Patient-Adaptive Population-Based Modeling of Arterial Input Functions

Area of Science:

• Nuclear Medicine
• Radiochemistry
• Pharmacokinetics

Background:

• Accurate arterial input function (AIF) estimation is crucial for dynamic PET kinetic modeling.
• Arterial blood sampling for AIF is invasive and labor-intensive.
• Existing methods for non-invasive AIF estimation have limitations.

Purpose of the Study:

• To develop and validate a patient-adaptive method for estimating individual AIFs using population-derived time course profiles.
• To improve the accuracy and efficiency of AIF estimation in dynamic PET studies.

Main Methods:

• A patient-adaptive mixture model fitting historical population time course profiles.
• Modeling tracer travel time and circulation time using Gamma distribution and subject-specific linear mixtures.
• Estimating individual AIFs by projection onto population profile components.
• Incorporating injection duration into the model for varying protocols.

Main Results:

• The proposed model outperformed reference techniques in analyses of 18F-FDG, 15O-H2O, and 18F-FLT clinical data.
• Statistically significant gains were observed when using population data for training basis components.
• Simulations demonstrated the reliability and potential benefits for estimating physiological parameters.
• Numerical simulations confirmed convergence and stability under varying training population sizes and noise levels.

Conclusions:

• The developed method offers a more accurate and less invasive approach to AIF estimation for dynamic PET.
• Leveraging population data significantly enhances the performance of AIF estimation.
• This technique shows promise for improving kinetic analyses and physiological parameter estimation in PET imaging.