Decoupling the dosimetric impact of setup errors and anatomical deformations in image-guided radiotherapy using diagnostic-quality FBCT and glmmLasso regularization

Evaluation of dosimetric impact and plan robustness of MLC motion and setup errors in SRS for multiple brain metastases

Abstract Background HyperArc is an effective technique for treating multiple small brain metastases using high‐definition, precise radiation. However, the small size of individual lesions makes treatment vulnerable to inherent inaccuracies from beam‐limiting devices such as multi‐leaf collimators (MLCs), as well as patient setup errors. These uncertainties must be thoroughly evaluated prior to administering a single‐fraction, high‐dose HyperArc treatment to ensure optimal clinical outcomes. Purpose To comprehensively evaluate the robustness of HyperArc VMAT plans for multiple brain metastases (MBT) against (i) high‐definition MLC leaf positional errors and (ii) patient rotational and translational setup variations. Methods and materials A retrospective analysis of 34 MBT plans from 30 patients with 123 lesions treated using HyperArc VMAT plans was performed. Prescription doses (21–24 Gy) targeted PTVs (GTV + 1 mm), with each plan addressing 2–7 brain metastases (< 2 cm). D100% for GTV and V12Gy for normal brain were re‐evaluated under simulated (i) 20 MLC positional errors (0.2–1 mm) and (ii) 26 patient rotational (Yaw/Pitch/Roll by 1°) and 26 translational errors (SI/LR/AP by 1 mm). The impact of target distance to isocenter (DTI), target volume, and plan modulation factor on plan robustness was assessed. Replans with larger PTV margins, 1.5 and 2 mm, were investigated for a proper PTV margin for the worst‐case scenarios. Results Per our cohort, D100% of GTV is highly sensitive to both MLC and patient errors. Their dosimetric effects follow linear regressions: 1 mm shift reduces GTV D100% by 8%; 1 mm symmetric MLC openings increase GTV D100% by > 22%, while 1 mm symmetric closings decrease it by > 25.8% per 1 mm; Patient rotation/translation impacts GTV D100% by 0.04–1.28% per 1° or ∼7% per 1 mm. Normal brain V12Gy is only sensitive to MLC errors, increasing slightly with shifts and significantly with symmetric openings (+5cc/1 mm), but decreasing with closings (−2.77cc/1 mm). DTI shows minimal correlation with MLC shifts but is affected by patient setup uncertainties. Large targets amplify dose sensitivity to symmetric MLC errors, whereas plan modulation factor shows no significant impact. A minimum 1 mm margin is required to absorb these uncertainties, with 2 mm margins recommended in the worst cases to ensure treatment accuracy. Conclusions Overall, the target dose and V12Gy for HyperArc VMAT plans are more significantly affected by symmetric MLC opening and closing errors than by patient rotational setup errors. The pronounced effects of these errors highlight the importance of rigorous plan robustness verification, accounting for target size, margins, and DTI.

Improved synthetic CT image quality does not guarantee dosimetric agreement in CBCT-based adaptive proton radiotherapy

Objective. Synthetic CT (sCT) generation from cone-beam CT (CBCT) has emerged as a promising enabler for adaptive radiotherapy. While deep learning (DL) methods have demonstrated substantial improvements in image quality, their impact on proton beam therapy (PBT) dose calculation remains unclear, particularly for CBCT acquired on proton gantries, which are prone to increased artifacts and noise. Approach. In this study, we evaluated sCT generation from proton gantry CBCT using a denoising diffusion probabilistic model (DDPM) and a cycle-consistent generative adversarial network, and compared them against non-learning-based correction methods (corrected CBCT (corrCBCT) and virtual CT (vCT)) implemented in a commercial treatment planning system. Image quality (mean absolute error, peak signal-to-noise ratio, normalized cross-correlation) and dosimetric accuracy (gamma passing rate/GPR, dose-volume histogram metrics) were assessed in head-and-neck (H&N) and lung cohorts using clinically approved PBT plans. Main Results. All methods improved image quality compared to CBCT, with vCT achieving the best quantitative metrics (mean H&N MAE = 36.3HU; lung MAE = 39.9HU). However, image quality metrics did not correlate with dosimetric accuracy. In H&N cases, corrCBCT demonstrated inferior image metrics (MAE = 52.7HU) but achieved mean GPR of 98.4%, comparable to vCT and superior to DL-based methods. In lung cases, DDPM-generated sCTs achieved the highest mean GPR of 93.6% despite not having the best image metrics (MAE = 41.1HU). Dosimetric accuracy was highly sensitive to anatomical and Hounsfield Unit (HU) fidelity along beam paths, with beam angle significantly affecting GPR. DL-based methods effectively reduced artifacts but occasionally introduced anatomically inconsistent features that impacted dose calculation. Significance. Improvements in conventional image quality metrics did not necessarily translate to improved dosimetric accuracy in PBT. Instead, accurate anatomical representation and HU fidelity along beam paths were critical determinants of dose agreement. These findings highlighted the limitations of global image-based evaluation alone and underscored the need for dosimetric validation frameworks for sCT commissioning in PBT.

