|
KMID : 1149020210230020021
|
|
Journal of Korean Society of Computed Tomographic Technology
2021 Volume.23 No. 2 p.21 ~ p.27
|
|
|
Quantitative analysis of image quality and radiation dose of ultra low dose chest CT using AI-based deep learning reconstruction algorithm ‘AiCE’
|
|
Ko Dong-Hee
Lee Soo-Young
Shim Hak-Joon
Jeon Pil-Hyeon
|
|
|
Abstract
|
|
|
|
With the development of new image reconstruction techniques, ultra-low-dose chest computed tomography (ULDCT) can acquire adequate image quality even in lower does than conventional low-dose chest CT (LDCT). Recently, advanced intelligent clear-IQ engine (AiCE), an artificial intelligence-based image reconstruction algorithm, has been commercialized in the medical field. In this study, AiCE, a state-of-the-art technology, was applied to ULDCT to reconstruct both phantom and human image. The lungman phantom was examined by using LDCT and ULDCT protocols. The difference in image quality which applied existing and AiCE reconstruction methods was quantitatively analyzed by computing signal to noise ratio (SNR), noise, and dose length product values. Also, under the application of same AiCE reconstruction methods, 30 patients were tested by using LDCT protocols and another 30 patients were tested by using ULDCT protocols.In the phantom study, the radiation dose of ULDCT was reduced by 49% compared to LDCT and ULDCT images with AiCE exhibited the highest SNR value and the lowest noise. In the patient study, the radiation dose of ULDCT was reduced by 55% compared to LDCT. In ULDCT images of thick upper lobe with AiCE, the SNR increases by about 2% and the noise decreased by 0.5% compared to LDCT. The selective usage of ULDCT within appropriate body mass index range might provide the reduction of exposure risk and the optimal images while relieving patient anxiety.
|
|
|
KEYWORD
|
|
|
Ultra low dose chest CT, artificial intelligence image reconstruction, deep learning reconstruction, AiCE
|
|
|
FullTexts / Linksout information
|
|
|
|
|
|
Listed journal information
|
|
|
|
|