Bayesian reconstruction of photon interaction sequences for high-resolution PET detectors

Area of Science:

• Medical Imaging
• Nuclear Physics
• Computational Science

Background:

• High-resolution positron emission tomography (PET) systems require precise event positioning.
• Accurate positioning relies on identifying the first interaction of annihilation photons.
• Photon multiple-interaction events (PMIE) complicate accurate positioning in PET.

Purpose of the Study:

• To evaluate a Bayesian estimation algorithm for accurate photon interaction positioning in high-resolution PET.
• To assess the impact of the algorithm on reducing mispositioned events and improving image quality.
• To compare the algorithm's performance against simpler positioning methods.

Main Methods:

• Utilized Bayesian estimation, specifically a maximum a posteriori (MAP) algorithm.
• Simulated a high-resolution PET system using cadmium zinc telluride detectors.
• Reconstructed the complete sequence of gamma-ray interactions for each photon.
• Quantified improvements in point-spread function, contrast, and spatial resolution.

Main Results:

• 93.8% of recorded coincidences in simulations involved at least one photon multiple-interactions event (PMIE).
• The MAP estimate accurately determined the first interaction for 85.2% of single photons.
• A two-fold reduction in mispositioned events was achieved compared to the minimum pair distance method.
• MAP estimation resulted in a point-spread function with lower tails and a higher peak value.

Conclusions:

• Bayesian estimation with MAP provides a robust framework for accurate photon interaction reconstruction in high-resolution PET.
• The developed algorithm significantly improves event positioning, reducing mispositioned events and enhancing image quality.
• This approach offers substantial gains in contrast and spatial resolution, advancing PET imaging capabilities.