[A robust method for automated detection and grading of retinal venous beading in diabetes mellitus]

Diabetes mellitus is a high prevalence disease in the world and retinal vessel changes such as venous beading are common findings in patients suffering from long-standing diabetes. Three methods (1. Combined use of the SED, ROSE, Modified LoG algorithm; 2. Gaussian fitting-based method; 3. Gradient-based method) for detection venous caliber were used in the first processing stage, then the magnitude of fast Fourier transform of variation function of venous caliber was used as a detective index of venous beading. The reliability of the results, justified by their low variability, makes feasible the standardization of quantitative parameters to be used both in the diagnosis and in the prognosis of diabetes, and also allows prospective studies based upon them.