Simulation-based evaluation of OSEM iterative reconstruction methods in dynamic brain PET studies.

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Radiochemistry

Background:

• Dynamic Positron Emission Tomography (PET) data reconstruction commonly uses Filtered Backprojection (FBP).
• FBP, while fast and robust, yields poor image quality (noise, artifacts, low contrast) with low-count data.
• This image degradation can impact quantification of physiological parameters like Binding Potential (BP).

Purpose of the Study:

• To investigate the performance of iterative reconstruction methods (UW-OSEM, ANW-OSEM) versus analytical methods for low-count dynamic PET ligand-receptor studies.
• To compare the quantitative accuracy of iterative and analytical reconstruction techniques.
• To assess the impact of reconstruction methods on Binding Potential (BP) estimation.

Main Methods:

• Simulated [18F]MPPF dynamic PET acquisitions were used for assessment.
• Quantitative accuracy of iterative (UW-OSEM, ANW-OSEM) and analytical reconstruction methods was compared.
• Reconstruction of real PET data was performed to validate findings.

Main Results:

• Analytical methods demonstrated greater robustness with low-count data compared to iterative methods.
• Iterative methods, due to positivity constraints, overestimated activity in low-concentration regions (e.g., cerebellum).
• This overestimation led to significant bias in Binding Potential (BP) estimates derived from iterative reconstructions.

Conclusions:

• Analytical reconstruction methods offer more reliable quantification for low-count dynamic PET studies.
• Iterative reconstruction methods, specifically MLEM-based algorithms, introduce bias in BP quantification.
• Filtered Backprojection (FBP) remains a more robust choice for quantitative analysis in challenging low-count dynamic PET scenarios.