fidoptSynthRAD2023, CBCT to CT, brain
Cone-beam CT of one patient's brain and the CT to synthesise from it, which the challenge brought onto the CBCT's grid (SynthRAD2023, Task 2).
Read before comparing (6)
- Cases. The challenge's pair as delivered, copied byte for byte from Task2.zip: the CBCT, the CT the organisers registered onto its grid, and their body mask, inside which the synthetic image is scored, as the challenge scores it.
- Registration. How the organisers registered the CT to the CBCT decides how far a perfect synthesis can match it: what the registration left misaligned counts as error for every method.
- Overlap. SynthRAD_BrainCBCTCT's patients are left out because their CT, the target here, is that registration set's public fixed image. Centre B brain patients carry one number in both tasks, 1BBnnn and 2BBnnn the same patient with the same CT (correlation 0.99 after rigid registration on the eight such pairs the two registration sets hold), so the twins of Task 1 cases are left out too.
- Other patients. The provider says its two tasks 'do not necessarily contain the same patients', so every drawn case's CT was registered rigidly to every CT of SynthRAD_BrainMRCT and SynthRAD_BrainMRCT_Syn: they correlate at most 0.91 inside the bodies, where one CT on two grids reads 0.98 or more and a centre-B patient's two CTs 0.998. No dataset outside the eight SynthRAD2023 sets draws on its record.
- Baseline. Inside the mask on the 13 test cases, with KonfAI's criteria at CT's range of 4095, a constant at the target's median scores an MAE of 399 (354 to 540) HU, a PSNR of 16.9 (16.1 to 17.1) dB and an SSIM of 0.32 (-0.02 to 0.38).
- Input. The CBCT taken as the CT is no floor: its values are not Hounsfield units at every centre, its MAE on the test cases 1023 to 1270 HU at 2BA, 239 to 293 at 2BB and 89 to 140 at 2BC.
No method has a result on every case of this dataset yet.
Under a licence that is not open
Methods whose code or weights are released for non-commercial use, or under share-alike or custom terms: shown for reference, ranked nowhere.
| Accuracy | Cost | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| # | Method | MAE | PSNR dB | SSIM | Anatomy Dice | Anatomy HD95 mm | Time s | RAM GB | GPU GB | Machine | Images |
| nnsyn · CBCT to CT, head and neckGPUapptrained on CC-BY-NC-4.0 dataolder run | 98.4 ±25.9 | 26.85 ±1.77 | 0.821 ±0.036 | 0.711 ±0.098 | 4.53 ±1.95 | 15 | 4.0 | 2.6 | Laptop | ||
MAE: mean absolute error of a synthetic image against its target, inside the body, in the image's units. PSNR: the same by signal-to-noise ratio, higher is better. SSIM: the same by structural similarity, 1 is perfect. Anatomy Dice and HD95: TotalSegmentator run on the synthetic and on the target image, the structures compared; shown, not ranked.
Case by case
| MAE | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 |
|---|
Click a score to look at that result: the target image, and over it the method's synthetic image. Underlined scores have their image; for the others you see the case without the result.
About the data
- Test cases
- 13 test patients drawn with seed 2026 by centre (2BA, 2BB, 2BC) in proportion, then 12 development patients, the last of the same seeded order, the centres in proportion (5 2BA, 3 2BB, 4 2BC), none of them a patient of SynthRAD_BrainCBCTCT or a Task 1 case's twin. The pool is the 136 of Task 2's 180 brain training patients left without draw.held_out's 44: 26 of SynthRAD_BrainCBCTCT's 34 patients and the 11 and the 7 whose Task 1 twin is a case of SynthRAD_BrainMRCT or of SynthRAD_BrainMRCT_Syn. SynthRAD_BrainCBCTCT's other 8, of centre B, are in the pool and drawn neither for test nor for development.
- Truth
- target image as released the patient's CT as the challenge delivers it, registered to the CBCT by the organisers, in Hounsfield units
- Licence
- CC BY-NC 4.0
12 cases for tuning, never tested
brain_2BA029, brain_2BC021, brain_2BB111, brain_2BA073, brain_2BC075, brain_2BB041, brain_2BA080, brain_2BC056, brain_2BA063, brain_2BC048, brain_2BB007, brain_2BA051