT Synthesis with IMPACT-Based Registration</i>, arXiv preprint arXiv:2510.21358, <a href=\"https://doi.org/10.48550/arXiv.2510.21358\">https://doi.org/10.48550/arXiv.2510.21358</a>.</cite>",
+  "tta": 2,
+  "mc_dropout": false,
+  "models": ["CV_0.pt"],
+  "inputs": {
+    "MR": {
+      "display_name": "MR or CBCT",
+      "volume_type": "VOLUME",
+      "required": true
+    }
+  },
+  "outputs": {
+    "sCT": {
+      "display_name": "sCT",
+      "volume_type": "VOLUME",
+      "required": true
+    }
+  },
+  "inputs_evaluations": {
+    "Image": {
+      "Evaluation.yml": {
+        "sCT": {
+          "display_name": "sCT",
+          "volume_type": "VOLUME",
+          "required": true
+        },
+        "CT": {
+          "display_name": "CT",
+          "volume_type": "VOLUME",
+          "required": true
+        }
+      }
+    }
+  }
+}

MR_CBCT/requirements.txt

@@ -0,0 +1 @@
+segmentation_models_pytorch

SAM2.1_Small.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10c2fb000e57fa4adcb5f1c23aa7dfe6f622e1f649643ceb7e43e8188f024899
+size 135971068

sam_channel_weights_optimized.json

@@ -0,0 +1,1622 @@
+{
+  "method": "adam_ordered_channel_weights_from_precomputed_features_v1",
+  "features_dir": "audit_CT_deformed_release_mae/features",
+  "patients": [
+    "AB/2ABA006",
+    "AB/2ABA024",
+    "AB/2ABA066",
+    "AB/2ABA077",
+    "AB/2ABA082",
+    "AB/2ABA103",
+    "AB/2ABA118",
+    "AB/2ABA135",
+    "AB/2ABB018",
+    "AB/2ABB039",
+    "AB/2ABB051",
+    "AB/2ABB059",
+    "AB/2ABB061",
+    "AB/2ABB090",
+    "AB/2ABB095",
+    "AB/2ABB117",
+    "AB/2ABC019",
+    "AB/2ABC049",
+    "AB/2ABC063",
+    "AB/2ABC093",
+    "AB/2ABC118",
+    "AB/2ABC150",
+    "AB/2ABC152",
+    "AB/2ABC222",
+    "AB/2ABE022",
+    "AB/2ABE124",
+    "AB/2ABE158",
+    "AB/2ABE169",
+    "AB/2ABE180",
+    "AB/2ABE184",
+    "AB/2ABE191",
+    "AB/2ABE197",
+    "HN/2HNA002",
+    "HN/2HNA005",
+    "HN/2HNA024",
+    "HN/2HNA032",
+    "HN/2HNA050",
+    "HN/2HNA058",
+    "HN/2HNA084",
+    "HN/2HNA098",
+    "HN/2HNA100",
+    "HN/2HNA110",
+    "HN/2HNB015",
+    "HN/2HNB019",
+    "HN/2HNB044",
+    "HN/2HNB047",
+    "HN/2HNB067",
+    "HN/2HNB075",
+    "HN/2HNB089",
+    "HN/2HNB099",
+    "HN/2HNB105",
+    "HN/2HNC021",
+    "HN/2HNC024",
+    "HN/2HNC028",
+    "HN/2HNC036",
+    "HN/2HNC043",
+    "HN/2HNC060",
+    "HN/2HNC077",
+    "HN/2HNC078",
+    "HN/2HNC099",
+    "HN/2HNE007",
+    "HN/2HNE010",
+    "HN/2HNE026",
+    "HN/2HNE041",
+    "HN/2HNE042",
+    "HN/2HNE054",
+    "HN/2HNE075",
+    "HN/2HNE079",
+    "HN/2HNE101",
+    "TH/2THA005",
+    "TH/2THA043",
+    "TH/2THA077",
+    "TH/2THA082",
+    "TH/2THA099",
+    "TH/2THA101",
+    "TH/2THA113",
+    "TH/2THA117",
+    "TH/2THA122",
+    "TH/2THB014",
+    "TH/2THB015",
+    "TH/2THB031",
+    "TH/2THB038",
+    "TH/2THB046",
+    "TH/2THB055",
+    "TH/2THB059",
+    "TH/2THB088",
+    "TH/2THC033",
+    "TH/2THC037",
+    "TH/2THC056",
+    "TH/2THC068",
+    "TH/2THC076",
+    "TH/2THC097",
+    "TH/2THC102",
+    "TH/2THC107",
+    "TH/2THC111",
+    "TH/2THE039",
+    "TH/2THE040",
+    "TH/2THE046",
+    "TH/2THE058",
+    "TH/2THE069",
+    "TH/2THE081",
+    "TH/2THE099",
+    "TH/2THE119"
+  ],
+  "channel_weights": [
+    [
+      0.0006171520799398422,
+      0.0005136821419000626,
+      0.0006172073190100491,
+      0.0005728057003580034,
+      0.0005712952697649598,
+      0.0005774051533080637,
+      0.0006076112622395158,
+      10.003186225891113,
+      0.005408514756709337,
+      0.0005605806945823133,
+      0.006068961229175329,
+      0.000585941132158041,
+      0.0005129298660904169,
+      0.000491302227601409,
+      0.0005069899489171803,
+      0.0006187129765748978,
+      0.0006240687216632068,
+      0.0005597768467850983,
+      0.0017875866033136845,
+      0.02446560002863407,
+      0.000584447814617306,
+      0.0005994382663629949,
+      0.0005213120603002608,
+      0.0005782971275039017,
+      0.0005136086838319898,
+      10.038290977478027,
+      0.0005242027691565454,
+      0.0005694732535630465,
+      0.00048017740482464433,
+      0.0007305329781956971,
+      0.002240255940705538,
+      0.0005427927826531231,
+      0.0005736321327276528,
+      0.0024858347605913877,
+      0.0005490012117661536,
+      0.0005289871478453279,
+      10.069931030273438,
+      0.0005890378961339593,
+      0.0005384367541410029,
+      0.0007900098571553826,
+      0.0007435446605086327,
+      0.0005673144478350878,
+      0.0005590940127149224,
+      0.0058243232779204845,
+      10.0343599319458,
+      0.001932881772518158,
+      5.439353942871094,
+      10.000136375427246,
+      0.0005080231931060553,
+      0.0013258099788799882,
+      0.0005624794284813106,
+      0.0005475050420500338,
+      0.004734368529170752,
+      10.000762939453125,
+      0.06523310393095016,
+      0.003011316293850541,
+      0.0005516970995813608,
+      10.149458885192871,
+      0.0005472758784890175,
+      0.0005593824898824096,
+      0.0005249620880931616,
+      0.0005184299079701304,
+      0.0005841492093168199,
+      0.0006099396268837154,
+      10.008008003234863,
+      0.0005502901040017605,
+      0.000583179178647697,
+      0.0005756898317486048,
+      0.0005968278273940086,
+      0.0019690925255417824,
+      0.0005693890270777047,
+      0.0004952430026605725,
+      0.0005309210973791778,
+      0.0005610283114947379,
+      0.0004867143288720399,
+      0.0005130193312652409,
+      0.0005250982940196991,
+      0.0005490839830599725,
+      0.000536899664439261,
+      0.0007697922992520034,
+      0.0005088182515464723,
+      0.0005231261602602899,
+      10.007970809936523,
+      0.000559445412363857,
+      0.0005369129357859492,
+      0.0005818804493173957,
+      0.0006135341245681047,
+      0.0006175528978928924,
+      0.0006266830605454743,
+      0.0005625695339404047,
+      0.0011138181434944272,
+      0.07479756325483322,
+      0.0063269007951021194,
