期刊名称：International Journal of Innovative Research in Computer and Communication Engineering
印刷版ISSN：2320-9798
电子版ISSN：2320-9801
出版年度：2014
卷号：2
期号：2
出版社：S&S Publications
摘要：To account for the Chemical Shift Displacement (CSD) artifact in the framework of unsupervised brain tumor tissue differentiation methods based on 2Dmagnetic resonance spectroscopic imaging (MRSI) data. The CSD object is specific to the localization protocol used for selecting a certain volume of interest inside the brain. For vendor - provided point - resolved spectroscopy (PRESS) localization of MRSI on 3T clinical scanners, the CSD object implies that spectra corresponding to voxels within ~2cm of the borders of the selected volume are distorted, which means that the peaks at outer spectral regions are heavily suppressed. This effect may be wrongly interpreted as lower concentration or lack of some metabolites in the outer regions of the localized volume and may also hinder the methods for unsupervised tissue different iation (e.g., non - negative matrix factorization). These methods can disentangle mixed tissue voxels in MRSI data acquired from brain tumors, and thus extract representative, tissue - specific spectra (called spectral sources), the spatial mixing coefficients and also the spectra assumed to be an approximate linear mixture of a few spectral sources. Here, we propose an approach to take the CSD artifact into account when evaluating the spatial distribution of tissue - specific spectral sources obtained from hiera rchical non - negative matrix factorization
