Patterns (Nov 2020)
Virtual Monoenergetic CT Imaging via Deep Learning
Abstract
Summary: Conventional single-spectrum computed tomography (CT) reconstructs a spectrally integrated attenuation image and reveals tissues morphology without any information about the elemental composition of the tissues. Dual-energy CT (DECT) acquires two spectrally distinct datasets and reconstructs energy-selective (virtual monoenergetic [VM]) and material-selective (material decomposition) images. However, DECT increases system complexity and radiation dose compared with single-spectrum CT. In this paper, a deep learning approach is presented to produce VM images from single-spectrum CT images. Specifically, a modified residual neural network (ResNet) model is developed to map single-spectrum CT images to VM images at pre-specified energy levels. This network is trained on clinical DECT data and shows excellent convergence behavior and image accuracy compared with VM images produced by DECT. The trained model produces high-quality approximations of VM images with a relative error of less than 2%. This method enables multi-material decomposition into three tissue classes, with accuracy comparable with DECT. The Bigger Picture: The physical process of X-ray CT imaging is described by an energy-dependent non-linear integral equation. However, this equation is noninvertible using an efficient solution and is practically approximated as a linear integral model in the form of the Radon transform. Because the approximation suppressed energy-dependent information, the resultant image reconstruction requires beam-hardening correction to suppress image artifacts. To overcome the limitation, dual-energy CT (DECT) uses two energy spectra and produces virtual monoenergetic (VM) and material-specific images at the cost of increased system complexity and additional radiation dose. In this work, we present a physics-based deep learning approach to map polyenergetic CT images to monoenergetic images at pre-specified energy levels, accurately and robustly approximating VM images from DECT using only single-spectrum images. This method enables multi-material decomposition with a performance comparable with that of DECT.
