Tensor Decomposition-Based Multi-Signal Matrix Pencil Method for Myelin Water Fraction Estimation

Area of Science:

• Magnetic Resonance Imaging
• Neuroimaging
• Biomedical Engineering

Background:

• Myelin water fraction (MWF) is a key biomarker for myelin integrity in the brain.
• Estimating MWF typically requires multi-echo gradient-recalled echo (mGRE) sequences.
• Limitations in echo-train length and spatial sampling can hinder MWF estimation accuracy.

Purpose of the Study:

• To develop a robust method for MWF estimation from mGRE data.
• To enable MWF estimation under acquisition constraints like limited echo-train length and higher spatial sampling.
• To improve the efficiency and accuracy of MWF quantification.

Main Methods:

• A tensor decomposition-based multi-signal matrix pencil (T-MP) framework was developed.
• The T-MP framework incorporates spatial information from neighboring voxels.
• The method reduces temporal sampling requirements, allowing stable parameter estimation with fewer echoes.

Main Results:

• Numerical simulations confirmed accurate MWF estimation with fewer temporal samples.
• In vivo experiments demonstrated consistent MWF maps across various spatial resolutions.
• The T-MP method showed improved estimation consistency in white and gray matter compared to voxel-wise fitting.
• Per-slice computation time was substantially reduced.

Conclusions:

• The T-MP method offers a robust approach for MWF estimation.
• It integrates spatial information while reducing temporal sampling needs.
• The framework supports spatially efficient mGRE acquisitions, offering improved robustness and computational efficiency.