Combining HR-MAS and In Vivo MRI and MRSI Information for Robust Brain Tumor Recognition'

In this study we propose to classify short echotime brain MRSI data by using multimodal information coming from magnetic resonance imaging (MRI), magnetic resonance spectroscopic imaging (MRSI) and high resolution magic angle spinning (HR-MAS), and to develop an advanced pattern recognition method that could help clinicians in diagnosing brain tumors. We study the impact of using HR-MAS information in combination with in vivo information for classifying brain tumors and we investigate which parameters influence our classification results. To integrate HR-MAS, MRSI and MRI information a harmonization of all the input spaces is required due to the fact that we have to manage the use of very different information/ data, obtained with different measurement techniques, as well as the use of data coming from different clinical centers. The problem is overcome by extracting common characteristic features from all the different data types. The pattern recognition technique used in this study is Canonical Correlation Analysis (CCA), a statistical method developed to assess the relation between two sets of variables. The method has recently been successfully applied to prostate and brain data and is able to simultaneously exploit the spectral as well as the spatial information characterizing the MRSI data. Here, the performance of CCA when making use of different feature vector approaches is analyzed and compared. © 2009 Springer Berlin Heidelberg.

ISBN:9783540892076

Publication year:2008

BOF-keylabel:yes

IOF-keylabel:yes

Authors from:Higher Education