Answers are grounded in this paper and cite their sources.
Objectives: Accurate setup margin (SM) estimation is critical in radiotherapy to ensure target coverage while sparing organs at risk. Traditional methods like van Herk’s formalism assume that setup errors follow a normal or Gaussian distribution, which may not hold for small or skewed datasets and may result in under- or over-estimated SMs.This study evaluates an unconventional method based on the non-Gaussian percentile approach against the van Herk formalism for SM estimation across multiple tumor sites using electronic portal imaging device (EPID) setup data. Material and Methods: Eighty patients (20 per site: brain, head and neck, thorax, pelvis) treated with external beam radiotherapy were analyzed retrospectively. Setup errors were measured via EPID-based imaging. Conventional SMs were calculated using van Herk’s method (2.5Σ + 0.7σ), while the unconventional method derived SMs directly from the setup-error histogram, i.e., the 5th–95th percentiles (90% range of setup errors, RSE) of per-patient mean daily shifts.Normality was assessed using Shapiro-Wilk(S-W) tests, skewness/kurtosis analysis, and Q-Q plots. Methods were compared using the Wilcoxon signed-rank test (p<0.05) and effect sizes (Cohen’s d)reported without claiming clinical significance. Results: Unconventional SMs were consistently larger than conventional SMs (overall medians: 4.21 vs 2.71 mm; p<0.05), with the largest differences in thorax (Δ up to 3.63 mm) and pelvis (Δ up to 2.17 mm). Several datasets were non-normal by S–W and/or skewness/kurtosis. As a practical benchmark, margin differences ≥2 mm were considered clinically meaningful for interpretation, given typical planning target volume (PTV) rounding and published cone-beam computed tomography (CBCT)-era margins. Conclusion: In non-Gaussian or limited-data scenarios, the percentile-based method yields more conservative SMs than van Herk. Given EPID’s inability to capture organ motion/rotation, validation with CBCT or 4D imaging is recommended before clinical implementation.
No takes yet. Share an insight, caveat, or question.
Mondal et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: