Application of a deep learning image reconstruction algorithm to improve the measurement accuracy of vessel wall thickening in pediatric patients with Takayasu arteritis
10.3760/cma.j.cn112149-20201203-01275
- VernacularTitle:应用深度学习图像重建算法提升多发性大动脉炎患儿增强CT血管壁测量精度的研究
- Author:
Jihang SUN
1
;
Lixin YANG
;
Xiaolu TANG
;
Haoyan LI
;
Yun PENG
Author Information
1. 国家儿童医学中心 首都医科大学附属北京儿童医院影像中心 100045
- Keywords:
Tomography, X-ray computed;
Takayasu arteritis;
Child;
Radiation dosage;
Deep learning
- From:
Chinese Journal of Radiology
2021;55(12):1308-1312
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To demonstrate whether image quality and measurement accuracy of vessel wall thickening could be improved using a deep learning image reconstruction (DLIR) algorithm in children with Takayasu arteritis.Methods:From September 2019 to April 2020, 32 patients with Takayasu arteritis underwent low-dose contrast-enhanced CT with 100 kVp in Beijing Children′s Hospital were enrolled retrospectively. The raw data were reconstructed at 0.625 mm slice thickness using the filtered back projection (FBP), 50% adaptive statistical iterative reconstruction-V (ASIR-V) at 50% (50%ASIR-V), ASIR-V at 100% (100%ASIR-V) and DLIR. Subjective evaluation including the image quality of vessel wall identification, overall image noise and diagnostic confidence were evaluated using a 5 points scales by 2 observers. Objective evaluation including the thickness and standard deviation of vessel wall were measured, then the coefficient of variation (CV) was calculated. The CT value and noise of aorta were measured to calculate the contrast to noise ratio (CNR) of image. Friedman test was used to compare the differences of subjective scores among the four groups, and the analysis of variance of random blocks was used to compare the differences of objective measurement indices.Results:In terms of subjective score results, there was no significant difference between 100%ASIR-V and DLIR of overall image noise ( P>0.05), and the image scores of the two groups were higher than those of FBP and 50%ASIR-V (all P<0.05). The vessel wall identification and diagnostic confidence of DLIR were higher than those of other images (all P<0.05). The objective measurement results showed that the standard deviation and CV of vessel wall thickness in DLIR were significantly lower than those in other images (all P<0.05). There was no significant difference in vascular noise, muscle noise and CNR between 100%ASIR-V and DLIR (all P>0.05), which were lower than those in FBP and 50%ASIR-V (all P<0.05). Compared with 50%ASIR-V, the CV of DLIR was reduced by 22.9%, and the CNR was increased by 46.8%. Conclusion:DLIR can improve the overall image quality of CECT in children with Takayasu arteritis and the measurement accuracy of vascular wall, making it possible to further reduce the radiation dose.
