A Novel Method in PET Image Reconstruction Using MRI Anatomical Priors

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Radiology

Background:

• Current whole-body PET/MR spatial resolution is limited to 3-4 mm.
• Higher-resolution MRI can improve PET reconstruction but is sensitive to anatomical mismatches.
• Mismatches between MRI priors and actual tracer distribution can compromise PET imaging accuracy.

Purpose of the Study:

• To introduce a novel PET reconstruction method, Magnetic Resonance-guided Block Sequential Regularized Expectation Maximum (MRgBSREM).
• To develop a PET reconstruction technique robust to mismatches between anatomical MRI priors and true tracer distribution.
• To evaluate the performance of MRgBSREM in diverse clinical settings and with various tracers.

Main Methods:

• Developed and applied the Magnetic Resonance-guided Block Sequential Regularized Expectation Maximum (MRgBSREM) algorithm.
• Evaluated MRgBSREM in clinical studies using ¹⁸F-florbetaben (dementia), ¹⁸F-FDG (stroke), ¹⁸F-NaF (knee), and ¹⁵O-water (Moyamoya disease).
• Assessed improvements in spatial resolution, image quality, and robustness to MR-tracer distribution mismatches.

Main Results:

• MRgBSREM visually enhanced spatial resolution and image quality across all evaluated studies and tracers.
• The method effectively reduced white-matter spill-in and gray-matter spill-over in ¹⁸F-florbetaben PET imaging.
• Demonstrated robustness to mismatches between MR priors and actual tracer distribution.

Conclusions:

• MRgBSREM offers improved PET spatial resolution and image quality, potentially enhancing diagnostic accuracy.
• The novel reconstruction method is robust to anatomical mismatches, a common challenge in PET/MR imaging.
• MRgBSREM shows promise for more accurate quantitative analysis in various clinical applications.