A progressive optic nerve condition called glaucoma causes irreversible eyesight loss. To diagnose retinal diseases, retinal fundus imaging has been used in recent years. Analyzing these images effectively requires pinpointing the areas of interest, which can be tricky, due to the anatomy and vascular patterns in fundus images. Different image segmentation techniques are used to extract the area of interest from the fundus images. This paper explores the various segmentation methodologies, emphasizing conventional and modern retinal fundus image segmentation approaches. Evaluation measures such as the Disc damage likelihood scale, Inferior superior temporal region, Dice similarity coefficient, Jaccard index, Sensitivity and Specificity are used to measure the effectiveness of segmentation algorithms, which detect small structural changes that indicate glaucomatous damage. Furthermore, this paper also provides a detailed analysis of deep learning algorithms used for optic cup and optic disc segmentation. This detailed analysis demonstrates that accurate disc and cup segmentation remains a significant challenge and suggests that effective segmentation strategies and deep learning approaches are required for vast and complex datasets.
Read full abstract