# PANTHER Challenge: Public Training Dataset

## Reference

Please cite the following article, if you are using the PANTHER Challenge: Public Training Dataset:

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

## License

[CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/)

## Dataset Characteristics

The PANTHER Challenge dataset has been created to support the development of automated pancreatic tumor segmentation models and is divided into two distinct tasks. Task 1 includes diagnostic MR images acquired for tumor segmentation in the diagnostic setting. In these images, filenames that contain _0001 indicate the arterial phase scans, which are the only ones with available ground truth labels; images with different suffix numbers correspond to other sequences. Task 2 is based on MR-Linac images acquired for radiotherapy treatment planning.

## Folder Structure

The dataset is provided as two separate zip files:

* PANTHER_Task1.zip

```

ImagesTr

├── 10000_00001_0000.mha (T1W CE arterial phase MRIs)

├── ...

LabelsTr

├── 10000_00001.mha (segmentation labels for arterial phase MRIs)

├── ...

ImagesTr_unlabeled

├── 10000_00002_0000.mha (additional training images from other sequences without labels)

├── ...

```

* PANTHER_Task2.zip

```

ImagesTr

|── 10303_0000.mha (T2W MR-Linac MRI)

|── ...

LabelsTr

|── 10303.mha (segmentation labels for T2W MRIs)

|── ...

```

## Managed By

Diagnostic Image Analysis Group,

Radboud University Medical Center,

Nijmegen, The Netherlands

## Contact Information

- Amparo Betancourt : amparosoeli.betancourttarifa@radboudumc.nl