Low-dose single-energy material decomposition in radiography using a sparse-view computed tomography scan

Abstract

Dual-energy material decomposition (DEMD) is a well-established theoretical x-ray technique that uses low- and high-kilovoltage radiographs to separate soft tissue and bone in radiography and computed tomography (CT). However, it requires double exposures that result in increased patient radiation doses, causes increases in the execution time, and generates errors due to misregistration attributed to the patient motion between two scans. In this study, we investigated a low-dose, single-energy material decomposition (LSEMD) method in radiography using a sparse-view CT scan where the attenuation length in the object was estimated from the CT image. We performed a systematic simulation and an experiment to demonstrate the feasibility of use of the LSEMD method in radiography. Only 60 projections, far fewer than those required by the Nyquist sampling theory, were acquired at an x-ray tube voltage of 80 kVp, and were used to reconstruct a sparse-view CT image with a state-of-the-art dictionary-learning (DL) algorithm. We investigated the image performance of the LSEMD and compared the elicited results with those obtained with the use of DEMD (80 kVp and 120 kVp were used). Our results indicate that the DL algorithm produced high-quality sparse-view CT images. Accordingly, the LSEMD method yielded material decomposition results that were very similar to the results elicited by the conventional DEMD method in radiography.