A preliminary study on a multi-resolution-level inverse planning algorithm for Gamma Knife radiosurgery
Manual forward planning for GK radiosurgery is complicated and time-consuming, particularly for cases with large or irregularly shaped targets. Inverse planning eases GK planning by solving an optimization problem. However, due to the vast search space, most inverse planning algorithms have to decouple the planning process to isocenter preselection and sector duration optimization. This sequential scheme does not necessarily lead to optimal isocenter locations and hence optimal plans. In this study, we attempt to optimize the isocenter positions, beam shapes and durations simultaneously by proposing a multi-resolution-level (MRL) strategy to handle the large-scale GK optimization problem. In our approach, several rounds of optimizations were performed with a progressively increased spatial resolution for isocenter candidate selection. The isocenters selected from last round and their neighbors on a finer resolution were used as new isocenter candidates for next round of optimization. After plan optimization, shot sequencing was performed to group the optimized sectors to deliverable shots supported by GK treatment units. We have tested our algorithm on 6 GK cases previously treated in our institution (2 meningioma cases, 3 cases with single metastasis and 1 case with 6 metastases). Compared with manual planning, achieving same coverage and similar selectivity, our algorithm improved the gradient index averagely from 3.1 to 2.9 and reduced the maximum dose of brainstem from 8.0Gy to 5.6Gy. The average beam-on time was also reduced by from 103.8 mins to 87.4 mins. Our method was also compared with the inverse planning algorithm provided in Leksell GammaPlan planning system, and outperformed it with better plan quality for all the 6 cases.This preliminary study has demonstrated the effectiveness and feasibility of our MRL inverse planning approach for GK radiosurgery.