Development and validation of a deep learning system to screen vision-threatening conditions in high myopia using optical coherence tomography images

Abstract

Background/aims: To apply deep learning technology to develop an artificial intelligence (AI) system that can identify vision-threatening conditions in high myopia patients based on optical coherence tomography (OCT) macular images.

Methods: In this cross-sectional, prospective study, a total of 5505 qualified OCT macular images obtained from 1048 high myopia patients admitted to Zhongshan Ophthalmic Centre (ZOC) from 2012 to 2017 were selected for the development of the AI system. The independent test dataset included 412 images obtained from 91 high myopia patients recruited at ZOC from January 2019 to May 2019. We adopted the InceptionResnetV2 architecture to train four independent convolutional neural network (CNN) models to identify the following four vision-threatening conditions in high myopia: retinoschisis, macular hole, retinal detachment and pathological myopic choroidal neovascularisation. Focal Loss was used to address class imbalance, and optimal operating thresholds were determined according to the Youden Index.

Results: In the independent test dataset, the areas under the receiver operating characteristic curves were high for all conditions (0.961 to 0.999). Our AI system achieved sensitivities equal to or even better than those of retina specialists as well as high specificities (greater than 90%). Moreover, our AI system provided a transparent and interpretable diagnosis with heatmaps.

Conclusions: We used OCT macular images for the development of CNN models to identify vision-threatening conditions in high myopia patients. Our models achieved reliable sensitivities and high specificities, comparable to those of retina specialists and may be applied for large-scale high myopia screening and patient follow-up.

Keywords: diagnostic tests/investigation; imaging; retina.

Figure 1

Workflow of our AI system. Vertical and horizontal macular OCT images from a high myopia eye were independently subjected to the AI system as the input. After four rounds of categorisation, the positive diagnoses and corresponding heat maps were given as the output. Illustrated by Feng. AI, artificial intelligence; OCT, optical coherence tomography; PMCNV, pathological myopic choroidal neovascularisation.

Figure 2

Comparison of the AI system and ophthalmologists using ROC curves. (A) The performance of the AI system and ophthalmologists for retinoschisis. (B) The performance of the AI system and ophthalmologists for macular hole. (C) The performance of the AI system and ophthalmologists for retinal detachment. (D) The performance of the AI system and ophthalmologists for pathological myopic choroidal neovascularisation (PMCNV). AI, artificial intelligence; AUC, area under the curve; ROC, receiver operating characteristic.

Figure 3

Heatmaps for retinoschisis, macular hole, retinal detachment and pathological myopic choroidal neovascularisation (PMCNV). (A) An example of a retinoschisis lesion detected by our AI system. (B) An example of a macular hole lesion detected by our AI system. (C) An example of a retinal detachment lesion detected by our AI system. (D) An example of a PMCNV lesion detected by our AI system. AI, artificial intelligence; PMCNV, pathological myopic choroidal neovascularisation.