Evaluation of quantitative, efficient image reconstruction for VersaPET, a compact PET system.

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Reconstruction

Background:

• Developed a high-resolution positron emission tomography (PET) system, VersaPET, with a geometry prone to parallax blurring.
• Existing reconstruction methods often use idealized system geometries, potentially limiting image quality.

Purpose of the Study:

• To evaluate a graphic processing unit (GPU)-accelerated maximum-likelihood by expectation-maximization (MLEM) image reconstruction framework for VersaPET.
• To incorporate accurate system geometry and projection space point-spread-function (PSF) modeling into the reconstruction process.

Main Methods:

• Combined STIR's ray-tracing with VersaPET's exact geometry for the system matrix.
• Utilized custom Monte Carlo simulations for PSF modeling (crystal penetration, scattering).
• Parallelized reconstruction on GPU, leveraging system symmetry for PSF computation; tested various PSF kernels.

Main Results:

• Accurate geometry reconstruction improved image quality over idealized STIR.
• PSF modeling enhanced contrast recovery and reduced bias in hot-sphere phantoms compared to no PSF.
• PSF modeling improved contrast recovery and quantification for 5-10 mm cold spheres.
• Real phantom data showed optimal PSF kernel choice depends on imaging task and may require further investigation.

Conclusions:

• A practical MLEM reconstruction framework for VersaPET was fully evaluated.
• Different PSF kernels can optimize results for specific imaging tasks.
• Further research is needed to understand variations in optimal PSF kernels between simulation and real-world studies.