Early detection and treatment of retinal eye diseases is critical to avoid preventable vision loss. Digital retinal imaging is widely used to identify patients with retinal disease in primary care. With the advent of the latest screening technology, the advantage of using ultra wide field scanning-laser ophthalmoscope (SLO) is its wide field of view of the retina making it a valuable tool in the management of patients with retinal disease. One consequence of the wide field imaging process is that artefacts such as eyelashes, eyelids and dust on optical surfaces are also imaged along with the retinal area.
To the best of our knowledge, there is no existing work related to automated differentiation between the true retinal area and these artefacts in an SLO image. The purpose of performing this study is to develop a method that can exclude artefacts in SLO images so as to improve automatic detection of retina diseases. In this paper, we propose a novel approach to automatically extract true retinal area from an SLO image based on image processing and machine learning approaches. To reduce the computational complexity of the image processing tasks and to provide a convenient primitive image pattern, we have grouped pixels into regions based on their regional size and compactness, called super pixels. Image based features representing textural and structural information are calculated and are used to classify the super pixels as retinal area and artefacts. The experimental evaluation results have shown good performance with an overall accuracy of 92% on both healthy and diseased retinal scans compared to clinical annotations. The proposed approach enables effective analysis of retinal area and would have applications that include registering multi-view images into a montage and automated disease diagnosis.