A Research On The Processing And Analyzing Method For In Vivo MRS Signal

According to the state that the great value of Magnetic Resonance Spectroscopy (MRS) can not be brought into full play in realities, the study of methods for in vivo MRS signal processing and analysis has been carried out, which aims at putting forward some more effective methods on the basis of existing methods and creating the essential condition for the application of MRS.There are six chapters in this paper. The first chapter gives the main ideas of this paper, i.e., the background and meaning of this paper, the development state about this research at home and abroad, and our main research work. The last chapter summarizes the research work in this paper and introduces our future research work. The four other chapters are the main contents of this paper, which include the principle and application of MRS, the non-interest components suppression methods, the parameters correction methods and the parameters estimation methods of MRS signal, etc.In the second chapter, the principle and application of MRS is discussed comprehensively, thus the complex situation of in vivo MRS signal can be known and the meaning of our research in this thesis can be understood.The third chapter studies different non-interest peaks suppression methods for in vivo MRS. For noise suppression, one noise adaptive suppression method by wavelet analysis is presented and an integrating method combined with apodization method, average method and wavelet analysis method is introduced; As to water suppression, one residual water suppression technique by quick power method is proposed.In the fourth chapter, parameters correction is studied. A new phase correction method, which is based on minimum information entropy, and nonlinear parameters of frequency and damping factor correction method, which takes advantage of complex principal component analysis, are proposed. The experiments on both simulate data and in vivo MRS data show that both methods have better performance.The fifth chapter, the key part of the paper, investages how to estimate the parameters of MRS signal. Through studying several classical methods, two quantification methods are presented in this part. At first, according to the fact that the interesting resonance peaks usually are part of the whole spectrum peaks, an effective frequency-selective quantification method, which is based on total least squares and forward-backward adaptive filtering approach, is put forward. In turn, another better quantification method based on decimation and prior knowledge is presented. As a result, the problem of overlapped resonance peaks quantification can be better overcome.