PET-ABC

PET-ABC performs Bayesian likelihood-free inference for kinetic models of dynamic Positron Emission Tomography (PET) data as an R package, enabling parameter estimation and model selection for single-subject PET studies.

Key Features:

• Bayesian Statistical Inference: Employs Bayesian likelihood-free inference to produce posterior distributions and point estimates for kinetic model parameters and associated uncertainty (probability density functions).
• Model Selection and Evaluation: Evaluates reliability of kinetic parameter estimates by simulating dynamic PET data using one-tissue and two-tissue compartmental models across varying noise levels and performs model selection, demonstrating higher statistical power than weighted non-linear least squares (WNLS) in simulated ligand displacement studies.
• Application to Pharmacological PET Data: Quantifies probabilities of physiological changes in pharmacological studies, exemplified by analysis of non-steady-state dynamic [^11C] raclopride data from conscious rats administered amphetamine or saline to assess dopamine release in the striatum.

Scientific Applications:

• Neuroimaging Kinetic Modeling: Parameter estimation and model selection for dynamic PET kinetic models (one-tissue and two-tissue compartmental models) in neuroimaging studies.
• Pharmacological PET Studies: Analysis of ligand displacement and drug-induced changes such as dopamine release using tracers like [^11C] raclopride in preclinical and single-subject settings.
• Single-Subject / Personalized Analysis: Statistical inference and model selection tailored to single-subject dynamic PET data to support individualized conclusions about physiological responses.

Methodology:

Simulates dynamic PET datasets with varying noise conditions and one- and two-tissue compartmental models, applies Bayesian likelihood-free inference for posterior estimation and model selection, and compares model-selection performance to weighted non-linear least squares (WNLS) in simulated ligand displacement studies.