chehablaborg/BraTS_PED_2024

This dataset is derived from the BraTS PED 2024 (Brain Tumor Segmentation Challenge 2024 – Pediatric track). The original dataset consists of multi-institutional, multi-parametric MRI scans of pediatric brain tumor patients, curated for the task of automated tumor segmentation. Each subject includes the following MRI modalities:

• T1-weighted (T1)
• T1-weighted post-contrast (T1c)
• T2-weighted (T2w)
• T2-FLAIR (T2f)

Additionally, expert-annotated segmentation masks are provided with voxel-wise tumor labels:

• Background (0)
• Tumor core / non-enhancing / cystic (1)
• Edema (2)
• Enhancing tumor (3)

This repository provides a 2D slice-based version of the BraTS PED 2024 dataset, designed for efficient training of deep learning models.

📦 Dataset Structure

Each entry in the dataset corresponds to a single 2D slice and contains:

• volume_id → Unique identifier for the patient/volume
• slice_id → Index of the slice within the volume
• t1c → T1-weighted contrast-enhanced slice
• t1n → T1-weighted slice
• t2f → T2-FLAIR slice
• t2w → T2-weighted slice
• seg → Segmentation mask slice

Use

from datasets import load_dataset
bratsped24 = load_dataset("chehablaborg/BraTS_PED_2024", split="train")
sample_id = 314
t1c = bratsped24[sample_id]["t1c"]
t1n = bratsped24[sample_id]["t1n"]
t2f = bratsped24[sample_id]["t2f"]
t2w = bratsped24[sample_id]["t2w"]
seg = bratsped24[sample_id]["seg"]
plt.imshow(t1c, cmap="grey")
plt.show()

📚 Citation (Original Dataset)

If you use this dataset, please mention us in an acknowledgement (chehablab.com) and cite the original BraTS PED paper:

@misc{BraTS_PEDs,
  title = {The Brain Tumour Segmentation in Pediatric Magnetic Resonance Imaging (BraTS-PEDs) Dataset and Challenge},
  author = {
    Fathi Kazerooni, Anahita and Khalili, Nastaran and Liu, Xinyang and Jiang, Zhifan and Gandhi, Deep and Pavaine, Julija and Shah, Lalit M. and Jones, Blaise V. and Sheth, Nakul and Prabhu, Sanjay P. and McAllister, Aaron S. and Tu, Wenxin and Nandolia, Khanak K. and Rodriguez, Andres and Shaikh, Ibraheem Salman and Sanchez-Montano, Mariana and Lai, Hollie Anne and Haldar, Debaditya and Anderson, Hannah and Bagheri, Saba and Zapaishchykova, Anna and Familiar, Ariana M. and Gottipati, Anurag and Maleki, Nazanin and Poussaint, Tina and Storm, Paul B. and Bornhorst, Miriam and Packer, Roger and Hummel, Trent and de Blank, Peter and Hoffman, Lindsey and Aboian, Mariam and Nabavizadeh, Ali and Ware, Jeffrey B. and Kann, Benjamin H. and Bakas, Spyridon and Rood, Brian and Resnick, Adam and Vossough, Arastoo and Linguraru, Marius George
  },
  year = {2025},
  howpublished = {\url{https://doi.org/10.7937/DX5C-TJ86}},
  doi = {10.7937/DX5C-TJ86},
  eprint = {2404.15009},
  archivePrefix = {arXiv},
  primaryClass = {cs.CV},
  url = {https://arxiv.org/abs/2404.15009}
}

License

This work is licensed under a Creative Commons CC BY-NC 4.0 License.

Chehab Lab @ 2026