Machine Learning Approach for Detection of Diabetic Retinopathy with Improved Pre-Processing

Diabetic Retinopathy has been one of the common reason why diabetic patients lose sight. It is caused by the damaging of the blood vessel of the light sensitive tissue at the back of the eye or the retina. The traditional methods for the detection of the disease requires lots of time and efforts hence the proposed system aims to detect diabetic retinopathy using Image Processing and Machine Learning. The input provided to the proposed system are the retinal images from the standard databases DIARETDB0 and DIARETDB1. The system aims to automate the process of detection using a combination of basic image processing steps with more focus on pre-processing to obtain clear image for feature extraction, further Machine Learning algorithms are applied for classification. The statistical features such as area and perimeter are obtained from blood vessels and exudates after pre-processing stage, different pre-processing techniques such as grayscale conversion, binarization, canny edge detection and some morphological operations including image dilation and erosion are applied to obtain clear image. The Machine Learning algorithms that are used for result analysis are Weighted KNN, Cubic SVM and Simple Tree which provides an accuracy of 85.8%, 87.2% and 88.6% respectively. The sensitivity and specificity parameters are also obtained, Simple tree yields better results for all the three result parameters.