skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: WE-AB-209-01: A Monte Carlo-Based Method to Include Random Errors in Robust Optimization

Abstract

Purpose: To develop an efficient method to implement random set-up errors and organ motion in robust optimization for proton therapy treatment planning. Methods: The plans were created with an in-house robust optimizer, coupled with a super-fast Monte Carlo (MC) engine (less than 1 minute for final dose). MC simulates random errors by shifting randomly the starting point of each particle, according to their probability distribution. Such strategy assumes a sufficient number of treatment fractions. Two strategies are presented: 1) Full robust optimization with beamlets that already include the effect of random errors and 2) Mixed robust optimization, where the nominal beamlets are involved but a correction term C modifies the prescription. Starting from C=0, the method alternates optimization of the spot weights with the nominal beamlets and updates of C, with C=Drandom-Dnominal and where Drandom results from a regular MC computation (without pre-computed beamlets) that simulates random errors. Updates of C can be triggered as often as necessary by running the MC engine with the last corrected values for the spot weights as input. The method was applied to lung and prostate cases. For both patients the range error was set to 3%, systematic setup error to 5mm and standardmore » deviation for random errors to 5 mm. Comparison between full robust optimization and the mixed strategy (with 3 updates of C) is presented. Results: Target coverage was far below the clinical constraints (D{sub 95} > 95% of the prescribed dose) for plans where random errors were not simulated, especially for lung case. However, by using the proposed strategies the plans achieved a target coverage above clinical constraints. Conclusion: Full robust optimization gives better results than the mixed strategy, but the latter can be useful in cases where a MC engine is not available or too computationally intensive for beamlets calculation.« less

Authors:
; ; ;  [1]
  1. Universite catholique de Louvain, B-1200 Bruxelles (Belgium)
Publication Date:
OSTI Identifier:
22654131
Resource Type:
Journal Article
Resource Relation:
Journal Name: Medical Physics; Journal Volume: 43; Journal Issue: 6; Other Information: (c) 2016 American Association of Physicists in Medicine; Country of input: International Atomic Energy Agency (IAEA)
Country of Publication:
United States
Language:
English
Subject:
60 APPLIED LIFE SCIENCES; 61 RADIATION PROTECTION AND DOSIMETRY; ERRORS; MONTE CARLO METHOD; OPTIMIZATION; PLANNING; PROTON BEAMS; RADIOTHERAPY; RANDOMNESS; SIMULATION

