Overview

Schema Version

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

Name

Multicenter MRI dataset for automated 3D vestibular schwannoma segmentation (Leiden University Medical Center)

Link

https://pubmed.ncbi.nlm.nih.gov/35923375/

Indexing

Keywords: vestibular schwannoma, MRI, contrast-enhanced T1-weighted, T2-weighted, segmentation, nnU-Net, multicenter, multivendor

Content: HN, MR

RadLex: RID35976, RID10312, RID4473, RID49531, RID10796

SNOMED: 126949007, 15188001

Author(s)

Olaf M. Neve

Yunjie Chen

Qian Tao

Stephan R. Romeijn

Nick P. de Boer

Willem Grootjans

Mark C. Kruit

Boudewijn P. F. Lelieveldt

Jeroen C. Jansen

Erik F. Hensen

Berit M. Verbist

Marius Staring

Organization(s)

Leiden University Medical Center

Delft University of Technology, Knowledge Driven AI Lab

Contact

Corresponding author: Olaf M. Neve (see article)

Funding

Supported by a strategic fund of the Leiden University Medical Center. Y.C. supported by the China Scholarship Council (grant 202008130140).

Ethical review

Institutional review board approved (Leiden University Medical Center; protocol G19.115); informed consent requirement waived.

Comments

Retrospective, multicenter, multivendor MRI dataset used to develop and test nnU-Net models for vestibular schwannoma segmentation on contrast-enhanced T1-weighted and T2-weighted MRI.

Date

Published: 2022-06-22

Created: 2020-01-01

References

[1] Neve OM, Chen Y, Tao Q, et al.. "Fully Automated 3D Vestibular Schwannoma Segmentation with and without Gadolinium-based Contrast Material: A Multicenter, Multivendor Study". Radiology: Artificial Intelligence. 2022-07-01. doi:10.1148/ryai.210300. PMID: 35923375. PMCID: PMC9344213.

[2] Shapey J, Kujawa A, Dorent R, et al.. "Segmentation of Vestibular Schwannoma from Magnetic Resonance Imaging: An Open Annotated Dataset and Baseline Algorithm [data set]". The Cancer Imaging Archive. 2021-01-01. PMID: 34711849. PMCID: PMC8553833. Available from: https://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=70229053

Dataset

Motivation

Develop and evaluate automated 3D segmentation of vestibular schwannoma (whole, intrameatal, extrameatal) on contrast-enhanced T1- and T2-weighted MRI for robust clinical use across centers and vendors.

Sampling

214 consecutive patients imaged for hearing loss (134 VS-positive adults with unilateral tumors and at least one contrast-enhanced T1 MRI; 80 VS-negative hearing loss controls). Excluded: post-surgery or post-irradiation scans.

Partitioning scheme

Random split at patient level: 80% training/validation from 26 centers with fivefold cross-validation and model ensembling; 20% independent test from 11 unseen centers. T1 model additionally evaluated on an external public dataset.

Missing information

Exact per-partition counts (patients/exams) by class; complete demographics for negative cases; image file formats; detailed acquisition parameters per site.

Relationships between instances

For positive cases, paired T1-weighted and high-resolution T2-weighted exams were available in a subset (112 patients). Intra- and extrameatal components and whole tumor masks are linked per case.

Noise

Multicenter variability in scanners and protocols; challenging cases include tumors with large peripheral cystic components.

External data

Publicly available dataset by Shapey et al. (TCIA) used as additional external test for the T1-weighted model (n=242; 47 postsurgery scans omitted).

Confidentiality

Retrospective clinical imaging data; negative cases were previously anonymized and provided without demographic data.

Re-identification

Images were anonymized for analysis; negative cases lacked demographics due to prior anonymization.

Sensitive data

Contains patient health information (medical imaging).