Probabilistic Quality Assurance of Biopsy DVH Dose Thresholds In HDR Prostate Brachytherapy Using Monte Carlo Localization Uncertainty
Abstract
Purpose
To perform cohort-level probabilistic quality assurance of biopsy dose thresholds in HDR prostate brachytherapy under a fixed localization-uncertainty model and relate robustness to nominal distance from threshold.
Methods
Twenty-seven MR-US guided biopsies from 15 HDR monotherapy patients were reconstructed as chains of 1 mm voxels and sampled on the clinical HDR dose lattice. Localization uncertainty was modeled with 10,000 Monte Carlo trials per biopsy using isotropic 3D Gaussian translations (σ = 1.25 mm) plus an axial tissue-deficit offset. For each voxel we compared nominal dose and dose-gradient magnitude with Monte Carlo propagated summaries. For each biopsy we evaluated DVH metrics D2%, D50%, D98%, and V150%, and for four robustness rules modeled Monte Carlo pass probabilities as functions of nominal distance from threshold, with geometric, spatial, and radiomics-style predictors as secondary covariates. Intra-core dose decorrelation was estimated from voxel-pair dose differences versus axial separation.
Results
Nominal voxel doses tended to exceed Monte Carlo samples, with a pooled mean absolute per-trial dose difference of about 7.6 Gy. Margin-only logistic models explained most of the variability in pass probabilities. The nominal buffers required to achieve 95 percent Monte Carlo pass probability were approximately 6 Gy for D98%, 21 Gy for D50%, 49 Gy for D2%, and 49 percentage points for V150%. Adding geometric, spatial, or radiomics predictors changed these margins modestly. Along biopsy cores, mean absolute dose differences between voxel pairs approached about 10 Gy at separations of 1 to 1.5 cm, indicating decorrelation on that length scale.
Conclusion
Under an explicit millimetre-scale localization-uncertainty model, nominal biopsy DVH metrics systematically overestimate Monte Carlo propagated doses and do not by themselves convey robustness. For the thresholds studied, high confidence required sizeable nominal margins above rule values. These findings support routine Monte Carlo based, threshold-focused reporting when interpreting biopsy dosimetry in HDR prostate brachytherapy.