+      0.0004964643740095198,
+      0.00048434798372909427,
+      0.000495799642521888
+    ],
+    [
+      0.0005621173186227679,
+      0.000542987312655896,
+      10.01394271850586,
+      0.0012214280432090163,
+      0.0005677255103364587,
+      0.00048470147885382175,
+      0.0007504619425162673,
+      0.0005487996386364102,
+      0.0005906096193939447,
+      0.0005769628332927823,
+      0.0005940243718214333,
+      0.0005744381924159825,
+      0.0005644538323394954,
+      0.00046894172555767,
+      0.0005684405332431197,
+      0.0005300413467921317,
+      0.000591580756008625,
+      0.00046666801790706813,
+      0.0005354383029043674,
+      0.000598234822973609,
+      0.0005216854624450207,
+      0.0005441929679363966,
+      0.0007805791683495045,
+      0.0005233970587141812,
+      0.0005406272248364985,
+      0.0004826509393751621,
+      0.0005638781585730612,
+      0.0005587889463640749,
+      0.0005350647261366248,
+      0.000587750575505197,
+      10.006255149841309,
+      0.00047420067130587995,
+      10.072598457336426,
+      0.0005762212094850838,
+      0.0005271918489597738,
+      0.0005750772543251514,
+      0.0005721912602894008,
+      0.000587361108046025,
+      0.0005489524337463081,
+      0.0025994298048317432,
+      0.0004844565992243588,
+      0.001561364741064608,
+      0.0006109864916652441,
+      0.0005420183297246695,
+      10.0402193069458,
+      0.0005400177324190736,
+      0.0005147869233042002,
+      0.006317995022982359,
+      0.0005255004507489502,
+      0.0005244576605036855,
+      0.0004646525194402784,
+      0.0005824363906867802,
+      0.0004867714596912265,
+      0.0005557461990974844,
+      0.0005057422677055001,
+      0.0005254909629002213,
+      0.000599151651840657,
+      0.0005210645031183958,
+      10.082661628723145,
+      0.0007720187422819436,
+      10.044072151184082,
+      0.0011559127597138286,
+      0.0005602064775303006,
+      0.00046678821672685444,
+      0.0016796010313555598,
+      0.0004937916528433561,
+      0.0005241597536951303,
+      10.029358863830566,
+      0.0005839963559992611,
+      0.0006193576846271753,
+      0.0009723914554342628,
+      0.0005918092210777104,
+      0.0005875190836377442,
+      0.0017954474315047264,
+      0.0008775803726166487,
+      0.0005406360141932964,
+      0.0005206969799473882,
+      10.01317310333252,
+      0.0005843874532729387,
+      0.0011016099015250802,
+      0.0005850297166034579,
+      0.0005902323755435646,
+      0.000548135896679014,
+      0.0005076120723970234,
+      0.0005436255596578121,
+      0.0005254538846202195,
+      0.0005274402792565525,
+      0.0005633257096633315,
+      0.0005938782705925405,
+      0.0005936977104283869,
+      0.0005849828594364226,
+      0.0005949331098236144,
+      0.0035920189693570137,
+      10.655861854553223,
+      0.0005447214934974909,
+      0.0005196082638576627,
+      0.0004789779777638614,
+      0.0005394428735598922,
+      0.0004643251886591315,
+      0.0021162861958146095,
+      0.0009356873342767358,
+      10.206332206726074,
+      10.026381492614746,
+      0.0022207959555089474,
+      0.00053284881869331,
+      10.211456298828125,
+      0.0004990393063053489,
+      0.0014011792372912169,
+      0.0005580487777478993,
+      0.007704720366746187,
+      10.142542839050293,
+      0.0030247452668845654,
+      0.0006027077324688435,
+      0.0006467207567766309,
+      0.0005387827404774725,
+      0.0005848963628523052,
+      0.004775334615260363,
+      0.0005931250634603202,
+      0.0013550586299970746,
+      0.0010190688772127032,
+      0.0005919289542362094,
+      0.0014830176951363683,
+      0.000552180630620569,
+      10.512600898742676,
+      0.0005839006043970585,
+      0.0005409923614934087,
+      0.0004664745065383613,
+      0.0012861782452091575,
+      0.0017628960777074099,
+      0.0005704039358533919,
+      0.0005538602126762271,
+      0.0006659552454948425,
+      10.052324295043945,
+      0.0005768710398115218,
+      0.000870930845849216,
+      0.0005076972884126008,
+      0.0005908309831283987,
+      0.0005683695781044662,
+      0.0004898171173408628,
+      0.0005864237900823355,
+      9.105240821838379,
+      0.0006085096392780542,
+      0.0005710151162929833,
+      0.0005880325334146619,
+      0.001469354028813541,
+      0.0006030596559867263,
+      0.0005873247282579541,
+      0.0006034472608007491,
+      0.0005145351751707494,
+      0.0005247168010100722,
+      0.0004840031615458429,
+      10.509066581726074,
+      0.0005313510773703456,
+      0.0005198619910515845,
+      0.0005307099199853837,
+      0.0005896692164242268,
+      0.00046692261821590364,
+      0.0009715628693811595,
+      0.0005040898686274886,
+      0.0005859939265064895,
+      0.005087240133434534,
+      0.0006153369904495776,
+      0.0004717831325251609,
+      0.000556363957002759,
+      0.0005156879778951406,
+      0.0006008697673678398,
+      0.000811468344181776,
+      0.0022218378726392984,
+      0.0004706512263510376,
+      0.0005339342751540244,
+      10.017330169677734,
+      0.03879345953464508,
+      0.00049905403284356,
+      0.0006391300121322274,
+      0.0012749260058626533,
+      0.000555241946130991,
+      0.0005254383431747556,
+      0.0005824147374369204,
+      0.0005457525257952511,
+      0.0005445500719361007,
+      0.000503110233694315,
+      0.0009111007675528526,
+      0.0005556418327614665,
+      0.0005850469460710883,
+      0.0005731945275329053,
+      0.000586525013204664,
+      0.0026170951314270496,
+      10.07252025604248,
+      0.0005264890496619046,
+      0.0004849955439567566,
+      0.0016158267389982939,
+      0.0005844671395607293
+    ],
+    [
+      0.0007311372319236398,
+      0.018451722338795662,
+      0.00037708497256971896,
+      0.0003466402122285217,
+      0.0005144589231349528,
+      0.0005112045910209417,
+      0.00042771751759573817,
+      0.0003855444665532559,
