The Fusion of Features for Detection of Cotton Wool Spots in Digital Fundus Images

The Cotton Wool Spots are lesions seen on the surface of the retina among diabetic patients. It indicates a pre-proliferative Diabetic Retinopathy state and needs to be treated. This paper describes a method based on morphological operations and thresholding for detection of Cotton Wool Spots. Feature level fusion is a technique used in image processing and pattern recognition that can improve the accuracy of object recognition and computer vision. Hence, in this experiment, LBP and HOG features have been combined as it improves the ability to detect affected areas in a retinal image considerably. Also, Decision tree based classification method is used to classify and distinguish between healthy and diseased images. The proposed method yielded Precision 0.92, Recall 0.88 and F-Score 0.90 for IDRiD database. For the Kaggle database, Precision, Recall and F-score were computed as 0.95 each respectively. The performance evaluation measures comprising Precision, Recall and F-Score for STARE database were computed as 0.98, 0.93 and 0.96 respectively.