Session Item

Analysis of modulation complexity has never been performed for Robotic MLC-based plans. In this study several complexity metrics, mostly used for IMRT and VMAT plans, were computed adapting definitions to CyberKnife (CK) plans. The purposes were i) to compare the complexity of plans by two optimization algorithms, ii) to analyse relationships between metrics and iii) correlations between metrics and patient-specific quality assurance (PSQA) results.

An in-house program was developed in R to compute 5 metrics for CK MLC plans exported in xml format: Modulation Complexity Score (MCS), Edge Metrics (EM), Plan Modulation (PM), Plan Irregularity (PI) and weighted Leaf Gap (LG).  MCS and PM definitions were adapted to CK plan characteristics, i.e. several non-coplanar beams and mostly 1-2 segments per beams. For MCS, two different solutions (a-b) were examined. By definition MCS and LG values decrease with increasing modulation, whereas the relation is reversed for EM, PI, and PM. All Modulation indices were computed over a total of 80 clinically acceptable plans, created for Liver, Pancreas, Prostate and Spine SBRT. Among these plans, 24 Sequential and 24 VOLO plans, created for the same cases using the same clinical protocols, were compared in terms of modulation complexity. Pearson’s r was used to explore dependencies between modulation indices over the totality of VOLO plans (56). Relationships of each metric with PSQA gamma passing rates for 32 plans were also analysed. Correlation was regarded as weak for absolute r values in the range 0.2 - 0.39, moderate 0.4 - 0.59, strong 0.6 -0.79 and very strong 0.8-1.

Average beam number per plan and segment number per beam were respectively 44.0 and 1.68. Mean values and standard deviations of modulation metrics over all treatment sites and for both optimizers are shown in table1. When compared to VOLO, Sequential plans exhibited a higher complexity showing lower MCS(a,b) and LG values and higher EM, PM and PI. Differences were significant for 4/5 metrics (Wilcoxon p<0.01). Among treatment sites, the lowest modulation degree was found for liver plans whereas the highest complexity was found for spinal plans. A very strong significant negative correlation (p<0.01) was observed between MCS(a,b) and PM, as well as between EM and LG (table2). (2%,1mm) gamma pass-rate (range 83%-99.1%) did not correlate significantly with any metrics and Pearson’s r absolute values were below 0.4.

For the first time several complexity indices were computed for CK MLC plans. CK plan characteristics required to adapt the computation of two metrics. The computed indices permitted to compare modulation complexity of plans created by two optimizing systems. Further data are needed to examine correlations between complexity metrics and PSQA pass-rates. This study set a basis to compare CK plans modulation in a multicentre, multi-platform context.