Ce projet vise la construction d'un prototype de scanner PET pour petits animaux utilisant des capteurs monolithiques à pixels en silicium comme senseurs. Avec l'aide de méthodes de reconstruction avancées basées sur l'apprentissage automatique, le scanner permettra d'atteindre des résolutions volumétriques dix fois meilleures qu'actuellement.
Ces développements dans l'instrumentation et les logiciels nous permettront d'ouvrir le champ à l'imagerie moléculaire à ultra-haute résolution.
Dans une première application de ces nouveaux outils, nous étudierons l'apparition et la progression des plaques athéroscléreuses chez la souris afin de mieux comprendre, surveiller et traiter l'athérosclérose.
With this project, we aim to pioneer ultra-high-resolution molecular imaging and make this technology available for novel cutting-edge medical applications. We propose to develop an ultra-high resolution small-animal PET scanner in combination with advanced image-reconstruction algorithms, and perform ultra-high-resolution molecular imaging of the onset and progression of atherosclerosis in ApoE-/- mice as a first biomedical application.The ultra-high-resolution scanner will use a multi-layer of monolithic silicon pixel sensors as detection medium. In contrast to typical PET scanners, in our concept photons are converted inside 50µm lead foils and the electrons produced are measured in very precise 100µm thick monolithic silicon sensors. A stack of 60 lead+silicon detection layers ensures already high efficiency, with a sensitivity peak of 4%. The scanner granularity of 100×100×200µm3 will provide ground-breaking spatial resolution and depth-of-interaction measurements for high-quality imaging over the entire field of view: full simulation and a simple filtered back-projection result in an upper limit of the point-spread function of 350µm FWHM in both the radial and tangential directions. The use of fast silicon sensors permits noise-equivalent count rate larger than 2 Mcps at 150 MBq. Several prototypes of the monolithic silicon sensor were produced, while the engineering of the scanner components and the construction procedures and methods were already developed within the SINERGIA grant CRSII2_160808. The scanner will provide an outstanding volumetric spatial resolution of 0.06 mm3 and therefore require the development of advanced image-reconstruction algorithms able to exploit the several millions of detection channels.
The main challenge for the reconstruction software will be to handle the large quantity of sensor data provided by the100µPET scanner, to reduce the measurement noise and produce image reconstructions with the highest possible spatial resolution in reasonable computation time. This will be accomplished by taking advantage of the latest developments in image reconstruction technology; in particular, the use of sophisticated regularization schemes and machine learning. Since these techniques proceed by iterative refinement, they tend to be computationally too expensive. We propose to overcome this bottleneck via an adequate reformatting of the data using multidimensional splines in conjunction with neural networks.
In addition, we shall create a virtual ultra-high-resolution PET model to guide the design of the reconstruction software. These developments in instrumentation and software will enable us to overcome the current resolution limit of PET scanners and open the field to ultra-high-resolution molecular imaging. As a first application of these new tools, we will study the onset and progression of atherosclerotic plaques in mice to better understand, monitor and treat atherosclerosis.
Specifically, we will perform four imaging series in ApoE-/- mice to establish the value of ultra-high-resolution molecular imaging in monitoring (i) individual atherosclerotic plaque development, (ii) metabolic changes under different diets, (iii) response to therapy with statins and fibrates, and (iv) macrophage metabolism and microcalcification.
Last updated:14.05.2022
