Dual Energy CBCT Methods to Produce RED, Zeff and Proton RSP Images Using the On-Board Imager of a Linear Accelerator

Abstract

To present a feasibility study for using dual-energy (DE) cone beam computed tomography (CBCT) to produce relative electron density (RED), effective atomic number (Zeff) and proton relative stopping power (RSP) images, using the on-board imager (OCI) of a commercial linear accelerator. Low- and high-energy (80 kVp and 130 kVp, respectively) CBCT scans of the Catphan 604 phantom were acquired sequentially using the OBI of a commercial Linac. Each acquired projection pair was subsequently decomposed into basis material equivalent thicknesses using a calibration obtained from an aluminum (Al) and polymethyl methacrylate (PMMA) wedge phantom. The Al and PMMA thicknesses projections where then recombined using the linear attenuation coefficient (μ) for each basis material at the desired energy and reconstructed to produce virtual monoenergetic (VM) CBCT images. RED values for each insert of the phantom were obtained by converting the measured μ values for 150 keV VM-CBCT images directly to RED values based on a theoretical μ-RED linear correlation at that energy. The Zeff numbers for each insert were generated using the correlation between the ratio of 40 and 150 keV μ values and the reference Zeff. After obtaining RED and Zeff images, RSP images were obtained using the parametric method proposed by Hunemohr et al. Our theoretical analysis showed good linear correlation (0.9937) between 150 keV μ-values with the reference RED for elements with Z from 6 to 20, with a mean prediction error of 0.000 ± 0.073.The mean RED percentage error for phantom inserts was -0.24% ± 0.75%. The 50% bone insert had the largest percentage error (-1.49%). A logarithmic fit that transforms the ratio of 150/40 keV μ theoretical values to Zeff values from 6 to 20 was also defined. This fit resulted in a correlation coefficient of 0.9979, with a mean prediction error of 0.000 ± 0.021. Applying this data to the phantom inserts, the mean percentage error found for Zeff was 0.08% ± 2.09%. The Teflon insert had the largest percentage error for Zeff estimation (3.44%). The mean percentage RSP error value was -0.19% ± 1.78% with the 50% bone insert having the highest percentage error (-3.74.%). The presented method successfully obtained RED, Zeff and RSP images from VM-CBCT images produced using a sequential DE approach by scanning the phantom directly with the OBI of a commercial linear accelerator. Such an approach may allow for direct verification of RSP values at the time of treatment for proton systems that are equipped with an OBI.