fidoptSynthRAD2025, MR to CT, thorax and abdomen
The MR of one patient's thorax or abdomen and the CT to synthesise from it, the CT brought onto the MR's grid by the challenge (SynthRAD2025, Task 1).
30 test casesthorax, abdomensource MR, target CTresearch only
Read before comparing (5)
- Case. The challenge's pair as delivered, copied byte for byte from the organisers' training archive synthRAD2025_Task1_Train.zip, its size and MD5 pinned: the MR, the CT the organisers registered onto its grid, and their body mask. The synthetic image is scored inside the mask, as the challenge scores it. The archive is given after registration on the challenge's site.
- Alignment. How the organisers registered the CT to the input decides how far a perfect synthesis can match it: what the registration left misaligned counts as error for every method.
- Baseline. Inside the mask on the 30 test cases, with KonfAI's criteria at CT's range of 4095, a constant at the target's median scores an MAE of 294 (157 to 445) HU, a PSNR of 19.2 (17.1 to 22.8) dB and an SSIM of 0.16 (-0.04 to 0.40). On the 28 development cases its MAE is 290.
- Source. The MR taken as the CT scores an MAE of 505 (306 to 1142) HU, a PSNR of 16.2 (10.5 to 19.4) dB and an SSIM of -0.05 (-0.14 to 0.15), better than the constant on 0 of the 30; 577 on the development cases.
- Overlap. Besides SynthRAD2025_MRCT, only DoseRAD2026_MRCT_Syn holds Task 1's thorax and abdomen patients, its centre B images under the same ids: its 20 cases are held out, and 11 other patients of its pool are cases here. The provider's two tasks may hold one patient under two numbers, as SynthRAD2023's centre B brain patients are, so every case's CT was registered rigidly to every CT of the corpus's Task 2 sets of the same region (SynthRAD2025_CBCTCT's 20, SynthRAD2025_CBCTCT_Syn's 60): they correlate inside the bodies at most 0.50, where one CT cropped and moved reads 0.95 or more.
No method has a result on every case of this dataset yet.
Case by case
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
- 30 test patients drawn with seed 2026 by region and centre in proportion, and 28 development patients from the end of the same seeded order. The pool is the 317 of Task 1's 357 abdomen and thorax training patients left once SynthRAD2025_MRCT's 10 test and 10 development patients, whose CT, this target, is public there as a fixed image, and DoseRAD2026_MRCT_Syn's, centre B images, are held out (draw.held_out).
- Truth
- target image as released the patient's CT as the challenge delivers it, registered to the MR by the organisers, in Hounsfield units
- Licence
- CC BY-NC 4.0 (centres A, B, C, E)
28 cases for tuning, never tested
AB_1ABA029, AB_1ABA065, AB_1ABA108, AB_1ABA060, AB_1ABA039, AB_1ABB161, AB_1ABB145, AB_1ABB039, AB_1ABB143, AB_1ABB020, AB_1ABB047, AB_1ABB025, AB_1ABC016, TH_1THA017, TH_1THA288, TH_1THA050, TH_1THA244, TH_1THA224, TH_1THA283, TH_1THA026, TH_1THA016, TH_1THB200, TH_1THB207, TH_1THB179, TH_1THB191, TH_1THB126, TH_1THB029, TH_1THB019