PANTHER Challenge Banner
Join

Data Splits 🗂️

Data is divided into 3 splits , for each task, with the following use-cases:

  • Public Training Dataset : Available for all paticipants and researchers to train and develop AI segmentation models. All data is fully anonymized and made available under a non-commercial CC BY-NC-SA 4.0 license.

  • Development Dataset: Allows participants to familiarize themselves with the Grand Challenge platform and submission process. Participants can evaluate their models on a small set of unseen cases, with results displayed on a live leaderboard. Please note: The Development Dataset is not publicly released.

  • Hidden Testing Cohort: Used for the final evaluation to identify the top performing AI algorithms at the conclusion of the Open Development Phase. This dataset remains unseen during development, ensuring a fair and unbiased assessment.

Dataset Overview 🔎

☑️TASK 1: Pancreatic Tumor Segmentation on Diagnostic MRIs

  • Public Training Dataset :
    • 92 annotated cases (T1-weighted contrast-enhanced arterial phase)
    • 367 unannotated cases (various phases, with/without contrast, DWI).
  • Development Dataset: 5 cases
  • Hidden Testing Cohort: 30 cases

☑️TASK 2: Pancreatic Tumor Segmentation on MR-Linac MRIs

  • Public Training Dataset: 50 cases (T2 weighted MR-Linac image acquired during a radiotherapy session)
  • Development Dataset: 4 cases
  • Hidden Testing Cohort: 30 cases

Annotations ☑️

For each task, both pancreas and tumor annotations will be available in the public training dataset, allowing participants to familiarize themselves with organ anatomy and tumor locations.

❗During the Open Development and Testing Phases, only the MR images will be provided as input, and evaluation will focus solely on the tumor segmentation output. While you may use the provided labels at your discretion, the final expected output for both tasks is a binary tumor segmentation mask (0 = background, 1 = tumor).

▶️ Imaging data and annotations have been released via: https://zenodo.org/records/15192302

Imaging Data 🏥

The PANTHER challenge dataset consists of a total of 489 diagnostic pancreas MRIs from Radboud University Medical Center (RUMC) in The Netherlands and 84 MR-linac images from Odense University Hospital (OUH) in Denmark.

  • RUMC Cases:
    All cases come from patients with pathologically confirmed Pancreatic Ductal Adenocarcinoma (PDAC) who underwent diagnostic MRI as part of their clinical care. All RUMC cases were acquired using Siemens scanners (Skyra, Prisma Fit and Avanto).

  • OUH Cases:
    These cases originate from patients with confirmed primary pancreatic tumors who received radiotherapy using MR-Linac systems at various stages of treatment. All OUH cases were acquired using the Elekta Unity MR-Linac system (Philips Marlin scanner).