Purpose: To generate synthetic CT phantoms of Sun Nuclear Corporation (Melbourne, FL) phantoms: WL-QA, and MultiMet-WL Cube with zero positional error, a known isocenter position, and no artifacts to identify systemic errors in CT simulation and improve CBCT-based Winston-Lutz results on an Elekta Synergy S (Stockholm, Sweden) linac.
Methods: The synthetic phantoms SNC WL-QA and SNC MultiMet-WL were generated using Python libraries: pydicom and NumPy. The SNC WL-QA phantom was sampled into a blank CT volume of 512x512 pixels with 256 slices. Geometric equations for sphere and squares were used to create Boolean masks to represent the real phantom with appropriate Hounsfield units. The synthetic CTs were successfully imported into Varian Eclipse (Palo Alto, CA) and compared with real CT scans of phantoms. Using rigid image registration techniques, systemic errors from manual setup of the phantoms were identified and eliminated. Reference images were generated with zero positional error to perform CBCT-based Winston-Lutz on an Elekta Synergy S.
Results: The analytical phantoms were generated successfully with zero positional errors and sub-millimeter geometrical accuracy. Rigid registration techniques show sub-millimeter (±0.3 mm) and rotational (±0.2°) discrepancies in our Winston-Lutz Phantom reference images. The rigid transformation matrix was used to resample the real CT phantom into its ideal space. The synthetic and resampled CTs were used to perform CBCT-based Winston-Lutz on our stereotactic treatment unit. Trending the first 3 months of collected data since implementation, the measured mean total delta with portal imaging and DoseLab v7.0 decreased from (0.55 ± 0.08) mm to (0.46 ± 0.09) mm.
Conclusion: Use of synthetic SNC WL-QA and SNC MultiMet-WL phantoms show improvement in CBCT phantom alignment for Winston-Lutz. The synthetic CTs and Python code are made publicly available for the medical physics community.
Quality Assurance, Cone-beam CT, Localization
TH- External Beam- Photons: Quality Assurance - Linear accelerator