Early detection of diabetic retinopathy from big data in hadoop framework

In this article, we have designed a fast and reliable Diabetic Retinopathy (DR) detection technique in Hadoop framework, which can identify the early signs of diabetes from eye retinal images. In the proposed scheme the retinal images are classified into five categories: No Diabetic Retinopathy (DR), Mild DR, Moderate DR, Severe DR and Proliferative DR. The proposed scheme follows three distinct steps for classification of the diabetic retinopathy images: feature extraction, feature reduction and image classification. In the initial stage of the algorithm, the Histogram of Oriented Gradients (HOG) is used as a feature descriptor to represent each of the Diabetic Retinopathy images. Principal Component Analysis (PCA) is used for dimensional reduction of HOG features. In the final stage of the algorithm, K-Nearest Neighbors (KNN) classifier is used, in a distributed environment, to classify the retinal images to different classes. Experiments have been carried out on a substantial number of high-resolution retinal images taken under an assortment of imaging conditions. Both left and right eye images are provided for every subject. To handle such large datasets, Hadoop platform is used with MapReduce and Mahout framework for programming. The results obtained by the proposed scheme are compared with some of the close competitive state-of-the-art techniques. The proposed technique is found to provide better results than the existing approaches in terms of some standard performance evaluation measures.