Related Experiment Video
Updated: Sep 16, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.9K
Deep learning for predicting myopia severity classification method
WangMeiYu Xing1,2, XiaoNa Li1,2, JingShu Ni2
1College of Biomedical Engineering, Anhui Medical University, Hefei, 230011, China.
Biomedical Engineering Online
|July 9, 2025
Summary
A novel deep learning model, X-ENet, accurately classifies myopia severity using enhanced fundus images. This efficient method improves vision impairment screening by combining depthwise separable and dynamic convolutions for precise feature extraction.
Area of Science:
- Ophthalmology
- Computer Science
- Artificial Intelligence
Background:
- Myopia is a leading cause of vision impairment globally.
- Current myopia screening methods require improvement in efficiency and accuracy.
Purpose of the Study:
- To develop and evaluate X-ENet, a deep learning model for classifying myopia severity.
- To enhance the efficiency and accuracy of myopia screening through advanced image analysis.
Main Methods:
- Fundus images were preprocessed and enhanced for improved feature extraction.
- X-Net, utilizing depthwise separable and dynamic convolutions, was trained for myopia severity classification.
- Grad-CAM was used for model interpretability, and a GUI was developed for usability.
Main Results:
- The proposed X-ENet model achieved high performance metrics: 0.9104 accuracy, 0.8154 precision, 0.8177 recall, 0.8147 F1-score, and 0.9376 specificity.
- The model demonstrated effective feature extraction from fundus images for myopia classification.
Conclusions:
- X-ENet significantly outperforms conventional deep learning models in myopia severity classification.
- The model shows strong effectiveness and reliability for practical application in vision impairment screening.
