Session Item

Good fine-tuning of TPS modelling parameters is crucial for creating the optimal beam model, particularly with the ever-increasing accuracy required for advancing techniques. Challenges arise in balancing the trade-off between multiple parameters. How well this is done will often depend on the experience of the physicist, particularly in vendor neutral TPS such as Pinnacle and RayStation where there can be numerous parameters to optimise. The actual parameter values used have been shown to vary widely between centres [1]. We investigated whether using community generated standardised parameters could produce a good quality model in comparison to a well-benchmarked model.

This study focussed on Philips Pinnacle TPS v16 and Elekta VersaHD with Agility MLC. Beam modelling parameters investigated were: Effective source size perpendicular & parallel to gantry axis, MLC transmission, tongue & groove width, flattening filter scatter source Gaussian height & width, and Rounded Leaf Tip Radius. Based on published data from 71 cancer centres [1], 50^th percentile values for each parameter were used to generate a new beam model (Median Model) to compare with the Original Model (see Table 1). 11 step&shoot IMRT and VMAT plans of varying geometrical complexity were created using the Original Model and recalculated with fixed plan settings using the Median Model. These included head & neck, prostate, and C-shape geometric cases designed with varying modulation. Dose calculations used the CCC algorithm and 2mm dose grid spacing. Measurements were performed using: Octavius4D (1500Detector), OctaviusII (729Detector), anthropomorphic CIRS phantom with EBT3 film, and SNC ArcCHECK phantom. Analysis of the datasets was done using gamma index (γ) with standardised calculation settings; global γ with dose difference relative to max dose, 20% lower dose threshold, 3%/3mm and 2%/2mm criteria. γ passing rate as % of points with γ<1 was recorded. Statistical tests for differences between the Median and the Original Models were based on Wilcoxon signed-rank test.

Table 2 shows the median 2%/2mm and 3%/3mm γ pass rates for both models for all 11 plans. For the CIRS phantom, OctaviusII and Octavius4D, the median γ pass rates were similar with no statistically significant difference between the Median and Original Models (p>0.05 in all cases). The ArcCHECK results showed a small but statistically significant increase in the γ pass rate between the Median and Original Models.

The results of this study indicate that beam modelling parameters based on 50^th percentile community-generated data compare well against a well-benchmarked model. These parameters may potentially provide an easier starting point for the input parameters or pave the way for standardisation in beam modelling and thereby improving multicentre dose calculation accuracy which is of particular importance in clinical trials.

References

[1] Glenn et al, 2019. DOI: 10.3252/pso.eu.ESTRO38.2019