Deconvolution-based partial volume correction of PET images with parallel level set regularization.

Area of Science:

• Medical Imaging
• Biophysics
• Image Processing

Background:

• Partial volume effect (PVE) in Positron Emission Tomography (PET) degrades image quality due to limited spatial resolution.
• Magnetic Resonance (MR) imaging offers anatomical information crucial for Partial Volume Correction (PVC) methods.
• Existing MR-guided PVC methods often rely on segmented MR tissue maps and assume uniform PET activity within regions, limiting their application.

Purpose of the Study:

• To develop a novel post-reconstruction Partial Volume Correction (PVC) method for PET images.
• To incorporate anatomical information from MR imaging without requiring segmentation or uniformity assumptions.
• To improve the quantitative accuracy and image quality of PET scans.

Main Methods:

• A post-reconstruction PVC method based on deconvolution with parallel level set (PLS) regularization was developed.
• The method frames the problem as iterative deconvolution, incorporating anatomical guidance.
• An efficient algorithm using the split Bregman framework was implemented for non-smooth optimization, enabling application to 3D images.

Main Results:

• The proposed method demonstrated enhanced quantitative performance with realistic MR guidance.
• It effectively reduced image noise while preserving structural details in human PET data.
• The method showed potential for improved differentiation between amyloid-positive and amyloid-negative scans.

Conclusions:

• The developed MR-guided PVC method offers superior performance compared to existing techniques.
• It successfully addresses limitations of previous methods by avoiding MR segmentation and uniformity assumptions.
• The approach shows significant promise for enhancing clinical PET imaging applications.