A variational bayesian inference method for parametric imaging of PET data
| Autor: | Federico Turkheimer, Mattia Veronese, Alessandra Bertoldo, Michael A. Chappell, Gaia Rizzo, Marco Castellaro, Matteo Tonietto |
|---|---|
| Rok vydání: | 2017 |
| Předmět: |
L[1-C]leucine
Mean squared error Computer science Cognitive Neuroscience Bayesian probability [F]FDG Models Neurological Signal-To-Noise Ratio Machine learning computer.software_genre Bayesian inference Noise (electronics) Synthetic data 030218 nuclear medicine & medical imaging 03 medical and health sciences 0302 clinical medicine Voxel Voxel-wise kinetic analysis Prior probability Image Processing Computer-Assisted Humans Parametric statistics Brain Mapping business.industry [C]WAY-100635 [11C]WAY-100635 Bayes Theorem Models Theoretical Neurology Positron-Emission Tomography L[1-11C]leucine Positron Emission Tomography Variational Bayes [18F]FDG Artificial intelligence business computer Algorithm 030217 neurology & neurosurgery Algorithms |
| Zdroj: | Castellaro, M, Rizzo, G, Tonietto, M, Veronese, M, Turkheimer, F E, Chappell, M A & Bertoldo, A 2017, ' A Variational Bayesian inference method for parametric imaging of PET data ', NeuroImage, vol. 150, pp. 136-149 . https://doi.org/10.1016/j.neuroimage.2017.02.009 |
| DOI: | 10.1016/j.neuroimage.2017.02.009 |
| Popis: | In dynamic Positron Emission Tomography (PET) studies, compartmental models provide the richest information on the tracer kinetics of the tissue. Inverting such models at the voxel level is however quite challenging due to the low signal-to-noise ratio of the time activity curves. In this study, we propose the use of a Variational Bayesian (VB) approach to efficiently solve this issue and thus obtain robust quantitative parametric maps. VB was adapted to the non-uniform noise distribution of PET data. Moreover, we propose a novel hierarchical scheme to define the model parameter priors directly from the images in case such information are not available from the literature, as often happens with new PET tracers. VB was initially tested on synthetic data generated using compartmental models of increasing complexity, providing accurate (%bias |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|