Room: AAPM ePoster Library
Purpose: To develop and test a system to monitor deep-inspiration breath hold (DIBH) during whole breast radiotherapy using real-time EPID imaging for IMRT and VMAT.
Method: Treatment accuracy was assessed by automatically extracting anatomical features from transit EPID images frames acquired every 130ms during treatment. A method was first developed and tested for tangential IMRT, where the chest wall (CW) contour was extracted from each EPID image frame and compared to the CW contour from the planning CT. When the CW was obstructed by the MLC, the breast edge (BE) contour was used to predict the CW using a correlation model. This method was tested offline for 12 IMRT patients (147 fractions and 47,000 image frames) treated with DIBH and gated using the Varian RPM system. The techniques have been adapted and tested for VMAT breast treatments where there is additional obstruction by the MLC and gantry rotation.
Results: The mean and standard deviation of errors was 0.2±2 mm for CW position and 0.1±1.5 degrees for CW angle. Of the 12 patients in this study, 3 patients were identified with less than acceptable accuracy with the CW measured to be >±5 mm from planned position >5% of the time. For VMAT, a portion of the CW was unobstructed by the MLC in the EPID images for 69% of the treatment. The mean and standard deviation of CW error for the one VMAT fraction analysed was 4.1±1.0 mm.
Conclusion: A system has been developed and tested to assess the accuracy of DIBH breast and nodal radiotherapy for IMRT and VMAT using EPID imaging. The methods have been tested offline and demonstrate that real-time monitoring was required for 25% of patients. Current work is being undertaken to translate these methods into a real-time monitoring system.