MRS-Sim: Open-Source Framework for Simulating In Vivo-Like Magnetic Resonance Spectra

Area of Science:

• Medical Physics
• Biophysics
• Data Science

Background:

• Realistic synthetic data is crucial for developing and validating magnetic resonance spectroscopy (MRS) methods.
• Current simulation tools lack the ability to model complex in vivo scenarios accurately.
• Inconsistencies in MRS protocols hinder reproducibility and accessibility.

Purpose of the Study:

• To introduce MRS-Sim, an open-source framework for simulating realistic in vivo-like MRS data.
• To provide a unified platform addressing inconsistencies in current MRS simulation protocols.
• To enhance the development, validation, and reproducibility of MRS methods.

Main Methods:

• Developed a modular simulation framework incorporating novel physical models for B0 field inhomogeneity and semi-parametric signal generation.
• Integrated tools for analyzing in vivo data fitting parameters to tailor simulations.
• Included supplemental information to assist non-expert users in MRS data simulation.

Main Results:

• MRS-Sim can simulate diverse MRS data, from raw transients to processed averages, including complex B0 field variations and baseline artifacts.
• The framework enables tailoring simulations to specific datasets by analyzing real in vivo data.
• Facilitates verification of spectral fitting protocols and reproducibility analyses.

Conclusions:

• MRS-Sim promotes reproducibility and accessibility in magnetic resonance spectroscopy research.
• The framework supports various MRS applications, including deep learning research requiring large datasets.
• Its modularity and community-driven development foster customization for diverse in vivo scenarios.