Stopping power ratio estimation for proton therapy dose calculations using projection- vs. image-based dual-energy CT methods

International audienceBackground:Several strategies have been proposed to calculate stopping power ratios (SPRs) from dual-energy CT (DECT) to improve the accuracy of proton dose calculations. DECT methods can be categorized into two groups: projection-based methods, where material decomposition is performed prior to image reconstruction, and image-based methods, where decomposition takes place after image reconstruction. This simulation-based study aimed at comparing the SPR accuracy of projection- vs. image-based DECT methods.Materials and Methods:X-ray CT projections of the female ICRP phantom were simulated using two different X-ray spectra with realistic detector response and noise levels. Four ICRP slices were selected: head, sternum, breast and pelvis. The projection-based SPR estimation consisted of two steps: estimation of the relative electron density (RED) through a two-material decomposition method and conversion of RED into SPR through poly-lines relations. The image-based estimation was based directly on parametrization of SPR through a calibration process (Med.Phys. 2016;43:5547). The SPR comparison was based on percentage deviation inside ROIs and difference maps for which the Radon transform was calculated to find the mean range errors over a whole slice.Conclusions:SPR estimations with the image-based method and the projection-based method produced comparable results. Projection-based methods require access to raw image data and low and high energy projections sampled at the same angle, while image-based methods require a beam-hardening correction