Related Experiment Video
Updated: Sep 15, 2025
05:49
Author Spotlight: Deciphering Electrical Networks Behind Complex Brain Activities and Disorders
Published on: November 1, 2024
969
A deep learning-based image analysis model for automated scoring of horizontal ocular movement disorders
Xiao-Lu Jin1, Yu-Fei Liu1, Bing-Bing He1
1Ocular Motility Disorder Treatment and Rehabilitation Center, Department of Acupuncture, Harbin Medical University, Harbin, Heilongjiang Province, China.
Frontiers in Neurology
|July 17, 2025
Summary
A new deep learning method, RetinaEye, accurately scores horizontal ocular movement disorders. This automated approach shows high consistency with manual scoring, offering potential for improved diagnosis and treatment selection.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Horizontal ocular movement disorders affect patient quality of life.
- Traditional manual scoring methods can be subjective and time-consuming.
Purpose of the Study:
- To develop and evaluate a deep learning-based image analysis method for automated scoring of horizontal ocular movement disorders.
- To compare the performance of the automated method against traditional manual scoring.
Main Methods:
- A deep learning model, RetinaEye, was trained on 2,565 ocular images.
- A test set of 184 binocular gaze images from patients with limited horizontal ocular movement was used.
- ImageJ (manual) and RetinaEye (automated) were used for scoring, and consistency was assessed.
Main Results:
- RetinaEye accurately identified key ocular landmarks like pupils and canthi.
- Automated scoring showed high consistency (κ=0.860) and correlation (ρ=0.897) with manual scoring.
- The model demonstrated particular accuracy in identifying lateral canthi.
Conclusions:
- The RetinaEye automated scoring method is highly consistent with manual scoring.
- This objective method has significant potential for diagnosing and selecting treatments for horizontal ocular movement disorders.
