Image derived arterial input function for PET quantitative analysis
Résumé
Introduction: Dynamic PET images provide information on the pharmacokinetics of the tracer in the organs and/or in functional regions used to derive biological parameters provided the arterial and the plasmatic input function are known. The aim of the work is to derive image
arterial input function and to identify the relevant pharmacokinetics of the tracer in each voxel.
Methods: Under few conditions the kinetics mixture can be assumed linear and the blind source separation methods can be applied to dynamic PET data. In this work, we compare Non-negative Matrix Factorization, NMF, algorithm with a recent adaptation of the projected gradient technique to the NMF problem on PET 3D dynamic images of rats. The method was evaluated on a rat study of the pharmacokinetics of the [18F]-A85 tracer.
Results: Figure 1 shows the spatial distribution of the component 1 corresponding to the tracer distribution in the heart, the aorta and the kidneys for a model including 5 components. This component can be seen as the input function for the pharmacokinetics of the tracer.
Figure 2 shows the time distribution of the component 1 corresponding to the spatial distribution shown on figure1 estimated for a model including 3, 4 and 5 components.
Conclusion: The arterial input function estimation is robust with respect to the model order. Extraction of other pharmacokinetics and their corresponding spatial distribution can be done using the same approach.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|---|
| licence |
