Optimization of Q.Clear reconstruction for dynamic ¹⁸F PET imaging

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Quantitative PET

Background:

• Q.Clear, a Bayesian penalized likelihood algorithm, enhances PET quantitation accuracy.
• The algorithm uses a β penalization factor to control noise during iterative reconstruction.

Purpose of the Study:

• Determine the optimal β-factor for accurate quantitation in dynamic PET scans.
• Evaluate the impact of β-factors on pharmacokinetic analysis in clinical PET data.

Main Methods:

• Utilized a PET phantom with varying sphere sizes and acquisition times.
• Reconstructed data using Q.Clear with 18 β-factors (100-3500).
• Assessed recovery coefficient, coefficient of variation, and RMSE in phantom data; calculated AUC and K_i in patient data.

Main Results:

• Increasing β-factor generally decreased recovery coefficient and variability.
• Low β-factors led to overestimation in small spheres, while high β-factors caused underestimation.
• Higher β-factors increased K_i values in clinical data, with a 25.5% change observed between β 300 and 1000.

Conclusions:

• The optimal β-factor balances accuracy and precision in dynamic PET imaging.
• A β-factor of 300-500 is recommended for quantitating small structures.
• Larger structures may benefit from higher β-factors for improved quantitation.