Computer estimation of the AVR parameter in diabetic retinopathy

Abstract

The Arteriolar-to-Venular diameter Ratio (AVR), a parameter derived from vessel caliber measurements in a specific region of retinal images, is used as a descriptor of generalized arteriolar narrowing, an eye fundus sign often seen in patients affected by hypertensive or diabetic retinopathies. The manual computation of AVR is a tedious process, involving repeated measurements of the diameters of all arterioles and venules in the region by human graders. To facilitate large-scale clinical use, an accurate, rapid and efficient system to compute AVR is required. We developed a semiautomatic computer procedure to achieve this goal. At first the vessel network is traced by a vessel tracking algorithm and vessel caliber is estimated. In the region of interest for estimation of AVR, vessels are classified as either arteries or veins and AVR parameter is eventually computed. The algorithm has shown good performances in 38 images extracted from two large datasets (DCCT and ETDRS studies). Results provided by the proposed system have been compared with manually derived AVR values on the fundus images, resulting in a final correlation coefficient of 0.95 for the ETDRS subset and of 0.91 for the DCCT subset. User interaction is limited to correct tracking errors in 2 vessels per image and classification errors in 1 vessel per image on average.