A fast Bayesian reconstruction algorithm for emission tomography with entropy prior converging to feasible images

Area of Science:

• Medical Imaging
• Computational Science

Background:

• Iterative reconstruction algorithms are crucial for accurate emission tomography.
• Bayesian statistical concepts offer a robust framework for image reconstruction.

Purpose of the Study:

• To develop and test a novel iterative reconstruction algorithm for emission tomography.
• To leverage Bayesian statistical concepts, specifically entropy priors, for enhanced image quality.

Main Methods:

• Developed an iterative reconstruction algorithm using entropy as a prior distribution.
• Implemented the fast maximum a posteriori with entropy (FMAPE) algorithm via successive substitution.
• Utilized adjustable parameters for uniform convergence and contrast control.

Main Results:

• The algorithm converges uniformly to feasible images, consistent with initial data.
• Demonstrated the effectiveness of the entropy prior and adjustable parameters.
• Identified a uniform field as the optimal initial image for maximum likelihood estimator (MLE) and FMAPE algorithms without prior knowledge.

Conclusions:

• The developed FMAPE algorithm offers an efficient and accurate method for emission tomography reconstruction.
• Bayesian approaches with entropy priors enhance image fidelity and consistency.
• Standardized initial image selection improves iterative reconstruction outcomes.