A Novel Approach for Diagnosis and Severity Grading of Diabetic Maculopathy

摘要

Diabetic maculopathy is a severe complication of retinopathy and is growing as a major threat to vision acuity. One such cause of maculopathy is due to diabetic macular edema where damaged retinal vasculature starts leaking fluid and protein onto the macula and its accumulation results in vision loss. This paper presents an effective automated system for detection and severity grading of maculopathy. For this, centre of optic disc is detected and the fovea region is localized using superior and inferior vascular arcades within the retina. Macular regions are then marked based on international grading standard. Next, the lesions in scaled macular regions are detected using morphological techniques. Features are extracted and selected and are passed to the multiclass SVM for the severity grading of maculopathy into normal, mild, moderate and severe. The proposed method was tested on publicly available databases and achieved a sensitivity of 96.89% and a specificity of 97.15% which is really competitive with the state-of-the-art in this area.