ESTRO 2025 - Abstract Book

S2710

Physics - Dose calculation algorithms

ESTRO 2025

Conclusion: Encoder-decoder structures offer a versatile framework for sequence-based DNN models, suitable for precise millisecond dose predictions. Complexity reduction can significantly reduce computation times while maintaining high accuracy.

Keywords: Proton dose calculation, deep learning

References: 1 Hochreiter, Sepp and Jürgen Schmidhuber (Nov. 1997). ”Long Short-Term Memory“. In: Neural Computation 9 (8), pp. 1735–1780. 2 Vaswani, Ashish et al. (2017). ”Attention Is All You Need“. In: CoRR abs/1706.03762. arXiv: 1706.03762. 3 Sun, Yutao et al. (2023). ”Retentive Network: A Successor to Transformer for Large Language Models“. In: ArXiv abs/2307.08621. 4 Neishabouri, Ahmad et al. (Mar. 2021). ”Long short-term memory networks for proton dose calculation in highly heterogeneous tissues“. In: Medical Physics 48.4, pp. 1893–1908. 5 Pastor-Serrano, Oscar and Zolán Perkó (May 2022). ”Millisecond speed deep learning based proton dose calculation with Monte Carlo accuracy“. In: Physics in Medicine & Biology 67 (10), p. 105006.