Benchmarking of 4D CBCT Based Dose Calculation Against Measurements in a Lung Phantom
Dumas M, Rusu S, Chetty I, and Wen N. Benchmarking of 4D CBCT Based Dose Calculation Against Measurements in a Lung Phantom. J Med Phys 2019; 46(6):e146-e147.
J Med Phys
Purpose To quantify 4DCBCT dose calculation accuracy of lung SBRT plans by comparing the calculated dose to the measured dose on an anthropomorphic phantom. Methods: 4DCBCTs were taken of a CIRS lung phantom using the RPM system (Varian). Scans were acquired with our clinical lung scan protocol, which compromises between SNR and imaging dose (gantry rotation speed of 3 degrees/s). Images were binned into 10 phases, and dose was calculated on each phase using the AAA algorithm in Eclipse v15.5(Varian). Computed dose for each phase was summed to give the total dose per beam. Breathing motion was simulated by moving the couch 1.5 cm superior/inferior and 0.5 cm anterior/posterior. Developer mode was used to transpose the couch while the beams were delivered. For comparison, 4DCBCTs and measurements were acquired without couch motion to determine if motion affects the dose accuracy. Doses were measured with an IBA CC01 chamber. Patient plan prescriptions ranged from 48 to 56 Gy in 4 fractions. Results: Mean measured versus calculated dose differences were-1.0 ± 1.6% for deliveries with motion and-0.1 ± 1.0% for static deliveries. For each plan, dose calculations were within 3% of measurement. The overall mean coefficient of variation of calculated dose over the 10 phases was 3.8 ± 1.4% per beam, which gives insight into the dose uncertainty caused by limited projection data per phase. Conclusion: This study demonstrated the overall 4DCBCT dose calculation accuracy on an anthropomorphic phantom is within 4.1% (95% CI), with per phase dose calculation uncertainty of 6.5% (95% CI). Calculation and measurement dose differences between the phantom moving and static were within delivery uncertainty. Future work will examine using iterative CBCT (iCBCT) reconstruction to decrease noise and artifacts, with the hypothesized goal of decreasing calculation uncertainty.