Deep Learning-based Classification of Eye Laterality in Optical Coherence Tomography Images
10.21561/jor.2024.9.2.177
- Author:
Richul OH
1
;
Eun Kyoung LEE
;
Kunho BAE
;
Un Chul PARK
;
Kyu Hyung PARK
;
Chang Ki YOON
Author Information
1. Department of Ophthalmology, Seoul National University Hospital, Seoul, Korea
- Publication Type:ORIGINAL ARTICLE
- From:
Journal of Retina
2024;9(2):177-183
- CountryRepublic of Korea
- Language:English
-
Abstract:
Purpose:To develop a deep learning model classifying the laterality of optical coherence tomography (OCT) images.
Methods:The study included two-dimensional OCT images (horizontal/vertical macular section) from Seoul National University Hospital. A deep learning model based on ResNet-18 was developed and trained to classify whether OCT images were horizontal or vertical sections and to predict the laterality of the images. Analysis of the results included calculating a mean area under the receiver operating characteristic curve (AUROC) and evaluating accuracy, specificity, and sensitivity. Gradient-weighted class activation for mapping visualization highlighted critical regions for classification.
Results:A total of 5,000 eyes of 2,500 patients (10,000 images) was included in the development process. The test dataset consisted of 1,000 eyes of 500 patients (590 eyes without macular abnormalities, 208 epiretinal membranes, 111 age-related macular degenerations, 56 central macular edemas, 23 macular holes, and 12 other macular abnormalities). The deep learning model predicted the OCT section of the eyes in the test dataset with a mean AUROC of 0.9967. The accuracy, sensitivity, and specificity were 0.9835, 0.9870, and 0.9800, respectively. The model predicted the laterality of the eyes in horizontal OCT images with a mean AUROC of 1.0000. The accuracy, sensitivity, and specificity were 0.9970, 1.0000, and 0.9940, respectively. Using vertical OCT images, deep learning models failed to demonstrate any predictive performance in laterality classification.
Conclusions:We developed a deep learning model to classify the horizontal/vertical sections of OCT images and predict the laterality of horizontal OCT images with high accuracy, sensitivity, and specificity.