+      0.000624240143224597,
+      0.0004546527052298188,
+      0.00031219416996464133,
+      0.0007199030369520187,
+      0.00042389275040477514,
+      0.0005298629403114319,
+      10.002569198608398,
+      0.0003413397935219109,
+      0.0004696752002928406,
+      0.005424083210527897,
+      0.0006970877293497324,
+      10.047492027282715,
+      0.0004245472664479166,
+      0.0007049469859339297,
+      10.040411949157715,
+      0.0004533033643383533,
+      0.0025565021205693483,
+      0.00035739451413974166,
+      0.0003920213785022497,
+      0.000492763239890337,
+      0.0003489721566438675,
+      0.0013251446653157473,
+      0.00045049685286358,
+      0.00042435177601873875,
+      0.000664695689920336,
+      10.405211448669434,
+      0.0003390120400581509,
+      0.00071729626506567,
+      0.0006643909146077931,
+      0.0016874028369784355,
+      0.0006392354262061417,
+      0.0003157537430524826,
+      0.0004995980416424572,
+      0.0005886242142878473,
+      0.0015843729488551617,
+      7.864404678344727,
+      0.000386081199394539,
+      0.0004012832068838179,
+      10.120689392089844,
+      0.0004748956416733563,
+      0.0009038617718033493,
+      0.001181477913632989,
+      10.075601577758789,
+      0.0006273853941820562,
+      0.000360877369530499,
+      0.00029294032719917595,
+      0.003350179409608245,
+      10.027373313903809,
+      0.0003325899306219071,
+      0.0012259186478331685,
+      0.0003620064817368984,
+      0.000319186452543363,
+      0.0002985498576890677,
+      0.00038418362964875996,
+      0.005886268336325884,
+      0.0018260183278471231,
+      0.0002930400660261512,
+      0.00030045854509808123,
+      0.00039229809772223234,
+      0.0003344342112541199,
+      0.0005069856415502727,
+      0.00043127735261805356,
+      0.00037683514528907835,
+      0.00033420787076465786,
+      0.0003591009008232504,
+      0.0003843543818220496,
+      0.003150220727548003,
+      10.467209815979004,
+      0.003165316302329302,
+      0.000400889286538586,
+      0.0005466477596201003,
+      0.0005099494592286646,
+      0.0004652427160181105,
+      0.000291653152089566,
+      0.0005974334198981524,
+      0.0005703107453882694,
+      0.00029463920509442687,
+      0.0007033650763332844,
+      10.417888641357422,
+      0.0004389546811580658,
+      0.0003934471169486642,
+      0.00033044061274267733,
+      0.001580407959409058,
+      0.0017257046420127153,
+      0.00048730982234701514,
+      0.0006821677088737488,
+      0.0018689573043957353,
+      0.00042173892143182456,
+      0.0003434238024055958,
+      0.0022140638902783394,
+      0.0021442838478833437,
+      0.0005488994065672159,
+      0.0005371235893107951,
+      0.0017518562963232398,
+      0.0002913873759098351,
+      0.00036266815732233226,
+      0.000559527485165745,
+      0.000340393278747797,
+      0.0008635531412437558,
+      0.0003989318502135575,
+      0.0004108989960514009,
+      0.0004829848767258227,
+      10.140016555786133,
+      0.00031836508424021304,
+      0.00030010603950358927,
+      0.00032989372266456485,
+      0.0003294215421192348,
+      0.00039481325075030327,
+      0.010307556949555874,
+      0.0014580106362700462,
+      0.00033411261392757297,
+      0.0009240186773240566,
+      0.0005106828175485134,
+      0.0005491015035659075,
+      0.0023148120380938053,
+      0.000518068321980536,
+      0.000515514228027314,
+      0.00036347066634334624,
+      0.003751461859792471,
+      0.00048413898912258446,
+      0.00034356562537141144,
+      0.000453006912721321,
+      0.0005914806388318539,
+      0.0008201510645449162,
+      0.00037608452839776874,
+      0.0002918312093243003,
+      0.00046267942525446415,
+      0.0004873307188972831,
+      0.0005965811433270574,
+      0.0003848292981274426,
+      0.0003214057069271803,
+      0.0007379876915365458,
+      0.0004314369289204478,
+      0.0004023963410872966,
+      0.0003371183993294835,
+      0.0003631138533819467,
+      0.000463931035483256,
+      0.0005956825916655362,
+      0.0008362497319467366,
+      0.0006926229107193649,
+      0.00029585405718535185,
+      0.0004866722447331995,
+      10.00960922241211,
+      0.002083508064970374,
+      0.00258965278044343,
+      0.0006968206143938005,
+      0.0014616177650168538,
+      0.0005853811162523925,
+      0.0011825022520497441,
+      0.0004054480523336679,
+      0.001422431319952011,
+      0.015821930021047592,
+      0.0008529867045581341,
+      0.000323706422932446,
+      10.176154136657715,
+      0.0013034536968916655,
+      0.000512251746840775,
+      0.0025632502511143684,
+      0.00038034567842260003,
+      0.00029222609009593725,
+      0.0005778924096375704,
+      10.2711763381958,
+      0.0004394957795739174,
+      0.00048093884834088385,
+      0.000428747822297737,
+      0.00038113640039227903,
+      0.0003312640474177897,
+      0.020908623933792114,
+      0.0005429444136098027,
+      0.025768309831619263,
+      0.0005995655083097517,
+      10.188761711120605,
+      0.0003578412579372525,
+      0.0008831569575704634,
+      0.0004099773068446666,
+      0.0017696823924779892,
+      0.005343940574675798,
+      0.004370040725916624,
+      0.0006189042469486594,
+      0.0004469147534109652,
+      0.00030286950641311705,
+      0.0004075505130458623,
+      10.052818298339844,
+      0.00040609537973068655,
+      0.00029145798180252314,
+      0.00034905903157778084,
+      0.000332639116095379,
+      0.0003306953003630042,
+      10.344036102294922,
+      0.00032636921969242394,
+      10.019112586975098,
+      0.0017434933688491583,
+      10.030306816101074,
+      0.000565306399948895,
+      0.0005145104369148612,
+      0.0003189329872839153,
+      0.0007079012575559318,
+      0.0002950330963358283,
+      0.0005121096037328243,
+      10.055734634399414,
+      0.0004004499060101807,
+      10.019464492797852,
+      0.002155934227630496,
