Automated verification of multileaf collimation (MLC) using a Siemens EPID

Here novel applications involving a Siemens EPID are presented: determination of radiation central axis, and MLC leaf position verification and calibration. The radiation central axis (cax) was defined using portal images of a field at different collimator angles. The location of the car in terms of pixel number on the imaging vidicon camera was found to be stable within ±3 pixels (1.5 mm) over 2 weeks of repeated extensions/ retractions of the EPID. The variation in the location of the cax as a function of gantry angle was found to be repeatable for the same gantry angle to within I pixel. A number of edge detection algorithms were tested: Sobel, Roberts, Prewitt, Laplace, morphological and Canny. These algorithms were tested on simulated, EPID and portal film images, and found to agree with each other to within 1 pixel. Using only a single one time calibration procedure, which involved measuring the pixel dimension associated with the EPID at isocenter, and the determination of a virtual machine coordinate system, automated edge detection of the collimation boundary was successfully carried out for both in air and phantom images, allowing for comparison with intended leaf positions from the treatment planning system. The location of the MLC leaves determined from the EPID images was found to agree within 1 mm of the corresponding values measured using film and an ionization chamber. Applications of this work include a fast pre-treatment QA of MLC/cerrobend fields, dosimetric calibration of the MLC, and improved quantification of linac QA