An Objective Method for Automated Classification of Brain Tumors using Proton MR Spectroscopy

Yu Zhang¹, Sanjeev Chawla¹, Sumei Wang¹, Sangeeta Chaudhary¹, Jaroslaw Krejza¹, E. R. Melhem¹, Harish Poptani¹

We investigated whether LC Model and linear discriminant analysis of the 1H MRS data from the contrast enhancing and peri-tumoral regions of the tumor can be used as an objective method for better classification of brain tumors. Linear discriminant analysis correctly classified 85% of 138 patients from six different brain tumor subtypes. Classification accuracies of glioblastomas versus metastases and astrocytomas versus oligodendrogliomas were 83% and 94% respectively. In comparison to previous 1H MRS studies, our study proposes a fully automated 1H MRS data analysis approach with minimum operator intervention and high diagnostic accuracy in brain tumor classification.