+      0.0004941442748531699,
+      0.00043852668022736907,
+      0.0019740823190659285,
+      0.0018054062966257334,
+      10.030364990234375,
+      0.010627523064613342,
+      0.0005926916492171586,
+      0.00032784402719698846,
+      0.0004604457935784012,
+      0.0004263249284122139,
+      0.00032164575532078743,
+      0.0005133289378136396,
+      0.00034067872911691666,
+      0.0006663633976131678,
+      0.0005500264232978225,
+      10.044910430908203,
+      0.0006980166072025895,
+      0.000546457536984235,
+      0.000353804585756734,
+      10.059196472167969,
+      0.00033812320907600224,
+      0.0005303638172335923,
+      0.004803686402738094,
+      10.019658088684082,
+      0.0003035414556507021,
+      0.00035930026206187904,
+      10.035774230957031,
+      0.00042563682654872537,
+      0.0006066757487133145,
+      0.00035554671194404364,
+      0.00042861088877543807,
+      0.00044286594493314624,
+      0.0003059499431401491,
+      0.0003453281242400408,
+      0.0004220261762384325,
+      0.0005285650258883834,
+      0.0004023576038889587,
+      10.8136625289917,
+      0.0019428523955866694,
+      0.0005612850072793663,
+      0.0004508616984821856,
+      0.00031340550049208105,
+      0.00033547967905178666,
+      0.0003754689241759479,
+      0.00032991639454849064,
+      0.0005444239359349012,
+      0.00035601318813860416,
+      0.0003134906874038279,
+      0.0002923414867836982,
+      0.0008749580010771751,
+      0.0056645432487130165,
+      0.0003222797531634569,
+      0.0003207501722499728,
+      0.0004882823559455574,
+      0.00034534226870164275,
+      0.0009857872501015663,
+      0.00042117145494557917,
+      0.0010314974933862686,
+      0.0002927738823927939,
+      0.00041998716187663376,
+      0.0003898536378983408,
+      0.00032511059544049203,
+      0.000389434106182307,
+      0.0009048377396538854,
+      0.0003942365583498031,
+      0.000588548369705677,
+      10.090950012207031,
+      0.0018781975377351046,
+      0.00029254419496282935,
+      0.00034292310010641813,
+      0.0003419907880015671,
+      10.25858211517334,
+      0.0006253079627640545,
+      10.016085624694824,
+      0.0005762295913882554,
+      0.010386484675109386,
+      0.0003205046523362398,
+      0.0006462051533162594,
+      0.0002930155023932457,
+      0.000382954312954098,
+      0.0003509865782689303,
+      0.00040533594437874854,
+      0.0004258020780980587,
+      0.0003770752518903464,
+      0.001102495240047574,
+      0.00031998337362892926,
+      0.0003507981891743839,
+      0.0009064788464456797,
+      0.00035019058850593865,
+      0.0005615135887637734,
+      0.00102073117159307,
+      10.566134452819824,
+      0.00043831593939103186,
+      10.324061393737793,
+      0.001936902990564704,
+      0.00032934616319835186,
+      0.0004713886883109808,
+      0.0005602877354249358,
+      0.0004042578220833093,
+      0.0029741525650024414,
+      10.04542064666748,
+      0.0003568530664779246,
+      0.0004240354464855045,
+      0.0013539856299757957,
+      0.0004768518265336752,
+      0.00031117230537347496,
+      0.0003027883358299732,
+      0.0003153950092382729,
+      0.00045615219278261065,
+      0.0006188482511788607,
+      0.004981611389666796,
+      0.00032865218236111104,
+      0.000770034734159708,
+      0.0004963420215062797,
+      0.000400298711610958,
+      0.0005654956912621856,
+      0.0006664120010100305,
+      0.0003440928412601352,
+      10.088988304138184,
+      0.0004696729884017259,
+      0.00029815465677529573,
+      0.0005272934213280678,
+      0.01289332378655672,
+      0.0005431452882476151,
+      0.00046691513853147626,
+      0.00039813530747778714,
+      0.0003440452564973384,
+      0.0003933308180421591,
+      0.00040043730405159295,
+      0.0007764056790620089,
+      0.0004039360792376101,
+      0.0005812906892970204,
+      0.00043331951019354165,
+      0.00033642983180470765,
+      0.000294335070066154,
+      0.0003935063723474741,
+      0.0020609316416084766,
+      0.0004751936940010637,
+      0.0003316575603093952,
+      10.155426025390625,
+      0.0018567381193861365,
+      10.234187126159668,
+      0.0009013076196424663,
+      0.0004926148103550076,
+      0.00034426350612193346,
+      0.0005556219257414341,
+      0.0004031299031339586,
+      0.00041533506009727716,
+      0.00043933236156590283,
+      0.00075289118103683,
+      0.0008942292188294232,
+      0.0015565321082249284,
+      0.007293062750250101,
+      0.0002915374643635005,
+      0.0007046320824883878,
+      0.0007343485485762358,
+      0.00040875861304812133,
+      0.000583234999794513,
+      0.00033361531677655876,
+      0.0010239399271085858,
+      10.032794952392578,
+      0.0003367697645444423,
+      0.000503973220475018,
+      0.0013150075683370233,
+      0.0005609854124486446,
+      0.0003362518618814647,
+      0.0003422876470722258,
+      0.007716118823736906,
+      10.014554023742676,
+      0.0003766275185626,
+      0.0021088426001369953,
+      0.0003628442354965955,
+      0.0003146500384900719
+    ],
+    [
+      0.00045236063306219876,
+      0.00026434651226736605,
+      0.0008864013361744583,
+      0.001711830380372703,
+      10.057476043701172,
+      0.0007005424704402685,
+      1.434910535812378,
+      0.00038871640572324395,
+      10.030904769897461,
+      10.16816520690918,
+      0.0003838920092675835,
+      0.0004295888065826148,
+      0.00044763810001313686,
+      0.0011478974483907223,
+      0.006260661408305168,
+      0.0003324053541291505,
+      0.00041602272540330887,
+      0.0002756445319391787,
+      0.0018442587461322546,
+      0.00032781806658022106,
+      0.0009630565182305872,
+      0.0005134132807143033,
+      0.000508687982801348,
+      0.0003382336290087551,
+      0.0002615417761262506,
+      0.0004970525624230504,
+      0.0005240546306595206,
+      10.038270950317383,
+      10.08368968963623,
+      0.0003739022940862924,
