SMuK 2023 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
ST: Fachverband Strahlen- und Medizinphysik
ST 8: AI Topical Day – AI in Medicine (joint session ST/AKPIK)
ST 8.3: Vortrag
Donnerstag, 23. März 2023, 14:30–14:45, ZEU/0146
Event identification in the SiFi-CC Compton camera for imaging prompt gamma rays in proton therapy via deep neural networks — •Alexander Fenger1, Ronja Hetzel1, Jonas Kasper1, George Farah1, Achim Stahl1, and Aleksandra Wrońska2 — 1III. Physikalisches Institut B, RWTH Aachen University — 2M. Smoluchowski Institute of Physics, Jagiellonian University Kraków, Poland
One of the biggest challenges in proton therapy is ensuring that the dose is delivered to the right position. A promising approach for online monitoring of the beam range is the detection of prompt gamma rays using a Compton camera, as it provides the possibility to reconstruct the 3D distribution of the deposited dose.
The SiFi-CC (SiPM and scintillating Fiber-based Compton Camera) project is a joint collaboration of the RWTH Aachen University, the Jagiellonian University in Kraków and the University of Lübeck. The two modules of the SiFi-CC, the scatterer and the absorber, both consist of stacked LYSO fibres and are read out by SiPMs. Deep neural networks are employed to separate valid Compton events from background and reconstruct the direction and energy of prompt gamma rays. First implementations of neural networks show promising results in classification of Compton events as well as full reconstruction of the event topology and kinematics. The next step is to further optimize the current neural network implementation to gain sensitivity towards a detectable range shift in the source position. Different neural network designs as well as an evaluation of their performance are presented.