Anatomy-guided PET reconstruction usingl₁bowsher prior.

Area of Science:

• Medical Imaging
• Image Reconstruction
• Radiophysics

Background:

• Simultaneous positron emission tomography/magnetic resonance imaging (PET/MRI) is advancing PET image reconstruction.
• Existing anatomy-guided methods, like Bowsher's, can over-smooth fine details.

Purpose of the Study:

• To develop an improved Bowsher prior for PET/MRI image reconstruction.
• To overcome the over-smoothing limitations of the original second-order smoothing prior.

Main Methods:

• Proposed a novel Bowsher prior using the l1-norm and an iterative reweighting scheme.
• Derived a closed-form solution for iterative image reconstruction with the non-smooth prior.
• Compared the proposed l1-Bowsher prior with the original l2-Bowsher prior using simulations and real human data.

Main Results:

• The l1-Bowsher prior improved detection of small lesions with abnormal PET uptake.
• Enhanced contrast between tumors and surrounding tissues due to edge-preserving properties.
• Demonstrated lower bias and reduced hyper-parameter dependency in regions with matched PET/MRI boundaries.

Conclusions:

• The proposed l1-Bowsher prior methods significantly improve PET image quality in PET/MRI.
• These methods offer better lesion detectability and contrast, crucial for accurate diagnosis.
• The approach effectively utilizes anatomical information from MRI to refine PET reconstruction.