SynthRAD2025 Benchmark Tests Synthetic CT for Radiotherapy
The SynthRAD2025 challenge assessed synthetic CT (sCT) generation techniques using data from 2,362 patients across five European centers, focusing on the head and neck, thorax, and abdomen. Two specific tasks were analyzed: MRI-to-CT involving 890 cases and CBCT-to-CT with 1,472 cases. Evaluation metrics comprised image similarity (MAE, PSNR, MS-SSIM), segmentation accuracy (Dice, HD95), and dosimetric precision from both photon and proton treatment plans. Task 1 saw participation from 803 individuals, with 12 out of 13 submissions deemed valid, showcasing top performance. This challenge builds upon SynthRAD2023, emphasizing the conversion of MRI or CBCT into CT-equivalent images with precise Hounsfield Unit (HU) values for MRI-only RT and CBCT-based adaptive workflows.
Key facts
- SynthRAD2025 benchmarked sCT methods on 2,362 patients from five European centers.
- Two tasks: MRI-to-CT (890 cases) and CBCT-to-CT (1,472 cases).
- Evaluated via image similarity (MAE, PSNR, MS-SSIM), segmentation (Dice, HD95), and dosimetric metrics.
- 803 participants with 12/13 valid submissions.
- Task 1 top performance achieved.
- Addresses MRI-only RT and CBCT-based adaptive workflows.
- Builds on SynthRAD2023.
- Covers head and neck, thorax, and abdomen.
Entities
Locations
- Europe