Cancer area characterization by non-parametric clustering

The application of machine learning techniquesto open problems in different medical researchfields appears to be stimulating and fruitful, especiallyin the last decade. In this paper, a newmethod for MRI data segmentation is proposed,which aims at improving the support of medicalresearchers in the context of cancer therapy. Inparticular, our effort is focused on the processingof raw output obtained by Dynamic Contrast-Enhanced MRI (DCE-MRI) techniques. Here,morphological and functional parameters are extracted,which seem indicate the local developmentof cancer. Our contribute consists in organizingautomatically these output, separatingMRI slice areas with different meaning, in a histologicalsense. The technique adopted is basedon the Mean-Shift paradigm, and it has recentlyshown to be robust and useful for different andheterogeneous segmentation tasks. Moreover,the technique appears to be predisposed to numerousextensions and medical-driven optimizations.