[Relationship between Image Quality and Reconstruction FOV in Deep Learning Reconstructed Images of CT].

In this study, we compared the image quality of deep learning reconstruction (DLR) with that of conventional image reconstruction methods under the same conditions of reconstruction FOV and acquisition dose assuming abdomen computed tomography (CT) in children. Standard deviation (SD) of the CT value, noise power spectrum (NPS), and task-based modulation transfer function (TTF) were evaluated. DLR reduced image noise while maintaining sharpness, and the noise reduction effect showed a different characteristic depending on the size of reconstruction FOV from the conventional image reconstruction methods. The SD of CT value increased gradually in the range from 320 mm to 240 mm, but there was almost no change from 240 mm to 200 mm. The NPS showed completely different characteristics. The low-frequency component increased, and the high-frequency component decreased at 240 mm. However, the frequency component below 0.5 cycle/mm decreased at 200 mm and the peak frequency moved to the lower side at 320 mm. DLR showed the highest TTF value compared to the conventional reconstruction methods.

Accuracy of virtual non-contrast images with different algorithms in dual-energy computed tomography.

We investigated the accuracy of the computed tomography (CT) numbers of virtual non-contrast (VNC) images for two different material decomposition algorithms using the same image data for different diluted contrast agent concentrations. A container filled with contrast agents was inserted into a cylindrical phantom and scanned with dual-energy protocols (80/Sn140 kV, 100/Sn140 kV) using a dual-source CT. VNC images were generated by the 2-material decomposition (MD) algorithm using the energy of each tube voltage and the linear attenuation coefficient, calculated from the theoretical spectral curve of the agent and the CT number of the image, respectively. Furthermore, VNC images using 3-material decomposition (3-MD) algorithm were produced by applying LiverVNC, an analysis parameter implemented in the scanner. The robustness of both the algorithms was verified by investigating the CT numbers of the agents in the VNC. The closer the CT number is to 0 HU, the more robust the algorithm. Without beam-hardening correction, the CT numbers increased with an increase in concentration in both the algorithms, maximal at 50 mg/ml concentration, with CT numbers of 38 HU for 2-MD, 86 HU for 3-MD. With correction, CT numbers were ± 10 HU or less for both the algorithms up to 30 mg/ml concentration, whereas, for concentrations above 40 mg/ml, the maximal averaged CT number was 12 HU for 2-MD, 22 HU for 3-MD. For both the algorithms, the accuracy of the CT numbers was maintained in the low-concentration range; parameter adjustment was necessary to maintain the accuracy at concentrations higher than clinically expected.

Accuracy of virtual monochromatic images generated by the decomposition of photoelectric absorption and Compton scatter in dual-energy computed tomography.

The decomposition of the linear attenuation coefficient into photoelectric absorption and Compton scattering provides virtual monochromatic images (VMIs). The accuracy of the computed tomography (CT) number of VMI, which is obtained by decomposing the linear attenuation coefficient into photoelectric absorption and Compton scattering, was verified in the energy range of 40-200 keV. The possibility of improving the accuracy of CT numbers by using pre-energy-calibrated images as input was also investigated. The VMIs were generated in two groups of images: (i) dual-energy scanned images and (ii) high- and low-energy images generated by two-material decomposition (i.e., pre-energy-calibrated images). The object for analysis was solid iodine rods inserted in the center of the multi-energy CT phantom. The VMIs were generated from the dual-energy scanned images and pre-energy-calibrated images, and the theoretical and measured CT numbers of solid iodine rods were compared. Furthermore, the absolute error (AE) and relative error (RE) were calculated. With both images, the accuracy of the CT numbers was extremely high for regions close to the high- and low-tube-voltage X-ray energy or the high and low energy of the input images. By using the pre-energy-calibrated images, the maximum AE was reduced from 133 to 96 HU at an energy of 40 keV. Similarly, the maximum RE was reduced from 325 to 50% at an energy of 200 keV. The pre-energy-calibrated images reduced the overall error of the CT numbers and controlled the energy region where accurate CT numbers could be obtained.

Accuracy of spectral curves at different phantom sizes and iodine concentrations using dual-source dual-energy computed tomography.

To validate the accuracy of spectral curves obtained by an image-data-based algorithm and clarify the error factors that reduce accuracy. Iodine rods of known composition and different concentrations were inserted into a cylinder or elliptic-cylinder phantom and scanned according to the dual-energy protocol. Spectral curves were obtained by (i) theoretical calculation, (ii) image-data-based 2-material decomposition, and (iii) using a dedicated workstation. Accuracy was verified by comparing the spectral curve obtained by theoretical calculations with those obtained by the image-data-based algorithms or the dedicated workstations. For a quantitative evaluation, the error and relative error (RE) were calculated. In the image-data-based calculation, the errors with respect to the theoretical CT number ranged from - 8.3 to 71.1 HU. For all 192 combinations, 80.7% of the errors were under ± 15 HU, and 97.9% of the REs were under 10%. In the dedicated workstation, the errors ranged from - 94.7 to 26.8 HU. For all combinations, 68.8% of the errors were under ± 15 HU, and 68.2% of the REs were under 10%. By appropriately setting the effective energy corresponding to the CT number of the basis materials, an accurate spectral curve can be obtained. The beam-hardening effect is canceled by the 2-material decomposition process even without beam-hardening correction. Accuracy is primarily reduced by scattered radiation rather than the beam-hardening effect.

[Relationship between Quantitativeness of Iodine and Accuracy of Virtual Non-contrast Image in Dual-energy Computed Tomography].

Dual-energy computed tomography (DE-CT) is the promising technology, such as enabling material decomposition, generation of the virtual monochromatic image, and measurement of effective atomic numbers. There are reports that utilization of the virtual non-contrast (VNC) image, the iodine map image, and the virtual monochromatic image can contribute to the improvement of lesion detection and its characterization, compared with conventional contrast CT by single-energy computed tomography (SE-CT). In addition, acquisition of the VNC images makes it possible to skip scanning of true non-contrast CT, which is also expected to reduce exposure. However, a reliable evaluation of the accuracy of the VNC image has not been established, and only a few reports have verified their accuracy. In this study, we evaluated the relationship between the quantitativeness of iodine and the CT value of VNC image. As a result of our study, when the iodine volume was overestimated, the CT value of the VNC image was lower than the reference value, and when the iodine volume was underestimated, the CT value was upper than the reference value. Moreover, we clarified that the CT value of the VNC image greatly diverges as the iodine volume increases.