Revised CT/CBCT for Task 1 & Task 2 – Now Available

Dear all,

We have uploaded a revised version of the CT for both Task 1 and Task 2 to address an inconsistency in intensity clipping that affected some of the previously released data.

🔍 What happened?

An outdated version of our processing script was mistakenly used for part of the pipeline, which resulted in missing or inconsistent clipping in a subset of CT volumes. In some cases, unusually high intensity values originated from the clinical source data and were not correctly constrained in the final release.

✅ What has been fixed?

• All CTs (deformed and non-deformed) in the datasets have now been consistently clipped to the correct intensity range: [-1024, 3071] HU. N.B. in some cases, the CT range may be even smaller.
• The issue has been resolved for both Task 1 and Task 2 datasets.
• CBCTs have also been re-clipped accordingly.
• MRI remains unchanged.

📦 Where to find the updated data?
The updated datasets have been uploaded to Zenodo, and the corresponding scripts are available on GitHub:

• Zenodo: [https://doi.org/10.5281/zenodo.14918089] (https://doi.org/10.5281/zenodo.14918089)
• GitHub: clip_data.py script

If you have already downloaded the previous version and prefer to update locally, please apply the script above to your CT and CBCT volumes. However, we strongly suggest re-downloading the whole dataset at https://doi.org/10.5281/zenodo.14918213.

Note: Data from Center D is still distributed separately from the main dataset. The updated version (v2) is available for download at the following link: Download

🙏 Apologies

We sincerely apologize for the inconvenience and confusion this may have caused. We’re grateful for the community's vigilance and feedback, which helped us identify and resolve this issue promptly. To err is human, and in research, it's through transparency and swift correction that we build trust and move forward together.

Thank you for being so understanding, and please don’t hesitate to reach out if you have questions.

Warm regards,
Matteo and the SynthRAD2025 Team

In the updated SynthRAD2025 Grand Challenge dataset (11.05.2025, v2) (https://zenodo.org/records/15373853) are the data from Center D already included in the folders synthRAD2025_Task1_Train.zip and synthRAD2025_Task2_Train.zip?

Best regards

Dear angeldid,

Thank you for your message.

Please note that the data from Center D continues to be distributed separately from the main dataset. The updated version (v2) is available for download at the following link: Download

We have now also updated the original post to include this download link. Thank you for bringing this to our attention.

Best regards, Adrian and the SynthRAD2025 Team

Dear Matteo,

Could you please clarify it the training set was changed? Also, was that change only clipping the HU values?

Best regards!

Dear andreferreira,

Yes, I can confirm that the training has been updated and clipping has been fixed. To ensure things are consistent throughout the datasets, we suggest re-downloading the data once more, instead of re-applying the clipping.

Warm Regards,

Matteo