2D spectral-temporal fitting of functional MRS improves the fitting precision and noise robustness

Abstract

Functional magnetic resonance spectroscopy (fMRS) is a powerful technique for detecting endogenous neurochemical changes in the brain over time. However, its widespread application is hindered by the inherently low signal-to-noise ratio (SNR) of fMRS data, leading to low temporal resolution, long acquisition time, and the need for large cohort sizes. A promising approach to overcoming these limitations is two-dimensional (2D) spectral-temporal fitting. Recent studies have demonstrated that 2D fitting improves quantification precision, enabling a reduction in cohort size. Building on these findings, this study investigates the robustness of 2D fitting against noise, demonstrating its potential for reliable quantification even in low-SNR data. This advancement enables the acquisition of fewer transients per spectrum, thereby enhancing temporal resolution and reducing acquisition time. We implemented a 2D spectral-temporal fitting framework for fMRS and evaluated its performance across synthetic and in vivo datasets. Two synthetic datasets and a previously published in vivo dataset were employed to assess noise robustness and generalizability. The results indicate that 2D fitting improves fitting precision and noise robustness across both types of data, suggesting its potential to improve temporal resolution and decrease acquisition time in fMRS studies. When combined with reduced cohort sizes, 2D spectral-temporal fitting could boost the sensitivity of fMRS, facilitating its broader adoption in neuroscience research.