Analysis and improvement of quantification algorithms for magnetic resonance spectroscopy

Magnetic Resonance Spectroscopy (MRS) is a technique used in fundamental research and in clinical environments. During recent years, clinical application of MRS gained importance, especially as a non-invasive tool for diagnosis and therapy monitoring of brain and prostate tumours. The most important asset of MRS is its ability to determine the concentration of chemical substances non-invasively. To extract relevant signal parameters, MRS data have to be quantified. This usually doesn?t prove to be straightforward since in vivo MRS signals are characterized by poor signal-to-noise ratios, overlapping peaks, acquisition related artefacts and the presence of disturbing components (e.g. residual water in proton spectra). The work presented in this thesis aims to improve the quantification in different applications of MRS in vivo. To obtain the signal parameters related to MRS data, different approaches were suggested in the past. Black-box methods, don?t require user interaction, but deliver statistically suboptimal parameter estimates. Interactive methods, such as AMARES, fit a model function to the MRS data by optimizing parameter estimates using a non-linear least squares minimization method. Incorporating information about mutual parameter relations (so called prior knowledge) results in more accurate parameter estimates. This principle, initially used to constrain parameters within the spectrum of one volume element voxel was extended to the quantification of multivoxel (CSI or MRSI) data, incorporating prior knowledge between different voxels. Multivoxel processing was shown to improve the parameter estimation, depending on the prior knowledge used. The need for automated processing protocols emerges, when large CSI datasets are to be processed. An aspect of automating CSI processing is the choice of the number of resonances to model. Because the number of spectral components can vary through a CSI, choosing a fixed number of resonances could introduce errors in the parameter estimation. By investigating the influence of under- and overmodeling, it was shown that only slight undermodeling had a negative effect on estimation accuracy. MRSI can be used as diagnostic tool in prostate cancer. The performance of two quantification methods was compared in diagnosing prostate cancer. To this end, simulated, in vitro and in vivo MRS prostate data were quantified in both time and frequency domain and the accuracy of the outcomes of both methods were compared. It was found that a Time Domain Fitting algorithm (based on AMARES) and a Frequency Domain Analysis method performed equally well, with a slightly better performance of the Time Domain Fitting algorithm on well resolved spectra. A method optimized for the quantification of short echo time MRS data is presented. Short echo time spectra require a special approach to deal with the baseline and the high complexity of the spectral patterns. To use as much prior knowledge as possible, a database of measured metabolite model spectra is used. A two-step method is suggested which increases robustness of the quantification. To bring advanced signal processing algorithms to the clinical environment, a GUI was developed to display both Magnetic Resonance Images and MRSI data. The GUI incorporated preprocessing and time domain quantification using AMARES.