Polynomial modeling of retinal vessels for tortuosity measurement

Abstract Tortuosity is one of the micro vascular change that is observed in many retinopathies. Its early detection can prevent the progression of various retinopathies to a critical stage at which a person may become blind. Here, we propose a novel method for the measurement of tortuosity by polynomial modeling of retinal vessels for the analysis of hypertensive retinopathy. The proposed method is tested on a set of 30 arteries and 30 veins vessel images collected from the Retinal Vessel Tortuosity Dataset (RET-TORT). Also, 90 vessel segments from Digital Retinal Images for Vessel Extraction (DRIVE) and 149 vessel segments from High Resolution Fundus (HRF) databases are used for tortuosity evaluation. The experimental results demonstrate that the order of the polynomial increases with the increase in the tortuosity of the blood vessels. Hence, the order of the polynomial can be used as a parameter to classify vessels as non-tortuous and tortuous. The results of the method are also evaluated subjectively and the inter-rater agreement analysis is made by using Fleiss Kappa index. The Spearman's rank order correlation coefficient is used to analyze the correlation between the ranking provided by the expert in the RET-TORT database and the ranking obtained by the proposed method. The results demonstrate that this method is capable of evaluating the tortuosity and classify vessel segments into non-tortuous or tortuous effectively.