Automatic Unsupervised Segmentation of Retinal Vessels Using Self-Organizing Maps and K-Means Clustering Carmen Alina Lupas ¸cu and Domenico Tegolo Dipartimento di Matematica e Informatica Universit` a degli Studi di Palermo Palermo, Italy lupascu@math.unipa.it, domenico.tegolo@unipa.it Abstract. In this paper an automatic unsupervised method for the segmentation of retinal vessels is proposed. A Self-Organizing Map is trained on a portion of the same image that is tested and K-means clustering algorithm is used to divide the map units in 2 classes. The entire image is again input for the Self-Organizing Map, and the class of each pixel will be the class of the best matching unit on the Self-Organizing Map. Finally, the vessel network is post-processed using a hill climbing strategy on the connected components of the segmented image. The experimental evaluation on the publicly available DRIVE database shows accurate extraction of vessels network and a good agreement between our seg- mentation and the ground truth. The mean accuracy, 0.9459 with a standard de- viation of 0.0094, is outperforming the manual segmentation rates obtained by other widely used unsupervised methods. A good kappa value of 0.6562 is inline with state-of-the-art supervised and unsupervised approaches. Keywords: Retinal Vessels, Self-Organizing Map, K-means. 1 Introduction Automatic analysis of retinal vasculature is important in the diagnosis of many eye pathologies. Once the vessel tree is extracted from retinal images, it is useful not only for diagnosis purposes, but also in the registration of retinal images. Branching and crossover points in the vasculature structure are used as landmarks for image registra- tion. Image registration is needed to integrate information from several images, but also to observe the progression of diseases over time. Finally, automatically generated vessel maps have been used to guide the identification of retinal landmarks like the optic disc and the fovea. 1.1 Related Work Many different approaches for automated vessel segmentation have been proposed. Some of them are rule-based methods (those based on vessel tracking ([4]), those based on matched filter responses ([3], [6]) and other ones are based on mathematical mor- phology ([14], [22])). The methods listed above are unsupervised. R. Rizzo and P.J.G. Lisboa (Eds.): CIBB 2010, LNBI 6685, pp. 263–274, 2011. c  Springer-Verlag Berlin Heidelberg 2011