Dynamic MLC tracking of moving targets with a single kV imager for 3D conformal and IMRT treatments

Background. Tumor motion during radiotherapy is a major challenge for accurate dose delivery, in particular for hypofractionation and dose painting. The motion may be compensated by dynamic multileaf collimator (DMLC) tracking. Previous work has demonstrated that a single kV imager can accurately localize moving targets for DMLC tracking during rotational delivery, however this method has not been investigated for the static gantry geometry used for conformal and IMRT treatments. In this study we investigate the accuracy of single kV-imager based DMLC tracking for static-gantry delivery. Material and methods. A 5-field treatment plan with circular field shape and 200 MU per field was delivered in 20 s per field to a moving phantom with an embedded gold marker. Fluoroscopic kV images were acquired at 5 Hz perpendicular to the treatment beam axis during a 120° pre-treatment gantry rotation, during treatment delivery, and during inter-field gantry rotations. The three-dimensional marker position was estimated from the kV images and used for MLC adaptation. Experiments included 12 thoracic/abdominal tumor trajectories and five prostate trajectories selected from databases with 160 and 548 trajectories, respectively. The tracking error was determined as the mismatch between the marker position and the MLC aperture center in portal images. Simulations extended the study to all trajectories in the databases and to treatments with prolonged duration of 60 s per field. Results. In the experiments, the mean root-mean-square (rms) tracking error was 0.9 mm (perpendicular to MLC) and 1.1 mm (parallel to MLC) for thoracic/abdominal tumor trajectories and 0.6 mm (perpendicular) and 0.5 mm (parallel) for prostate trajectories. Simulations of these experiments agreed to within 0.1 mm. Simulations of all trajectories in the databases resulted in mean rms tracking errors of 0.6 mm (perpendicular) and 0.9 mm (parallel) for thorax/abdomen tumors and 0.4 mm (perpendicular) and 0.2 mm (parallel) for prostate for both 20 s and 60 s per field. Conclusion. Single kV imager DMLC tracking, which is fully compatible with IMRT, was demonstrated for static fields. The mean tracking error was sub-2 mm for most tumor trajectories with respiratory motions and sub-1 mm for most prostate trajectories.