Glaucomatous optic neuropathy is one of the leading causes of irreversible blindness, with a worldwide prevalence of 3.5% and as many as 76 million people affected globally in 2020. Diagnosis and treatment early can preserve vision in individuals, especially in the early stages of the disease. Diagnosis and monitoring of the disease involves integrating information from the clinical examination with subjective data from visual field testing and objective biometric data acquired through pachymetry, corneal hysteresis and imaging of the optic nerve and retina. This complex process is further complicated by the lack of unambiguous definitions of the presence and progression of glaucomatous optic neuropathy, resulting in the possibility of misinterpretation by the clinician.
Artificial intelligence in medicine
In this context, artificial intelligence (AI) technology has been proposed as a possible option for the diagnostic process. Indeed, applications developed in this field of informatics can improve the quality and robustness of information obtained from clinical data, which can, in turn, improve the physician's approach to patient care.
In particular, one of the applications of AI that has shown great promise in medicine is machine learning (ML), which has been developed since the 1980s. Out of ML, around 2010, came deep learning (DL), whose algorithms, in a short time, surpassed pre-existing algorithms in medicine and other disciplines.
The potential of DL in the medical field, to date, has been applied, for example, to the evaluation of tumours, the detection of atrial fibrillation and the timing of stroke onset. In ophthalmology, artificial intelligence and, in particular, DL have proved useful in the diagnosis of diabetic retinopathy (DR), age-related macular degeneration and retinopathy of the premature, right up to the assessment of glaucoma.
Artificial intelligence and glaucoma management
The use of optical coherence tomography (OCT), visual field assessment (VF) and clinical examination of the optic disc form the basis for the diagnosis of glaucomatous optic neuropathy in the clinical setting. The availability of automated systems based on artificial intelligence and algorithms can help improve diagnostic efficiency and automate the diagnosis process. The detection of glaucoma progression is also a key component of the clinical management of patients in order to identify those individuals at risk of developing glaucoma-related visual impairment. In this sense, therefore, artificial intelligence has several potentials applications in glaucoma. These include:
- DL and glaucomatous optic disc detectionGlaucomatous disc detection from images of the fundus of the eye can be difficult and these difficulties can be compounded by factors affecting the image acquisition platform (exposure, focus, magnification), mydriasis status and the presence of non-glaucomatous diseases. Deep learning has made significant advances in the detection of glaucomatous disc damage from digital photography images. In addition, DL has also been applied to images obtained from optical coherent radiation tomography (OCT).
- AI in the examination of the visual field: computerised automated visual field testing has been a real advancement that has enabled visual field testing to be a milestone in the diagnosis and monitoring of glaucoma, thanks to the various platforms and algorithms developed. Several studies have shown that DL algorithms are able to identify glaucoma from visual field analysis more efficiently than ophthalmologists.
- Clinical forecasting and AI: clinical prediction makes it possible to anticipate the evolution of the disease and formulate the prognosis. Studies have shown that prediction models using AI can be updated using data from the various clinical visits, leading to more accurate predictions, as well as greater timeliness.
Future applications
In the future, AI could have a great impact on outpatient glaucoma screening, the management of glaucomatous optic neuropathy and its remote monitoring. Indeed, AI-based methods could be applied to teleretinal screening programmes in the context of primary care, enabling effective initial patient triage. Remote patient monitoring could be facilitated by home field of view testing, which could be performed using virtual reality. Of course, tools for remote disease monitoring will require the ability of DL algorithms to synthesise remotely acquired data.
- Ittoop SM, Jaccard N, Lanouette G, Kahook MY. The Role of Artificial Intelligence in the Diagnosis and Management of Glaucoma. J Glaucoma. 2022 Mar 1;31(3):137-146.
- Mayro EL, Wang M, Elze T, Pasquale LR. The impact of artificial intelligence in the diagnosis and management of glaucoma. Eye (Lond). 2020 Jan;34(1):1-11.
- Mursch-Edlmayr AS, Ng WS, Diniz-Filho A, Sousa DC, Arnold L, Schlenker MB, Duenas-Angeles K, Keane PA, Crowston JG, Jayaram H. Artificial Intelligence Algorithms to Diagnose Glaucoma and Detect Glaucoma Progression: Translation to Clinical Practice. Transl Vis Sci Technol. 2020 Oct 15;9(2):55.