Overview

Schema Version

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

Name

NeoLUNet dataset

Link

https://github.com/SchubertLab/NeoLUNet

Indexing

Keywords: neonatal, preterm, lung MRI, bronchopulmonary dysplasia, quiet breathing, U-Net, segmentation, 3D morphologic features, BPD severity, machine learning

Content: CH, MR, PD, RS

RadLex: RID1347, RID28915, RID34494, RID11434, RID10312, RID10782

SNOMED: 395507008, 67569000

Author(s)

Benedikt Mairhörmann

Alejandra Castelblanco

Friederike Häfner

Vanessa Koliogiannis

Lena Haist

Dominik Winter

Andreas Flemmer

Harald Ehrhardt

Sophia Stöcklein

Olaf Dietrich

Kai Förster

Anne Hilgendorff

Benjamin Schubert

Organization(s)

Computational Health Center, Helmholtz Zentrum München

Institute for Lung Health and Immunity and Comprehensive Pneumology Center, Helmholtz Zentrum München

Institute of AI for Health, Helmholtz Zentrum München

Perinatal Center, Hospital of the Ludwig-Maximilian University, Munich

Department of Radiology, Hospital of the Ludwig-Maximilian University, Munich

Center for Comprehensive Developmental Care (CDeCLMU), Dr. von Hauner Children’s Hospital, LMU Munich

Department of General Pediatrics & Neonatology, Justus-Liebig-University, Giessen

University Medical Center Ulm

Department of Mathematics, Technical University of Munich

License

Text: CC BY 4.0

URL: https://creativecommons.org/licenses/by/4.0/

Funding

Helmholtz Foundation/Helmholtz Zentrum München (Young Investigator Grant NWG VH-NG-829), German Center for Lung Research (DZL; BMBF), DFG Research Training Group GRK2338, Stiftung AtemWeg (LSS AIRR), Helmholtz Zentrum München Postdoctoral Fellowship, BMBF-funded de.NBI Cloud.

Ethical review

Informed parental consent obtained; ethics approvals: cohort 1 EC#195-07 and cohort 2 EC#135-12.

Comments

Prospective two-site neonatal study of quiet-breathing lung MRI in preterm infants with and without BPD; includes multirater manual lung annotations, deep learning segmentations, and derived 3D morphologic MRI features.

Date

Published: 2023-10-25

References

[1] Mairhörmann B, Castelblanco A, Häfner F, et al.. "Automated MRI Lung Segmentation and 3D Morphologic Features for Quantification of Neonatal Lung Disease". Radiology: Artificial Intelligence. 2023-11-01. doi:10.1148/ryai.220239. PMID: 38074782. PMCID: PMC10698600.

Dataset

Motivation

Enable robust, standardized, radiation-free assessment of neonatal lung disease using MRI through automated segmentation and 3D morphologic feature extraction.

Sampling

Prospective enrollment of 107 preterm infants <32 weeks GA at birth across two sites; MRI at near-term age during quiet or lightly sedated breathing.

Partitioning scheme

Cohort 1 used for leave-one-patient-out cross-validation training; Cohort 2 used as independent external test set. Subcohorts used for lung injury scoring (n=58) and infant lung function testing (n=33).

Missing information

Exact imaging file formats and public availability of raw images not specified in the article; only code/data generated by the authors indicated at repository link.

Relationships between instances

Each participant underwent near-term age quiet-breathing lung MRI; multirater manual annotations per image were used to train multiple models and an ensemble.

Noise

Neonatal MRI subject to motion, blurring, ghosting; cohort 2 exhibited lower image quality affecting segmentation performance.

External data

No external imaging datasets reported; analyses confined to two prospectively enrolled clinical cohorts.

Confidentiality

Pseudonymization of images and clinical information was performed for all participants.

Re-identification

Images and clinical data were pseudonymized; no identifiers reported.