Analysis of penalized likelihood image reconstruction for dynamic PET quantification

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Biophysics

Background:

• Dynamic positron emission tomography (PET) is vital for quantifying physiological and biochemical processes.
• Image reconstruction significantly influences the accuracy of kinetic parameter estimation from time activity curves.
• Penalized likelihood reconstruction is a common method with adjustable parameters.

Purpose of the Study:

• To analyze the impact of penalized likelihood image reconstruction on tracer kinetic parameter estimation in dynamic PET.
• To derive theoretical expressions for bias, variance, and mean squared error of kinetic parameters.
• To develop a practical method for selecting optimal regularization parameters.

Main Methods:

• Theoretical analysis of bias, variance, and ensemble mean squared error.
• Computer simulations to validate theoretical predictions.
• Derivation of formulae relating reconstruction parameters to kinetic parameter statistics.

Main Results:

• Penalized likelihood reconstruction parameter choice significantly affects kinetic parameter estimation accuracy.
• Theoretical formulae accurately predict changes in bias, variance, and MSE with regularization.
• Simulations confirm the impact of regularization on tracer kinetic analysis.

Conclusions:

• Proper selection of regularization parameters in penalized likelihood reconstruction is essential for accurate dynamic PET studies.
• The developed theoretical framework and practical method can guide optimal parameter selection.
• This work enhances the reliability of quantitative analysis in dynamic PET imaging.