Challenge outcomesΒΆ

PANTHER evaluated fully automatic pancreatic tumor segmentation in two settings: Task 1 on diagnostic T1-weighted contrast-enhanced arterial-phase MRI and Task 2 on T2-weighted MR-Linac MRI. Submissions were assessed on a hidden test set using a multi-metric segmentation protocol combining overlap (Dice), boundary agreement (5 mm Surface Dice), boundary distances (MASD, HD95), and tumor burden / volume error (RMSE), aggregated into a single ranking score.

The winning solutions are publicly available and can be found in the GitHub repositories linked below.


Task 1 - Diagnostic MRI (T1W CE arterial Phase)ΒΆ

Winner Team GitHub GC Algorithm
πŸ₯‡ 1st MIC-DKFZ GitHub repository bauchspeichel-einzel-duse
πŸ₯ˆ 2nd SpaceCY GitHub repository BluNet
πŸ₯‰3rd sjtu_lab426 GitHub repository version2

Task 2 - MR-Linac MRI (T2W)ΒΆ

Winner Team GitHub Algorithm (short)
πŸ₯‡ 1st MIC-DKFZ GitHub repository bauchspeichel-doppel-duse
πŸ₯ˆ 2nd BreizhSeg GitHub repository docker_Task2
πŸ₯ˆ 2nd LiboZhang GitHub repository Task2Baseline
πŸ₯‰3rd Baseline GitHub repository Baseline