Overview

Schema Version

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

Name

Pediatric Brain MRI for Myelin Maturation Age Estimation (Basel cohort)

Link

https://zenodo.org/record/8055666

Indexing

Keywords: pediatric brain MRI, myelin maturation, myelination age, T1-weighted, T2-weighted, deep learning, CNN, ensemble model, Basel cohort, neonates, infants, young children

Content: MR, NR, PD

RadLex: RID10796

Author(s)

Tugba Akinci D’Antonoli

Ramona-Alexandra Todea

Nora Leu

Alexandre N. Datta

Bram Stieltjes

Friederike Pruefer

Jakob Wasserthal

Organization(s)

University Children’s Hospital Basel, Department of Pediatric Radiology

University Children’s Hospital Basel, Department of Pediatric Neurology and Developmental Medicine

Institute of Radiology and Nuclear Medicine, Cantonal Hospital Basel

University Hospital Basel, Department of Neuroradiology, Clinic of Radiology and Nuclear Medicine

University Hospital Basel, Department of Research and Analysis, Clinic of Radiology and Nuclear Medicine

Funding

Authors declared no funding for this work.

Ethical review

Approved by the Ethics Committee Northwest and Central Switzerland (EKNZ BASEC ID 2021–01682) with waiver of informed consent.

Comments

Retrospective single-center pediatric brain MRI dataset (0–3 years) labeled with radiologist-determined myelin maturation age; released with the associated publication.

Date

Published: 2023-07-26

References

[1] Akinci D’Antonoli T, Todea RA, Leu N, Datta AN, Stieltjes B, Pruefer F, Wasserthal J. "Development and Evaluation of Deep Learning Models for Automated Estimation of Myelin Maturation Using Pediatric Brain MRI Scans". Radiology: Artificial Intelligence. 2023-09-01. doi:10.1148/ryai.220292. PMID: 37795138. PMCID: PMC10546368.

Dataset

Motivation

Provide a large, diverse open-access dataset to develop and evaluate models that estimate myelin maturation age from pediatric brain MRI.

Sampling

Retrospective retrieval from PACS using PACS/RISCrawler based on brain MRI, age 0–3 years, exams from Jan 1, 2011 to Mar 17, 2021; excluded external transfers, age >36 months, and severe motion or near-global pathology.

Partitioning scheme

Random split with subsequent minor manual adjustments to balance pathology: 85% (710) for training/cross-validation, 15% (123) for internal testing; fivefold cross-validation used for hyperparameter optimization.

Missing information

Sex, race/ethnicity, and exact image resolutions not reported.

Relationships between instances

Follow-up examinations for the same patient were excluded to maintain one examination per patient.

Noise

Includes cases with minor pathologic features and motion artifacts to reflect real-world variability; acquisition protocols varied over a decade (same scanner).

External data

Two public datasets (NIH pediatric brain MRI database and Developing Human Connectome Project) were used only for external testing, not included in this released dataset.

Confidentiality

Images were de-identified and skull stripped to remove nonbrain tissue.

Re-identification

Risk mitigated by de-identification and skull stripping (SynthStrip) to avoid facial features and PHI.

Sensitive data

Pediatric neuroimaging data (infants and young children).