Automatic Detection of Exudates in Digital Retinal Images

Exudate is one of the serious complications and a major cause of blindness in diabetic retinopathy patients. Nowadays, the digital retinal image is frequently used to follow-up and diagnoses eye diseases. Therefore, the retinal image is crucial and essential for expert to detect of exudates. Unfortunately, the retinal images in Thailand are poor quality image. Detecting exudates in a large number of poor quality retinal images generated by screening programs, is very expensive in professional time and opens to human error. This paper proposes a part of a larger effort to develop a new method for detection of exudates in poor quality retinal image. The retinal images are segmented using fast mean shift algorithm following key preprocessing steps, i.e., color normalization, contrast enhancement, noise removal and color space selection. On the difficult data set, the method can achieves accuracy, specificity and sensitivity with 93.8%, 95.3%, 94.9% for the detection of exudates on a set of 1,220 retinal images. General Terms Medical Image Processing