+      0.0049607111141085625,
+      0.0004002635250799358,
+      0.0003931771498173475,
+      0.00032089377054944634,
+      0.0010314795654267073,
+      0.0028975275345146656,
+      0.0003872410161420703,
+      0.0005181181477382779,
+      0.0008440011297352612,
+      0.00047900795470923185,
+      0.00046656353515572846,
+      10.004620552062988,
+      0.0007849450339563191,
+      0.0003173289878759533,
+      0.0003634571621660143,
+      0.00029651098884642124,
+      0.0004966465057805181,
+      0.004313073121011257,
+      0.0019392479443922639,
+      0.0004028423281852156,
+      0.00035777786979451776,
+      0.00026109497412107885,
+      0.00032590804039500654,
+      0.0006112653063610196,
+      10.042451858520508,
+      0.000638986995909363,
+      10.055750846862793,
+      0.00046334616490639746,
+      0.0004154003690928221,
+      0.0023212414234876633,
+      0.00031076042796485126,
+      0.0004265430325176567,
+      0.0003328458988107741,
+      0.0008590770303271711,
+      0.00041366234654560685,
+      0.0003569454129319638,
+      0.00044870650162920356,
+      10.01990032196045,
+      10.042254447937012,
+      0.0003601451753638685,
+      0.0006295751081779599,
+      0.000521352281793952,
+      0.0003992305137217045,
+      0.0010338019346818328,
+      10.055455207824707,
+      0.00026881828671321273,
+      0.001187217072583735,
+      0.0002880852553062141,
+      0.00026447841082699597,
+      0.0036014134529978037,
+      0.013429212383925915,
+      0.00032549246679991484,
+      0.09957699477672577,
+      0.0003742047701962292,
+      0.0004718436684925109,
+      0.0005349419661797583,
+      10.003296852111816,
+      0.0002614918921608478,
+      0.0003087951918132603,
+      10.119909286499023,
+      0.00046462431782856584,
+      0.0004029268166050315,
+      0.001108837896026671,
+      0.00040926766814664006,
+      0.00047136563807725906,
+      0.00043870590161532164,
+      0.0003176174941472709,
+      0.026243913918733597,
+      0.00030478090047836304,
+      0.0010167401051148772,
+      0.0004554153128992766,
+      0.0003770523180719465,
+      0.00036848511081188917,
+      0.000425183039624244,
+      0.8741932511329651,
+      0.0004736449627671391,
+      0.00046012987149879336,
+      0.00032978455419652164,
+      0.0004783210752066225,
+      0.00038585803122259676,
+      0.00037302481359802186,
+      0.00027984235202893615,
+      0.0011277773883193731,
+      0.0005622119060717523,
+      0.003907905891537666,
+      0.0002635745331645012,
+      0.005420098081231117,
+      0.00031461394974030554,
+      0.0004709141794592142,
+      0.003419435117393732,
+      0.0003397240652702749,
+      0.00042593656689859927,
+      0.0063278935849666595,
+      0.00028410463710315526,
+      10.050003051757812,
+      0.0004256641259416938,
+      0.0003922185569535941,
+      0.0004836572043132037,
+      0.0003838645643554628,
+      0.0015652719885110855,
+      0.004742186516523361,
+      0.0004876220191363245,
+      0.0004248438053764403,
+      0.0038093070033937693,
+      0.00033023644937202334,
+      0.0015951022505760193,
+      0.0003408155753277242,
+      10.1242094039917,
+      0.00031874768319539726,
+      0.0005425748531706631,
+      0.0002808866265695542,
+      0.00031687752925790846,
+      0.0004948450950905681,
+      0.00037159552448429167,
+      0.00026093865744769573,
+      0.000550218450371176,
+      0.00041740952292457223,
+      0.0002981287252623588,
+      0.00028866439242847264,
+      0.0003482587053440511,
+      10.014415740966797,
+      0.0003142970090266317,
+      0.000500442401971668,
+      0.0004394293646328151,
+      0.0006125950021669269,
+      0.0005542658036574721,
+      0.05545002594590187,
+      0.0005925896693952382,
+      0.0008266771328635514,
+      0.0036152235697954893,
+      0.00045928609324619174,
+      0.0013326734770089388,
+      10.004836082458496,
+      0.0005118509870953858,
+      0.0004062942462041974,
+      0.000388296291930601,
+      10.053632736206055,
+      10.003183364868164,
+      0.0004931460716761649,
+      10.031455039978027,
+      0.0003921134630218148,
+      0.0032148314639925957,
+      0.00035972267505712807,
+      10.153971672058105,
+      0.0004517157794907689,
+      0.0002911971532739699,
+      0.00035088538425043225,
+      0.0006350344629026949,
+      10.077688217163086,
+      0.0008487915620207787,
+      0.0005215026903897524,
+      0.0005466376896947622,
+      10.20242691040039,
+      0.0004356153658591211,
+      0.0005112101207487285,
+      0.0003291781758889556,
+      0.0005995286628603935,
+      10.047139167785645,
+      0.0003613749868236482,
+      0.001562827150337398,
+      0.0007706008036620915,
+      0.0003917636349797249,
+      0.006060006562620401,
+      0.00048660897300578654,
+      0.0004078915808349848,
+      0.00039058865513652563,
+      0.00034510495606809855,
+      0.0006456668488681316,
+      0.005363712087273598,
+      0.0003359052643645555,
+      0.0003827789914794266,
+      0.0012280065566301346,
+      10.069463729858398,
+      0.0003723733825609088,
+      0.0004262498114258051,
+      0.00238229357637465,
+      0.0009440483408980072,
+      0.0009715179912745953,
+      0.003928378690034151,
+      10.028955459594727,
+      0.0004727858176920563,
+      0.0017734908033162355,
+      0.00036543348687700927,
+      0.0003785387089010328,
+      0.0010415856959298253,
+      0.00036853112396784127,
+      10.040223121643066,
+      0.0015290983719751239,
+      0.0004387678054627031,
+      0.00035236572148278356,
+      0.00039861773257143795,
+      0.0008882658439688385,
+      10.022110939025879,
+      0.0036263572983443737,
+      0.0004184719582553953,
+      10.044540405273438,
+      0.000276342238066718,
+      0.0005081915878690779,
+      0.00031893164850771427,
+      0.0037350808270275593,
+      0.00040461093885824084,
+      0.0003330525942146778,
+      0.0016244553262367845,
