Diabetic Retinopathy Lesions Identification in the Color Fundus Images using Multi-Layer Perceptron

The prognosis of Diabetic Retinopathy (DR) is characterized by Non-Proliferative Diabetic Retinopathy (NPDR) and Proliferative Diabetic Retinopathy (PDR). The early stage of DR is known as NPDR. Detecting NPDR in the early stage becomes crucial to avoid blindness. The purpose of this study is to perceive NPDR lesions using image processing techniques and classification methods. The detection of lesions is carried out by pre-processing, feature extraction, feature vector construction, and classification. The vessel network is extracted for feature extraction in the pre-processing stage. Apart from the regular statistical image features, the color layer features are extracted from the smoothened input image. A clustering-based feature extraction method is introduced to capture features from each color layer. The filtered features, which produce the desired output, are combined and fed into Multi-Layer Perceptron (MLP) classifier. The proposed algorithm achieves 100% accuracy in detecting DR. Hence, this study shows that the proposed method can able to find the DR lesions in the early stage itself.