Meeting Banner
Abstract #3368

Supervised Non-Negative Matrix Factorization Based Classification of Multiparametric MR Imaging of Gliomas at 3T

Fusun Citak Er 1 , Zeynep Firat 2 , Basar Sarikaya 2 , Ugur Ture 3 , and Esin Ozturk-Isik 4

1 Department of Genetics and Bioengineering, Yeditepe University, Istanbul, Turkey, 2 Department of Radiology, Yeditepe University Hospital, Istanbul, Turkey, 3 Department of Neurosurgery, Yeditepe University Hospital, Istanbul, Turkey, 4 Department of Biomedical Engineering, Yeditepe University, Istanbul, Turkey

This study aims to evaluate the performance of non-negative matrix factorization (NMF) for supervised classification of brain tumor grade using quantitative multiparametric MR imaging at 3T. Fractional anisotropy, cerebral blood volume, mean transit time, cerebral blood flow, apparent diffusion coefficient and peak height ratios of N-acetyl aspartate over creatine (NAA/Cr) and choline over creatine (Cho/Cr) of thirty newly diagnosed glioma patients were calculated, and used as predictors for classification of tumor grade. NMF results were compared with k-nearest neighbor (kNN) algorithm. This study showed that non-negative matrix factorization performed better than kNN in glioma grading using multiparametric MRI at 3T.

This abstract and the presentation materials are available to members only; a login is required.

Join Here