Abstract #3319

Classification on Ex-Vivo MRS Signals of Glioma Samples

Bernd Merkel¹, Frauke Nehen², Yasemin Oezdemir¹, Markus Thorsten Harz¹, Dieter Leibfritz², Rudolf Fahlbusch³, Horst Karl Hahn¹

¹Fraunhofer MEVIS, Bremen, Germany; ²Institute of Organic Chemistry, University of Bremen, Bremen, Germany; ³International Neuroscience Institute, Hannover, Germany

The goal of this work is the automated classification of glioma samples with high-resolution ex-vivo MR-spectroscopy. HR-MRS is a sensitive method to detect metabolite changes in different tumor and tissue types. Altogether 47 biopsates of healthy, tumor margin and tumor center tissue, measured on a 600 Mhz spectrometer, were analyzed. For further analysis, the lipophilic compounds were omitted and only the hydrophilic ones were analyzed. By the application of ICA and further classification and feature reduction techniques, we show that the tumor margin is distinctively different from the tumor center.