Fast unsupervised hot-spot detection in 1 H-MR spectroscopic imaging data using ICA Markus T Harz 1 ,Volker Diehl 2 , Bernd Merkel 1 , Burkhardt Terwey 2 , and Heinz-Otto Peitgen 1 1 Fraunhofer MEVIS, Center for Medical Image Computing, Bremen, Germany 2 MR and PET Center Bremen Mitte, Bremen, Germany ABSTRACT Independent Component Analysis (ICA) is a blind source separation technique that has previously been applied to various time-varying signals. It may in particular be utilized to study 1 H-MR spectroscopic imaging (MRSI) data. The work presented firstly investigates preprocessing and parameterization for ICA on simulated data to assess different strategies. We then ap- plied ICA processing to 2D/3D brain and prostate MRSI data obtained from two healthy volunteers and 17 patients. We con- ducted a correlation analysis of the mixing and separating matrices resulting from ICA processing with maps obtained from metabolite quantitations in order to elucidate the relationship between quantitative and ICA results. We found that the mixing matrices corresponding to the estimated independent components highly correlate with the metabolite maps for some cases, and for others differ. We provide explanations and speculations for that and propose a scheme to utilize the knowledge for hot-spot detection. From our experience, ICA is much faster than the calculation of metabolic maps. Additionally, water and lipid contaminations are on the way removed from the data; the user needs not manually exclude spectroscopic voxels from processing or analysis. ICA results show hot spots in the data, even where quantitation-based metabolic maps are difficult to assess due to noisy data or macromolecule distortions. Keywords: MR spectroscopy, spectroscopic imaging, MRSI, chemical shift imaging, CSI, Independent Component Analysis, ICA, unsupervised analysis 1. INTRODUCTION Blind source separation techniques have initially been devised for speech analysis. The ICA procedure was then devel- oped and has successfully been applied e.g. to fMRI data 4,5,10 , EEG data 25 , and several other domains 16,17,31 . We suggest to apply it to 1 H MR spectroscopic imaging (MRSI) data 15,27 . The purpose of the work presented is twofold: firstly, to assess various processing options, which to the best of our knowledge has not been described before, in order to set up a processing procedure and parameterization for ICA on 2D and 3D MRSI data. Secondly, we elaborate a procedure to qualitatively and quantitatively assess hot-spot detection results for in vivo data. A clinical validation of the approach, however, is out of scope of the present work. It will be conducted subsequently, based on the results provided herein. 2. METHODOLOGY 2.1 Post-processing of spectroscopic data All in vivo data was preprocessed with the Siemens Syngo Spectroscopy software 27 according to a defined protocol that encompasses time-domain filtering, zero-filling, and water referencing. Spectroscopic data was Fourier transformed only in the spatial directions. The data is of different lengths, ranging from 512 to 2048 complex data points in the time domain (free induction decay data; FIDs). Metabolic maps were then calculated for choline (Cho) and n-acetylaspartate (NAA) for brain data, and Cho and citrate (Ci) for prostate data. The raw FID data was also exported for custom post- processing and ICA. Medical Imaging 2009: Image Processing, edited by Josien P. W. Pluim, Benoit M. Dawant, Proc. of SPIE Vol. 7259, 72591X · © 2009 SPIE CCC code: 1605-7422/09/$18 · doi: 10.1117/12.808122 Proc. of SPIE Vol. 7259 72591X-1