The Magnetic Resonance Signal Processing Method Based On The Characteristics Of Decomposition

As the only measurement, functional magnetic resonance imaging (fMRI) with highly spatial resolution and no-invasion is mainly based on blood oxygenation level dependent (BOLD), and has the potential to localize tiny foci and detect wreck signal fluctuate. Effective and robust algorithms are crucial to catch this goal.In this thesis, three algorithms to detect active sources have been studied based on singular value decomposition. The main comments are shown as follows:1. A novel modified principle component analysis algorithm is proposed and applied to detect glioma active sources. In this algorithm, introduce time delay to depress noise, and initially propose a weighted correlation coefficient and conduct statistic test on it. The validity is confirmed by computer stimulation, and the results of localizing glioma active sources are conformed to the clinical diagnosis and the significance of physiology.2. A united SPM-ICA-PCA algorithm is proposed. In this algorithm, an active strength formula is defined to index the relative active intensity; a criterion is provided to select independent components, which still is an open issue. What's more, the novel algorithm extends the SPM application score from stimulus state to involve non-stimulus state. The SPM is widely used to analysis fMRI data. Simulated cases and the real glioma data are investigated to demonstrate the feasibility and validity of the novel algorithm, respectively.3. Based on the fMRI theory and MUSIC algorithm, find that the inherent connection is phase difference that bridges fmri signal processing and MUSIC algorithm, and also, deduce the orientation vector, which plays an important role for the performance of the MUSIC algorithm for fMRI signal...