Diabetic Retinopathy Pathological Signs Detection using Image Enhancement Technique and Deep Learning

The screening of diabetic retinopathy (DR) affects the visual inspection of retina images taken by ophthalmologists to detect the specific signs of pathology such as exudate, hemorrhage (HEM) and microaneurysm (MA). However, this process is currently conducted manually in many hospitals. Therefore, it is time-wasting and risky for humans to make mistake. In general, this paper introduces an automated machine learning algorithm for detecting diabetic retinopathy (DR) in fundus images. It also involves an image pre-processing enhancement technique to support accuracy on deep learning for DR classification. For the image enhancement process, high-pass filter, histogram equalization and de-haze algorithm are applied to improve the visual quality of fundus images. By using four convolution layers, a CNN architecture is set up to classify the three pathological signs; HEM, MA and exudate. Two public online datasets, e-Ophtha and DIARETDB1 are used to evaluate the performance of this system. From training and testing results using enhanced DR images, a slight improvement in classification accuracy is revealed, compared to those original images with no enhancement for both datasets