Accuracy of dose calculations on kV cone beam CT images of lung cancer patients

Purpose: To develop a clinically feasible method for dose calculations on cone beam CT (CBCT) images of two different vendors, and to determine the accuracy of these dose calculations for lung cancer patients. Methods: Lung cancer patients with CBCT imaging (n = 10 for Elekta, n = 6 for Varian) and a repeated planning CT scan on the same day were selected. For CBCT dose calculations, an adapted Hounsfield units-to-mass density table (HU table) was used which was obtained by comparing CT values of corresponding points on the CBCT and the repeated planning CT scan. Dose calculations with three different HU tables were compared: a patient-specific, a general thorax-CBCT, and the standard CT HU table. Planning CT data were used to compensate for the limited field of view (FOV) (Elekta) or scan length (Varian) of the CBCT. For evaluation, clinically relevant dose metrics were compared between the repeated CT and CBCT to assess the accuracy of dose calculations on CBCT for both vendors. Results: For both vendors, isodose lines and dose volume histograms were very similar between dose calculation on CBCT and CT. For Varian, average differences between CT and CBCT dose calculations were 2%-3% for most dose metrics when the standard CT HU table was used. A better agreement was observed when a thorax-CBCT HU table was used, with differences of 1%-2%. No added value was found by using a patient-specific HU table, showing similar results as the general thorax-CBCT HU table. For Elekta, the dose metrics showed large deviations when the CT HU table was used, but using a patient-specific HU table resulted in similar accuracy as for Varian CBCT dose calculations, with average differences between repeated CT and CBCT dose metrics below 3%, and for most dose metrics even below 2%. Conclusions: Differences between Elekta and Varian CBCT, including hardware, reconstruction software, HU calibration, FOV, and scan length, resulted in different challenges for CBCT dose calculations for the different vendors. For Elekta CBCT scans, the procedure with a patient-specific HU table resulted in similar accuracy as for Varian CBCT dose calculations with a general HU correction for all thorax patients. The vendor-specific corrective methods used in this study resulted in dose calculations feasible for treatment re-evaluation for both Elekta and Varian CBCT scans.