Fully Automatic Processing of In Vivo Proton Spectra and Spectroscopic Images

Magnetic resonance spectroscopic imaging (MRSI) provides simultaneous observation of MR spectra from multiple volumes over the observed region, and is a powerful tool for in vivo studies of metabolism. The quantification of MRSI data is widely used [ 1,2] in in vivo biomedical studies, however, traditional quantification methods of M R spectra data are time consuming and unsuitable for processing large data sets. Therefore, images of individual metabolite distributions are usually created by integrating over the regions of the M R spectrum at each voxel in space. Such methods are limited, due to low signal-to-noise ratio, spatially dependent shifts of resonance frequency, and overlap between metabolite resonances. The parametric estimation of MR spectra is capable of extracting the parameters of M R spectra from noisy data and generating better spectroscopic images. The results are more reliable and require less operator interaction than traditional spectral analysis methods. Automated parametric MR spectra estimation is imperative for routine processing of multidimensional MRSI without human involvement [3]. A fully automated spectral analysis method is proposed in this paper. This is used in combination with several additional processing steps to provide totally automated formation of l H metabolite images from the M R spectral data of human brain. The M R images obtained by the new method show great improvement compared with the images obtained by the traditional method. The new method could provide a very promising solution to massive routine processing of MRSI data for clinical and biomedical studies.