Visualization of simulated small vessels on computed tomography using a model-based iterative reconstruction technique

This article describes a quantitative evaluation of visualizing small vessels using several image reconstruction methods in computed tomography. Simulated vessels with diameters of 1–6 mm made by 3D printer was scanned using 320-row detector computed tomography (CT). Hybrid iterative reconstruction (hybrid IR) and model-based iterative reconstruction (MBIR) were performed for the image reconstruction.


a b s t r a c t
This article describes a quantitative evaluation of visualizing small vessels using several image reconstruction methods in computed tomography. Simulated vessels with diameters of 1-6 mm made by 3D printer was scanned using 320-row detector computed tomography (CT). Hybrid iterative reconstruction (hybrid IR) and model-based iterative reconstruction (MBIR) were performed for the image reconstruction.
& 2017 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). This data shows the superiority of MBIR for visualization of small vessels.

CT Scanning protocol
CT scans were obtained on a 320-detector scanner (Aquilion ONE ViSION Ed., Toshiba Medical Systems Corp., Tokyo, Japan) with ECG-triggering. Scanning was at a tube voltage of 120 kV in volume-scanning mode. The tube current was set at 60 mA to emulate the routine imaging conditions (noise index 25 SD) for coronary CT angiography (CCTA). The other scanning parameters were 0.275 s rotation time, 0.5 mm slice thickness, 40 mm (0.5 mm Â 80 slices) z-coverage range, and 100-mm field of view.

CT Image reconstruction
Adaptive iterative dose reduction 3D (AIDR 3D, Toshiba Medical Systems Corp.) was used as the hybrid IR. To obtain AIDR 3D images we used three kinds of standard soft tissue convolution kernels (FC13, FC14 and FC15); they are recommended by the vendor for routine CCTA studies. Forward projected model-based iterative reconstruction solution (FIRST, Toshiba Medical Systems Corp.) with CCTA mode was used as the MBIR.

Data analysis and quantitative evaluation
To evaluate the visibility of simulated vessels we recorded their CT attenuation profile curves (APCs) on CTA images. We used ImageJ software [1] and its particle analysis tool (Plot Profile) to generate APCs and recorded 36 APCs radially around the vessel centers at 5°intervals. As the objective edge response index on APCs we determined the 10 -90% edge rise distance (ERD) [2] and the 10-90% edge rise slope (ERS) at the vessel boundaries (Fig. 2). The distance and slope values were examined on both sides of the simulated vascular wall, hence, a total of 72 of ERDs and 72 ERSs was obtained and they were averaged in every simulated vessel. We also recorded the peak CT attenuation number in Hounsfield units (HU, PCT-HU) on the APCs and the CT attenuation value and the standard deviation (image noise) for the simulated aorta. The signal-to-noise ratio (SNR) for each vessel was calculated by dividing PCT-HU by the image noise.   3. Results of attenuation profile curve analysis. On images reconstructed with FIRST, the ERD was smaller and the ERS was larger than on vessel images reconstructed with the other methods.

Data
The ERDs and ERSs measured on vessels with different diameters are shown in Figs. 3a and 3b. With FIRST the ERD was smaller and the ERS was larger than with AIDR 3D. The PCT-HU value of the vessels is shown in Fig. 3c. For 6-mm vessels the PCT-HU obtained with the different reconstruction methods was similar. On images reconstructed with AIDR 3D and any of the convolution kernels-, the smaller the vessel diameter, the lower was the PCT-HU value. On FIRST images, the PCT-HU value were constantly stable and highest at the vessel diameter of 2-5 mm. The aortic PCT-HU of images acquired with AIDR 3D-FC13, AIDR 3D-FC14, and AIDR 3D-FC15 and with FIRST was 412.3, 410.6, 410.8, and 411.1, respectively. Representative APCs are shown in Fig. 4. The image noise on those images was 21.5, 25.8, 30.6, and 24.2, respectively (Fig. 5); it was comparable on AIDR 3D-FC14-and FIRST images (24.2 and 25.8 HU, respectively).
As shown in Fig. 6, the SNR on AIDR 3D-FC14 and AIDR 3D-FC15 images of vessels was lower than on FIRST scans for vessels of all diameters examined. We obtained higher SNR values for vessels whose diameter exceeded 4 mm when AIDR 3D-FC13 rather than FIRST was applied.    6. Signal-to-noise ratios (SNRs). The SNR of simulated vessels with a diameter of up to 3 mm was higher on FIRST-than on AIDR 3D-FC14 and AIDR 3D-FC15 images. For vessels with a diameter greater than 4 mm the SNR value was higher on AIDR 3D-FC13-than FIRST images.