4. Dual energy CT: Iterative metal artifact reduction for radiotherapy

Introduction Metal implants are a well-known source of artifact in computed tomography (CT). To improve image quality, several solutions currently exist: metal artifact reduction algorithms such as iMAR® (Siemens) using an iterative method for single energy CT (SECT) or the use of dual energy CT (DECT) as another solution with reconstruction of virtual monoenergetic images (VMIs). VMIs are a linear combination of the images obtained from the two sets of data for each energy scan [1] . The goal of this work is to study the different combinations of MAR techniques to obtain a valid and optimal clinical protocol for conventional radiotherapy and protontherapy treatment. Methods Metals of different densities are centered in a homogeneous phantom. Data acquisitions were obtained with a CT Siemens Definition AS OpenRT. Two scans were performed, one with single energy at 120 kV and then another with double energy at 80 kV and 140 kV. A filtered back projection reconstruction and an iterative one using SAFIRE® with and without iMAR were obtained for each scan. Dual energy images are reconstructed into VMIs ranging from 70 keV to 180 keV. Quantitative analysis is achievable using a home-made code on Matlab® that classifies pixels by their Hounsfield Units (HU). Results Preliminary results indicate an underestimation of HU values for pixels near the metallic object while reconstructing the dual energy images without iMAR when a raise in HU values by 15 to 20 UH for the reconstruction with iMAR is observed (cf.image). Moreover, HU values are better estimated near the metal with iMAR reconstruction. However, artifacts generated by iMAR appear far from the metal, potentially introducing uncertainties on dose calculation. Conclusions Dual energy images admit fewer uncertainties on HU values for each voxel and grant a reconstruction of the image without much loss of information around the metal. The images with iMAR have the most accurate HU values. The importance of dual energy images being already known in the reduction of uncertainties in the calibration for proton treatment can lead this study to also determine the influence of the overestimation of HU values with iMAR on calibration curves and dose calculation.