Denoising filters for fast Monte Carlo and genetic algorithm-based optimization

Objective. To evaluate the suitability of denoising filters for accelerating Monte Carlo (MC) dose calculation within genetic algorithm (GA) optimization frameworks for external beam radiation therapy treatment planning through preservation of optimization-relevant information in filtered low-statistic simulations.Approach. Gaussian smoothing, adaptive anisotropic diffusion (AAD), and wavelet threshold denoising (WTD) were applied to Geant4-generated three-dimensional MC dose distributions. Filter parameters were tuned across varying simulation conditions and evaluated using voxel-based dose agreement metrics. Filter performance within a GA framework was then assessed using multiple realistic objective functions, with optimization robustness quantified using Spearman rank correlation, winner identification, and winner consistency across repeated filtered MC dose calculations.Main results. MC particle statistics and voxel resolution had the strongest influence on optimal parameter selection for all three denoising filters, while beam energy, field size, and evaluation metric exhibited comparatively weaker effects. Filtering improved the ranking stability of candidate system configurations for low-statistic MC dose distributions, supporting faster evaluation within stochastic optimization algorithms. AAD filtering provided the highest ranking fidelity for objectives dependent on localized dose extrema, while WTD consistently identified the optimal configuration for dose-volume objectives across simulation statistics despite lower overall rank correlation. In contrast, Gaussian smoothing frequently degraded ranking fidelity for these objectives due to oversmoothing of important dose features.Significance. Denoising filters can accelerate MC-driven optimization by enabling lower-statistic dose calculations while preserving optimization-relevant ranking information. Because optimal tuning depended primarily on particle statistics and voxel resolution rather than beam configuration, filter parameters can remain applicable across a wide range of candidate treatment setups. AAD and WTD showed the greatest promise for integration into MC-driven optimization workflows, with relative performance depending on the underlying MC noise level.

Authors

Institutions

Publication Details

Journal
Physics in Medicine and Biology
Published
2026-11-19
DOI
https://doi.org/10.1088/1361-6560/aec4b6
Primary Topic
Advanced Radiotherapy Techniques
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Denoising filters for fast Monte Carlo and genetic algorithm-based optimization

A O’Brien, Rebecca J. Abergel
Physics in Medicine and Biology
Advanced Radiotherapy Techniques
article

Denoising filters for fast Monte Carlo and genetic algorithm-based optimization

A O’Brien, Rebecca J. Abergel
article en

Abstract

Objective. To evaluate the suitability of denoising filters for accelerating Monte Carlo (MC) dose calculation within genetic algorithm (GA) optimization frameworks for external beam radiation therapy treatment planning through preservation of optimization-relevant information in filtered low-statistic simulations.Approach. Gaussian smoothing, adaptive anisotropic diffusion (AAD), and wavelet threshold denoising (WTD) were applied to Geant4-generated three-dimensional MC dose distributions. Filter parameters were tuned across varying simulation conditions and evaluated using voxel-based dose agreement metrics. Filter performance within a GA framework was then assessed using multiple realistic objective functions, with optimization robustness quantified using Spearman rank correlation, winner identification, and winner consistency across repeated filtered MC dose calculations.Main results. MC particle statistics and voxel resolution had the strongest influence on optimal parameter selection for all three denoising filters, while beam energy, field size, and evaluation metric exhibited comparatively weaker effects. Filtering improved the ranking stability of candidate system configurations for low-statistic MC dose distributions, supporting faster evaluation within stochastic optimization algorithms. AAD filtering provided the highest ranking fidelity for objectives dependent on localized dose extrema, while WTD consistently identified the optimal configuration for dose-volume objectives across simulation statistics despite lower overall rank correlation. In contrast, Gaussian smoothing frequently degraded ranking fidelity for these objectives due to oversmoothing of important dose features.Significance. Denoising filters can accelerate MC-driven optimization by enabling lower-statistic dose calculations while preserving optimization-relevant ranking information. Because optimal tuning depended primarily on particle statistics and voxel resolution rather than beam configuration, filter parameters can remain applicable across a wide range of candidate treatment setups. AAD and WTD showed the greatest promise for integration into MC-driven optimization workflows, with relative performance depending on the underlying MC noise level.

Physics in Medicine and Biology
University of California, Berkeley (US)
Advanced Radiotherapy Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon