Quantification in magnetic resonance spectroscopy based on semi-parametric approaches.

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Spectroscopy

Background:

• Magnetic Resonance Spectroscopy (MRS) enhances Magnetic Resonance Imaging (MRI) for disease diagnosis and monitoring.
• High magnetic fields, hyperpolarized nuclei, and short echo-time acquisition significantly boost MRS potential.
• Accurate quantification of metabolite concentrations in complex spectra remains a significant challenge.

Purpose of the Study:

• To review and discuss semi-parametric techniques for quantifying short echo-time proton MRS spectra.
• To evaluate methods based on their handling of macromolecule signals and metabolite signal decay (lineshape).
• To detail error estimation and real-world application compromises, focusing on the bias-variance trade-off.

Main Methods:

• Review of semi-parametric quantification approaches: QUEST, AQSES, TARQUIN, LCModel, and SiToolsFITT.
• Analysis of methods' performance in managing macromolecule signals and spectral lineshape.
• Examination of noise-related error estimation and bias-variance trade-offs in practical applications.

Main Results:

• Semi-parametric methods are essential due to incomplete model functions in MRS signal analysis.
• The review highlights differences in handling spectral complexities among the discussed quantification techniques.
• QUEST and AQSES are specifically applied to quantify MRS, HRMAS, and MRSI data.

Conclusions:

• Semi-parametric methods provide robust solutions for challenging MRS spectral quantification.
• Understanding the bias-variance trade-off is crucial for reliable metabolite concentration estimation.
• The reviewed methods offer valuable tools for advancing MRS applications in research and diagnostics.