HU-to-density calibration curves using diagnostic CT for radiotherapy treatment planning: a feasibility study

Objective.Accurate radiotherapy (RT) dose calculation critically depends on reliable conversion of Hounsfield units (HUs) to density, traditionally achieved through scanner-specific calibration using dedicated simulation computed tomography (CT) systems. This study aimed to derive and validate proposed HU-to-density calibration curves (PCTs) from multi-vendor diagnostic CT (dCT) data, independent of scanner and acquisition parameters.Approach.Both HU-to-relative electron density (RED) and mass density (MD) calibration curves were generated using the CIRS 062 M phantom scanned on eight multi-vendor CT systems at 80-140 kilovoltage-peak (kV_p), including dual-energy modes. PCT (PCT-RED, PCT-MD) were derived by piecewise regression across the multi-kV_pdataset. 20 palliative 6 MV flattening-filter-free (FFF) volumetric modulated arc therapy and intensity-modulated RT plans were recalculated in Eclipse v18 using all these curves with Acuros XB (dose-to-medium, MD-based) and the anisotropic analytical algorithm (RED-based). Dosimetric comparisons assessed: (1) PCT validation versus scanner-kV_p-specific calibrations; (2) intra-scanner effects of varying kV_p(versus 120/110 kV_pper scanner). Evaluation included dose-volume histogram (DVH) metrics-planning target volume, organs at risk (OARs), body maximum dose, conformity index-and 3Dγanalysis (2%/2 mm, 1%/1 mm) between calculated dose distributions. Statistical differences were assessed using the Wilcoxon signed-rank test (α= 0.05) with effect sizes.Main results.Differences between plan dose distributions obtained with PCT and those from scanner-specific curves were minimal. Although several DVH metrics were statistically different (p< 0.05), all differences remained within clinical acceptability. Across both PCT validation and intra-scanner kV_pvariation scenarios, median dosimetric differences were  < 1.0% for both targets and OARs, regardless of HU variability. All plans achievedγpassing rates  > 99% (1%/1 mm), confirming robustness.Significance.We propose a CT calibration model based on dCT for RT. This multi-vendor approach provides default curves when scanner-specific calibration is impractical, facilitating resource-efficient, simulation-free RT workflows via dCT-based treatment planning, while maintaining dosimetric accuracy across the evaluated dataset.