Using spatial prior knowledge in the spectral fitting of MRS images

Area of Science:

• Medical Imaging
• Biophysics
• Computational Neuroscience

Background:

• Magnetic Resonance Spectroscopic Imaging (MRSI) is crucial for non-invasive metabolic profiling.
• Spectral fitting of MRSI data often faces challenges with parameter estimation accuracy and spatial resolution.
• Existing methods may struggle to resolve overlapping metabolite peaks, limiting diagnostic capabilities.

Purpose of the Study:

• To introduce a novel Bayesian smoothness prior for enhancing spectral fitting in MRSI.
• To improve the accuracy and spatial resolution of metabolic maps derived from MRSI data.
• To demonstrate the efficacy of the proposed prior in resolving complex spectral features and reducing parameter variance.

Main Methods:

• A Bayesian smoothness prior, incorporating a Gaussian Markov random field, was integrated into the spectral fitting process.
• A new optimization objective was formulated to encourage smooth parameter maps while preserving spatial details.
• The method was evaluated using simulated MRSI data and real patient data (brain tumor, varying SNR).

Main Results:

• The Bayesian smoothness prior significantly reduced the variance of estimated parameter maps, even surpassing the Cramér-Rao lower bound.
• The approach successfully resolved overlapping choline and creatine peaks in a brain tumor case where single-voxel methods failed.
• High-spatial-resolution, short-echo time (TE) MRSI at 3 Tesla yielded improved metabolic maps.
• The method demonstrated general benefit across various signal-to-noise ratios in long-TE brain MRSI data.

Conclusions:

• The proposed Bayesian smoothness prior is an effective tool for improving spectral fitting in MRSI.
• This method enhances the accuracy and reliability of metabolic maps, aiding in the diagnosis and study of neurological conditions.
• The approach offers a valuable advancement for both research and clinical applications of MRSI.