This study develops a probabilistic framework to map high-dose-rate (HDR) brachytherapy dose and dose-gradient magnitude onto MR-informed, TRUS-guided prostate core-needle biopsies while explicitly propagating millimetre-scale localization uncertainty. The aim is to generate probability-weighted voxel- and core-level dose descriptors that separate nominal (deterministic) dose estimates from variability introduced by biopsy localization and reconstruction uncertainties.
Two representative second-fraction biopsy cores were reconstructed and voxelized, then registered to the clinical dose lattice. The registration and mapping workflow used the clinical treatment-dose lattice as the reference space onto which biopsy-derived voxels were mapped. Spatial uncertainty arising from several sources — image registration, segmentation differences, and mismatch between biopsy length and mapped segment length — was modelled and explicitly incorporated into the mapping process.
The study focused on method development and mechanistic exemplars rather than cohort-level statistics. Two exemplar cores were selected from a representative patient to demonstrate the method in contrasting dose-gradient environments: one core sampled a steep-gradient region and the other sampled a low-gradient region.
Spatial uncertainty was modelled as rigid translations sampled from a distribution (details of the distributional parameters were used in the study but are not reported in the abstract) plus an independent axial shift along the biopsy core. This composite perturbation model was applied to each voxel within the reconstructed core.
A Monte Carlo sampling approach produced trialwise perturbed voxel placements. For each Monte Carlo trial the clinical dose lattice was sampled at the perturbed voxel positions to produce distributions of per-voxel dose and dose-gradient magnitude. These per-trial samples yield empirical distributions for dose-related quantities at voxel and core levels, capturing propagated localization uncertainty.
From the Monte Carlo–derived distributions the authors computed several probability-aware summaries and diagnostic metrics:
The two exemplar second-fraction cores showed distinct dosimetric behavior consistent with their dose-gradient environments. In the core sampling a steep dose-gradient environment, trialwise dose distributions in high-gradient segments were right-skewed, showed widened uncertainty bands, and produced larger discrepancies between nominal (single-position) summaries and the probability-weighted summaries obtained from Monte Carlo sampling.
In contrast, the core sampling a low-gradient environment exhibited narrower uncertainty bands and smaller nominal-versus-probability-weighted discrepancies. The abstract states that the typical per-trial perturbation in mapped dose was substantially larger in the steep-gradient core than in the low-gradient core; however, the abstract does not report the specific numeric median or comparative values.
Axial length-scale curves computed from the trialwise data increased with axial separation and then plateaued, indicating centimetre-scale along-core spatial structure in these exemplar cores. This suggests that dose variability and spatial contrast along a biopsy core are not purely local at millimetre scale but show coherent structure over centimetre distances in the shown exemplars.
Voxel-pair contrast maps highlighted both near-uniform subsegments—regions where dose and gradient remain relatively homogeneous across voxels—and voxel pairs that meet prescribed inter-voxel contrast thresholds. These spatial summaries support targeted selection or pairing of subsegments for downstream assays that require known dose homogeneity or contrast.
The framework converts biopsy localization uncertainty into interpretable, probability-weighted dose descriptors at voxel and core scales and provides metrics that separate nominal bias from propagated variability. Potential applications identified in the abstract include:
The authors note that this publication is exemplar-focused: cohort-level dosimetric characterization and robustness analyses across a patient population were explicitly outside the scope of the paper as described in the abstract. Where the abstract references specific numeric comparisons (for example median per-trial perturbations), those numeric details are not reported in the abstract and would require consultation of the full text for exact values and distributional parameters.
This work presents a Monte Carlo–based probabilistic mapping pipeline that propagates millimetre-scale localization uncertainty into voxel-level and core-level dosimetric estimates for prostate HDR brachytherapy mapped onto MR-informed, TRUS-guided core-needle biopsies. The method yields distributional DVHs and spatially resolved metrics that can help separate nominal mapping bias from uncertainty-driven variability and enable more informed dose–biology correlation studies and targeted subsegment selection for downstream assays.