MR Elastography With Optimization-Based Phase Unwrapping and Traveling Wave Expansion-Based Neural Network (TWENN).

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Soft Tissue Mechanics

Background:

• Magnetic Resonance Elastography (MRE) is crucial for characterizing soft tissue biomechanics in disease diagnosis and treatment planning.
• Challenges in MRE include complex wavefields and noise, hindering accurate displacement extraction and modulus estimation.

Purpose of the Study:

• To develop an advanced pipeline for processing MRE images, improving accuracy in displacement extraction and biomechanical property estimation.
• To address noise and wavefield complexity issues in MRE data.

Main Methods:

• Proposed a pipeline combining optimization-based displacement extraction with a Traveling Wave Expansion-based Neural Network (TWENN) for modulus estimation.
• Utilized an objective function with Dual Data Consistency (Dual-DC) for accurate phase unwrapping and displacement extraction.
• Developed a complex-valued neural network for complex wavenumber estimation, trained on simulated MRE data with varying noise levels.

Main Results:

• The pipeline successfully estimated biomechanical properties with minimal root-mean-square errors on simulated brain and liver images, outperforming state-of-the-art methods.
• Processing of phantom, brain, and liver MRE images revealed clear anatomical features and demonstrated robustness to noise.
• The method showed good generalizability across different MRE datasets.

Conclusions:

• The proposed MRE processing pipeline offers a robust and accurate method for estimating soft tissue biomechanical properties.
• This approach has significant potential for improving disease diagnosis and treatment planning through enhanced MRE analysis.
• The pipeline's noise resilience and generalizability make it a valuable tool for clinical and research applications.