Physics in Medicine & Biology

Dosimetric and spatial impact of high-resolution SPACE MRI on target delineation for glomus tumor radiosurgery: a retrospective cohort study

Dosimetric correlation between exit dose maps and anatomical changes in head and neck cancer treatment on Halcyon‐E: A gamma analysis approach

Abstract Background Head and neck cancer (HNC) has a high incidence in Brazil. Although Image‐Guided Radiation Therapy (IGRT) improves treatment precision, anatomical changes during the treatment course may compromise dose distribution and require replanning. Purpose This study investigated the correlation between exit dose fluence and anatomical volumetric variations during HNC radiotherapy. Subsequently, a decision support methodology was proposed and validated to flag potential dosimetric deviations and guide objective replanning strategies. Methods A retrospective study was conducted with 11 patients, totaling 312 fractions treated on a Halcyon‐E linear accelerator. Volumetric variations were derived from cone‐beam computed tomography (CBCT) and exit dose fluence maps from each fraction were compared with the reference from the first treatment day using gamma analysis (γ_(1%/1 mm) to γ_(5%/5 mm); 10% threshold). Correlations between normalized volumetric variation (ΔV) and gamma indices were assessed using Kendall's tau coefficient (α = 0.05). A decision support methodology for replanning (γ_(1%/1 mm) < 80% and γ_(2%/2 mm) < 90%) was proposed and retrospectively validated in an independent cohort of 20 patients, comprising 618 treatment fractions. Results Strong negative correlations were observed between ΔV and gamma indices, with γ_(1%/1 mm) showing the highest sensitivity (τ = −0.83), followed by γ_(2%/2 mm) (τ = −0.75). Volumetric variations exceeding 5% were associated with significant dosimetric degradation. The validation correctly identified all replanned patients and detected potential indicators of dosimetric divergence not recognized in routine clinical practice. Conclusions The thresholds γ_(1%/1 mm) < 80% and γ_(2%/2 mm) < 90%, combined with ΔV > 5%, proved effective for guiding replanning decisions. The proposed methodology demonstrated high sensitivity for detecting dosimetric deviations, with potential to enhance radiotherapy safety and effectiveness.

Dosimetric analysis along core-needle biopsies in prostate HDR brachytherapy

Objective. Develop a probabilistic framework that maps high-dose-rate brachytherapy dose and dose-gradient magnitude onto MR-informed, TRUS-guided prostate core-needle biopsies, explicitly propagating millimetre-scale localization uncertainty. Approach. We reconstructed two representative second-fraction biopsy cores, voxelized them at 1mm3, and registered them to the clinical 1×1×0.5mm3 dose lattice. Spatial uncertainty from registration, segmentation, and biopsy length mismatch was modelled as NT=10 000 rigid translations sampled from N(0,σ2I) with σ=1.25mm, plus an independent axial shift along the core. For each voxel we obtained Monte Carlo distributions of dose D and dose-gradient magnitude G=‖∇D‖, distributional DVHs, Δ-metrics comparing nominal to probability-weighted summaries, and spatial summaries including axial length-scale dose-difference curves and voxel-pair contrast maps. Main results. Two exemplar second-fraction cores from a representative patient, selected as mechanistic exemplars of steep- versus low-gradient dose environments, showed right-skewed trialwise dose distributions in high-gradient segments, with widened uncertainty bands and larger nominal-summary discrepancies in the steep-gradient core. In that core, the typical per-trial perturbation in mapped dose was substantially larger (median |ΔDb,v,t|≈7Gy versus ≈1Gy in the low-gradient core). Length-scale curves increased with axial separation and then plateaued, indicating centimetre-scale along-core spatial structure in these exemplars; voxel-pair maps highlighted both near-uniform subsegments and voxel pairs achieving prescribed inter-voxel contrast. Significance. The framework converts biopsy localization uncertainty into probability-weighted voxel- and core-level dose descriptors and interpretable Δ-metrics that separate nominal bias from propagated variability. These estimands can be used to assess feasibility for dose-biology studies, guide selection or pairing of subsegments with specified homogeneity or contrast, and provide dose and gradient covariates for Raman, biochemical, and genomic assays, including settings where ex vivo biopsy orientation or integrity is partially lost. Cohort-level dosimetric characterization and robustness analyses are outside the scope of this exemplar-focused paper.