ESTRO 2023 - Abstract Book

S642

Monday 15 May 2023

ESTRO 2023

à Figure 1 - image quality (resolution, contrast) for the integrated mode proton radiography system

Figure 2 - WET accuracy for the integrated mode proton radiography system

Conclusion This work illustrates that high quality/clinically ready proton radiographs can be obtained with clinical beam settings using a fast, low-cost scintillation-based system, and shows benefits of 2D Bragg curve data as opposed to a range telescope for integrated mode pRads. Future work includes rapid fluoroscopic imaging for real-time treatment adaptation, and potential use for proton therapy QA.

OC-0775 Comprehensive proton dose prediction with Bayesian LSTMs L. Voss 1,2,3 , A. Neishabouri 3,4,5 , T. Ortkamp 2,3,6,7 , N. Wahl 2,3

1 Ruprecht Karl University of Heidelberg, Institute of Computer Science, Heidelberg, Germany; 2 German Cancer Research Center (DKFZ), Department of Medical Physics in Radiation Oncology, Heidelberg, Germany; 3 Heidelberg Institute for Radiation Oncology (HIRO), Department of Medical Physics in Radiation Oncology, Heidelberg, Germany; 4 Heidelberg Ion Therapy Center, HIT, Heidelberg, Germany; 5 German Cancer Research Center (DKFZ), Clinical Cooperation Unit Radiation Oncology, Heidelberg, Germany; 6 Karlsruhe Institute of Technology (KIT), Steinbuch Centre for Computing (SCC), Heidelberg, Germany; 7 Helmholtz Information and Data Science School for Health, HIDSS4Health, Heidelberg, Germany Purpose or Objective Deep learning-based proton dose calculation engines may achieve near Monte Carlo (MC) prediction accuracy while maintaining millisecond prediction time per pencil beam (PB). However, such networks tend to be overconfident in their prediction and face a lack of interpretability in the prediction finding process. For clinical translation and quality assurance, quantifying the prediction model uncertainty would be desirable - the same way an MC algorithm provides statistical noise. Materials and Methods This approach uses a Bayesian long-short term memory network (B-LSTM) to incorporate such an uncertainty estimate by replacing deterministic weights and biases in the network architecture with probability distributions. Our B-LSTM (see Figure 1) with corresponding Bayesian back-end network is based on a previous deterministic PB dose prediction model (Neishabouri et al. 2021): It was tested and trained on MC PB doses with initial beam energies of 104.25 MeV, once in a constructed phantom case with block heterogeneities and once on a lung case. The B-LSTM’s mean prediction was compared to MC doses, its equivalent deterministic structure, and the original LSTM by Neishabouri et al. 2021 with a [1%, 3 mm] global γ -analysis. The accuracy of the predicted standard deviation ( σ ) was quantified with a significance test ( ≥ 3 σ ) against dose differences.