Meeting Banner
Abstract #0994

Brain Tumor Classification Using a Novel H1 HRMAS MRS Method and Robust Algorithmic Classifiers

Dionyssios Mintzopoulos1,2, Ovidiu C. Andronesi1,2, Konstantinos D. Blekas3, Loukas G. Astrakas1,4, Peter M. Black5, A. Aria Tzika1,2

1NMR Surgical Laboratory, MGH & Shriners Hospitals, Harvard Medical School, Boston, MA, USA; 2Radiology, Athinoula A. Martinos Center for Biomedical Imaging, Boston, MA, USA; 3Computer Science, University of Ioannina, Ioannina, Greece; 4Medical Physics, University of Ioannina, Ioannina, Greece; 5Neurosurgery, Brigham and Womens Hospital, Harvard Medical School, Boston, MA, USA

We developed a novel approach that combines robust classification strategies with a 2D, solid-state, H1 HRMAS MRS method, TOBSY (TOtal Through-Bond SpectroscopY), which maximizes the advantages of HRMAS. We employed a linear Support Vector Machine (SVM) classifier combined with the powerful and robust minimum redundancy/maximum relevance (MRMR) feature-selecting method resulting in highly accurate classification. A robust classification approach and a sensitive multidimensional MRS technique at high magnetic fields should improve in vivo characterization, typing, and prognostication of brain tumors, and assist in stratifying patients for appropriate therapeutic protocols and for monitoring new therapies.