Revisiting stopping rules for iterative methods used in emission tomography.

Area of Science:

• Medical Imaging
• Computational Science

Background:

• Expectation maximization (EM) algorithm is standard for positron emission tomography (PET) sinogram reconstruction.
• Current methods often stop EM iterations before convergence, potentially affecting image quality.

Purpose of the Study:

• To design and validate a new, quantitative, and reproducible stopping rule for EM algorithm in PET image reconstruction.
• To improve computational efficiency and accuracy over existing methods.

Main Methods:

• Developed a novel stopping rule based on achieving the most probable emission density estimate.
• Validated the rule using Monte Carlo simulations with Poisson noise on sinogram data.
• Compared performance against the standard Pearson's χ(2) test stopping rule.

Main Results:

• The new stopping rule terminates iterations at the optimal solution for emission density estimation.
• Demonstrated superior accuracy in image reconstruction compared to the Pearson's χ(2) test.
• Showcased enhanced computational efficiency.

Conclusions:

• The proposed stopping rule offers a more accurate and efficient approach for PET image reconstruction.
• This method enhances the reliability of EM algorithm application in medical imaging.