Abstract. The characterisation of tumours from Magnetic Resonance (MR) images of the brain is still a challenging task. In this paper we present an approach based on a K-Means clustering algorithm combined with textural feature information as opposed to intrinsic MR parareenters T1,T2 and PD. This is due to the fact that MR parameters may exhibit significant alterations in the presence of pathological conditions and, therefore lead to incorrect classification. We also address two important aspects of clustering: the selection of the optimum number of classes (Cluster Validity) and the most effective features (reduction of the feature space).