Overview

Schema Version

https://atlas.rsna.org/schemas/2025-11/dataset.json

Name

Pediatric medulloblastoma multiparametric MRI with expert subcompartment annotations

Link

https://doi.org/10.1148/ryai.230115

Indexing

Keywords: Pediatrics, MRI, Medulloblastoma, Segmentation, nnU-Net, Transfer learning, FLAIR, T2-weighted, Gadolinium-enhanced T1-weighted

Content: NR, MR, PD, OI

RadLex: RID10795, RID35806, RID1568, RID10312, RID10783, RID4408, RID49531

SNOMED: 443333004, 1156923005

Author(s)

Rohan Bareja

Marwa Ismail

Douglas Martin

Ameya Nayate

Ipsa Yadav

Murad Labbad

Prateek Dullur

Sanya Garg

Benita Tamrazi

Ralph Salloum

Ashley Margol

Alexander Judkins

Sukanya Iyer

Peter de Blank

Pallavi Tiwari

Organization(s)

University of Wisconsin–Madison

University Hospitals, Cleveland, Ohio

Case Western Reserve University

Children’s Hospital Los Angeles

Nationwide Children’s Hospital

Keck School of Medicine of USC / Children’s Hospital Los Angeles

Cincinnati Children’s Hospital Medical Center

Children’s Hospital of Philadelphia (CBTN)

License

Contact

Corresponding author: Pallavi Tiwari (email provided in article)

Funding

Supported by multiple sources listed in Acknowledgments, including NCI grants (1U01CA248226, 1R01CA264017, 3U01CA248226-S1, 1R01CA277728, P30CA014520), DoD PRCRP (W81XWH-18-1-0404), VA (1 I01 BX005842), and others.

Ethical review

HIPAA-compliant; IRB approval at each site (identification no. 2022–1683). Waiver of informed consent due to retrospective design.

Comments

Retrospective, multi-institutional pediatric medulloblastoma MRI dataset with expert consensus segmentations of tumor habitat and subcompartments (ET, ED, NET+CC). Data from hospitals A and B are institutionally protected; hospital C data accessed via CBTN. Segmentations from CBTN studies will be released back to CBTN.

Date

Published: 2024-08-21

Created: 2000-01-01

References

[1] Bareja R, Ismail M, Martin D, et al.. "nnU-Net–based Segmentation of Tumor Subcompartments in Pediatric Medulloblastoma Using Multiparametric MRI: A Multi-institutional Study". Radiology: Artificial Intelligence. 2024;6(5):e230115. 2024-08-21. doi:10.1148/ryai.230115. PMID: 39166971. PMCID: PMC11427926.

Dataset

Motivation

Enable accurate, automated delineation of medulloblastoma subcompartments to support radiation therapy planning and quantitative analyses.

Sampling

Inclusion: axial Gd-T1w, T2w, FLAIR available; diagnosis of medulloblastoma; acceptable diagnostic quality per radiologists. Quality control with MRQy; poor-quality cases excluded.

Partitioning scheme

Cross-site evaluation: train on two sites, test on the third (A+B→C; A+C→B; B+C→A). Fivefold cross-validation used within training for model selection.

Missing information

T1 precontrast available for ~20% of cases only; some tumor subcompartments absent in a subset of patients (ED rarely present; CC scarce, combined with NET).

Relationships between instances

Each patient study includes aligned multiparametric MRI sequences (Gd-T1w, T2w, FLAIR) and corresponding expert consensus labels for ET, ED, NET+CC; tumor habitat label is union of subcompartments.

Noise

Multi-institutional variability (scanner vendor/field strength); retrospective quality variability. Poor-quality cases identified and excluded using MRQy.

External data

Adult glioma BraTS dataset used for pretraining in transfer learning experiments; not part of pediatric dataset distribution.

Confidentiality

Hospitals A and B data are protected via institutional compliance; sharing requires IRB approvals and data use agreements. Hospital C data accessed via CBTN membership and policies.

Sensitive data

Pediatric clinical MRI with potential PHI; controlled access via institutional/CBTN procedures.