Room: Karl Dean Ballroom A1
Purpose: Multi-leaf Collimators (MLCs) are a common reason for maintenance on linacs and thus machine downtime. Routine quality assurance (QA) can fail to identify MLCs that require maintenance until failure, whether it be patient specific, or machine QA suggested in TG-142. A method to identify suspect leaves and reduce machine downtime is proposed which analyzes trajectory log data from a custom MLC QA plan.
Methods: A custom MLC QA plan was developed using Python in order to produce MLC movements which result in near clinical maximum for leaf speeds, interleaf friction, momentum, and gravitational effects. A Python code that utilizes Pylinac was developed to find the maximum speed error for each leaf by analyzing the data collected from the trajectory logs. In order to test the sensitivity of the proposed method, a "T-nut" test was performed which utilizes a frequent point of failure in the MLCs. Historic patient plan deliveries were also evaluated before and after a MLC failure as a retrospective analysis.
Results: A 15.0% decrease in leaf speed error was observed when a suspect MLC's T-nut was switched with a well-performing MLC's T-nut. The well performing MLC's speed error increased by 14.4% when its T-nut was switched with the suspect MLC's T-nut. The leaves then showed 8.8% and 5.7% decreases in speed error when their T-nuts were replaced with new ones. The retrospective analysis showed a 34.1% decrease in speed error from 0.276 cm/s to 0.182 cm/s after a failing leaf had a T-nut replacement.
Conclusion: This method has greater sensitivity in finding underperforming leaves than traditional QA and can be used to quantify MLC errors and identify issues prior to terminal failure. Periodic tests could be run in order to establish a threshold which MLC maintenance is required, thus reducing downtime.