A Curvature based Approach to Tortuosity Measurement of Retinal Blood Vessels

Georg Maier, Peter Barth, Georg Pisinger, Wadim Bowl, Monika Andrassi-Darida, and Birgit Lorenz


Retinal vessel, tortuosity, curvature, arc spline, retinopathy of prematurity (ROP)


Retinopathy of prematurity (ROP) is an eye disease that affects prematurely-born babies. It may even lead to blindness in serious cases. Among others, ROP can be recognized by evaluating the tortuosity of retinal vessels. An automatic detection and evaluation of tortuous vessels can support ophthalmologists when judging retinal images. There already exist various methods for computing vessel tortuosity. However, they do not always cover the medical properties and requirements. Thus, we propose an efficient measure, which is based to the total curvature and the number of curvature sign changes (twists) of the vessel. In order to achieve a stable detection of twists we represent the extracted vessels by curves composed of circular arcs and line segments: The vessels center pixels are approximated by smoothly joined segments of constant curvature. For any maximum tolerance we obtain the minimum number of segments. This way vessels are represented by preferably long curve segments of constant curvature. Together with the proposed measure this representation enables an effective and stable method for evaluating vessel tortuosity. Its performance is examined by means of both synthetic data and vessels of real retinal images.

Important Links:

Go Back