fidoptnnsyn
nnsyn, CSIRO AEHRC's nnU-Net ResEnc-L that regresses the CT (team KoalAI, first in SynthRAD2025's MR-to-CT task), with its authors' released SynthRAD2025 models, one per source modality and region, run as their challenge docker runs them on the one fold they published.
SynthesisGPUversion nnsyn af36483; fold 0 of huggingface.co/aehrc/Synthrad2025_task1 (ee17646) and aehrc/Synthrad2025_task2 (30da517), SHA-256 checked; CT de-normalisation of aehrc/synthrad2025_docker (25cbfe7)every datasetApache-2.0
Run it
With its KonfAI app, at the benchmark's settings (the fidopt app repositories are private for now: a Hugging Face token with access opens them):
pip install konfai-apps
| model | Command | |
|---|---|---|
| CBCT to CT, head and neck | Get the app | konfai-apps infer fideus-labs/fidopt-synthesis:nnsyn_CBCT_HN -i image.nii.gz -o out --gpu 0 |
| CBCT to CT, thorax | Get the app | konfai-apps infer fideus-labs/fidopt-synthesis:nnsyn_CBCT_TH -i image.nii.gz -o out --gpu 0 |
| CBCT to CT, abdomen | Get the app | konfai-apps infer fideus-labs/fidopt-synthesis:nnsyn_CBCT_AB -i image.nii.gz -o out --gpu 0 |
| MR to CT, head and neck | Get the app | konfai-apps infer fideus-labs/fidopt-synthesis:nnsyn_MR_HN -i image.nii.gz -o out --gpu 0 |
| MR to CT, thorax | Get the app | konfai-apps infer fideus-labs/fidopt-synthesis:nnsyn_MR_TH -i image.nii.gz -o out --gpu 0 |
| MR to CT, abdomen | Get the app | konfai-apps infer fideus-labs/fidopt-synthesis:nnsyn_MR_AB -i image.nii.gz -o out --gpu 0 |
Results
| Dataset | model | # | Read by | Score | Failed | Time s |
|---|---|---|---|---|---|---|
| Head Neck CBCT to CT (HUST) | CBCT to CT, head and neck | MAE | 201.7 | 0 | 13 | |
| HaN-Seg, T1 MR to CT | MR to CT, head and neck | MAE | 101.6 | 0 | 34 | |
| ICASSP-2024 3D-CBCT challenge, simulated low-dose CBCT to CT | CBCT to CT, thorax | MAE | 311.6 | 0 | ||
| 4D-Lung, CBCT to CT | CBCT to CT, thorax | MAE | 66.8 | 0 | ||
| Paired head CT and MR, MR to CT | MR to CT, head and neck | MAE | 142.7 | 0 | ||
| Pelvic-Reference-Data, CBCT to CT | CBCT to CT, abdomen | MAE | 45.4 | 0 | ||
| Simulated CBCT to planning CT, LCTSC | CBCT to CT, thorax | MAE | 353.3 | 0 | 22 | |
| SynthRAD2023, CBCT to CT, brain | CBCT to CT, head and neck | MAE | 98.4 | 0 | 15 | |
| SynthRAD2023, MR to CT, brain | MR to CT, head and neck | MAE | 131.8 | 0 | ||
| SynthRAD2023, CBCT to CT, pelvis | CBCT to CT, abdomen | MAE | 100.5 | 0 | ||
| SynthRAD2023, MR to CT, pelvis | MR to CT, abdomen | MAE | 92.9 | 0 | 32 |
- Reference
- Rogowski et al., SynthRAD2025 challenge report, arXiv:2605.13555 (2026), team KoalAI