New method for correcting beam-hardening artifacts in CT images via deep learning

Proceedings of the 16th Virtual International Meeting on Fully 3D Image Reconstruction in Radiology and Nuclear Medicine, 19-23 July 2021, Leuven, Belgium.Beam-hardening is the increase of the mean energy of an X-ray beam as it traverses a material. This effect produces two artifacts in the reconstructed image: cupping in homogeneous regions and dark bands among dense areas in heterogeneous regions. The correction methods proposed in the literature can be divided into post-processing and iterative methods. The former methods usually need a bone segmentation, which can fail in low-dose acquisitions, while the latter methods need several projections and reconstructions, increasing the computation time. In this work, we propose a new method for correcting the beamhardening artifacts in CT based on deep learning. A U-Net network was trained with rodent data for two scenarios: standard and low-dose. Results in an independent rodent study showed an optimum correction for both scenarios, similar to that of iterative approaches, but with a reduction of computational time of two orders of magnitude.This work has been supported by project "DEEPCT-CMUC3M", funded by the call "Programa de apoyo a la realización de proyectos interdisciplinares de I+D para jóvenes investigadores de la UC3M 2019-2020, Convenio Plurianual CAM - UC3M" and project "RADCOV19", funded by CRUE Universidades, CSIC and Banco Santander (Fondo Supera). The CNIC is supported by the Ministerio de Ciencia, Innovación y Universidades and the Pro CNIC Foundation, and is a Severo Ochoa Center of Excellence (SEV-2015-0505)