Author: Min Cheol Han, Hojin Kim, Jin Sung Kim, Hyejin Lee, Sac Lee, Yongdo Yun 👨🔬
Affiliation: Department of Radiation Oncology, Yonsei Cancer Center, Heavy Ion Therapy Research Institute, Yonsei University College of Medicine 🌍
Purpose: Carbon ion radiotherapy (CIRT) is characterized by high relative biological effectiveness (RBE), which must be integrated into the treatment planning process. This study aims to develop a novel beam selection algorithm in multi-field optimization of CIRT, based on a group sparse algorithm.
Methods: Selecting effective beam angles must be a challenging step in multi-field optimization of CIRT as RBE weighted dose distribution is known to be unpredictable. Group sparse algorithm was designed to sparsify any user-defined groups, enabling the identification of optimal beam angles that satisfy the RBE dose constraints of multi-field CIRT without requiring an exhaustive search. The proposed optimization algorithm was formulated with three components: dose fidelity for RBE dose constraints, approximated L-norm constraint for minimizing the number of active scanning spots, and L2,1/2-norm group sparse constraint for selecting beam angles. This was initially applied to a simple prostate cancer case with clinical target volume of 73 cc and four organs-at-risk (OARs) (bladder, rectum, right- and left-femoral heads). Dose influence matrices for plan optimization were generated for 12 beam angles (0o to 330o) across all available energy layers, using in-house Monte Carlo simulations.
Results: The group-sparse algorithm successfully reduced the number of beam angles from 12 to 7 angles, compared to the unconstrained method. It also reduced the active scanning spots by 76.4 %, relative to the unconstrained optimization. Additionally, the proposed algorithm decreased mean dose to the bladder and rectum by 9.38 and 14.7 %, respectively, and maximum dose to the femoral heads by 81.5%.
Conclusion: This feasibility study demonstrated that incorporating a group sparse algorithm into multi-field CIRT can help attain optimal RBE weighted dose distribution with a reduced number of critical beam angles.