scholarly journals DeepBeam – A Machine Learning Framework For Tuning The Primary Electron Beam of The PRIMO Monte Carlo Software

2021 ◽  
Author(s):  
Zbisław Tabor ◽  
Damian Kabat ◽  
Michael Waligórski
Keyword(s):  
Machine Learning ◽  
Monte Carlo ◽  
Electron Beam ◽  
Regression Models ◽  
Primary Electron ◽  
Support Vector ◽  
Linear Accelerators ◽  
Primary Electron Beam ◽  
Measured Dose ◽  
Beam Parameters

Abstract BackgroundAny Monte Carlo simulation of dose delivery using medical accelerator-generated megavolt photon beams begins by simulating electrons of the primary electron beam interacting with a target. Because the electron beam characteristics of any single accelerator are unique and generally unknown, an appropriate model of an electron beam must be assumed before MC simulations can be run. The purpose of the present study is to develop a flexible framework with suitable regression models for estimating parameters of the model of primary electron beam in simulators of medical linear accelerators, basing on real reference dose profiles measured in a water phantom. MethodsAll simulations were run using PRIMO MC simulator. Two regression models for estimating the parameters of the simulated primary electron beam, both based on machine learning, were developed. The first model applies Principal Component Analysis to measured dose profiles in order to extract principal features of the shapes of the these profiles. The PCA-obtained features are then used by Support Vector Regressors to estimate the parameters of the model of the electron beam. The second model, based on deep learning, consists of a set of encoders processing measured dose profiles, followed by a sequence of fully connected layers acting together, which solve the regression problem of estimating values of the electron beam parameters directly from the measured dose profiles. Results of the regression are then used to reconstruct the dose profiles, basing on the PCA model. Agreement between the measured and reconstructed profiles can be further improved by an optimization procedure resulting in the final estimates of the parameters of the model of the primary electron beam. These final estimates are then used to determine dose profiles in MC simulations.ResultsAnalysed were a set of actually measured (real) dose profiles of 6 MV beams from a real Varian 2300 C/D accelerator, a set of simulated training profiles, and a separate set of simulated testing profiles, both generated for a range of parameters of the primary electron beam of the Varian 2300 C/D PRIMO simulator. Application of the two-stage procedure based on regression followed by reconstruction-based minimization of the difference between measured (real) and reconstructed profiles resulted in achieving consistent estimates of electron beam parameters and in a very good agreement between the measured and simulated photon beam profiles.ConclusionsThe proposed framework is a readily applicable and customizable tool which may be applied in tuning virtual primary electron beams of Monte Carlo simulators of linear accelerators. The codes, training and test data, together with some trained models and readout procedures, are freely available at the site: https://github.com/taborzbislaw/DeepBeam.

scholarly journals DeepBeam: a machine learning framework for tuning the primary electron beam of the PRIMO Monte Carlo software

2021 ◽  
Vol 16 (1) ◽  
Author(s):  
Zbisław Tabor ◽  
Damian Kabat ◽  
Michael P. R. Waligórski
Keyword(s):  
Machine Learning ◽  
Monte Carlo ◽  
Electron Beam ◽  
Regression Models ◽  
Primary Electron ◽  
Support Vector ◽  
Linear Accelerators ◽  
Primary Electron Beam ◽  
Measured Dose ◽  
Beam Parameters

Abstract Background Any Monte Carlo simulation of dose delivery using medical accelerator-generated megavolt photon beams begins by simulating electrons of the primary electron beam interacting with a target. Because the electron beam characteristics of any single accelerator are unique and generally unknown, an appropriate model of an electron beam must be assumed before MC simulations can be run. The purpose of the present study is to develop a flexible framework with suitable regression models for estimating parameters of the model of primary electron beam in simulators of medical linear accelerators using real reference dose profiles measured in a water phantom. Methods All simulations were run using PRIMO MC simulator. Two regression models for estimating the parameters of the simulated primary electron beam, both based on machine learning, were developed. The first model applies Principal Component Analysis to measured dose profiles in order to extract principal features of the shapes of the these profiles. The PCA-obtained features are then used by Support Vector Regressors to estimate the parameters of the model of the electron beam. The second model, based on deep learning, consists of a set of encoders processing measured dose profiles, followed by a sequence of fully connected layers acting together, which solve the regression problem of estimating values of the electron beam parameters directly from the measured dose profiles. Results of the regression are then used to reconstruct the dose profiles based on the PCA model. Agreement between the measured and reconstructed profiles can be further improved by an optimization procedure resulting in the final estimates of the parameters of the model of the primary electron beam. These final estimates are then used to determine dose profiles in MC simulations. Results Analysed were a set of actually measured (real) dose profiles of 6 MV beams from a real Varian 2300 C/D accelerator, a set of simulated training profiles, and a separate set of simulated testing profiles, both generated for a range of parameters of the primary electron beam of the Varian 2300 C/D PRIMO simulator. Application of the two-stage procedure based on regression followed by reconstruction-based minimization of the difference between measured (real) and reconstructed profiles resulted in achieving consistent estimates of electron beam parameters and in a very good agreement between the measured and simulated photon beam profiles. Conclusions The proposed framework is a readily applicable and customizable tool which may be applied in tuning virtual primary electron beams of Monte Carlo simulators of linear accelerators. The codes, training and test data, together with readout procedures, are freely available at the site: https://github.com/taborzbislaw/DeepBeam.

