A hybrid optimization approach for the inverse problem of radiotherapy

The dose optimization problem of intensity modulated radiotherapy (IMRT) is considered to be an ill-posed inverse problem. Although IMRT technique offers considerable advantage over conventional methods of radiotherapy treatment planning (RTP), it becomes difficult to realize in practice. Different methods applied to determine the beam intensity solution are not able to find the optimum solution, as they are prone to be trapped in local optima. However, the stochastic based genetic algorithm (GA) is capable of escaping local optima, thus able to arrive at a global optimum solution. But this stochastic based algorithm is inherently slow to converge as it employs a random search technique. Moreover, IMRT planning is a computationally intensive process. Thus, it becomes almost impractical to implement GA for the dose optimization problem of IMRT. Here, the authors utilize the advantage of GA combined with an analytic technique that would significantly reduce the computation costs of the treatment planning and would result in a more optimized solution for RTP. Here, the authors propose a hybrid approach with a faster technique, steady-state Kalman filter-type algorithm to obtain a seed for GA to determine optimum intensity modulated beams. They adopted the distributed genetic search approach to obtain increased genetic diversity. The results of computer simulations show that the proposed approach is successful to achieve the optimization objectives of RTP