Abstract Book

S1013

ESTRO 37

Conclusion AP for VMAT APBI was proven to be at least equivalent and overall superior to manual planning. The use of automated planning reduces the planning workload while allowing improvements in plan quality. EP-1874 Treatment time reduction in robotic stereotactic body radiation therapy (SBRT) S. Breedveld 1 , L. Rossi 1 , M. Keijzer 2 , B. Heijmen 1 1 Erasmus Medical Center Rotterdam Daniel den Hoed Cancer Center, Radiation Oncology, Rotterdam, The Netherlands 2 Delft University of Technology, Applied Mathematics, Delft, The Netherlands Purpose or Objective In robotic SBRT, a treatment plan is generated to optimally irradiate the tumour with maximal avoidance of OARs, using a large number of non-coplanar beams/nodes. The overall travel time of the robot between the nodes can be substantial. Currently, travel time is not optimised in the planning process. In this study, we investigate opportunities to generate for each patient an alternative plan with similar quality but a substantially reduced travel time. Material and Methods The test cohort consisted of 30 prostate cancer patients and 15 lung cancer patients, previously treated in our center with the CyberKnife robotic treatment unit (Accuray Inc, Sunnyvale, USA). For this study, 25-node treatment plans were first generated for each patient using our in-house developed TPS for automated multi- criterial plan generation. An alternative (fast) plan was then generated with a minimised travel time. For all plans (original and fast), a Travelling Salesman Problem (TSP) approach was used to establish the order of the 25 selected nodes for a minimal travel time, consistent with the commercial CyberKnife software. The travel time reduction strategy investigated for each node if a nearby node (angle < 15 degrees) leads to a significantly shorter travel time. Per node, 3-7 nearby nodes were explored. If the travel time could be reduced by more than 3.5 seconds, the original node was replaced with the faster alternative. This was sequentially repeated for each node. For the new node-set, the TSP is solved again. Finally, a new plan optimisation was performed for the alternative, faster travel path using the automated multi- criteria treatment planning system. Results The average travel time per patient was reduced with 22% (range 11%-30%) for the prostate cases and 20% (range 10%-30%) for the lung cases. Computation time required to find the faster node-set was 21 seconds on average. The upper panels in the two figures show for prostate and lung cancer patient specific reductions in travel time. Differences in plan quality between slow and fast plans were clinically negligible (lower panels of the figures).

No statistically significant differences in organs at risk doses were observed, although a trend in the reduction of ipsilateral breast V 15Gy (considered clinically significant by the scoring physician) was observed (average reduction in V 15Gy 3%, range 1-11%). Data are presented in Figure 1.

Overall, the modulation degree was reduced with AP compared to manual (25% improvement in modulation complexity score). MUs were also reduced (about 30% on average), while treatment time was slightly increased (on average by 29 s). Concerning dosimetric verifications, no statistically significant differences were observed in γ passing rates. In the blind scoring, for 10/20 plan comparisons, the AP plan was considered superior to the manual plan, with high clinical relevance. In 8/20 the AP plan was considered better but with minor clinical relevance, while in 2/20 the plans were considered to be equivalent.