Abstract:Dual-energy CT (DECT) is a very powerful and valuable image modality and has been extensively used in clinical practice. However, it adds an extra layer of complexity on top of the widely used single-energy CT (SECT) and increases the system cost which hinder the application of DECT imaging, especially in less developed regions. To overcome this issue, we purposed a data-driven strategy of obtaining DECT images from SECT images. We developed a deep learning (DL) model to map a low-energy CT image to a high-energy CT image and trained the model using thousands of paired clinical DECT images. A fully convolutional network (FCN) was first employed to denoise the standard DECT images, then the residual images between the denoised low- and high-energy CT images were used to train a dual-energy mapping network together with the low-energy CT images. During the inference procedure, the trained network used a testing low-energy image to generate a residual image which was added to the input low-energy image to yield the final corresponding high-energy image. We assessed the approach using images from 22 patients who received contrast-enhanced abdomen DECT scan with a popular DE application: virtual non-contrast (VNC) imaging and contrast quantification. The DL-predicted high-energy CT images and the original high-energy CT images were quantitatively compared and analyzed using clinically relevant metrics. The absolute Hounsfield unit (HU) differences between the predictions and the ground truth high-energy images are 1.3 HU, 1.6 HU, 1.8 HU, and 1.3 HU (corresponding maximum absolute HU differences are 3.0 HU, 2.9 HU, 3.1 HU, and 3.0 HU) for the region-of-interests in the aorta, liver, spine and stomach, respectively. For the VNC imaging, the overall mean percentage difference of the iodine concentrations obtained using the conventional DECT and DL-DECT images is 0.9% for the testing cases. Compared to that reconstructed from conventional DECT images, the noise levels in the VNC images and iodine maps reconstructed from DL-DECT images are reduced by 7-fold. This study demonstrates that highly accurate and superior DECT imaging and material-specific maps are achievable by using only SECT data via deep learning. Compared to the standard DECT techniques, the proposed method has potential to significantly simplifies the system design, reduce the patient imaging dose, and enhance the signal-to-noise ratio of final material decomposition. The DL strategy allows us to obtain high-quality DECT images without paying the overhead of conventional hardware-based DECT solutions and thus lead to a new paradigm of spectral CT imaging.

Deep-Cleansing: Deep-Learning Based Electronic Cleansing in Dual-Energy CT Colonography

Electronic Cleansing in CT Colonography:Colon Enhancement Filtering

An Integrated Electronic Colon Cleansing for Ct Colonoscopy Via Map-Em Segmentation and Scale-Based Scatter Correction

Integration Of 3d Scale-Based Pseudo-Enhancement Correction And Partial Volume Image Segmentation For Improving Electronic Colon Cleansing In Ct Colonograpy

An Improved Electronic Colon Cleansing Method for Detection of Colonic Polyps by Virtual Colonoscopy

A New Electronic Colon Cleansing Method for Virtual Colonoscopy

Electronic cleaning of intestinal contents tagging by diatrizoate meglumine for single-source dual-energy CT colonography with sequential acquisitions and volume scanning

A Deep Learning Approach for Dual-Energy CT Imaging Using a Single-Energy CT Data

An Em Approach To Map Solution Of Segmenting Tissue Mixture Percentages With Application To Ct-Based Virtual Colonoscopy

Dual-energy CT imaging using a single-energy CT data is feasible via deep learning

Dual-energy CT Imaging Using a Single-energy CT Data Via Deep Learning: A Contrast-enhanced CT Study

Dual-energy Computed Tomography Imaging from Contrast-enhanced Single-energy Computed Tomography

DeepECG: Image-based Electrocardiogram Interpretation with Deep Convolutional Neural Networks

Comparison of the performance of conventional and spectral-based tagged stool cleansing algorithms at CT colonography

Obtaining dual-energy computed tomography (CT) information from a single-energy CT image for quantitative imaging analysis of living subjects by using deep learning

CT-based synthetic contrast-enhanced dual-energy CT generation using conditional denoising diffusion probabilistic model

An Automatic Colon Cleansing Method for Virtual Colonoscopy

Improving diagnostic confidence in low-dose dual-energy CTE with low energy level and deep learning reconstruction

Estimating dual-energy CT imaging from single-energy CT data with material decomposition convolutional neural network

Computer Aided Annotation of Early Esophageal Cancer in Gastroscopic Images based on Deeplabv3+ Network

Pseudo Dual Energy CT Imaging using Deep Learning Based Framework: Initial Study