Synergistic PET and SENSE MR Image Reconstruction Using Joint Sparsity Regularization

Area of Science:

• Medical Imaging
• Image Reconstruction
• Regularization Techniques

Background:

• Conventional independent reconstructions of PET and MRI data limit image quality.
• Synergistic reconstruction aims to leverage complementary information for improved outcomes.
• Existing joint regularization methods like joint Total Variation (TV) have limitations in preserving common boundaries.

Purpose of the Study:

• To propose a generalized joint sparsity regularization prior and reconstruction framework for synergistic PET-MR image reconstruction.
• To improve image quality beyond conventional independent reconstruction methods.
• To enhance preservation and formation of common boundaries in PET-MR images.

Main Methods:

• Developed a novel non-convex joint sparsity regularization prior.
• Utilized an Alternating Direction Method of Multipliers (ADMM) optimization framework.
• Implemented alternating scaling for gradient vector distributions to address signal intensity dependency.

Main Results:

• The proposed joint sparsity prior significantly improved PET-MR image quality compared to unregularized and separately regularized methods.
• Demonstrated superior performance in edge preservation and detail recovery over joint TV and linear Parallel Level Sets (PLS) methods.
• Clinical data showed improved synergistic reconstruction but highlighted limitations due to current PET scanner resolution.

Conclusions:

• The proposed joint sparsity regularization within the ADMM framework is a promising technique for synergistic PET-MR reconstruction.
• The method effectively preserves common boundaries and recovers details impaired by noise and artifacts.
• Clinical applicability is currently limited by the lower resolution of PET images in existing scanners.