Dual-Energy Material Discrimination using Beam Spectra Obtained from Depth Dose Analysis

In dual-energy X-rays imaging and analysis, high-energy X-rays are used to penetrate the dense materials and then discriminate the major elements of the subject. However, the low dependence of the attenuation coefficient of materials to the atomic number at high X-ray energies makes material discrimination challenging for large subjects. Recent studies demonstrated that characterizing the energy spectra of transmitted photons provides valuable discriminative information to dist inguish the material types compared to the approaches that rely only on counting the transmitted photons. However, the long dead-time of spectrometers at high X-ray flux hinders the accurate measurement of the transmitted X-ray spectrum. In this light, we set out to address this issue through deriving the energy spectrum of transmitted X-ray photons using a semi-empirical method. The suggested technique is based on the spectrums of high- and low-energy X-ray sources, as well as the X-ray beam transmitted through materials, which are acquired using simulated percentage depth dose (PDD) data in a water phantom. First, we created a square field with high- and low-energy poly-energetic spectra. Thereafter, through inserting sheets from different materials with varied thicknesses in front of beam, we obtained the PDD values and depth doses for monoener-getic beam and determined the photon beam spectra in each case by solving the linear matrix system for calculated doses. Finally, we used experimental data to validate this method for materials discrimination. The normalized root mean square error (NRMSE) and mean absolute percentage error (MAPE) errors were calculated 0.88 % and 0.56 %, at high- and low-energy, respectively, demonstrating good agreement between the simulated and experimentally measured incident X-ray spectrum. Furthermore, a more detailed analysis of the Z variations with the Intercepts of fitted lines, confirms that the intercepts’ maximum inaccuracy is around 0.25. Therefore, material discrimination with a unit resolution will be achievable.