Abstract
Simple Summary As the technology of the pencil beam scanning system advances, achieving higher accuracy in converting CT Hounsfield to proton-stopping power becomes imperative. Recently, the dual-energy computer tomography (DECT) technique has emerged as a superior technique to single-energy computer tomography in estimating SPR. However, such a technique is not widely adopted clinically. One of the reasons is that the optimization process on the DECT scanners involves a wide range of parameters, one of which is spectral pair. It is generally accepted that the larger separation of the spectra pair could yield a better estimation of SPR. However, it is not validated under a scenario of scanning objects deviating from calibration conditions. In this simulation study, we are examining the performance of variations of spectra pairs on SPR prediction.Abstract Our study aims to quantify the impact of spectral separation on achieved theoretical prediction accuracy of proton-stopping power when the volume discrepancy between calibration phantom and scanned object is observed. Such discrepancy can be commonly seen in our CSI pediatric patients. One of the representative image-domain DECT models is employed on a virtual phantom to derive electron density and effective atomic number for a total of 34 ICRU standard human tissues. The spectral pairs used in this study are 90 kVp/140 kVp, without and with 0.1 mm to 0.5 mm additional tin filter. The two DECT images are reconstructed via a conventional filtered back projection algorithm (FBP) on simulated noiseless projection data. The best-predicted accuracy occurs at a spectral pair of 90 kVp/140 kVp with a 0.3 mm tin filter, and the root-mean-squared average error is 0.12% for tissue substitutes. The results reveal that the selected image-domain model is sensitive to spectral pair deviation when there is a discrepancy between calibration and scanning conditions. This study suggests that an optimization process may be needed for clinically available DECT scanners to yield the best proton-stopping power estimation.