+      0.0005055454093962908,
+      0.00026299760793335736,
+      0.0003513568372000009,
+      0.0004657460667658597,
+      10.001442909240723,
+      0.00028667450533248484,
+      0.0002869176387321204,
+      0.0005332306027412415,
+      0.0015379067044705153,
+      0.0018616487504914403,
+      0.0003905078046955168,
+      0.0002892052580136806,
+      0.000485336990095675,
+      0.0003678991342894733,
+      0.00033464626176282763,
+      10.002801895141602,
+      0.00026380637427791953,
+      0.0004976163618266582,
+      0.0003418565320316702,
+      0.00048808997962623835,
+      0.0005647801444865763,
+      0.00046344910515472293,
+      0.0003708093718159944,
+      0.019066395238041878,
+      0.00037850730586797,
+      0.0009338136878795922,
+      0.0004245720920152962,
+      0.0009805912850424647,
+      0.00043317608651705086,
+      0.00048458046512678266,
+      0.0035631104838103056,
+      0.00044162123231217265,
+      0.0006442029844038188,
+      0.0004648454487323761,
+      0.00043976132292300463,
+      0.00026341297780163586,
+      0.0005045581492595375,
+      0.0003308220475446433,
+      0.00048298665205948055,
+      10.031598091125488,
+      0.0005117455730214715,
+      0.014150155708193779,
+      0.000307958252960816,
+      0.0002647781802807003,
+      0.00029669623472727835,
+      0.0013072789879515767,
+      10.290255546569824,
+      0.00034950824920088053,
+      0.0003285995335318148,
+      0.000527145282831043,
+      0.0003717808867804706,
+      0.0004774930130224675,
+      0.0038030820433050394,
+      0.0004222691059112549,
+      0.005138902459293604,
+      0.00039598136208951473,
+      0.00039201698382385075,
+      0.0004371970717329532,
+      0.0004004467627964914,
+      10.037381172180176,
+      0.0003462711756583303,
+      0.0013971107546240091,
+      0.00043510686373338103,
+      0.0003444562607910484,
+      0.00044281326700001955,
+      0.0010520829819142818,
+      0.0005298705073073506,
+      0.0003575954178813845,
+      10.033256530761719,
+      0.0016501095378771424,
+      0.00026211203658021986,
+      0.0005568803753703833,
+      0.0003134347207378596,
+      0.0011769433040171862,
+      0.000261956884060055,
+      0.00042029187898151577,
+      0.0006826920434832573,
+      0.00045724501251243055,
+      0.0005074806977063417,
+      0.00042397293145768344,
+      0.0004341060121078044,
+      0.0003440587315708399,
+      0.00029924866976216435,
+      0.00033520162105560303,
+      0.0003076377615798265,
+      0.0010939016938209534,
+      0.0023290610406547785,
+      10.01021671295166,
+      0.017415495589375496,
+      0.0003256927302572876,
+      0.0004484733217395842,
+      0.0016153306933119893,
+      0.00034979963675141335,
+      0.00031333102378994226,
+      0.00027290667640045285,
+      0.0005125371972098947,
+      0.0002825883566401899,
+      0.003605470759794116,
+      0.0012340872781351209,
+      0.00047288095811381936,
+      0.00026149788754992187,
+      0.0005520069389604032,
+      0.00038508576108142734,
+      0.004057307727634907,
+      0.003945962525904179,
+      10.07349681854248,
+      0.006008333992213011,
+      0.00035063186078332365,
+      0.0005126105388626456,
+      10.029349327087402,
+      0.0005106932367198169,
+      0.00026318198069930077,
+      0.0003790851915255189,
+      0.0005518961697816849,
+      10.048177719116211,
+      10.012214660644531,
+      0.0009972188854590058,
+      0.0007358682923950255,
+      0.0012961068423464894,
+      0.0005427963333204389,
+      0.0002886778675019741,
+      0.0030647125095129013,
+      0.0004184723657090217,
+      0.0002660441678017378,
+      0.016546141356229782,
+      10.025076866149902,
+      0.0003411117650102824,
+      0.00036914279917255044,
+      0.00033806433202698827,
+      0.0002840044617187232,
+      0.00026621317374520004,
+      0.000665654253680259,
+      0.00045907459571026266,
+      0.000462042837170884,
+      0.0004582816909532994,
+      0.00030912927468307316,
+      0.00038153817877173424,
+      0.0003302210243418813,
+      0.0004742609162349254,
+      0.000413451372878626,
+      0.0004354293632786721,
+      10.073020935058594,
+      0.0003763202694244683,
+      10.534000396728516,
+      0.0004871693381574005,
+      0.0012926476774737239,
+      0.002233613980934024,
+      0.0005095258820801973,
+      0.000996661838144064,
+      0.0009777010418474674,
+      0.00042154412949457765,
+      10.001173973083496,
+      0.00047262624138966203,
+      0.0006145129445940256,
+      0.00048034716746769845,
+      10.02835750579834,
+      0.000695948489010334,
+      0.00029664786416105926,
+      0.0020451308228075504,
+      0.0006336750811897218,
+      0.002389391651377082,
+      0.0004073136660736054,
+      0.00028576425393112004,
+      0.00034054595744237304,
+      10.022526741027832,
+      0.00032267809729091823,
+      0.0004460986237972975,
+      0.0002780169015750289,
+      10.043218612670898,
+      0.0003236444899812341,
+      10.024738311767578,
+      0.0003406226169317961,
+      0.0005086249439045787,
+      0.0004492998996283859,
+      0.000444147881353274,
+      0.00028702407144010067,
+      0.0004104471590835601,
+      0.0002846885472536087,
+      0.007252264767885208,
+      0.0005518832476809621,
+      0.0004044481029268354,
+      0.0004156432405579835,
+      0.0007334100664593279,
+      0.00031312103965319693,
+      0.002092991955578327,
+      0.0003405277675483376,
+      0.00029072933830320835,
+      0.0004889297997578979,
+      0.0004463224031496793,
+      0.00038210689672268927,
+      0.00027639491599984467,
+      0.00032436559558846056,
+      0.0002751169668044895,
+      0.009890764951705933,
+      10.160792350769043,
+      0.00028398819267749786,
+      0.0005616586422547698,
+      0.00027579779271036386,
+      0.07472816109657288,
+      0.0004234218504279852,
+      0.003044073237106204,
+      0.0024112756364047527,
