Regularized Joint Reconstruction and Slab Combination for Accelerated Three-Dimensional Multi-Slab Diffusion-Weighted...

Area of Science:

• Medical Imaging
• Magnetic Resonance Imaging (MRI)
• Image Reconstruction

Background:

• Diffusion MRI is crucial for neuroimaging.
• 3D multi-slab acquisitions accelerate data acquisition but introduce slab-boundary artifacts.
• Existing methods struggle to preserve anatomical detail while correcting artifacts.

Purpose of the Study:

• To develop a method for joint reconstruction of high-resolution diffusion-weighted volumes.
• To eliminate slab-boundary artifacts in undersampled 3D multi-slab k-space acquisitions.
• To preserve fine anatomical detail during accelerated MRI reconstruction.

Main Methods:

• Formulated a bilinear forward model for 3D multi-slab acquisition.
• Employed a maximum a posteriori framework with a CNN-based deep energy prior.
• Utilized alternating minimization to solve the non-convex optimization problem.

Main Results:

• The Energy-based Profile Encoding (EPEN) framework significantly reduced slab-boundary artifacts.
• Achieved improved structural consistency and contrast preservation.
• Demonstrated effectiveness across various acceleration factors and slab configurations.

Conclusions:

• EPEN enables robust joint 3D multi-slab diffusion MRI reconstruction.
• The method effectively corrects slab profiles using deep energy-based priors.
• EPEN offers a promising solution for high-quality diffusion MRI.