Analysis and comparison of two methods for motion correction in PET imaging

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Reconstruction

Background:

• Positron emission tomography (PET) is a crucial imaging modality.
• Motion artifacts significantly degrade PET image quality and quantitative accuracy.
• Existing motion correction methods lack sufficient comparative evidence for practical application.

Purpose of the Study:

• To characterize and compare the performance of two primary PET motion correction techniques: reconstruct-transform-average (RTA) and motion-compensated image reconstruction (MCIR).
• To evaluate convergence properties and image quality metrics of RTA and MCIR.

Main Methods:

• Both RTA and MCIR were implemented using the ordered subsets expectation maximization (OSEM) algorithm.
• Motion data was derived from dynamic MRI of a human volunteer.
• PET data was simulated using the dynamic MRI information, and multiple realizations were used for statistical assessment.

Main Results:

• MCIR accurately recovered true regional values, whereas RTA exhibited significant bias due to interpolation errors and limited count statistics.
• RTA produced low and stable noise, while MCIR noise increased with iterations.
• MCIR, when combined with post-filtering for noise reduction, achieved up to 42% lower Mean Squared Error (MSE) compared to RTA.

Conclusions:

• MCIR demonstrates superior performance over RTA, particularly when noise is effectively managed through post-filtering.
• RTA remains a viable, simpler option for PET motion correction in scenarios where precise quantification is not the primary goal.