Iterative Approach for Automatic Beam Angle Selection in Intensity Modulated Radiation Therapy Planning

Introduction: Beam-angle optimization (BAO) is a computationally intensive problem for a number of reasons. First, the search space of the solutions is huge, requiring enumeration of all possible beam orientation combinations. For example, when choosing 4 angles out of 36 candidate beam angles, C36 = 58905 possible combinations exist.  Second, any change in a   beam 4 configuration requires re-calculation and re-optimization of intensity maps, itself a time- consuming process. Third, many local minima (maxima) will appear in the objective function. This paper introduces a new framework to beam-angle selection in intensity modulated radiation therapy (IMRT).   Materials and Methods: In the proposed algorithm for beam angle selection, the problem is solved in two iterative steps: (1) For all beam that given in search space, the fluence map optimization is done to minimize an objective function that is based on clinical requirement and then used to scoring each beam direction. (2) the proposed algorithm is eliminated an insignificant beam angle (with the lowest score in step one) from search poll in each iteration until the beam selection algorithm is received to the predefined number of angles to be used for the treatment. To accelerate each iteration of dose calculation, dose matrix was precalculated in which the dose deposited in voxel i by an IMRT beam is given by Di = ∑ Kimxm where xm is the weight for the mth beamlet. the efficiency of the algorithm was examined in two clinical cases (TG- 119 phantom and prostate case) in terms of DVH and dose distribution. In all cases, the judgment of the algorithm’s efficiency was based on the comparison between plans with equi-spaced beams and plans with beams obtained using the algorithm.   Results: the beam selecting framework has been tested for both a simulated and clinical case(prostate), and the results show the target coverage and dose uniformity remains essentially unchanged, the mean and maximum dose of OARs is significantly decrease as a result of beam selection. For example, in prostate case, the mean dose of rectum and bladder reduced 10.2% and 7.6% respectively. Moreover, D95 of bladder decreased from 10.13 Gy in equiAngular beam selection to 20.3 Gy by the automatic beam angle selection.   Conclusion:This study demonstrates that the algorithm can be effectively applied to IMRT scenarios to get case specific beam angle configurations.