On the use of a convolution-superposition algorithm for plan checking in lung stereotactic body radiation therapy

Abstract

Stereotactic body radiation therapy (SBRT) aims to deliver a highly conformal ablative dose to a small target. Dosimetric verification of SBRT for lung tumors presents a challenge due to heterogeneities, moving targets, and small fields. Recent software (M3D) designed for dosimetric verification of lung SBRT treatment plans using an advanced convolution-superposition algorithm was evaluated. Ten lung SBRT patients covering a range of tumor volumes were selected. 3D CRT plans were created using the XiO treatment planning system (TPS) with the superposition algorithm. Dose was recalculated in the Eclipse TPS using the AAA algorithm, M3D verification software using the collapsed-cone-convolution algorithm, and in-house Monte Carlo (MC). Target point doses were calculated with RadCalc software. Near-maximum, median, and near-minimum target doses, conformity indices, and lung doses were compared with MC as the reference calculation. M3D 3D gamma passing rates were compared with the XiO and Eclipse. Wilcoxon signed-rank test was used to compare each calculation method with XiO with a threshold of significance of p < 0.05. M3D and RadCalc point dose calculations were greater than MC by up to 7.7% and 13.1%, respectively, with M3D being statistically significant (s.s.). AAA and XiO calculated point doses were less than MC by 11.3% and 5.2%, respectively (AAA s.s.). Median and near-minimum and near-maximum target doses were less than MC when calculated with AAA and XiO (all s.s.). Near-maximum and median target doses were higher with M3D compared with MC (s.s.), but there was no difference in near-minimum M3D doses compared with MC. M3D-calculated ipsilateral lung V20 Gy and V5 Gy were greater than that calculated with MC (s.s.); AAA- and XiO-calculated V20 Gy was lower than that calculated with MC, but not statistically different to MC for V5 Gy. Nine of the 10 plans achieved M3D gamma passing rates greater than 95% and 80%for 5%/1 mm and 3%/1 mm criteria, respectively. M3D typically calculated a higher target and lung dose than MC for lung SBRT plans. The results show a range of calculated doses with different algorithms and suggest that M3D is in closer agree-ment with Monte Carlo, thus discrepancies between the TPS and M3D software will be observed for lung SBRT plans. M3D provides a useful supplement to verification of lung SBRT plans by direct measurement, which typically excludes patient specific heterogeneities.

Figure 1

Isodose lines for (a) XiO, (b) AAA, (c) M3D, and (d) Monte Carlo for an example patient. The ITV (beige) and PTV (teal) are shown as colourwash. Note in the M3D and Monte Carlo images, the higher dose in the target (32.5 Gy isodose line) and the larger distance between the 26 Gy isodose line and the PTV contour.

Figure 2

Point doses relative to Monte Carlo for each target at the calculation point in the center of the PTV.

Figure 3

PTV D2% (top), median (middle), and D98% (bottom) values for each of the three calculation methods relative to Monte Carlo calculated dose.

Figure 4

Ipsilateral lung V20 Gy (left) and V5 Gy (right) for each of the three calculation methods relative to Monte Carlo.

Figure 5

Conformity indices for 100% (left) and 50% (right) isodose lines.

Figure 6

The gamma distribution map comparing M3D with XiO (left) and Eclipse (right) for Patient 2 at $5\%/1 \text{mm}$