Click here to


Are you sure ?

Yes, do it No, cancel

Feasibility of Virtual Monochromatic Spectral Imaging Using Single-Energy CT Data

W Zhao1, T Lv2*, y chen3, L Xing4, (1) Stanford University, Palo Alto, CA, (2) Stanford University, ,,(3) Southeast University, Nanjing, ,CN, (4) Stanford Univ School of Medicine, Stanford, CA


(Thursday, 7/16/2020) 1:00 PM - 2:00 PM [Eastern Time (GMT-4)]

Room: Track 1

Purpose: To provide noise correlated pseudo dual-energy CT (DECT) from single energy CT (SECT) using a deep learning (DL) approach and take advantage of the noise correlation of the pseudo DECT image to generate superior virtual monochromatic images (VMIs).

Methods: We designed a predenoising and difference learning mechanism to generate DECT images from SECT data. The method encompasses two parts: the first is image denoising using a fully convolutional network (FCN). The FCN is trained on the AAPM Low-Dose CT Grand Challenge dataset and applied to a set of contrast-enhanced DECT data. The second is dual-energy difference learning using a U-Net type network. In this part, the denoised low-energy CT images together with the difference image between the low-energy image and its corresponding high-energy counterpart image, are used as the network input and output, respectively. Finally, the predicted difference image is added to the input low-energy image to generate noise correlated high-energy image. VMIs are reconstructed using the pseudo DECT images and compared to those obtained from raw DECT images. The approach was evaluated using 5087 slices of routine contrast-enhanced abdominal DECT images. Clinically relevant metrics were used for quantitative assessment.

Results: The noise levels of low- and high-energy images are reduced by 3.9- and 3.7-fold, respectively. The maximum absolute HU differences between the DL-predicted and the ground truth 140 kV images 3.0 HU, 2.9 HU, 3.1 HU, and 3.0 HU for the region-of-interests on the aorta, liver, spine, and stomach, respectively. The HU accuracy of VMIs reconstructed from pseudo DECT is highly consistent with that obtained from raw DECT image, while the iodine contrast-to-noise ratio of the former surpasses the latter.

Conclusion: This study shows contrast-enhanced DECT imaging using SECT data is feasible. The noise correlation between the DL-predicted DECT images can be employed to reconstruct superior VMI images.


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email