Purpose: With many plan variables to determine, manual forward planning for Gamma Knife (GK) radiosurgery is very challenging. Inverse planning eases GK planning by determining the variables via solving an optimization problem. However, due to the vast search space, most inverse planning algorithms, including the one provided in Leksell GammaPlan (LGP) treatment planning system, have to predetermine the isocenter locations using some geometric methods and then optimize the shot shapes and durations at these preselected isocenters. This sequential planning scheme does not necessarily lead to optimal isocenter locations and hence globally optimal plans. In this study, we proposed a multiresolution-level (MRL) inverse planning approach, attempting to approach this large-scale GK optimization problem via an iterative method.
Methods: In our MRL approach, several rounds of optimizations were performed with a progressively increased resolution used for isocenter candidates. At each round, an optimization problem was solved to optimize the beam-on time for each collimator and sector at each isocenter candidate. The isocenters that obtained nonzero beam-on times at the previous round and their neighbors on a finer resolution were used as new isocenter candidates for the next round of optimization. After plan optimization, shot sequencing was performed to group the optimized sectors to deliverable composite shots.
Results: We have tested our MRL approach on six GK cases previously treated in our institution. For the five cases that have a single target, with similar target coverage obtained, our MRL inverse planning approach achieved better plan quality compared to manual forward planning and LGP inverse planning, with higher selectivity (0.73 ± 0.07 vs 0.72 ± 0.08 and 0.62 ± 0.10), lower gradient index (2.71 ± 0.25 vs 2.78 ± 0.24 and 3.00 ± 0.29), lower brainstem D0.1cc dose (6.10 ± 4.46 Gy vs 8.87 ± 4.82 Gy and 9.17 ± 3.80 Gy), and shorter total beam-on time (62.1 ± 22.9 min vs 83.6 ± 28.2 min and 70.7 ± 16.7 min). For the case that have six targets, compared with manual planning and LGP inverse planning, our MRL approach achieved higher selectivity (0.68 vs 0.57 and 0.47) and lower gradient index (3.77 vs 4.51 and 5.11). The beam-on time of our plan was slightly longer than manual planning and LGP inverse planning (206.4 min vs 204.7 min and 199.3 min). We have also performed sector duration optimization at the isocenters determined by manual planning or the LGP inverse planning, and the resulting plan qualities were found to be inferior to our MRL approach for all the six cases.
Conclusions: This preliminary study has demonstrated the efficacy and feasibility of our MRL inverse planning approach for GK radiosurgery.
Keywords: Gamma Knife radiosurgery; inverse planning; multiresolution-level strategy.
© 2020 American Association of Physicists in Medicine.