STS 2026 : The 4th Semi-supervised Teeth Segmentation Challenge on Metal Artifact Reduction and Beyond

MICCAI 2026, Strasbourg, France

September 27-October 1, 2026

Task 2 Winners: CBCT-IOS Registration

Task 2 evaluates rigid registration between intraoral scans (IOS) and CBCT volumes. Teams submit upper- and lower-jaw transformation matrices that align the high-resolution IOS crown surfaces with the CBCT roots, bone and surrounding anatomy, and are assessed with the Mean Translation Error (MTE, mm) and Mean Rotation Error (MRE, degree); lower is better for both.

The final ranking is based on the official Docker test evaluation published on the Codabench competition page; the Codabench validation leaderboard is not used for the final ranking. Failed or missing pairs receive fixed worst-case penalties, and the top three teams are highlighted below.

Team MTE (mm, ↓) MRE (deg, ↓) Successful / Total Pairs Status Final Rank Links
Useravalanchezy 7.902523 2.862457 198/198 Success 🥇 1
Userqingfeng 39.334334 11.696383 198/198 Success 🥈 2
Useraimagelabzip 10114.828872 36.122669 160/198 Completed with Errors 🥉 3

Additional Analysis: Left-Handed Coordinate Samples Excluded

Because of coordinate-system orientation inconsistencies in the dataset, the organizers also evaluated the submissions after excluding samples with left-handed coordinate systems. The results are shown below for reference only; the final ranking and awards are determined by the main table above.

Rank Team Weighted Translation Error (mm, ↓) Weighted Rotation Error (deg, ↓) Successful / Total Pairs Failed Pairs Status
1 Useraimagelabzip 6.752031 1.976201 160/162 2 Completed with Errors
2 Useravalanchezy 8.543497 2.612932 162/162 0 Success
3 Userqingfeng 45.399751 11.994508 162/162 0 Success

MTE is the primary ranking metric, with MRE used as the secondary metric. Both are computed over the upper and lower jaws of every evaluation case.

“Report” opens the team’s technical report, “Slides” the presentation deck shared for the STS 2026 winners’ session, and “Code” the team’s public repository. Greyed-out items were not released by the team.

Official results are published on the Codabench competition page. See the evaluation protocol for metric definitions and penalty values.