Impact of optimization algorithms on the intensity map in IMRT treatment planning

In inverse planning of IMRT, optimum intensity maps are generated using an optimization algorithm. In this paper, impacts of two different optimization algorithms on the intensity map in IMRT treatment planning were evaluated. These were from the steepest descent (SD) and simulated annealing (SA) methods. The following five patterns were compared: [1] SD with calculation time of 5 min; [2] SD with the terminal criterion based on cost function; [3] SA with calculation time of 5 min; [4] SA with the terminal criterion; and [5] SA with the terminal criterion using a smoothing filter. Differences of D(95%) for the planning target volume, V(70Gy) for the rectum wall and the bladder wall were up to 0.5, 1.8 and 3.2 %, respectively in all patterns. The dosimetric impact was negligible. In contrast, generated intensity maps were sensitive to the algorithms. Intensity maps generated by SA tended to have much fluctuation due to numerical artifacts compared to those generated by SD. The difference in the profile was over 7 % between the algorithms. The smoothing filter decreased the fluctuation in intensity maps of SA. In conclusion, it is important to understand impacts of optimization algorithms on the intensity map and the dose distribution.