PANTHER Challenge: Public Training Dataset

This dataset represents the PANTHER Challenge Training Dataset, comprising 509 anonymized MR images. The dataset is divided into two distinct tasks.

For Task 1, which focuses on diagnostic MR images, the dataset includes 459 scans acquired at Radboud University Medical Center. Within these, 92 images are T1-weighted contrast-enhanced arterial-phase scans with annotations, supplemented by 367 scans acquired using multiple MRI sequences. These images are intended to support the development of methods for pancreatic tumor segmentation in a diagnostic setting.

For Task 2, which focuses on treatment planning using MR-Linac images, the dataset contains 50 MR-Linac scans acquired at Odense University Hospital. These images are intended to support the development and evaluation of AI models capable of segmenting pancreatic tumors in the context of radiotherapy planning.

Access to the dataset files is restricted. Researchers may request access through Zenodo for non-commercial research use. Requests should include the requester’s institutional affiliation, intended use of the dataset, and confirmation that the data will be used in accordance with the dataset terms, including no attempt to re-identify individuals and no redistribution of the data.

Citation

If you use the PANTHER dataset in your research, please cite both the challenge report and this Zenodo dataset record:

Betancourt Tarifa, A. S., et al. (2026). PANTHER Challenge Report: Cross-Domain Pancreatic Tumor Segmentation in Magnetic Resonance Imaging. Medical Image Analysis, 113, 104186. https://doi.org/10.1016/j.media.2026.104186

Betancourt Tarifa, A. S., Mahmood, F., Bernchou, U., & Koopmans, P. J. (2025). PANTHER Challenge: Public Training Dataset (Version 1.0) [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.15192302