Purpose: To investigate dosimetric parameters of 4 algorithms for lung cancer patients treated with flattening filter free stereotactic body radiotherapy using HU as a surrogate for PTV density.
Methods: We studied 15 lung cancer patients treated with SBRT using 6xFFF. Dose regimen was 12 Gy/fx in 4 fractions. Treatment plans were created using the pencil beam convolution in iPlan with IMRT and recalculated using identical parameters with iPlan’s Monte Carlo Algorithm (MC), Analytical Anisotropic Algorithm (AAA), AcurosXB (AXB), and a collapsed cone convolution algorithm (Sun Nuclear Dose Check (SDC)). These lung-island tumors had ITV volumes of 12.44 ± 13.90 cm3. Data sets were indexed by average PTV density and several dosimetric indices for the target and normal lung tissue were evaluated. % difference relative to MC calculation is reported.
Results: Average tumor densities were -594.3 ± 146.7 (-828.9 to -189.9 HU). The ΔD95 was -8.35 ± 3.22% (-13.00% to -2.99%), 1.16 ± 2.85% (6.47% to -1.93%), and -1.52 ± 1.55% (-4.58% to 1.44%) for SDC, AAA, and AXB respectively across all densities. Agreement with MC was moderately (0.576 ≤ r ≤ 0.713) correlated with density with a p < 0.05. A threshold for clinical agreement between the algorithms was noted around -450 HU. Below -600 HU ΔD95 was -11.39 ± 1.63%, 3.25 ± 3.29%, and -1.41 ± 1.71% for SDC, AAA, and AXB respectively. Of note, AAA and AXB were closer to MC than SDC but AXB and SDC both predicted under coverage relative to MC with AAA predicting over coverage. Lung V20 and V5 showed a tight spread with max difference within 2.0% of MC.
Conclusion: Preliminary results suggest that target coverage agreement has moderate correlation with target density and a threshold for clinical agreement might be established. Lung V20 and V5 showed a much tighter spread of agreement.
Collapsed Cone Superposition, Monte Carlo, Inhomogeneity Corrections