Room: AAPM ePoster Library
Purpose: To develop a simple and quick method for automated quality control of the asymmetric jaws positioning test for medical linear accelerators (Canadian Partnership for Quality Radiotherapy ML14).
Methods: Two half-beam blocked EPID images are acquired for both upper (Y) and lower (X) jaws with the junction located at the zero jaw position. Two additional images with the junction at +/-5cm (Y) and 4 images incorporating collimator rotation are also taken. Individual dose profiles are extracted and combined into a composite profile. Both the composite dose profile at the junction of two abutting fields and individual dose profiles are used to evaluate relative jaw positioning.
In the current process the gap/overlap of abutted fields is manually estimated by displacing one of the profiles with respect to the other until the composite dose profile is homogeneous.
An alternative automated analysis was implemented in a Python script. Based on the intensity pattern, it selects the proper abutting fields automatically. It then minimizes the area between an upsampled composite profile, and a smoothed version of the same composite profile resulting in an homogeneous profile at the junction. The gap/overlap is calculated as the product of the number of iterations and the composite profile resolution.
84 junction profiles from a single linac were analyzed manually and automatically. The time to process the images was tracked for both processes.
Results: The difference between the manual and automatic methods averaged 1.16e-6 mm with a maximum difference of 0.16 mm.
The average time for the project assistant to complete the manual process, was 165 seconds; while the algorithm time was 26 seconds, an average 80% faster analysis.
Conclusion: This method provides a quantitative, quick method for asymmetric jaws quality control with equivalent performance. Further work is ongoing to process data from additional machines at our centre.