一种基于全局最小乘方法的磁共振波谱目标频率选择的量化方法

One Quantitative Method Based on Total Least Squares for Frequency-Selective Magnetic Resonance Spectroscopy

磁共振波谱中,感兴趣成分的单独量化不但符合许多实际应用的要求,而且同时也提高了量化的速度和精度.为此本文提出了一种基于频率选择的波谱信号量化方法,该方法通过全局最小乘方法来估计波谱信号的非线性参数,又结合前向和后向的APES方法来估计其线性参数.文中的模拟和活体数据实现表明,本文方法在计算速度和参数估计的稳健性上要明显优于经典的HSVD和HTLS方法,同时其综合性能比Stoica等提出的频率选择方法SELF-SVD还好.

Frequency-selective quantification can not only apply to many applications of magnetic resonance spectroscopy （MRS） but also improve the speed and accuracy of parameters estimation of selected spectral components. Thus, we presented a new frequency-selective method for MRS quantification, which estimated the nonlinear and linear parameters of the selected components baaed on total least squares and forward-backward APES respectively. From the experiments on both simulate data and MRS data in vivo, we can see that the computing speed and estimation stabilization of our method are obviously superior to those of the classical quantitative methods HSVD and HTLS;furthennore, its performance of parameters estimation is better than that of the frequency-selective method SELF_SVD proposed by Stoica et al.