Predicting Optical Coherence Tomography-Derived High Myopia Grades From Fundus Photographs Using Deep Learning

Abstract

Purpose: To develop an artificial intelligence (AI) system that can predict optical coherence tomography (OCT)-derived high myopia grades based on fundus photographs.

Methods: In this retrospective study, 1,853 qualified fundus photographs obtained from the Zhongshan Ophthalmic Center (ZOC) were selected to develop an AI system. Three retinal specialists assessed corresponding OCT images to label the fundus photographs. We developed a novel deep learning model to detect and predict myopic maculopathy according to the atrophy (A), traction (T), and neovascularisation (N) classification and grading system. Furthermore, we compared the performance of our model with that of ophthalmologists.

Results: When evaluated on the test set, the deep learning model showed an area under the receiver operating characteristic curve (AUC) of 0.969 for category A, 0.895 for category T, and 0.936 for category N. The average accuracy of each category was 92.38% (A), 85.34% (T), and 94.21% (N). Moreover, the performance of our AI system was superior to that of attending ophthalmologists and comparable to that of retinal specialists.

Conclusion: Our AI system achieved performance comparable to that of retinal specialists in predicting vision-threatening conditions in high myopia via simple fundus photographs instead of fundus and OCT images. The application of this system can save the cost of patients' follow-up, and is more suitable for applications in less developed areas that only have fundus photography.

Keywords: artificial intelligence; deep learning; fundus photographs; high myopia; optical coherence tomography.

Figure 1

Schematic representation of our approach for developing a high myopia predictive model. Retinal specialists labeled each fundus photograph according to the corresponding OCT images. The fundus photographs with ground true labels were used for model training. For clinical application, the model receives as input fundus photographs, then outputs the predicted ATN classification. OCT, optical coherence tomography; DL, deep learning; A, atrophy; T, traction; N, neovascularisation.

Figure 2

Representative fundus images and related OCT images. (A₁,A₂) Tessellated fundus; (B₁,B₂) Macular atrophy; (C₁,C₂) Retinal detachment; (D₁,D₂) PMCNV. OCT, optical coherence tomography; PMCNV, pathological myopic choroidal neovascularisation.

Figure 3

Framework of the deep learning methods.

Figure 4

Flow chart showing the AI system development and evaluation based on fundus photographs. DL, deep learning; OCT, optical coherence tomography.

Figure 5

Performance of AI system in A, T, and N categories. (A) The comparison of different methods and clinicians using ROC curves. AUC, area under the curve; ROC, receiver operating characteristic. (B) The t-SNE visualization of different methods. Red and green points represent negative and positive results, respectively.

Figure 6

Performance of AI system in category A with three sub-grades. (A) The ROC curves of different methods for category A with three sub-grades. AUC, area under the curve; ROC, receiver operating characteristic. (B) The t-SNE visualization of different methods for category A with three sub-grades. Red, blue, and green points represent the normal fundus (no myopic retinal lesions), tessellated fundus, and retinal atrophy, respectively.