A Benchmarking Application for Image-Based Machine QA
Abstract
Purpose
To demonstrate a novel software package that will generate two- dimensional radiation field images and dose profiles of known characteristics for calibration of radiation therapy QA devices.
Methods
An app, RadBenchmarkImage, was developed that can generate images of a radiation field with an associated dose profile of known character- istics. This application simulates geometric penumbra, flatness filtering, and wedge filtering. The user may adjust many parameters including the geomet- ric penumbra, flattening filter amplitude, flattening gaussian subtraction, and wedge angles. The dose profile and radiation field image are displayed and updated live as the user adjusts parameters. Upon completion, the user may save the image along with it’s metadata for QA device calibration. Calibration tests were conducted on a radiation QA device, n¨uFilm, that is programmed to measure flatness and symmetry of radiation field images. Images for testing were generated with increasing complexity. The first set of images differed only in their penumbra. The next set varied in flattening filter amplitude and gaus- sian subtraction. Then, an image set was tested with asymmetrical flattening filters. The final set of images had varying wedge filter angles to test symmetry measurements.
Results
For the first three tests, the mean deltas between known and measured flatness were 0.0342, -0.041, and 0.057 percent respectively. For the varied wedge angle tests, the mean delta in symmetry measurements was -0.839 percent, or 0.5926 percent for angles less than thirty degrees.
Conclusion
he RadBenchmarkImage app demonstrates the ability to cali- brate and test radiation therapy QA devices and software. It’s ability to generate known images and dose profiles allows users to test their devices’ accuracy in symmetry and flatness measurements. This application allows radiation therapy devices to report accurate error in their measurements.