In the past two decades, the ill‐conditioned problem of unfolding linac photon spectra from transmission measurements has been extensively investigated. Previous studies suffer from various limitations that affect the accuracy and robustness of the unfolding. The goal of this study is to address the limitations of unfolding in general, and using parameterization of the spectrum in particular. To this end, the following new and refined methods are implemented. Different attenuator materials and different buildup caps are simultaneously used for better energy differentiation in most of the megavoltage energy range. The spectra are described using a new flexible and physics‐based functional form validated against more than 70 realistic spectra. The radiation detection system is modeled accurately using the EGSnrc usercodes BEAMnrc and cavity. Forward Compton scatter in the attenuators is modeled and corrected for. The proposed methods are validated experimentally on NRC research linac whose incident electron beam parameters are accurately known to 0.5% and its spectra have been independently measured using a NaI detector. A linear system was built to automatically drive attenuator lengths one at a time into the beam, which reduces acquisition time and beam instability uncertainties. Causes for experimental Type B uncertainties (leakage, polarity and scatter) are being investigated, particularly for small transmission signals. Results show that the proposed methods significantly improve the accuracy and robustness of unfolding in the presence of realistic experimental noise. Improvements by factors of 4 and 8 have been achieved in the accuracy of spectral unfolding and maximum energy estimation, respectively.

Additional Metadata
Persistent URL
Journal Medical Physics
Ali, E. (Esm), Mcewen, M., & Rogers, D. (Dwo). (2010). Sci—Thur PM: YIS — 01: Computational and Experimental Methods to Address the Limitations of Reconstructing Linac Photon Spectra from Transmission Measurements. In Medical Physics (Vol. 37). doi:10.1118/1.3476096