Joint Optimization of Spot Position and Weight In Peak-Plane for Proton Radiation Therapy
Abstract
Purpose
Conventional intensity-modulated proton therapy (IMPT) planning commonly relies on fixed, uniformly distributed spot grids with optimization restricted to spot weights. Although adaptive spot placement has been explored to improve dose conformity, particularly for geometrically complex targets, most existing approaches are target shape-driven and depend on heuristic rules or extensive hyperparameter tuning, which limits general applicability. To better exploit the flexibility of spot placement while accounting for overall treatment objectives, a data-driven adaptive spot optimization strategy that jointly considers target coverage and organ-at-risk (OAR) sparing is desirable.
Methods
We propose a Joint Optimization of spot position and weight in the Peak-Plane (JOPP) framework for adaptive proton spot placement. For each energy layer, optimization is performed on the Bragg peak plane. This formulation reduces the optimization problem to two dimensions, resulting in substantially lower computational cost compared with three-dimensional optimization. In the peak plane, the lateral dose influence of each spot is approximated by a depth-dependent Gaussian model, enabling efficient joint optimization of spot positions and weights using analytical gradients. After optimized spot positions are determined, a conventional IMPT optimization is performed to refine spot weights using fixed optimized spot locations.
Results
Compared with conventional fixed uniform spot positions, JOPP achieves more efficient spot placement and improved dose conformity. For a similar number of spots, JOPP provides enhanced target coverage and improved OAR sparing relative to coarse uniform spots. Compared with fine uniform spots, JOPP achieves comparable plan quality using substantially fewer spots, demonstrating improved delivery efficiency.
Conclusion
The proposed peak-plane-based joint optimization framework enables more flexible and efficient proton spot placement than conventional uniform-spot IMPT planning. By integrating data-driven spot position optimization with weight optimization in a low-dimensional setting, JOPP offers a promising and computationally efficient approach for improving IMPT treatment planning.