Single-source dual-energy computed tomography for the detection of bone marrow lesions: impact of iterative reconstruction and algorithms.

Abstract

To compare the diagnostic performance of different reconstruction algorithms of single-source dual-energy computed tomography (DECT) for the detection of bone marrow lesions (BML) in patients with vertebral compression fracture using MRI as the standard of reference. Seventeen patients with an age over 50 who underwent single-source DECT of the spine were included. The raw data (RD) were reconstructed using filtered back-projection (FBP) and iterative reconstruction (IR) with three iteration levels (IR1-IR3). Bone marrow images were generated using a three-material decomposition (3MD) and a two-material decomposition (2MD) algorithm and an RD-based approach. Three blinded readers scored the images for image quality and the presence of bone marrow lesions (BML). Only vertebrae with height loss were included. The signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were calculated. The different reconstructions were compared using Dunn's multiple comparison test. Thirty-nine vertebrae were included. IR(1-3) showed superior sensitivity (87.5%) compared to FBP (75%) using 3MD but was comparable to RD (83.3%). All 2MD images were inferior (sensitivity < 38%). The image quality score was significantly higher for 3MD-IR(1-3) compared to 3MD-FBP (p < 0.0001) and all 2MD data sets (p < 0.03). This pattern was also supported by the SNR and CNR measurements. RD showed no significant improvement compared to IR. The image quality of bone marrow images acquired with DECT can be improved by using IR compared with FBP. RD-based reconstruction does not offer significant improvement over image data-based reconstruction. 2MD algorithms are not suitable for BML detection.