Optimization of penalization function in Bayesian penalized likelihood reconstruction algorithm for [¹⁸F]flutemetamol...

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Radiochemistry

Background:

• Point-spread-function (PSF) correction in amyloid PET imaging can cause Gibbs artifacts.
• Bayesian Penalized Likelihood (BPL) reconstruction without PSF correction reduces artifacts but may lower image contrast.

Purpose of the Study:

• To optimize time-of-flight (TOF) BPL reconstruction parameters for amyloid PET imaging without PSF correction.
• To recover image contrast lost during artifact reduction.

Main Methods:

• Phantom studies evaluated reconstruction parameters (iterations, β, γ factors) for TOF-BPL without PSF correction.
• Clinical studies with 71 participants using [¹⁸F]flutemetamol assessed image quality, standardized uptake value ratios (SUVr), and Centiloid scales (CL).

Main Results:

• Image contrast and coefficient of variation were independent of iteration count.
• A trade-off was observed between γ and β factors, leading to candidate parameters for clinical testing.
• Optimal parameters without PSF correction were identified as γ factor 10, 1 iteration, and β 800.

Conclusions:

• Optimized TOF-BPL reconstruction without PSF correction effectively recovers contrast in amyloid PET imaging.
• The identified parameters ensure stable SUVr and CL values, enabling reliable amyloid PET discrimination.
• The study provides optimal reconstruction settings for [¹⁸F]flutemetamol PET imaging.