Abstract
PURPOSE: DirectDensity enables tube voltage-independent, density-calibrated computed tomography (CT) images for treatment planning and is therefore increasingly used in radiotherapy facilities. In accelerated emergency workflows, diagnostic CT (dCT) data are often acquired without DirectDensity. It remains unclear whether using a DirectDensity calibration curve in this context leads to major dose deviations. This study therefore investigates the impact of different CT calibration curves on treatment planning for an artificial intelligence-supported emergency workflow in a phantom-based end-to-end test. To validate the transferability of phantom measurements to clinical scenarios, a clinically implemented emergency radiotherapy workflow using diagnostic CT data is retrospectively evaluated in four previously treated patients representing different electron density scenarios. METHOD: On Varian's Ethos 1.1 platform (Palo Alto, CA, USA), the entire treatment chain was simulated in an adaptive workflow using an Alderson-Rando phantom scanned on three diagnostic CTs. This simulation, which lacks a dedicated calibration curve for the treatment planning software, was conducted as an end-to-end test and was compared with the results of a radiotherapy-calibrated planning CT scan. The phantom featured a purpose-built insert for a Farmer 30013 chamber from PTW (Freiburg, Germany). The end-to-end test was performed using three different calibration curves (DirectDensity, 120 kV, and a custom hybrid of DirectDensity and 120 kV). Gamma analysis was performed for the retrospectively evaluated patient plans (calculated with the three different CT curves) using three different gamma criteria to assess the robustness of dose agreement under clinically realistic conditions. RESULTS: The point doses at the effective point of measurement of the ionization chamber, determined using the treatment planning software, were compared with the measured dose values. All tested approaches met the requirements of ICRU Report 24 (± 5% of the prescribed dose) and could be used in emergency workflows with dCTs. Despite minor differences (0.1%-0.7% better accuracy), the hybrid curve achieved the best results. Recalculations of four representative patients confirmed the trends observed in phantom measurements. 2D array evaluations showed that dose differences between calibration curves remained within clinically defined acceptance limits, with the hybrid curve consistently performing best. CONCLUSION: Using a DirectDensity calibration curve in emergency workflows with image data acquired without DirectDensity leads to clinically acceptable treatment plans. Dose deviations can be reduced by carefully evaluating the entire treatment chain. A self-generated hybrid CT density curve shows the best results, although deviations remain larger (< 1.5%) than when planning directly on a calibrated planning CT. The combined phantom- and patient-based analysis demonstrates that calibration curve selection can influence dose distributions under clinically realistic conditions, supporting the superior performance of the hybrid curve and highlighting clinical relevance beyond technical feasibility.