Poster Poster Program Therapy Physics

Spectral Reconstruction of High-Energy Photon Beams from Depth Dose Curves Using Neural Networks for Beam Characterisation In Radiotherapy

Abstract
Purpose

Precise knowledge of high-energy photon spectra from clinical linear accelerators (linacs) is important for dosimetry in radiotherapy. High-energy photon spectra cannot be measured directly, so alternative methods are required. In this work, the spectral fluence of high-energy photon beams was reconstructed from depth dose curves using neural networks.

Methods

The model developed to reconstruct spectral photon fluence from depth dose curves is a recurrent neural network based on a gated recurrent unit incorporating physical laws, implemented as a physics-informed neural network. The uncertainties of the reconstructed spectra were investigated regarding the accuracy of the neural network. Due to the limited availability of high-energy photon spectra and corresponding depth dose curves, the training data for the neural network were generated using a combination of Monte-Carlo simulations and analytical models. A depth dose curve corresponding to a high-energy photon spectrum from a clinical linac can be considered as a linear combination of monoenergetic depth dose curves, that are weighted by the spectral photon fluence emitted by the linac. A large training dataset was therefore created by summing monoenergetic depth dose curves weighted with analytically calculated spectra.

Results

The trained neural network was validated using measured depth dose curves. The spectra reconstructed from the depth dose curves were compared with Monte-Carlo simulated spectra from detailed accelerator head models. In general, the reconstructed spectra could reproduce all the essential features of realistic spectra and showed good agreement with them. The smallest spectral difference of 1.6% was obtained for 6 MV beams, while it increased for beams with higher energies (21.5% at 15 MV), the reason for which is still under investigation.

Conclusion

The neural network enables photon spectra to be quickly determined from simple measurements and will be used in the future to develop a method for unambiguous beam characterization of linacs.

People

Related

Similar sessions

Poster Poster Program
Jul 19 · 07:00
Python-Based Automation Framework for Annual Machine QA Data Archiving In Qatrack+

Annual water-tank measurements help ensure beam characteristics remain consistent with commissioning baselines. However, the lack of a standardized processing workflow and decentralized data storage makes it difficult to analyze...

Syed Bilal Ahmad, PhD
Therapy Physics 0 people interested
Poster Poster Program
Jul 19 · 07:00
User Expectations and Current Availability of HDR Brachytherapy Audits In Europe

The aim of this work was to evaluate the need to implement more dosimetric audits in high‐dose‐rate brachytherapy (HDR-BT) in Europe and to identify which characteristics such audits should meet according to users.

Javier Vijande, PhD Laura Oliver Cañamás
Therapy Physics 0 people interested