Purpose: To present initial results of DE-CBCT imaging on using the on-board imager (OBI) of a commercial linear accelerator. A method is described to generate virtual monoenergetic (VM)-CBCT and relative electron density (RED) images.
Methods: A step-wedge phantom consisting of aluminum (Al) and polymethyl methacrylate (PMMA) was created and used for calibration of the system. Planar projections were acquired at 130 kVp with the titanium foil and at 80 kVp without the foil. A scatter correction was applied to all projections before calibration and material decomposition. Using images of the step-wedge phantom, a calibration function was derived from the known thicknesses of aluminum and PMMA. Separately, a sequential DE-CBCT scan of the Catphan 604 was obtained and the projections were decomposed into the two basis materials (Al and PMMA) on a pixel-by-pixel basis using the pre-defined calibration function. The projections were summed and weighted by the energy specific attenuation coefficients (or relative electron density) for each basis material image, then CBCT and RED images were reconstructed using filtered back projection. A convolutional neural network filter was developed and applied to reduce noise on reconstructed images.
Results: VM-CBCT images were reconstructed at 40, 80, 100 and 150 keV. The measured Hounsfield units of the individual inserts were compared against the theoretical values. The average difference was of -5 Â± 25, 3 Â± 26, -5 Â± 19 and -2 Â± 12 HUVM for 40, 80, 100 and 150 keV, respectively. The RED images had a mean percent error of âˆ’0.19 Â±0.67%. The observed noise reduction was up to 25% on reconstructed VM-CBCT images.
Conclusion: VM-CBCT and RED images were produced using sequential DE-CBCT scanning on a commercial linac. These images increase the dynamic range of CBCT, and provide data that may be used for dose calculations in photon and particle beam therapy.
Funding Support, Disclosures, and Conflict of Interest: This work was funded by a research grant from Varian Medical System.