Citation Formats

Barragan Montero, A, Souris, K, Lee, J, and Sterpin, E. WE-AB-209-01: A Monte Carlo-Based Method to Include Random Errors in Robust Optimization. United States: N. p., 2016. Web. doi:10.1118/1.4957770.
Barragan Montero, A, Souris, K, Lee, J, & Sterpin, E. WE-AB-209-01: A Monte Carlo-Based Method to Include Random Errors in Robust Optimization. United States. doi:10.1118/1.4957770.
Barragan Montero, A, Souris, K, Lee, J, and Sterpin, E. 2016. "WE-AB-209-01: A Monte Carlo-Based Method to Include Random Errors in Robust Optimization". United States. doi:10.1118/1.4957770.
@article{osti_22654131,
title = {WE-AB-209-01: A Monte Carlo-Based Method to Include Random Errors in Robust Optimization},
author = {Barragan Montero, A and Souris, K and Lee, J and Sterpin, E},
abstractNote = {Purpose: To develop an efficient method to implement random set-up errors and organ motion in robust optimization for proton therapy treatment planning. Methods: The plans were created with an in-house robust optimizer, coupled with a super-fast Monte Carlo (MC) engine (less than 1 minute for final dose). MC simulates random errors by shifting randomly the starting point of each particle, according to their probability distribution. Such strategy assumes a sufficient number of treatment fractions. Two strategies are presented: 1) Full robust optimization with beamlets that already include the effect of random errors and 2) Mixed robust optimization, where the nominal beamlets are involved but a correction term C modifies the prescription. Starting from C=0, the method alternates optimization of the spot weights with the nominal beamlets and updates of C, with C=Drandom-Dnominal and where Drandom results from a regular MC computation (without pre-computed beamlets) that simulates random errors. Updates of C can be triggered as often as necessary by running the MC engine with the last corrected values for the spot weights as input. The method was applied to lung and prostate cases. For both patients the range error was set to 3%, systematic setup error to 5mm and standard deviation for random errors to 5 mm. Comparison between full robust optimization and the mixed strategy (with 3 updates of C) is presented. Results: Target coverage was far below the clinical constraints (D{sub 95} > 95% of the prescribed dose) for plans where random errors were not simulated, especially for lung case. However, by using the proposed strategies the plans achieved a target coverage above clinical constraints. Conclusion: Full robust optimization gives better results than the mixed strategy, but the latter can be useful in cases where a MC engine is not available or too computationally intensive for beamlets calculation.},
doi = {10.1118/1.4957770},
journal = {Medical Physics},
number = 6,
volume = 43,
place = {United States},
year = 2016,
month = 6
}
  • A split-step numerical method for calculating ultrafast free-electron dynamics in dielectrics is introduced. The two split steps, independently programmed in C++11 and FORTRAN 2003, are interfaced via the presented open source wrapper. The first step solves a deterministic extended multi-rate equation for the ionization, electron–phonon collisions, and single photon absorption by free-carriers. The second step is stochastic and models electron–electron collisions using Monte-Carlo techniques. This combination of deterministic and stochastic approaches is a unique and efficient method of calculating the nonlinear dynamics of 3D materials exposed to high intensity ultrashort pulses. Results from simulations solving the proposed model demonstrate howmore » electron–electron scattering relaxes the non-equilibrium electron distribution on the femtosecond time scale.« less
  • The purpose of this work was the development of a probabilistic planning method with biological cost functions that does not require the definition of margins. Geometrical uncertainties were integrated in tumor control probability (TCP) and normal tissue complication probability (NTCP) objective functions for inverse planning. For efficiency reasons random errors were included by blurring the dose distribution and systematic errors by shifting structures with respect to the dose. Treatment plans were made for 19 prostate patients following four inverse strategies: Conformal with homogeneous dose to the planning target volume (PTV), a simultaneous integrated boost using a second PTV, optimization usingmore » TCP and NTCP functions together with a PTV, and probabilistic TCP and NTCP optimization for the clinical target volume without PTV. The resulting plans were evaluated by independent Monte Carlo simulation of many possible treatment histories including geometrical uncertainties. The results showed that the probabilistic optimization technique reduced the rectal wall volume receiving high dose, while at the same time increasing the dose to the clinical target volume. Without sacrificing the expected local control rate, the expected rectum toxicity could be reduced by 50% relative to the boost technique. The improvement over the conformal technique was larger yet. The margin based biological technique led to toxicity in between the boost and probabilistic techniques, but its control rates were very variable and relatively low. During evaluations, the sensitivity of the local control probability to variations in biological parameters appeared similar for all four strategies. The sensitivity to variations of the geometrical error distributions was strongest for the probabilistic technique. It is concluded that probabilistic optimization based on tumor control probability and normal tissue complication probability is feasible. It results in robust prostate treatment plans with an improved balance between local control and rectum toxicity, compared to conventional techniques.« less
  • A new hierarchical method for the Monte Carlo simulation of random fields called the Fourier-wavelet method is developed and applied to isotropic Gaussian random fields with power law spectral density functions. This technique is based upon the orthogonal decomposition of the Fourier stochastic integral representation of the field using wavelets. The Meyer wavelet is used here because its rapid decay properties allow for a very compact representation of the field. The Fourier-wavelet method is shown to be straightforward to implement, given the nature of thenecessary precomputations and the run-time calculations, and yields comparable results with scaling behavior over as manymore » decades as the physical space multiwavelet methods developed recently by two of the authors. However, the Fourier-wavelet method developed here is more flexible and, in particular, applies to anisotropic spectra generated through solutions of differential equations. Simulation results using this new technique and the well-known nonhierarchical simulation technique, the randomization method, are given and compared for both a simple shear layer model problem as well as a two-dimensional isotropic GAussian random field. The Fourier-wavelet method results are more accurate for statistical quantities depending on moments higher than order 2, in addition to showing a quite smooth decay to zero on the scales smaller than the scaling regime when compared with the randomization method results. The only situation is which the nonhierarchical randomization method is more computationally efficient occurs when no more than four decades of scaling behavior are needed and the statistical quantities of interest depend only on second moments.« less
  • A model of triangulated random surfaces which is the discrete analog of the Polyakov string is considered. An algorithm is proposed which enables one to study the model by the Monte Carlo method in the grand canonical ensemble. Preliminary results on the determination of the critical index ..gamma.. are presented.