Advanced spectral quantification: Parameter handling, nonparametric pattern modeling, and multidimensional fitting

This article reviews spectral fitting topics that go beyond the time-or frequency-domain fitting methods described in Quantifying spectra in the frequency domain and Time-Domain Methods for Quantifying MR Spectra. First, a detailed review on different approaches for nonparametric modeling of line-shape and baseline distortions as well as unknown spectral components by pattern fitting and principle component analysis, is provided. Next, spectral fitting approaches are extended to multidimensional spectroscopy data that result from two-dimensional MRS (magnetic resonance spectroscopy), spectroscopic imaging, dynamic parameter studies, and functional spectroscopy experiments. In addition, principal component analysis and Bayesian approaches to spectral fitting are introduced and compared to the more established least-squares optimization methods. Finally, validation methods for testing the accuracy and precision of spectral fitting and quantification algorithms are described.