Meeting Banner
Abstract #0980

Multi-Source Feature Selection to Improve Multi-Class Brain Tumor Typing

Vangelis Metsis1, Dionyssios Mintzopoulos2,3, Heng Huang1, Michael N. Mindrinos4, Peter M. Black5, Filia Makedon1, A Aria Tzika2,3

1Computer Science, University of Texas, Arlington, TX, USA; 2NMR Surgical Laboratory, MGH & Shriners Hospitals, Harvard Medical School, Boston, MA, USA; 3Radiology, Athinoula A. Martinos Center for Biomedical Imaging, Boston, MA, USA; 4Biochemistry, Stanford University School of Medicine, Palo Alto, CA, USA; 5Neurosurgery, Brigham and Womens Hospital, Harvard Medical School, Boston, MA, USA

Recent trends in biomedical research have stressed the potential of combining more than one data sources to better understand a patients condition. We acquire state-of-the-art high-resolution magic angle spinning (HRMAS) proton (H1) Magnetic Resonance Spectroscopy (MRS) and gene expression date from the same brain tumor biopsies in order to identify and classify different profiles of brain tumors. We use a novel framework to perform heterogeneous data fusion on both MRS and gene expression datasets using machine learning algorithms. Our experimental results show that our framework outperforms any analysis using individual datasets.