PubMed1996
PURPOSE: Having the capability of fast previewing an expected optimization result for a defined combination of optimization parameters is essential to reach a better optimization result. In order to achieve this goal, the well-tried Cimmino optimization algorithm requires modification. MATERIALS AND METHODS: Usually, the speed of convergence of some iterative algorithms is restricted to a bounded relaxation parameter. Exploiting the special altering behavior of the weighting factors at each step, many iteration steps are avoided by overrelaxing this relaxation parameter. Therefore, the relaxation parameter is increased as long as the optimization result is improved. This can be performed without loss of accuracy. RESULTS: Our optimization technique is demonstrated by the case of a right lung carcinoma. The solution space for this case is 36 isocentric X-ray beams evenly spaced at 10 degree. Each beam is restricted to 23 MV X-ray fields with a planning target volume matched by irregular field shapes, similar to that produced by a multileaf collimator. Four organs at risk plus the planning target volume are considered in the optimization process. The convergence behavior of the optimization algorithm is shown by overrelaxing the relaxation parameter in comparison to conventional relaxation parameter control. CONCLUSION: The new approach offers the ability to get a fast preview of the expected final result. If the clinician is in agreement with the preview, the algorithm is continued and achieves the result proven by the Cimmino optimization algorithm. In the other case, if the clinician doesn't agree with the preview, he will be able to change the optimization parameters (e.g., field entry points) and to restart the algorithm.