Technical note: real-time MLC control and latency measurement optimization with external verification

Objective.Real-time tumor-tracked radiation therapy has the potential to enhance geometric treatment accuracy by adapting radiation delivery to tumor motion during irradiation. The multi-leaf collimator (MLC) plays an important role in real-time tracking, as it dynamically conforms the radiation beam to the moving target. However, an inherent time delay exists between the target and reached leaf positions. Accurate MLC latency characterization is essential for accurate target position prediction, facilitating reduced treatment margins. This study quantifies the mechanical latency of the Alberta Linac-MR (LMR) MLC under varying gantry angles (0°, 90°, 180°, 270°), maximum leaf velocities at the MLC (5-50 mm s^-1), and across synthetic and patient-derived tumor motion patterns.Approach.A QUASAR MRI^4Dmotion phantom was programmed to emulate 1-D abdominothoracic tumor motions. Phantom positions were transmitted to custom software which assigned the target MLC position using a custom-made proportional-integral-derivative controller tuned to minimize MLC latency. Both the phantom and current MLC positions were timestamped every 40 milli-seconds (ms). In addition, both MLC and phantom motions were externally verified using time-synchronized cameras. The assigned and actual MLC positions were plotted, and the temporal offset between the two curves was defined as the MLC latency.Main result.MLC latency was measured to be 73.0-80.1 ms across motion patterns. For sine motion, measured latencies for all gantry angles were between 74.3 and 74.9 ms, while measured latencies for maximum leaf velocities at 5, 10, and 30 mm s^-1were 111.4, 76.0, and 76.3 ms respectively.Significance.A technique for real-time MLC motion control and mechanical latency verification has been developed for a LMR MLC, representing one component of the total real-time tracking workflow latency. Gantry angle has no significant impact on the measured latency while selecting a maximum leaf velocity ⩾10 mm s^-1minimized latency.