Performance of dose calculation algorithms from three generations in lung SBRT: comparison with full Monte Carlo-based dose distributions

Abstract

The accuracy of dose calculation is a key challenge in stereotactic body radiotherapy (SBRT) of the lung. We have benchmarked three photon beam dose calculation algorithms--pencil beam convolution (PBC), anisotropic analytical algorithm (AAA), and Acuros XB (AXB)--implemented in a commercial treatment planning system (TPS), Varian Eclipse. Dose distributions from full Monte Carlo (MC) simulations were regarded as a reference. In the first stage, for four patients with central lung tumors, treatment plans using 3D conformal radiotherapy (CRT) technique applying 6 MV photon beams were made using the AXB algorithm, with planning criteria according to the Nordic SBRT study group. The plans were recalculated (with same number of monitor units (MUs) and identical field settings) using BEAMnrc and DOSXYZnrc MC codes. The MC-calculated dose distributions were compared to corresponding AXB-calculated dose distributions to assess the accuracy of the AXB algorithm, to which then other TPS algorithms were compared. In the second stage, treatment plans were made for ten patients with 3D CRT technique using both the PBC algorithm and the AAA. The plans were recalculated (with same number of MUs and identical field settings) with the AXB algorithm, then compared to original plans. Throughout the study, the comparisons were made as a function of the size of the planning target volume (PTV), using various dose-volume histogram (DVH) and other parameters to quantitatively assess the plan quality. In the first stage also, 3D gamma analyses with threshold criteria 3%/3mm and 2%/2 mm were applied. The AXB-calculated dose distributions showed relatively high level of agreement in the light of 3D gamma analysis and DVH comparison against the full MC simulation, especially with large PTVs, but, with smaller PTVs, larger discrepancies were found. Gamma agreement index (GAI) values between 95.5% and 99.6% for all the plans with the threshold criteria 3%/3 mm were achieved, but 2%/2 mm threshold criteria showed larger discrepancies. The TPS algorithm comparison results showed large dose discrepancies in the PTV mean dose (D50%), nearly 60%, for the PBC algorithm, and differences of nearly 20% for the AAA, occurring also in the small PTV size range. This work suggests the application of independent plan verification, when the AAA or the AXB algorithm are utilized in lung SBRT having PTVs smaller than 20-25 cc. The calculated data from this study can be used in converting the SBRT protocols based on type 'a' and/or type 'b' algorithms for the most recent generation type 'c' algorithms, such as the AXB algorithm.

Figure 1

The isodose distribution (a) calculated by the MC model for the plan with medium‐sized PTV (15.1 cc) in transversal isocenter plane. Dose levels from the outermost isodose curve to the innermost one are 5 Gy, 15 Gy, 25 Gy, 35 Gy, 45 Gy (the prescription dose), and 55 Gy. The two innermost contours represent the PTV and the GTV, respectively. A sagittal isocenter plane (b) showing colored areas, where the gamma calculation typically failed (threshold criteria $3\mathrm{\%}/3\hspace{0.17em}\text{mm}$). The graphs along the lines show the percentage dose difference between the MC‐recalculated AXB‐plan and the original AXB‐plan through the largest dose differences in orthogonal directions. Lines in the profiles represent the PTV and the GTV boundaries.

Figure 2

Comparison of the results for the algorithms as a function of the PTV size. Lines with triangles represent the differences for ${\mathrm{D}}_{95\mathrm{\%}},{\mathrm{D}}_{50\mathrm{\%}}$, and ${\mathrm{D}}_{5\%}$ between the AAA and the AXB‐recalculated AAA plan, and lines with circles represent the differences for ${\mathrm{D}}_{95\mathrm{\%}},{\mathrm{D}}_{50\mathrm{\%}}$, and ${\mathrm{D}}_{5\%}$ between the PBC algorithm and the AXB‐recalculated PBC plan. Solid lines are for ${\mathrm{D}}_{95\%}$, grey lines for ${\mathrm{D}}_{50\%}$, and dashed lines for ${\mathrm{D}}_{5\%}$.

Figure 3

Comparison of the results for the algorithms as a function of the PTV size. Solid line with triangles represents the difference in ${\mathrm{D}}_{2\text{cm}}$ between the AAA and the AXB‐recalculated AAA plan, and solid line with circles represents the differences in ${\mathrm{D}}_{2\text{cm}}$ between the PBC algorithm and the AXB‐recalculated PBC plan.

Figure 4

Comparison of the results for the algorithms as a function of the PTV size. Solid line with triangles represents the difference in ${\mathrm{V}}_{30\%}$ for the ipsilateral lung with the GTV subtracted between the AAA and the AXB‐recalculated AAA plan, and solid line with circles represents the difference for the same DVH parameter between the PBC algorithm and the AXB‐recalculated PBC plan.

Figure 5

Comparison of the results for the algorithms as a function of the PTV size. Solid line with triangles represents the difference in CI between the AAA and the AXB‐recalculated AAA plan, and solid line with circles represents the differences for the CI between the PBC algorithm and the AXB‐recalculated PBC plan. The reader should note the absolute scale used.