Segmentation of Retinal Blood Vessels and Optic Disc Using Deep Neural Networks: State-Of-The-Art Review

Medical image analysis places an intense focus on retinal blood vessel and optic disc segmentation due to the crucial role it plays in diagnosis, treatment planning, and clinical outcome evaluation in domains like ophthalmology and neurosurgery. Considering the many different medical imaging techniques, each of which has its own set of features, it is vital to have automated or semi-automatic vascular segmentation in order to provide support to clinicians in these activities. A growing number of computer vision applications have taken advantage of deep learning architectures, particularly for fundus image segmentation of the retinal blood vessels. This work provides a comprehensive review of recent advances in deep learning-based retinal blood vessel and optic disc segmentation systems, with a particular emphasis on their taxonomy and analysis of improvement approaches. The goals include analysing current methods, noticing trends in improvement strategies, identifying obstacles, and suggesting areas for future research. Optimization algorithms, regularization methods, pooling operations, activation functions, transfer learning, and ensemble learning approaches are some of the taxonomies that are being investigated. By scrutinizing 25 relevant papers spanning 2018 to 2023, this study aims to inform future research strategies and enhance the predictive accuracy of forthcoming models, ensuring optimal performance and generalization ability in automatic retinal blood vessel segmentation algorithms.