A generalised spatio-temporal registration framework for dynamic PET data: application to neuroreceptor imaging.

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Pharmacokinetics

Background:

• Dynamic Positron Emission Tomography (PET) imaging is crucial for studying tracer kinetics in vivo.
• Motion artifacts significantly degrade the quality and accuracy of dynamic PET data.
• Arterial blood sampling is traditionally required to determine the input function for pharmacokinetic modeling.

Purpose of the Study:

• To develop a novel motion correction algorithm for dynamic PET imaging.
• To eliminate the need for arterial blood sampling by deriving the input function from PET data.
• To improve the accuracy of pharmacokinetic modeling and image registration in dynamic PET studies.

Main Methods:

• A pharmacokinetic model-based registration algorithm was developed.
• A generalized model derives the input function from tomographic data to model tracer kinetics.
• Temporal and spatial constraints were integrated into a joint probabilistic model for iterative optimization of motion and kinetic parameters.
• A group-wise registration framework was established for motion-corrupted dynamic PET data.

Main Results:

• The algorithm was evaluated using simulated and measured human dopamine D3 receptor imaging data ([11C]-(+)-PHNO).
• Simulation-based validation demonstrated subvoxel registration accuracy for noisy data with simulated motion artifacts.
• Initial experiments with clinical [11C]-(+)-PHNO brain data showed reductions in motion.

Conclusions:

• The novel algorithm effectively corrects motion in dynamic PET images without arterial blood sampling.
• It enables accurate pharmacokinetic modeling and group-wise image registration.
• This approach enhances the quality of dynamic PET studies, particularly for receptor imaging.