A Deep Composite Predict Network for Eye Disease

Abstract

Different areas of eyes reflect different states of internal organs of human body. That can assist to detect the disease internal organs, facilitate the prevention and cure by analyzing the image of human eyes. In this paper, we propose a new algorithm, Deep Composite Predict Network (DCP-Net), combined deep learning technique with traditional Chinese medicine theory, which can segment interesting areas of the human eyes, detect the disease eyes texture and then predict internal lesion organ. Aimed at helping doctors to diagnose quickly. The proposed algorithm consists three sub-modules, i.e., multiple segmentation sub-module, fast detection sub-module and predict eye disease module. The multiple segmentation module is solved by the improved retraining U-network, which can obtain an accurate eye to segment interesting areas. The fast detection module is based on enhanced Yolo framework. The predication disease method combined Chinese traditional theory with computer vision technology. Finally, our network can accurately detect the disease of human eyes and predict lesions of internal organs, which can help doctors quickly diagnose lesion organs by human eyes.