Automated diagnosing primary open-angle glaucoma from fundus image by simulating human's grading with deep learning

Abstract

Primary open-angle glaucoma (POAG) is a leading cause of irreversible blindness worldwide. Although deep learning methods have been proposed to diagnose POAG, it remains challenging to develop a robust and explainable algorithm to automatically facilitate the downstream diagnostic tasks. In this study, we present an automated classification algorithm, GlaucomaNet, to identify POAG using variable fundus photographs from different populations and settings. GlaucomaNet consists of two convolutional neural networks to simulate the human grading process: learning the discriminative features and fusing the features for grading. We evaluated GlaucomaNet on two datasets: Ocular Hypertension Treatment Study (OHTS) participants and the Large-scale Attention-based Glaucoma (LAG) dataset. GlaucomaNet achieved the highest AUC of 0.904 and 0.997 for POAG diagnosis on OHTS and LAG datasets. An ensemble of network architectures further improved diagnostic accuracy. By simulating the human grading process, GlaucomaNet demonstrated high accuracy with increased transparency in POAG diagnosis (comprehensiveness scores of 97% and 36%). These methods also address two well-known challenges in the field: the need for increased image data diversity and relying heavily on perimetry for POAG diagnosis. These results highlight the potential of deep learning to assist and enhance clinical POAG diagnosis. GlaucomaNet is publicly available on https://github.com/bionlplab/GlaucomaNet .

Figure 1

Creation of the OHTS dataset.

Figure 2

The architecture of the proposed GlaucomaNet.

Figure 3

Examples of fundus photographs and their corresponding masked images: (a) POAG due to VF only (the left one is based on the optic disc mask and the right one is based on a random mask) and (b) POAG due to glaucomatous disc criteria (the left one is based on the optic disc mask and the right one is based on a random mask).

Figure 4

Comparison of different metrics (standard deviation) for different model architectures in the OHTS dataset. p-values are calculated between GlaucomaNet and other models. *p-value ≤ 0.05, **p-value ≤ 0.01.

Figure 5

Comparison of different metrics (standard deviation) for proposed and ensemble methods in the OHTS dataset.

Figure 6

Comparison of different metrics for different model architectures in the LAG dataset. The model reported in Li et al. was trained on 10,928 images, with 4528 having POAG. There is no reported F1-score for Li et al. P-values are calculated between GlaucomaNet and other models. *p-value ≤ 0.05, **p-value ≤ 0.01.

Figure 7

The t-SNE visualization of the proposed model on the OHTS dataset. Each point represents a fundus image. Red and gray dots represent the glaucomatous and non-glaucomatous images.