🧒🧠 BraTS-PEDs

English简体中文

Pediatric high-grade glioma segmentation · multiparametric MRI

Official dataset source Related challenge

📚 Publications and records

Radiology AI 2026 arXiv 2305.17033 TCIA DOI 10.7937/dx5c-tj86

Quick facts
Input T1n, T1c, T2w, T2-FLAIR
Total cohort 457 pediatric patients
Institutional origin This is a multi-consortium pediatric cohort assembled through the Children’s Brain Tumor Network (CBTN), the DIPG Registry (DIPGR), the CONNECT consortium, and additional pediatric institutions. ASNR collaborators contributed to expert annotations. Challenge and cohort article.
Official source TCIA collection · TCIA DOI

🧾 Dataset overview

BraTS-PEDs is a multi-institutional, expert-annotated MRI benchmark for pediatric high-grade gliomas. The dataset paper describes standardized preprocessing, including DICOM-to-NIfTI conversion, registration/resampling, and defacing. Its segmentation labels cover enhancing tumor (ET), nonenhancing tumor (NET), cystic component (CC), and edema (ED).

🗂️ Official split and label availability

Cohort Cases Official label status
Training 257 Expert-validated segmentation masks provided
Validation 91 Images provided without labels for leaderboard evaluation
Test 109 Hidden subset for final assessment

The counts and label status above come from the dataset paper’s cohort description. The official collection portal is the source of truth for current files and any later release changes.

🔒 Access and terms

CBTN–CONNECT–DIPGR–ASNR–MICCAI BraTS-PEDs 2023 introduced a pediatric brain tumor segmentation benchmark using data from multiple pediatric neuro-oncology consortia and institutions.

The cohort was also included in later BraTS challenge releases; see the TCIA collection page for current release notes.

📚 Publication and citation

Fathi Kazerooni et al. “Multi-Institutional Annotated Multiparametric MRI Dataset of Pediatric High-Grade Gliomas.” Radiology: Artificial Intelligence, vol. 8, no. 3, article e250902, 2026. DOI: 10.1148/ryai.250902.

The journal article describes the multi-institutional dataset; the linked challenge overview provides the benchmark context.


🕘 Sources last checked: 27 October 2026 · Back to the dataset catalog