Eccentricity Based Quantification of Retinal Vascular Tortuosity For Early Detection of Diabetes and Diabetic Retinopathy

Abstract

Diabetic retinopathy is the main cause for the loss of vision among working age adults. Therefore, early detection of diabetic retinopathy is vital for the prevention of blindness. Vascular tortuosity (amount of twists and turns) is an early indicator of diabetes and diabetic retinopathy compared to other signs observed in the retinal image. We propose a novel tortuosity index to quantify retinal vascular tortuosity including dilations and elongation due to vascular augmentation caused by diabetes in addition to conventional characteristics; curvature and twists. In the proposed method, the tortuosity of a blood vessel is quantified using the eccentricity of the pixels of its skeleton. Retinal images of 110 voluntary participants (72 healthy subjects, 5 Type I diabetic, 28 Type II diabetic and 5 diabetic retinopathy patients) were acquired using a retinal camera to study the capabilities of the proposed tortuosity index. The derived tortuosity values using the novel method were significantly different among healthy, Type II diabetic and diabetic retinopathy patients (p-value < 0.001). While tortuosity values of diabetic retinopathy patients were significantly higher than all other participants, the healthy group showed less tortuosity than others. When compared to age, diabetes was identified as the predominant factor which causes the increase in retinal vascular tortuosity.