Vessel masking and Hough transform for optic disc localisation from retinal images

Abstract

Background: Retinal images can be considered as one of the reliable indicators for symptoms of many ocular diseases such as diabetic retinopathy, macular degeneration and glaucoma. By analysing and tracking changes of important structures on a retinal image, symptoms of ocular diseases can be detected in a timely manner which helps physicians plan early treatment for better disease control. One of the important landmarks on a retinal image is the optic disc (OD), which must be localised to estimate retinal vessel parameters such as vessel width and tortuosity. This paper proposes a method for automatic OD localisation from a retinal image. Methods: A retinal image is first pre-processed and thresholded to produce a binary image that highlights most retinal vessels on the image. Next, a discrete cosine transform-based smoothing method is employed to replace the detected vessel pixel values on the pre-processed image with values closer to the surrounding neighbour pixel values, effectively masking most vessels on the image. Hough transform is then applied to the vessel-masked image to detect the circle representing the OD on the image, producing the estimated location of the OD center and its estimated diameter. Results: Applying the proposed method to three different public databases, namely Digital Retinal Images for Vessel Extraction (DRIVE), High-Resolution Fundus (HRF) and Methods to Evaluate Segmentation and Indexing Techniques in the field of Retinal Ophthalmology (MESSIDOR) resulted in an overall detection rate of 99.53%. Conclusions: The achieved performance by the proposed method is superior to many published methods of OD localization, with a processing time of less than one second for each image. While this has only been validated on one type of retinal images, future investigations may include validation on other types such as angiograms or scanning laser ophthalmoscopy.