Modeling Intershot Variability for Robust Temporal Subsampling of Dynamic, GABA-Edited MR Spectroscopy Data.

Area of Science:

• Neuroimaging
• Magnetic Resonance Spectroscopy
• Biophysics

Background:

• Quantifying GABA-edited MRS data is challenging due to variability in individual transients.
• This variability complicates analysis in functional MRS (fMRS) where discrete transient subsets are compared.
• Existing methods may not adequately address sources of variance like subject motion or spectral editing artifacts.

Purpose of the Study:

• To develop and validate a linear model for removing unwanted variance from GABA-edited MRS data.
• To preserve biologically relevant variance, such as metabolic responses to functional tasks.
• To improve spectral quality and quantification reliability in fMRS.

Main Methods:

• A linear model was applied to GABA-edited MRS data from 203 subjects (Big GABA collection).
• The model accounted for intrinsic, periodic, and movement-related variance, as well as spectral lineshape changes.
• Performance was benchmarked against uncorrected data and Spectral Improvement by Fourier Thresholding (SIFT), using synthesized functional task simulations.

Main Results:

• Composite models significantly improved signal-to-noise ratio (SNR) and reduced GABA+ estimate variability compared to uncorrected data.
• Individual model components showed varying performance, but composite models and individual components did not introduce bias.
• While SIFT reduced variance most effectively, it also reduced sensitivity to simulated functional changes.

Conclusions:

• The developed modeling approach effectively reduces unwanted variance in GABA-edited MRS data.
• The model retains sensitivity to temporal dynamics crucial for functional MRS applications.
• The study recommends incorporating this modeling approach into fMRS processing pipelines.