Statistical analysis of PET images

Abstract

In recent decades, neuroimaging techniques have become relevant and essential supports to the diagnosis and therapy of neurological diseases. Positron Emission Tomography (PET) is a method of functional investigation that measures brain metabolism and identifies the changes that occur at cellular and molecular level, supporting physicians for studying neurological diseases, and for the definition of the diagnosis. To support image studies, there exist automatic PET images analysis algorithms; e.g., voxel-byvoxel analysis technique allows to obtain statistical measurements that can be associated to functional neurological anomalies. Research groups, both physicians as well as computer scientists, have performed statistical analysis on images dataset to identify regions of variation of the glucose in the brain [1-2]. Although these studies demonstrate that a particular group of Regions Of Interest (ROIs) identifies specific pathologies, today systems for the automatic diseases classification are still not available.