Enhancing beam quality forecast with long short-term memory time series
Nuclear Technology and Radiation Protection · 2026 · 41 (2) · 121-129
DOI: 10.2298/ntrp2602121c ↗Kurz gefasst
Fachartikel in der Zeitschrift Nuclear Technology and Radiation Protection, 2026. Verfasst von Narmada Chinnakannan und 3 weiteren Autor:innen. Die Arbeit nennt 21 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
Ensuring consistent beam quality is vital in radiation therapy, as treatments typically span 15-20 days. Accurate forecasting of the beam quality factor enhances treatment precision. This study compares long short-term memory time series models optimized using Bayesian and hyperband optimization. Time series forecasting of 6 MV photon beam quality was performed using stacked timeseries model. Hyperparameter tuning was implemented in Python, with the two optimisation. Data was collected from three linear accelerators-Infinity, Synergy, and Truebeam-with approximately 1000 measurements per device. Bayesian optimization outperformed hyperband in both speed and predictive accuracy. For Infinity the root mean square error for Bayesian is 0.0155 achieved in 3 m 56 s against the hyperband with 0.0126 in 53 m 30 s. For Synergy the root mean square error for Bayesian is 0.0102 obtained in 4 m 52 s against hyperband with 0.0184 in 35 m 45 s. For TrueBeam the root mean square error for Bayesian is 0.0024 arrived in 3 m 27 s against hyperband with 0.0138 in 48 m 27 s. Bayesian optimization offers a faster and more accurate method for hyperparameter tuning in this time series based beam quality forecasting. Its novel application in radiotherapy quality assurance presents a promising advancement for improving treatment reliability.
Zitieren
APA
Chinnakannan, N., Nallamuthu, P., Ganesan, M., & Baladasan, J. (2026). Enhancing beam quality forecast with long short-term memory time series. Nuclear Technology and Radiation Protection, 41(2), 121-129. https://doi.org/10.2298/ntrp2602121c
Vancouver
Chinnakannan N, Nallamuthu P, Ganesan M, Baladasan J. Enhancing beam quality forecast with long short-term memory time series. Nuclear Technology and Radiation Protection. 2026;41(2):121-129. doi:10.2298/ntrp2602121c
BibTeX
@article{Chinnakannan2026,
author = {Chinnakannan, Narmada and Nallamuthu, Punithavelan and Ganesan, Malathi and Baladasan, Jaganathan},
title = {Enhancing beam quality forecast with long short-term memory time series},
journal = {Nuclear Technology and Radiation Protection},
year = {2026},
volume = {41},
number = {2},
pages = {121--129},
doi = {10.2298/ntrp2602121c}
}
EUVIMED