👁️ A-eye: Automated 3D MRI Segmentation and Morphometric Feature Extraction for Eye and Orbit Atlas Construction
Figure 1. Overview of the A-eye results for automated MRI eye segmentation and morphometry.
Figure 2. Example of axial length automatic extraction in 3D.
This repository accompanies the paper:
Barranco J., Luyken A., Jia Y., Kebiri H., Stachs P., Gordaliza P. M., Esteban O., Aleman Y., Sznitman R., Stachs O., Langner S., Franceschiello B.†, Bach Cuadra M.†
A-eye: Automated 3D MRI Segmentation and Morphometric Feature Extraction for Eye and Orbit Atlas Construction
† Equal last authorship 📧 Corresponding authors:
[email protected], [email protected], [email protected]
In this study, we introduce an automated 3D segmentation of the healthy human adult eye and orbit from Magnetic Resonance Images (MRI), aimed at advancing ophthalmic diagnostics and treatments.
Previous works have typically relied on small sample sizes and heterogeneous imaging modalities. Here, we leverage a large-scale dataset of T1-weighted MRI scans from 1,245 subjects and employ the deep learning-based nnU-Net for MR-Eye segmentation tasks.
Our results demonstrate robust and accurate 3D segmentations of the lens, globe, optic nerve, rectus muscles, and orbital fat. We further provide automated morphometric biomarkers such as axial length and volumetric measures, as well as benchmarking analyses correlating body mass index with eye structure volumes.
Quality control protocols ensure the reliability of large-scale segmentation, enhancing the applicability of our pipeline in clinical research.
A major outcome of this work is the first large-scale, unbiased eye atlases (female, male, and combined) to promote standardization of spatial normalization tools for MR-Eye.
MRI · Eye · MR-Eye · 3D segmentation · Large-scale dataset · Imaging biomarkers · Morphometry · Ophthalmology · Atlas · Benchmarking · Axial length
- CIBM Center for Biomedical Imaging, Lausanne, Switzerland
- Department of Radiology, Lausanne University Hospital (CHUV) and University of Lausanne (UNIL), Lausanne, Switzerland
- HES-SO University of Applied Sciences and Arts Western Switzerland
- The Sense Innovation and Research Center, Lausanne and Sion, Switzerland
- Department of Ophthalmology, Rostock University Medical Center, Rostock, Germany
- Karlsruhe Institute of Technology (KIT)
- ARTORG Center for Biomedical Engineering, University of Bern, Bern, Switzerland
- Department Life, Light & Matter, University of Rostock, Rostock, Germany
- Institute for Diagnostic and Interventional Radiology, Pediatric and Neuroradiology, Rostock University Medical Center, Rostock, Germany
† Equal last authorship
🚀 Coming soon!
An interactive web interface will allow exploration of the MR-Eye atlases, morphometric statistics, and automated segmentation examples.
If you use the A-eye pipeline, atlas, or derived datasets in your work, please cite the relevant resources below:
Barranco J, Luyken A, Stachs P, Esteban O, Aleman-Gomez Y, Stachs O, et al.
MR-Eye atlas: a large-scale atlas of the eye based on T1-weighted MR imaging [dataset].
Zenodo, 2024.
https://doi.org/10.5281/zenodo.13325371
Figure 3. Visualization of the MR-Eye population atlases (female, male, and combined) and in common VCS.
Barranco Hernandez J, Luyken A, Stachs O, Langner S, Franceschiello B, Bach Cuadra M.
A-eye: automated 3D segmentation of healthy human eye and orbit structures and axial length extraction.
2025.
https://doi.org/10.26039/TA7F-X088
Coming soon!
A-eye/models/— Trained nnU-Net weights and configs (when released)notebooks/— Example scripts for segmentation and morphometrydata/— Sample data or download scripts (if applicable)docs/— Documentation and supplementary materialLICENSE— License file
For questions or collaborations, please contact:
- Jaime Barranco — [email protected], [email protected], [email protected]
- Benedetta Franceschiello — [email protected]
- Meritxell Bach Cuadra — [email protected]
This repository and associated software are distributed under the
“Software License Agreement for Academic Non-Commercial Research Purposes Only”
between HES-SO Valais-Wallis and CHUV.
By downloading or using this software, you agree to the terms described in the LICENSE file.
Usage is strictly limited to academic and non-commercial research purposes.
For commercial or redistribution inquiries, please contact the licensors directly.
© 2025 CIBM Center for Biomedical Imaging SP CHUV-UNIL, Lausanne, Switzerland and HES-SO University of Applied Sciences and Arts Western Switzerland, Sion, Switzerland.