Room: Exhibit Hall
Purpose: Beam angle optimization (BAO) largely determines the performance of fixed-field intensity modulated radiation, but it is usually considered as an NP hard problem. In this work, we reformulate BAO into a framework of standard quadratic optimization, based on the theory of compressed sensing (CS).
Methods: We choose the maximum intensity of each incident field as the surrogate variable indicating whether one radiation field has been selected. The CS framework is used to enforce sparsity on the surrogate variable. Since the function of maximum value in the objective can be converted into a linear constraint, the optimization problem is finally solved as quadratic optimization with l1 norm minimization. A reweighting scheme is then implemented to improve the sparsity of the solution.
Results: The performance of the proposed BAO has been verified on a simulation phantom and a prostate patient. On the simulation phantom, six beam paths have been designed as the theoretical optimal angles, from which the radiation beams reach the planning target volume (PTV) without passing through the organs at risk (OARs). Our algorithm successfully finds all 6 optimal beam angles out of 180. On the prostate patient, Pareto frontiers of BAO and equiangular plans with 5, 7 and 9 incident fields and different sets of important factors are calculated and compared, showing that the BAO plans could significantly reduce the dose on OARs with a little sacrifice on the uniformity of dose coverage inside the PTV. Improved plan quality is demonstrated by comparing the dose distributions and DVH curves between a BAO plan and an equiangular plan.
Conclusion: A novel BAO algorithm in the form of standard quadratic optimization is proposed using the compressed sensing technique. This study demonstrates that the algorithm could generate beam orientations leading to improved plan quality for IMRT.
Inverse Planning, Radiation Therapy, Optimization
TH- External beam- photons: IMRT dose optimization algorithms