Automated quantification of a weak signal in Magnetic Resonance Spectroscopy data

The quantification of Magnetic Resonance Spectroscopy (MRS) signal remains challenging due to the low signal-to-noise ratio (SNR) of data. All time-domain quantification methods highly require user interactions, which reduce the reproducibility of the data quantification. The goal of our work is to design a systematic methodology for automated quantification of a weak signal in MRS data. We used Hankel Singular Value Decomposition (HSVD) algorithm in our signal estimation step, along with frequency selective preprocessing step using ER-filter to improve the computational efficiency. On the model order selection problem of HSVD, we investigated two strategies for optimal choice of model order K: 1) Gaussian line-shape fitting (GLF) method; 2) Kurtosis analysis (KA) method. The detection and estimation performance of both methodologies have been evaluated in terms of detection rate and relative root mean square error (RRMSE), in comparison to traditional strategies with fixed model order. The synthesis and semi-synthesis simulation results both show that GLF and KA outperform the traditional K fixed methods. Moreover, GLF works the best among all, and its performance is worsened when the signal damping factor increases.