Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
ST: Fachverband Strahlen- und Medizinphysik
ST 4: AI in Medicine
ST 4.2: Vortrag
Mittwoch, 20. März 2024, 09:45–10:00, PC 203
Fast Dose Prediction in Microbeam Radiation Therapy (MRT) Across Varied Field Angles. — •Tobias Cremer1, Cornelius Grunwald1, Kevin Alexander Kröninger1, Jens Weingarten1, Marco Schlimbach1, and Florian Mentzel2 — 1TU Dortmund, Germany — 2formally TU Dortmund, Germany
Microbeam radiation therapy stands as a promising preclinical approach to treating tumors, offering enhanced post-treatment outcomes. It uses a multi-slit collimator to create dosis peaks with high doserates and valleys in between with lower doserates. This spacial segmentation of the beam and high doserates over small amounts of time (FLASH-Therapy) leads to a higher survivability rate of the healthy tissue. For clinical implementation
This project builds upon prior research which demonstrated the suitability of Machine Learning for dose prediction. This work involves training a neural network capable of predicting doses for diverse field angles. Initially, training data are simulated for basic volumes as a proof of concept. Using this training data, a 3D U-Net will be trained to predict the corresponding dose deposition. Subsequently, the developed prediction model for basic volumes will be adapted to function with CT data imported into the Geant4 Simulations.
The presentation will showcase recent outcomes from the mentioned project, featuring initial results from the basic simulation and insights into the corresponding neural network utilized for dose prediction.
Keywords: dose prediction; microbeam radiation therapy