Automatic Detection of DR Lesion Using Active Contour and Tsallis Entropy Based Blended Approach

Diabetes has become a modern epidemic. Globally, the prevalence rate of diabetes is high and it tends to even higher. WHO reports, globally around 1.3 billion peoples suffer from some form of eye related diseases. Diabetic retinopathy (DR) is the leading cause of vision loss among working-age individuals and represents the most common repercussion of diabetes. Early DR diagnosis and therapy might stop or postpone its progression and subsequent vision loss. However, manual screening of DR is time-consuming and subject to inter-observer variability. Therefore, computer-aided methods for automated DR detection are of great interest. This article proposes the computer aided, automated DR detection approach. Initially anisotropic diffusion filtering is employed for pre-processing of retinal fundus images. In segmentation, the anatomical structures are identified and eliminated. The DR lesions segmentation is performed using active contour and modified Tsallis entropy approach. Active contour models, also known as snakes, have been extensively used for medical image segmentation and boundary detection. These models use an energy minimization approach to detect the contours of objects in an image. Finally feature extraction and classification of DR lesions is achieved with the help of Gray Wolf optimizer. Gray Wolf Optimizer (GWO) is an emerging optimization algorithm that draws inspiration from the hunting behavior of gray wolves. It has shown promising results in various optimization problems, including medical image segmentation. In this research work, we propose a novel method for DR detection using active contour and GWO. Our approach combines the strengths of these two techniques to improve the accuracy and efficiency of DR detection. We evaluate developed method using a publicly available dataset and compare its performance with recent DR detection approaches.