A comparison of Monte Carlo dose calculation denoising techniques

Abstract

Recent studies have demonstrated that Monte Carlo (MC) denoising techniques can reduce MC radiotherapy dose computation time significantly by preferentially eliminating statistical fluctuations ('noise') through smoothing. In this study, we compare new and previously published approaches to MC denoising, including 3D wavelet threshold denoising with sub-band adaptive thresholding, content adaptive mean-median-hybrid (CAMH) filtering, locally adaptive Savitzky-Golay curve-fitting (LASG), anisotropic diffusion (AD) and an iterative reduction of noise (IRON) method formulated as an optimization problem. Several challenging phantom and computed-tomography-based MC dose distributions with varying levels of noise formed the test set. Denoising effectiveness was measured in three ways: by improvements in the mean-square-error (MSE) with respect to a reference (low noise) dose distribution; by the maximum difference from the reference distribution and by the 'Van Dyk' pass/fail criteria of either adequate agreement with the reference image in low-gradient regions (within 2% in our case) or, in high-gradient regions, a distance-to-agreement-within-2% of less than 2 mm. Results varied significantly based on the dose test case: greater reductions in MSE were observed for the relatively smoother phantom-based dose distribution (up to a factor of 16 for the LASG algorithm); smaller reductions were seen for an intensity modulated radiation therapy (IMRT) head and neck case (typically, factors of 2-4). Although several algorithms reduced statistical noise for all test geometries, the LASG method had the best MSE reduction for three of the four test geometries, and performed the best for the Van Dyk criteria. However, the wavelet thresholding method performed better for the head and neck IMRT geometry and also decreased the maximum error more effectively than LASG. In almost all cases, the evaluated methods provided acceleration of MC results towards statistically more accurate results.

Image-Based Modeling of Normal Tissue Complication Probability for Radiation Therapy

“…CERR is free to use or modify for research use, and the current version is always maintained by WUSTL (currently at http://radium.wustl.edu/cerr). Several outcome analyses have been published to data which used CERR [67][68][69][70][71]. CERR has also been used for other types of radiotherapy research projects [72][73][74][75].…”

Current status and new horizons in Monte Carlo simulation of X-ray CT scanners

“…Some of the developed methods are inspired from classic image processing methods (e.g., filtering, anisotropic diffusion, wavelets, …etc), whereas other methods do not have direct relationship with image processing techniques (e.g., Kawrakow's locally adaptive method, Fipple and Nusslin's formulation, …etc) [67]. Although denoising seems to be the ideal solution to speed-up MC simulations, only one group reported on the use of dedicated curve fitting procedures inspired by the Richardson-Lucy deblurring algorithm to calculate scattered radiation projections in smallanimal cone-beam X-ray CT scanners [5].…”

“…The optimization and implementation of similar techniques for CT simulation remains to be explored. Recently, El Naqa et al [67] compared several denoising techniques including locally adaptive Savitzky-Golay filtering, content adaptive median hybrid filters, wavelet threshold denoising, anisotropic diffusion and noise reduction as an optimization problem. The results of denoising techniques effectiveness can be used for development of new denoising methods in the field of X-ray CT simulation.…”

Denoising of Monte Carlo dose calculations: Smoothing capabilities versus introduction of systematic bias

“…[6][7][8][9][10] However, in the present work it is illustrated that some evaluation methods ͑rms difference, 1% and 2% difference͒ sometimes underestimate the introduction of bias and rather illustrate the smoothing capability of the denoising algorithm, as possible bias effects are averaged out over the complete dose distribution. Therefore, a new evaluation method is introduced in the present work based on a sliding window superimposed on a difference dose distribution ͑reference dose-noisy/denoised dose͒, as the benefits of dose difference plots were already demonstrated by El Naqa et al 11 To illustrate the importance of this evaluation method, a new denoising technique ͑ANRT͒ is presented and compared with IRON for three challenging cases, 9 namely an electron and photon beam impinging on heterogeneous phantoms and two IMRT treatment plans of head-and-neck cancer patients to determine the clinical relevance of the obtained results.…”

“…Some groups focused on denoising of DVHs, 4,5 while others tried to denoise the whole threedimensional ͑3D͒ dose distribution map. 6-10 A comparison of different denoising techniques was presented by El Naqa et al, 11 illustrating that a reduction in calculation time can be obtained using denoising techniques by applying some of the test criteria described by Kawrakow,8 namely the visual inspection of isodose lines, improvement in root-mean-square differences, etc. However, as Deasy noticed in his work, 6 the denoised dose distribution will not necessarily coincide exactly with the corresponding "true" dose distribution, since a certain amount of systematic bias will be introduced ͑e.g., blurring of regions where dose information tends to vary sharply͒.…”