Abstract:
Computer vision has an extraordinary reputation in medical diagnostic in the current age and it covers a wide range of diseases related to human beings, animals, and what not. The human retina is a vital part of the eye for it is responsible for the vigorous vision. People with severe diabetes tend to develop symptoms of retinal damage in the later part of their lives if they are not taken care of. Though the retinal complications instigate from insignificant signs but can cover the whole retina in the advanced stage of the disease. This research work is proposed to identify the primary sign; exudates, from the retinal images of the patients who likely to develop retinal complications or already have developed. The presented work is forecasted to work for the early diagnosis of retinal exudates that most likely appear in the impediments of diabetic retinopathy and macular degeneration. We proposed an automated system for the screening of patients that is an ensemble of image analysis as well as computer vision methods. The presented work provides the computer aided diagnosing system for the patients of diabetic retinopathy and macular degeneration. In this research, we have presented the data models and clinical practice along with the computer aided diagnosing systems. Our proposed work first regulate the uniqueness in the input images, find the potential lesions with the support of anatomical knowledge with the help of image analysis techniques. Later we used machine learning algorithms to validate the spotted lesions. Experimental results show an improved overall performance with an average accuracy of 93% and 96% for the screening of diabetic retinopathy and macular degeneration respectively. We intend to incorporate a multi-lesion diagnostic solution to stimulate the clinical practice in an effort to provide better utility.
