Applications of deep learning in detection of glaucoma: A systematic review

Abstract

Glaucoma is the leading cause of irreversible blindness and disability worldwide. Nevertheless, the majority of patients do not know they have the disease and detection of glaucoma progression using standard technology remains a challenge in clinical practice. Artificial intelligence (AI) is an expanding field that offers the potential to improve diagnosis and screening for glaucoma with minimal reliance on human input. Deep learning (DL) algorithms have risen to the forefront of AI by providing nearly human-level performance, at times exceeding the performance of humans for detection of glaucoma on structural and functional tests. A succinct summary of present studies and challenges to be addressed in this field is needed. Following PRISMA guidelines, we conducted a systematic review of studies that applied DL methods for detection of glaucoma using color fundus photographs, optical coherence tomography (OCT), or standard automated perimetry (SAP). In this review article we describe recent advances in DL as applied to the diagnosis of glaucoma and glaucoma progression for application in screening and clinical settings, as well as the challenges that remain when applying this novel technique in glaucoma.

Normal‐tension glaucoma: Current concepts and approaches‐A review

“…In the past, such machine learning was limited by the need for engineered filters, in order to achieve pattern recognition. 106 Artificial neural networks (ANNs), by contrast, process input data through a series of interconnected nodes or "neurons", which are capable of transforming an input by ascribing a "weight" to it before passing it to the next neuron. 107 Throughout a learning process, a neural network will adjust these weights in order to develop a classification system capable of making a prediction.…”

“…DL have risen to the forefront of AI by providing nearly human-level performance, at times exceeding the performance of humans for detection of glaucoma on structural and functional tests. 106 In a recent Retinal Fundus Glaucoma Challenge (REFUGE), 108 a competition that was held as part of the Ophthalmic Medical Image Analysis (OMIA) workshop at MICCAI 2018, which aimed at a uniform evaluation framework to assess automated methods for Optic disc/optic cup segmentation and glaucoma classification from fundus photos, the best three performing algorithms are CUHKMED which used a DeepLabv3+ architecture based on atrous separable convolutions that are able to capture multiscale characteristics, together with adversarial learning 108,109 ; followed by Masker, using and ensemble of Mask-RNNs trained with bootstrap; followed by BUCT using a classical U-Net. 108 In a recent study, a deep-learning algorithm was developed to discriminate between glaucoma suspect and early NTG, on OCT, using the parameter Bruch's membrane opening-minimum rim width (BMO-MRW), peripapillary RNFL thickness, and colour classification of RNFL.…”

Deep Learning-Based Glaucoma Detection With Cropped Optic Cup and Disc and Blood Vessel Segmentation

“…MobileNet is a CNN architecture with a reduced number of parameters but high classification accuracy optimized for mobile computer vision models [35]. A typical network has a computational price of 300 million multiply-adds and utilizes 3.4 million parameters [14]. In this work, we use MobileNet v2 because it outperforms MobileNet v1 with significant model size and computational cost.…”

“…Narrower areas could indicate glaucomarelated damage. Glaucoma diagnostic methods focused on medical image processing are currently gaining traction over more traditional studies [13], [14]. Several characteristics of the ocular retinal structure must be noted in these instances, including the optic nerve head (ONH), cup, retinal nerve fiber surface, peripapillary atrophy, and so on.…”

Re: Xiong et al.: Multimodal machine learning using visual fields and peripapillary circular OCT scans in detection of glaucomatous optic neuropathy (Ophthalmology. 2022;129:171-180)

“…In recent years, many DL algorithms have been developed and proven to be promising in glaucoma classification for screening purposes, especially image-based ones. 2,3 In the study by Xiong et al, the proposed bimodal DL algorithm requires both OCT images and VFs for GON classification, which will potentially hinder its feasibility in a screening scenario because VF testing is relatively subjective and tedious for patients. 4 Besides, OCT and VF testing devices are not as widely available as fundus photography in lower resource areas.…”