A real-time dynamic-MLC control algorithm for delivering IMRT to targets undergoing 2D rigid motion in the beam's eye view.

An MLC control algorithm for delivering intensity modulated radiation therapy (IMRT) to targets that are undergoing two-dimensional (2D) rigid motion in the beam's eye view (BEV) is presented. The goal of this method is to deliver 3D-derived fluence maps over a moving patient anatomy. Target motion measured prior to delivery is first used to design a set of planned dynamic-MLC (DMLC) sliding-window leaf trajectories. During actual delivery, the algorithm relies on real-time feedback to compensate for target motion that does not agree with the motion measured during planning. The methodology is based on an existing one-dimensional (ID) algorithm that uses on-the-fly intensity calculations to appropriately adjust the DMLC leaf trajectories in real-time during exposure delivery [McMahon et al., Med. Phys. 34, 3211-3223 (2007)]. To extend the 1D algorithm's application to 2D target motion, a real-time leaf-pair shifting mechanism has been developed. Target motion that is orthogonal to leaf travel is tracked by appropriately shifting the positions of all MLC leaves. The performance of the tracking algorithm was tested for a single beam of a fractionated IMRT treatment, using a clinically derived intensity profile and a 2D target trajectory based on measured patient data. Comparisons were made between 2D tracking, 1D tracking, and no tracking. The impact of the tracking lag time and the frequency of real-time imaging were investigated. A study of the dependence of the algorithm's performance on the level of agreement between the motion measured during planning and delivery was also included. Results demonstrated that tracking both components of the 2D motion (i.e., parallel and orthogonal to leaf travel) results in delivered fluence profiles that are superior to those that track the component of motion that is parallel to leaf travel alone. Tracking lag time effects may lead to relatively large intensity delivery errors compared to the other sources of error investigated. However, the algorithm presented is robust in the sense that it does not rely on a high level of agreement between the target motion measured during treatment planning and delivery.