Related Experiment Video
Updated: Jun 13, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.7K
Development of Deep Learning Models to Screen Posterior Staphylomas in Highly Myopic Eyes Using UWF-OCT Images.
Yining Wang1, Changyu Chen1, Ziye Wang1
1Department of Ophthalmology and Visual Science, Institute of Science Tokyo, Tokyo, Japan.
Translational Vision Science & Technology
|June 12, 2025
Summary
Deep learning models accurately detect posterior staphyloma edges in highly myopic patients using ultra-widefield OCT images. This artificial intelligence system aids ophthalmologists in screening for staphylomas, improving clinical management.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Posterior staphylomas are a significant complication in highly myopic eyes.
- Early detection and management are crucial for preventing further vision loss.
Purpose of the Study:
- To develop and evaluate deep learning (DL) models for screening posterior staphylomas.
- Utilize ultra-widefield optical coherence tomography (UWF-OCT) images for staphyloma edge detection.
Main Methods:
- Retrospective analysis of 1428 UWF-OCT images from 438 highly myopic patients.
- Trained seven DL architectures (VGG16, VGG19, ResNet, DenseNet) to identify staphyloma edges.
- Evaluated model performance using the area under the receiver operating characteristic curve (AUC).
Main Results:
- DL models achieved AUCs ranging from 0.794 to 0.903 for staphyloma edge detection.
- The VGG19 model demonstrated high sensitivity (0.871) comparable to retina specialists.
- Heatmaps confirmed precise localization of staphyloma edges by the DL models.
Conclusions:
- Developed DL models reliably identify posterior staphyloma edges with high accuracy.
- These AI tools show promise for enhancing the clinical management of highly myopic patients.
- The system assists ophthalmologists in screening for posterior staphylomas in high myopia.
