Diabetic retinopathy is the most common reason for visual disability particularly in type II diabetes. Diabetic maculopathy is retinopathy affecting the macular region of the eye. Fundus Fluorescein Angiograms (FFA) is one of the modalities for detection of pathologies associated with diabetic maculopathy. This paper evaluates a set of machine learning algorithms like K-Means, Fuzzy C-Means, Rough Fuzzy C-Means and Expectation Maximization for segmenting the retinal vessels of FFA image, which is a preliminary step for automating the process of detection and classification of different forms of maculopathy. The ground truth of Fundus images were taken from the DRIVE database for evaluating the segmented results. The similarity measures like Jaccard and Dice coefficients were used to compare the segmented results with the ground truth images and we found that Fuzzy C-Means algorithm gives an accuracy of 96%. When comparing with existing methods in the literature, our proposed method has less user intervention, requires less number of parameters and does not require the identification of optic disc to locate the retinal vessels. This method also does not require any preprocessing like contrast enhancement and noise removal. © Research India Publications.
cited By 0
Sa Abirami, Swapna, T. Ra, Pulari, S. Ra, and Chakraborty, Cb, “Unsupervised segmentation of retinal vessels from fundus Fluorescein angiogram images”, International Journal of Applied Engineering Research, vol. 10, pp. 38-43, 2015.