Collimator Position Optimization for Proton Minibeam Radiation Therapy
Abstract
Purpose
Proton minibeam radiation therapy (pMBRT) employs spatially fractionated dose distributions to improve the therapeutic index by reducing normal tissue toxicity. A key component of pMBRT is multi-slit collimator (MSC), which shapes the beam into narrow, spatially separated minibeams. The lateral positioning of the MSC relative to the beam direction determines minibeam placement within the patient, and small lateral shifts can substantially affect peak-valley dose characteristics, target coverage, and organs-at-risk (OAR) sparing. Consequently, optimizing MSC positions across beam angles is critical to fully realize the dosimetric benefits of pMBRT. This work proposes a novel collimator position optimization (CPO) framework that enables independent lateral adjustment of the MSC at each beam angle to improve plan quality.
Methods
CPO is formulated as a mixed-integer programming problem that jointly optimizes MSC positions and spot intensities. For each beam angle, binary variables select one (CPO-S) or multiple (CPO-M) MSC positions from a predefined set of candidate shifts, while continuous variables represent the associated spot weights. The resulting non-convex problem is solved via iterative convex relaxation and alternating direction method of multipliers, which decomposes the optimization into tractable subproblems and ensures feasibility of integer and continuous decisions.
Results
Application to six clinical cases demonstrates that CPO-S consistently identifies near-optimal configurations with substantially reduced computation times compared to exhaustive enumeration. The results show improved target dose conformity and reduced mean doses to OAR relative to IMPT, conventional fixed-collimator configurations (Conv), and joint dose-PVDR optimization with fixed-collimator setups (JDPO). Additionally, the proposed methods achieve PVDR higher than IMPT and slightly lower than JDPO, as expected, since JDPO explicitly optimizes PVDR.
Conclusion
Optimizing MSC positions can enhance plan quality in pMBRT, especially when targets lie close to critical structures from multiple directions. The proposed framework provides a computationally efficient approach for incorporating this flexibility into clinical treatment planning.