+      0.0003274890477769077,
+      0.0020232798997312784,
+      0.00044398021418601274,
+      0.00037537128082476556,
+      0.0003540088946465403,
+      0.0010741617297753692,
+      0.00042996418778784573,
+      0.00036596530117094517,
+      0.00044163892744109035,
+      0.0003093639970757067,
+      0.00046565107186324894,
+      0.0004409858083818108,
+      0.00035996080259792507,
+      0.0004076120676472783,
+      0.0008286162628792226,
+      0.00037513975985348225,
+      0.0006538774468936026,
+      10.010176658630371,
+      0.008059469982981682,
+      0.00036956541589461267,
+      10.015402793884277,
+      0.00029724108753725886,
+      0.0004919606144540012,
+      0.00040708729648031294,
+      0.00046987354289740324,
+      0.0003084417257923633,
+      0.0005076825036667287,
+      0.0036273368168622255,
+      0.0003239801153540611,
+      0.00031752086943015456,
+      0.00033769445144571364,
+      0.0004511807346716523,
+      0.00035722425673156977,
+      0.0002743012155406177,
+      0.0005081580602563918,
+      10.021855354309082,
+      0.0004614471399690956,
+      0.0006006945623084903,
+      0.0002991924702655524,
+      0.00047367429942823946,
+      0.0005232951370999217,
+      0.0014212022069841623,
+      0.0002693606948014349,
+      0.00042253814171999693,
+      0.0021207884419709444,
+      0.0003562965721357614,
+      0.0013535849284380674,
+      0.0015020958380773664,
+      0.0005515968077816069,
+      0.00047423833166249096,
+      0.00029901647940278053,
+      0.00035930000012740493,
+      0.0004684054001700133,
+      10.086353302001953,
+      0.00044676268589682877,
+      10.090246200561523,
+      0.00026627996703609824,
+      0.00035170812043361366,
+      10.01407527923584,
+      0.00048677140148356557,
+      0.00027738558128476143,
+      0.00026119756512343884,
+      0.000262414658209309,
+      0.0008816419285722077,
+      0.0007772268727421761,
+      0.000443622178863734,
+      0.0002634293050505221,
+      0.00038450967986136675,
+      0.0017408885760232806,
+      0.0003101278271060437,
+      0.0014677151339128613,
+      0.00030267558759078383,
+      0.0026153759099543095,
+      0.006199484691023827,
+      0.006977311801165342,
+      0.00039906916208565235,
+      0.00043424099567346275,
+      10.148207664489746,
+      0.005416149739176035,
+      0.0004026053939014673,
+      0.0004896786413155496,
+      0.0005199830047786236,
+      0.0005945725715719163,
+      0.0004029364208690822,
+      0.0002626336645334959,
+      0.0007439365144819021,
+      0.0003423572634346783,
+      0.0014270850224420428,
+      0.0004400616744533181,
+      0.0003532444825395942,
+      0.0003134215949103236,
+      0.0002954319934360683,
+      0.0005240425816737115,
+      0.002389973495155573,
+      0.0005156450206413865,
+      0.0015695177717134356,
+      0.0008968707988969982,
+      0.00035924481926485896,
+      0.0004717325500678271,
+      0.0025384912732988596,
+      0.00034905894426628947,
+      0.0003656733315438032,
+      0.00036593040567822754,
+      0.0013010659022256732,
+      0.004445606376975775,
+      0.0004971013986505568,
+      0.00046819017734378576,
+      0.002315511694177985,
+      0.00028812920209020376,
+      0.0003960451576858759,
+      10.16661548614502,
+      10.001630783081055,
+      0.0004701967118307948,
+      0.0002726582461036742,
+      0.004596075974404812,
+      0.0003511119866743684,
+      0.0003656060143839568,
+      0.0007101536029949784,
+      0.0003479314036667347,
+      0.0024586482904851437,
+      0.005129736848175526,
+      0.0009435468236915767,
+      0.00032394117442891,
+      0.0003542510385159403,
+      0.0005115767708048224,
+      0.00033899792470037937,
+      0.0007561133825220168,
+      0.0002779159403871745,
+      0.0006340600666590035,
+      0.0012528580846264958,
+      0.0005685040960088372,
+      0.0005529580521397293,
+      0.006089525297284126,
+      0.0007664118893444538,
+      0.00027071856311522424,
+      10.008852005004883,
+      0.0004844141658395529,
+      0.0010349962394684553,
+      0.00044096686178818345,
+      0.0002611647068988532,
+      10.055500030517578,
+      10.055858612060547,
+      0.00033050516503863037,
+      0.0005192974931560457,
+      0.0005720185581594706,
+      0.0005669542588293552,
+      0.0006371092749759555,
+      0.00046939036110416055,
+      0.0013616826618090272,
+      0.00454803928732872,
+      0.00040911982068791986,
+      0.0004131313762627542,
+      0.0011453035986050963,
+      0.0002665101201273501,
+      0.002234218642115593,
+      0.0005613219109363854,
+      0.006465577520430088,
+      0.000703324971254915,
+      0.00034352700458839536,
+      0.0005037601804360747,
+      0.00029456612537615,
+      0.0005133178201504052,
+      0.0002708551473915577,
+      0.0005604282487183809,
+      0.0003447424678597599,
+      0.0005176517879590392,
+      0.00124424381647259,
+      0.0005791523144580424,
+      0.00033227098174393177,
+      0.0020616119727492332,
+      0.0004335471894592047,
+      0.0007982297101989388,
+      10.01159954071045,
+      0.0003997947205789387,
+      0.00048812999739311635,
+      0.0004908244009129703,
+      10.003552436828613,
+      0.0005129948258399963,
+      0.000319920654874295,
+      0.0002611430245451629,
+      0.00029885797994211316,
+      0.00035881789517588913,
+      0.0005458301166072488,
+      0.0003015753463841975,
+      0.0007425698568113148,
+      0.000275836733635515,
+      0.00034269789466634393,
+      0.0030087640043348074,
+      10.110722541809082,
+      0.0003798874677158892,
+      0.00056846858933568,
+      10.006515502929688,
+      0.0004861225315835327,
+      0.0005408516735769808,
+      0.0004896193277090788,
+      0.0028549744747579098,
+      0.00030383781995624304,
+      0.0002745943493209779,
+      0.0013526208931580186,
+      0.00048133800737559795,
+      0.0003159195766784251,
+      0.0006534355925396085,
