Automated Horizontal Strabismus Detection and Classification Using Deep Learning-based Analysis of Facial Images
Motahhareh Sadeghi1 *, Mahsa Yarkheir2 , Hamed Azarnoush2 , Elias Khalilipoor3 , Mohammad Reza Akbari4
- Assistant Professor of Pediatric Ophthalmology & Strabismus Farabi Eye Hospital, Tehran University of Medical
Science, Tehran, Iran
- Biomedical Engineering Department, Amirkabir University of Technology (Tehran Polytechnic), Tehran, Iran
- Assistant professor of ophthalmology & Vitreo retinal surgeon, Farabi Eye hospital, Tehran University of Medical Sciences, Tehran, Iran
- Professor of Pediatric Ophthalmology & Strabismus Farabi Eye Hospital, Tehran University of Medical Science, Tehran, Iran
Abstract: Strabismus, or eye misalignment, is a common condition affecting individuals of all ages. Early detection and accurate
classification are essential for proper treatment and avoiding long-term complications. This research presents
a new deep-learning-based way of automatically identifying and classifying strabismus from facial images
Methods: Our method
utilizes convolutional neural networks (CNNs) to achieve high accuracy in both binary classification (strabismus vs. normal)
and multi-class classification (eight-class deviation angle of two types: esotropia and exotropia). The dataset (for binary
classification) consisted of 4257 facial images, including (1599 normal and, 2658 strabismus cases) and 480 strabismic and
142 non-strabismic (for multi-class classification) labeled based on ophthalmologist measurements using the Alternate prism
cover test (APCT) or Modified Krimsky test (MK). Five-fold cross-validation was employed, and performance was evaluated
using sensitivity, accuracy, F1-score, and recall metrics
Results: The proposed deep-learning model achieved an accuracy of 86.38%
for binary classification and 92.7% for multi-class classification
Conclusion: . These results demonstrate the potential of our approach to
assist healthcare professionals in early strabismus detection and treatment planning, ultimately improving patient outcomes
