DCNN-based prediction model for detection of age-related macular degeneration from color fundus images

Age-related macular degeneration (AMD) is a degenerative disorder in the macular region of the eye. AMD is the leading cause of irreversible vision loss in the elderly population. With the increase in aged population in the world, there is an urgent need to develop low-cost, hassle-free, and portable equipment diagnostic and analytical tools for early diagnosis. As AMD detection is done by examining the fundus images, its diagnosis is heavily dependent on medical personnel and their experience. To remove this issue, computer-aided algorithms may be used for AMD detection. The proposed work offers an effective solution to the AMD detection problem. It proposes a novel 13-layer deep convolutional neural network (DCNN) architecture to screen fundus images to spot direct signs of AMD. Five pairs of convolution and maxpool layers and three fully connected layers are utilized in the proposed network. Extensive simulations on original and augmented versions of two datasets (iChallenge-AMD and ARIA) consisting of healthy and diseased cases show a classification accuracy of 89.75%, 91.69%, and 99.45% on original and augmented versions of iChallenge-AMD and 90.00%, 93.03%, and 99.55% on ARIA, using a 10-fold cross-validation technique. It surpasses the best-known algorithm using DCNN by 2%.