Deep Learning for Glaucoma Detection and Identification of Novel Diagnostic Areas in Diverse Real-World Datasets

Abstract

Purpose To develop a three-dimensional (3D) deep learning algorithm to detect glaucoma using spectral-domain optical coherence tomography (SD-OCT) optic nerve head (ONH) cube scans and validate its performance on ethnically diverse real-world datasets and on cropped ONH scans. Methods In total, 2461 Cirrus SD-OCT ONH scans of 1012 eyes were obtained from the Glaucoma Clinic Imaging Database at the Byers Eye Institute, Stanford University, from March 2010 to December 2017. A 3D deep neural network was trained and tested on this unique raw OCT cube dataset to identify a multimodal definition of glaucoma excluding other concomitant retinal disease and optic neuropathies. A total of 1022 scans of 363 glaucomatous eyes (207 patients) and 542 scans of 291 normal eyes (167 patients) from Stanford were included in training, and 142 scans of 48 glaucomatous eyes (27 patients) and 61 scans of 39 normal eyes (23 patients) were included in the validation set. A total of 3371 scans (Cirrus SD-OCT) from four different countries were used for evaluation of the model: the non overlapping test dataset from Stanford (USA) consisted of 694 scans: 241 scans from 113 normal eyes of 66 patients and 453 scans of 157 glaucomatous eyes of 89 patients. The datasets from Hong Kong (total of 1625 scans; 666 OCT scans from 196 normal eyes of 99 patients and 959 scans of 277 glaucomatous eyes of 155 patients), India (total of 672 scans; 211 scans from 147 normal eyes of 98 patients and 461 scans from 171 glaucomatous eyes of 101 patients), and Nepal (total of 380 scans; 158 scans from 143 normal eyes of 89 patients and 222 scans from 174 glaucomatous eyes of 109 patients) were used for external evaluation. The performance of the model was then evaluated on manually cropped scans from Stanford using a new algorithm called DiagFind . The ONH region was cropped by identifying the appropriate zone of the image in the expected location relative to Bruch's Membrane Opening (BMO) using a commercially available imaging software. Subgroup analyses were performed in groups stratified by eyes, myopia severity of glaucoma, and on a set of glaucoma cases without field defects. Saliency maps were generated to highlight the areas the model used to make a prediction. The model's performance was compared to that of a glaucoma specialist using all available information on a subset of cases. Results The 3D deep learning system achieved area under the curve (AUC) values of 0.91 (95% CI, 0.90–0.92), 0.80 (95% CI, 0.78–0.82), 0.94 (95% CI, 0.93–0.96), and 0.87 (95% CI, 0.85–0.90) on Stanford, Hong Kong, India, and Nepal datasets, respectively, to detect perimetric glaucoma and AUC values of 0.99 (95% CI, 0.97–1.00), 0.96 (95% CI, 0.93–1.00), and 0.92 (95% CI, 0.89–0.95) on severe, moderate, and mild myopia cases, respectively, and an AUC of 0.77 on cropped scans. The model achieved an AUC value of 0.92 (95% CI, 0.90–0.93) versus that of the h...

Applications of Artificial Intelligence and Deep Learning in Glaucoma

“…OCT volumetric scans, including ONH-centered and macula-centered ones, can potentially provide more comprehensive features, such as the changes in RNFL, GCIPL, Bruch membrane opening, neuroretinal rim area, the lamina cribrosa, and choroidal. Thus, several studies investigated the potential of 3D DL models in glaucoma detection 48–52. Maetschke et al48 developed a 3D DL model using ONH-centered volumetric OCT scans and compared its performance with 8 kinds of ML classifiers.…”

“…Thus, several studies investigated the potential of 3D DL models in glaucoma detection. [48][49][50][51][52] Maetschke et al 48 developed a 3D DL model using ONH-centered volumetric OCT scans and compared its performance with 8 kinds of ML classifiers. The DLbased approach achieved a peak test AUROC of 0.94, which was substantially higher (P < 0.05) than the best classic ML method (AUROC of 0.89) on segmentationbased features.…”

“…First, just as for fundus images, some anatomic morphologies such as myopic optic disc morphologies can influence the results, especially the RNFL thickness measurement for glaucoma detection from OCT. Groups have attempted to address this particular challenge. Russakoff et al 51 and Noury et al 52 developed 3D DL models based on macula-centered and ONH-centered scans, respectively. The models performed reasonably well across different myopic severity distributions.…”

Deep learning in optical coherence tomography: Where are the gaps?

“…Volumetric OCT scans, including ONH‐centred and macula‐centred ones, can potentially provide more comprehensive features such as the changes in the RNFL, GCIPL, Bruch's membrane opening, neuro‐retinal rim area, the lamina cribrosa, and choroid. Several studies have shown that DL models based on OCT 3D scans outperformed DL models based on 2D images and also showed non‐inferior performance to glaucoma specialists for detecting glaucomatous structural changes 26–31 …”

Artificial intelligence for glaucoma: state of the art and future perspectives

“…Braeu et al reported on a geometric DL model capable of extraction critical structural points from a 3D point cloud of the optic disc and peripapillary region which performed well on two distinct networks with high diagnostic accuracy and was superior to a 3D CNN model, while requiring fewer inputs [17,18]. Noury et al , conversely, identified the lamina cribosa as a region of interest in glaucoma diagnosis [19]. Such geometric models may prove advantageous comparing with class activation maps (CAM)-graded images in the identification of key structural locations within the posterior pole in glaucoma diagnosis and physiology.…”