|
KMID : 1149020210230020009
|
|
Journal of Korean Society of Computed Tomographic Technology
2021 Volume.23 No. 2 p.9 ~ p.19
|
|
|
Comparative Analysis and Usefulness by Quantitative Evaluation of Deep Learning Image Reconstruction and Adaptive Statistical Iterative Reconstruction-V in Aortic Vessels CT
|
|
Ko Chang-Su
Cho In-Wan
Kang Ji-Won
Jeong Woo-Jun
Song Hoon
|
|
|
Abstract
|
|
|
|
We studied that how the Deep Learning Image Reconstruction(DLIR) technique was useful through comparative evaluation after reconstructing the images by conventional ASIR-V technique and DLIR technique. We measured SSIM with PBU-60 Phantom and compared SNR and CNR by the each tissue equivalent material, using Tomotherapy Cheese Phantom. Also SNR and CNR were compared in the same way with phantoms, using the patients’ aortic vessels CT images. As a result of comparing average of SSIM from the images reconstructed by ASIR-V technique and True-Fidelity(TF) images which were reconstructed by DLIR, the average of SSIM from TF was higher than the average of SSIM from ASIR-V. In the using Phantom the result that based on SNR between two techniques had statistical significance.(p<0.05) The result that based on CNR was significant statistically in all the plugs except for the case of CNR from TF-H and ASIR 90% at the plug of brain. SNR result, there was statistical significance between the two techniques(p<0.05) in both cases of pre and post enhanced AVCT images from the patients. At the pre-enhanced images, CNR from every part had significant result statistically except for the part of descending aorta from TF-H and ASIR-90%. However, there was statistical significance in the post-enhanced images of every ROI.(p<0.05) DLIR technique help the radiologist and the clinicians diagnose the diseases accurately. Compared to the conventional technique reconstructing the images in AVCT, it would be good enough to get much more high-quality images.
|
|
|
KEYWORD
|
|
|
aortic vessel CT, deep learning image reconstruction, ASIR-V
|
|
|
FullTexts / Linksout information
|
|
|
|
|
|
Listed journal information
|
|
|
|