+      0.0002696490555536002,
+      0.00028676309739239514,
+      10.010702133178711,
+      0.0003703852999024093,
+      0.0005614926922135055,
+      0.00042022779234685004,
+      0.0011752699501812458,
+      0.0009338996023871005,
+      0.0003504543856251985,
+      0.0005143476882949471,
+      0.0003056420828215778,
+      0.004709989298135042,
+      0.0006328113959170878,
+      0.00038008674164302647,
+      0.0003764171851798892,
+      0.0010855176951736212,
+      10.014520645141602,
+      10.027713775634766,
+      0.0003179769264534116,
+      0.000882709224242717,
+      0.005627131555229425,
+      0.00029565943987108767,
+      0.00039411778561770916,
+      10.005242347717285,
+      0.00026786685339175165,
+      0.003190709976479411,
+      0.0006749660242348909,
+      0.00027011975180357695,
+      0.0003681851667352021,
+      0.00047062101657502353,
+      0.0004291331279091537,
+      0.0005281449994072318,
+      0.00046955287689343095,
+      0.000435910391388461,
+      0.003346869954839349,
+      0.0003354980726726353,
+      0.005164106842130423,
+      0.001352888299152255,
+      0.0017225475749000907,
+      0.00029368881951086223,
+      0.000288382638245821,
+      0.002311708638444543,
+      0.001357522327452898,
+      0.001160224201157689,
+      0.0005504758446477354,
+      0.004754437133669853,
+      0.005604547448456287,
+      0.0003223250387236476,
+      0.0003476405399851501,
+      0.003028896637260914,
+      0.00042883737478405237,
+      0.00038552298792637885,
+      0.00046039847075007856,
+      0.0003244707768317312,
+      0.000405994855100289,
+      0.00041504602995701134,
+      0.0002617600839585066,
+      0.00032877223566174507,
+      0.00040430965600535274,
+      0.0005325556267052889,
+      0.0014026184799149632,
+      0.08182022720575333,
+      0.0003344612196087837,
+      0.006220647599548101,
+      0.0004810526152141392,
+      0.0006830390775576234,
+      0.00026171415811404586,
+      0.0002978490956593305,
+      0.0005505485460162163,
+      0.00027523477911017835,
+      0.0015947798965498805,
+      0.0004622865526471287,
+      0.00037567957770079374,
+      0.0005279989563859999,
+      0.00026579262339510024,
+      10.060546875,
+      0.0008466070867143571,
+      0.0007698407862335443,
+      0.0013988431310281157,
+      0.00030481634894385934,
+      0.0004967573913745582,
+      0.0008642724133096635,
+      0.000261640059761703,
+      0.0013755100080743432,
+      0.00035870500141754746,
+      0.0004909826093353331,
+      10.026178359985352,
+      0.0005015162751078606,
+      0.0004715400864370167,
+      0.00029710421222262084,
+      0.0005013818736188114,
+      0.0002919464313890785,
+      0.00032576947705820203,
+      0.00041545068961568177,
+      0.00034420378506183624,
+      10.057757377624512,
+      10.062053680419922,
+      0.00040833171806298196,
+      0.0003668587014544755,
+      0.0002845970739144832,
+      0.02629697695374489,
+      10.145084381103516,
+      0.0007353850523941219,
+      0.00028168194694444537,
+      0.0028873307164758444,
+      10.063132286071777,
+      0.00036761403316631913,
+      0.00057459925301373,
+      0.0019123210804536939,
+      0.00029550414183177054,
+      0.0006746520521119237,
+      0.00032268394716084003,
+      0.01226389966905117,
+      0.01080909464508295,
+      0.0006055564736016095,
+      0.0075580524280667305,
+      0.00030399399111047387,
+      0.0002820332010742277,
+      0.0003649946884252131,
+      0.001027350896038115,
+      0.00039051900967024267,
+      0.00046115025179460645,
+      0.0003916616551578045,
+      0.00035237311385571957,
+      0.006484825164079666,
+      0.00026147966855205595,
+      0.0009768471354618669,
+      0.00040400135912932456,
+      10.120814323425293,
+      0.030100787058472633,
+      0.00043703202391043305
+    ]
+  ],
+  "layers": [
+    {
+      "layer": 0,
+      "channels": 96,
+      "ct_deformed_mean": 0.015019077807664871,
+      "release_mean": 0.014022364281117916,
+      "mae_impact_mean": 0.015791933983564377,
+      "release_minus_deformed": -0.0009967155056074262,
+      "mae_minus_release": 0.001769572263583541,
+      "ordered_fraction": 0.3106796145439148,
+      "max_weight": 10.149458885192871,
+      "min_weight": 0.00048017740482464433,
+      "channels_gt_1": 10.0
+    },
+    {
+      "layer": 1,
+      "channels": 192,
+      "ct_deformed_mean": 0.02139180339872837,
+      "release_mean": 0.02340662106871605,
+      "mae_impact_mean": 0.03371495008468628,
+      "release_minus_deformed": 0.0020148169714957476,
+      "mae_minus_release": 0.010308332741260529,
+      "ordered_fraction": 0.6893203854560852,
+      "max_weight": 10.655861854553223,
+      "min_weight": 0.0004643251886591315,
+      "channels_gt_1": 19.0
+    },
+    {
+      "layer": 2,
+      "channels": 384,
+      "ct_deformed_mean": 0.07979151606559753,
+      "release_mean": 0.07723046839237213,
+      "mae_impact_mean": 0.09813079982995987,
+      "release_minus_deformed": -0.0025610490702092648,
+      "mae_minus_release": 0.020900335162878036,
+      "ordered_fraction": 0.3689320385456085,
+      "max_weight": 10.8136625289917,
+      "min_weight": 0.0002913873759098351,
+      "channels_gt_1": 38.0
+    },
+    {
+      "layer": 3,
+      "channels": 768,
+      "ct_deformed_mean": 0.039202284067869186,
+      "release_mean": 0.038203440606594086,
+      "mae_impact_mean": 0.04993002489209175,
+      "release_minus_deformed": -0.0009988457895815372,
+      "mae_minus_release": 0.011726588010787964,
+      "ordered_fraction": 0.35922330617904663,
+      "max_weight": 10.534000396728516,
+      "min_weight": 0.00026093865744769573,
+      "channels_gt_1": 77.0
+    }
+  ],
+  "objective": {
+    "order": "CT_deformed < MAE_IMPACT_sCT",
+    "margin_def_rel": 0.01,
+    "margin_rel_mae": 0.01,
+    "lambda_separation": 0.1,
+    "lambda_l1": 0.0,
+    "lambda_range": 0.01,
+    "max_weight": 10.0
+  }
+}