Fully automated diabetic retinopathy screening using morphological component analysis.

Diabetic retinopathy is the major cause of blindness in the world. It has been shown that early diagnosis can play a major role in prevention of visual loss and blindness. This diagnosis can be made through regular screening and timely treatment. Besides, automation of this process can significantly reduce the work of ophthalmologists and alleviate inter and intra observer variability. This paper provides a fully automated diabetic retinopathy screening system with the ability of retinal image quality assessment. The novelty of the proposed method lies in the use of Morphological Component Analysis (MCA) algorithm to discriminate between normal and pathological retinal structures. To this end, first a pre-screening algorithm is used to assess the quality of retinal images. If the quality of the image is not satisfactory, it is examined by an ophthalmologist and must be recaptured if necessary. Otherwise, the image is processed for diabetic retinopathy detection. In this stage, normal and pathological structures of the retinal image are separated by MCA algorithm. Finally, the normal and abnormal retinal images are distinguished by statistical features of the retinal lesions. Our proposed system achieved 92.01% sensitivity and 95.45% specificity on the Messidor dataset which is a remarkable result in comparison with previous work.