14-15 December 2020
Virtuales
Europe/Madrid timezone
Verifique por favor que Zoom funciona correctamente: https://zoom.us/test y que tiene descargada la última versión https://support.zoom.us/hc/en-us/articles/201362393

Multi purpose Ultra fast Monte Carlo PET simulator

14 Dec 2020, 19:25
5m
Virtuales

Virtuales

Zoom
E-poster Main Session 2

Speaker

Pablo Galve Lahoz (Grupo de Física Nuclear, EMFTEL & IPARCOS Universidad Complutense de Madrid, CEI Moncloa, Madrid, Spain)

Description

Accurate Monte Carlo simulations play a major role in PET corrections and scanner design. We present the Ultra-fast Monte Carlo PET simulator (UMC-PET) developed for multiple purposes, an accurate, fast and flexible PET simulator. The UMC-PET includes all the relevant physics related to the emission, transport and detection of the radiation in a PET acquisition, such as positron range, scatter and attenuation inside the patient, photon interaction with the scanner, and detector response (energy resolution, time of flight, etc.). The simulator accuracy has been extensively tested against other MC PET simulators such as PeneloPET, obtaining similar results, while being more than 3000 times faster. The code can handle arbitrary scanner geometries with simple and intuitive input files. These features allow applying UMC-PET beyond standard MC uses. UMC-PET has been tested to accurately predict scatter and attenuation corrections during reconstruction, or to compute system response matrix (SRM) for adapting scanner geometries. Furthermore, the speed of UMC-PET allows for a 3D iterative reconstruction for complex scanners, with a projection step based on, on the fly, raw, MC calculations with UMC-PET and thus avoiding explicit storage of the SRM or physics simplifications. On a single common (500 USD) GPU these fully MC reconstructions require a few hours for a scanner with > 1 billion lines of response. This provides not only a useful and flexible gold standard method, but may become a practical reconstruction approach if it is combined with variance reduction methods and/or high performance multi-GPU systems.

Primary authors

Pablo Galve Lahoz (Grupo de Física Nuclear, EMFTEL & IPARCOS Universidad Complutense de Madrid, CEI Moncloa, Madrid, Spain) Fernando Arias Valcayo (Universidad Complutense de Madrid) Alejandro López Montes (Grupo de Física Nuclear , Facultad de Ciencias Físicas , C.E.I. Moncloa , E.28040 , Madrid , Spain) AMAIA VILLA ABAUNZA (Grupo de física nuclear, Facultad de Físicas. Universidad Complutense - CEI Moncloa, E-28040 Madrid, Spain) Paula Ibáñez García (Universidad Complutense de Madrid) JOAQUIN LOPEZ HERRAIZ (Universidad Complutense de Madrid. Grupo de Fisica Nuclear. UPARCO) Jose Udias (Universidad Complutense de Madrid)

Presentation Materials

Your browser is out of date!

Update your browser to view this website correctly. Update my browser now

×