Premium
Sensitivity of postplanning target and OAR coverage estimates to dosimetric margin distribution sampling parameters
Author(s) -
Xu Huijun,
Gordon J. James,
Siebers Jeffrey V.
Publication year - 2011
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.3544364
Subject(s) - dosimetry , sampling (signal processing) , margin (machine learning) , mathematics , monitor unit , isotropy , nuclear medicine , metric (unit) , physics , computer science , medicine , optics , detector , operations management , machine learning , economics
Purpose: A dosimetric margin (DM) is the margin in a specified direction between a structure and a specified isodose surface, corresponding to a prescription or tolerance dose. The dosimetric margin distribution (DMD) is the distribution of DMs over all directions. Given a geometric uncertainty model, representing inter‐ or intrafraction setup uncertainties or internal organ motion, the DMD can be used to calculate coverage Q , which is the probability that a realized target or organ‐at‐risk (OAR) dose metricD vexceeds the corresponding prescription or tolerance dose. Postplanning coverage evaluation quantifies the percentage of uncertainties for which target and OAR structures meet their intended dose constraints. The goal of the present work is to evaluate coverage probabilities for 28 prostate treatment plans to determine DMD sampling parameters that ensure adequate accuracy for postplanning coverage estimates. Methods: Normally distributed interfraction setup uncertainties were applied to 28 plans for localized prostate cancer, with prescribed dose of 79.2 Gy and 10 mm clinical target volume to planning target volume (CTV‐to‐PTV) margins. Using angular or isotropic sampling techniques, dosimetric margins were determined for the CTV, bladder and rectum, assuming shift invariance of the dose distribution. For angular sampling, DMDs were sampled at fixed angular intervals ω (e.g., ω = 1 ° , 2 ° , 5 ° , 10 ° , 20 ° ). Isotropic samples were uniformly distributed on the unit sphere resulting in variable angular increments, but were calculated for the same number of sampling directions as angular DMDs, and accordingly characterized by the effective angular incrementω eff. In each direction, the DM was calculated by moving the structure in radial steps of size δ ( = 0.1 , 0.2 , 0.5 , 1 mm )until the specified isodose was crossed. Coverage estimation accuracy Δ Q was quantified as a function of the sampling parameters ω orω effand δ . Results: The accuracy of coverage estimates depends on angular and radial DMD sampling parameters ω orω effand δ , as well as the employed sampling technique. Target| Δ Q | < 1 % and OAR| Δ Q | < 3 % can be achieved with sampling parameters ω orω eff = 20 ° , δ = 1 mm . Better accuracy (target| Δ Q | < 0.5 % and OAR| Δ Q | < ∼ 1 % ) can be achieved with ω orω eff = 10 ° , δ = 0.5 mm . As the number of sampling points decreases, the isotropic sampling method maintains better accuracy than fixed angular sampling. Conclusions: Coverage estimates for post‐planning evaluation are essential since coverage values of targets and OARs often differ from the values implied by the static margin‐based plans. Finer sampling of the DMD enables more accurate assessment of the effect of geometric uncertainties on coverage estimates prior to treatment. DMD sampling with ω orω eff = 10 ° and δ = 0.5 mm should be adequate for planning purposes.