Monte Carlo modelling of clinical accelerator beams and estimation of primary electron beam parameters

2018 ◽  
Vol 53 (1) ◽  
pp. 61-66
Author(s):  
S. Horová ◽  
L. Judas
Keyword(s):  
Monte Carlo ◽  
Electron Beam ◽  
Confidence Region ◽  
Primary Electron ◽  
Head Model ◽  
Primary Electron Beam ◽  
Regression Methods ◽  
Dosimetric Data ◽  
The Mean ◽  
Beam Parameters

The accuracy of Monte Carlo simulations of clinical photon beams in radiation oncology is dependent on the linac head model accuracy and on parameters of the primary electron beam. While the internal composition and geometry of the accelerator head are known precisely, at least in principle, the energy spectrum and the spatial characteristics of the primary electron beam are unknown and immeasurable. The mean energy and FWHM of the electron beam are commonly estimated by comparing the simulation results with measured dosimetric data. Percentage depth doses (PDDs) and dose profiles are sensitive to changes in the electron beam parameters and are therefore in general used for the comparison. In the published studies which deal with parameter estimation, the determination of electron beam parameters is typically performed through a trial and error process. As to the parameter optimization, there is no unified methodology agreed upon, and the uncertainty of the resulting parameter values is usually not quantified by the authors. The aim of our work was not only to estimate the mean energy and the FWHM of the primary electron beam, but also to determine the confidence region of the optimized values in a defined and repeatable way. A model of Varian Clinac 2100C/D linear accelerator 6 MV photon beam was built in the EGSnrc/BEAMnrc Monte Carlo system. PDDs and dose profiles for different field sizes and different depths were obtained from water phantom measurements. We show that an approach based on a large number of simulations, each with a relatively low number of primary particles, in combination with non-linear regression methods allows to find both the optimized values of the electron beam parameters and their common 95% confidence region.

Automatic determination of primary electron beam parameters in Monte Carlo simulation

2007 ◽  
Vol 34 (3) ◽  
pp. 1076-1084 ◽  
Author(s):  
Javier Pena ◽  
Diego M. González-Castaño ◽  
Faustino Gómez ◽  
Francisco Sánchez-Doblado ◽  
Guenther H. Hartmann
Keyword(s):  
Monte Carlo Simulation ◽  
Monte Carlo ◽  
Electron Beam ◽  
Primary Electron ◽  
Automatic Determination ◽  
Primary Electron Beam ◽  
Beam Parameters

PO-1352: Prediction of electron beam parameters of a Monte Carlo model using machine learning

2020 ◽  
Vol 152 ◽  
pp. S716-S717
Author(s):  
A. Wagner ◽  
K. Brou Boni ◽  
E. Rault ◽  
F. Crop ◽  
T. Lacornerie ◽  
...  
Keyword(s):  
Machine Learning ◽  
Monte Carlo ◽  
Electron Beam ◽  
Monte Carlo Model ◽  
Beam Parameters

A multivariate approach to determine electron beam parameters for a Monte Carlo 6 MV Linac model: Statistical and machine learning methods

2022 ◽  
Vol 93 ◽  
pp. 38-45
Author(s):  
Hye Jeong Yang ◽  
Tae Hoon Kim ◽  
Thomas Schaarschmidt ◽  
Dong-Wook Park ◽  
Seung Hee Kang ◽  
...  
Keyword(s):  
Machine Learning ◽  
Monte Carlo ◽  
Electron Beam ◽  
Multivariate Approach ◽  
Learning Methods ◽  
Machine Learning Methods ◽  
Beam Parameters

scholarly journals Machine learning enables discovery of DNA-carbon nanotube sensors for serotonin

2021 ◽  
Author(s):  
Payam Kelich ◽  
Sanghwa Jeong ◽  
Nicole Navarro ◽  
Jaquesta Adams ◽  
Xiaoqi Sun ◽  
...  
Keyword(s):  
Machine Learning ◽  
Carbon Nanotube ◽  
Dna Sequences ◽  
Regression Models ◽  
Near Infrared ◽  
Optical Response ◽  
Support Vector ◽  
Single Walled Carbon Nanotube ◽  
Nir Fluorescence ◽  
Selective Evolution

AbstractDNA-wrapped single walled carbon nanotube (SWNT) conjugates have remarkable optical properties leading to their use in biosensing and imaging applications. A critical limitation in the development of DNA-SWNT sensors is the current inability to predict unique DNA sequences that confer a strong analyte-specific optical response to these sensors. Here, near-infrared (nIR) fluorescence response datasets for ~100 DNA-SWNT conjugates, narrowed down by a selective evolution protocol starting from a pool of ~1010 unique DNA-SWNT candidates, are used to train machine learning (ML) models to predict new unique DNA sequences with strong optical response to neurotransmitter serotonin. First, classifier models based on convolutional neural networks (CNN) are trained on sequence features to classify DNA ligands as either high response or low response to serotonin. Second, support vector machine (SVM) regression models are trained to predict relative optical response values for DNA sequences. Finally, we demonstrate with validation experiments that integrating the predictions of ensembles of the highest quality CNN classifiers and SVM regression models leads to the best predictions of both high and low response sequences. With our ML approaches, we discovered five new DNA-SWNT sensors with higher fluorescence intensity response to serotonin than obtained previously. Overall, the explored ML approaches introduce an important new tool to predict useful DNA sequences, which can be used for discovery of new DNA-based sensors and nanobiotechnologies.

Sign in / Sign up

Export Citation Format

Share Document