Machine learning-assisted image analysis techniques for glaucoma detection

Glaucoma is an eye disease that damages the optic nerve which is necessary for clear vision. This injury is frequently caused by extremely high pressure inside the eye. Any damage to the optic nerve, which is responsible for transmitting visual information from the eye to the brain, can result in vision loss and, in the worst-case scenario, blindness. In most cases of glaucoma, structural changes occur first, followed by functional deficits. Ophthalmologists use three well-established tests to diagnose glaucoma: Intraocular pressure (IOP) measurements, visual field tests, and (stereoscopic) assessments of the optic nerve. These methods, when combined, provide information on both structural and functional defects. However, carrying out these tests is not simple. For the correct diagnosis of glaucoma, ophthalmologists must have many years of experience. As a result, even in the modern world, glaucoma is underdiagnosed or, paradoxically, overtreated. To address these issues, a combination of Artificial Intelligence, Image Processing, and computer vision-based models are being developed for easy and rapid glaucoma diagnosis. Many research experiments used traditional machine learning and deep learning techniques to detect glaucoma. These techniques mainly focused on standard retinal image datasets to train the model and demonstrated good detection accuracy. This article provides a comprehensive description of the numerous forms of glaucoma, their pathophysiology, distinct diagnosis procedures, available datasets, and machine learning, deep learning, and hybrid methods of glaucoma detection. This review evaluates a variety of published studies on glaucoma detection approaches ranging from conventional machine learning to complex neural networks and summarizes the benefits and drawbacks of each procedure in a table style. The discussed methodologies are further evaluated in terms of performance parameters, such as accuracy, sensitivity, specificity, and area under the curve. In addition, this research demonstrates hybrid methodologies for glaucoma detection and their advantages over machine learning and deep learning approaches. The comprehensive discussion offered in this review may assist researchers to choose the suitable dataset, extract anatomical features, decide on methods, and select performance evaluation metrics for glaucoma detection.