Room: Exhibit Hall
Purpose: To develop an artificial intelligent (AI) tool that can estimate the equivalent field size of electron cutouts and calculate the output factor with minimal human intervention.
Methods: This AI tool first reads the DICOM RT Plan files through Python Matplotlib and extracts the electron beam information, including cone size, field size, energy, and beam block (i.e., contour of the electron cutout). This tool then uses the Python Open CV computer vision module to refine the beam blockâ€™s contour through Gaussian blur and canny edge detection. Additionally, the program identifies the rectangular of minimum area and bounds it around the contour. The dimensions of the rectangle are then extracted and converted into field size units, and square field sizes are estimated. The output factor is computed by the square root method, interpolated using measured data (if necessary), and evaluated with stored data of previous contour shapes and areas. This tool was tested with 14 electron plans and compared to the measured values.
Results: All 14 DICOM RT files were successfully imported and the shape of electron cutout recognized by the computer vision module. Calculation of output factor using the Python application yielded output factors with an average error of 1.7% in comparison to the measured data. Seven of the 14 output factor values featured percent errors of less than 0.7%.
Conclusion: We have developed an AI tool that uses computer vision to mimic the operation of human physicist for estimating the electron output factors automatically. Preliminary tests indicate that this tool is accurate and reliable. This tool can potentially replace the tedious manual calculations performed by medical physicists.