深度学习图像算法用于改进大动脉炎CTA图像质量的研究
Study of deep learning image post-processing algorithm for CTA image improvement in patients with Takayasu arteritis
杨晰奥 1丁宁 1孔令燕 1王沄 1徐敏 2王志伟 1张大明 1王怡宁 1陈瑾 1金征宇 1冯逢1
作者信息
- 1. 100730 北京,中国医学科学院北京协和医院放射科
- 2. 100027 北京,佳能医疗系统(中国)有限公司
- 折叠
摘要
目的 探究深度学习重建算法对大动脉炎患者主动脉CT血管造影(CTA)图像质量的改善效果.方法 回顾性收集我院53例行主动脉CTA检查的大动脉炎患者,动脉期和延迟期图像分别经基于深度学习的高级智能Clear-IQ引擎(AiCE)重建和三维自适应迭代重建(AIDR 3D)得到4组图像.比较主观指标及客观指标[升主动脉、降主动脉、腹主动脉、左髂动脉、右髂动脉及同层面脊柱旁肌肉的CT值和标准差(SD)、计算信噪比(SNR)和对比噪声比(CNR)]的差异.结果 动脉期AiCE组各部位CT值均略高于AIDR 3D组,差异比较具有统计学意义(P＜0.001),延迟期两组图像左髂动脉CT值比较无统计学意义(P＞0.05),其余各血管的CT值比较均有统计学意义(P＜0.05).AiCE组SD值均低于AIDR 3D组,差异有统计学意义(P＜0.001).AiCE两组图像的SNR、CNR高于AIDR 3D两组,应用AiCE重建算法动脉期组SNR平均提高约55.7％,CNR平均提高约81.9％,延迟期组SNR平均提高约56.1％,CNR平均提高约75.7％,差异均有统计学意义(P＜0.001).两位医生主观评分的一致性较好,Kappa值在0.64～0.88.AiCE组总体图像质量主观评分优于AIDR 3D组,主观评分差异有统计学意义(P＜0.001).结论 AiCE算法与AIDR 3D算法相比,AiCE算法能有效改善主动脉CTA图像质量,提高主动脉CTA在大动脉炎临床诊断中的应用价值.
Abstract
Objective To explore the influence of deep learning reconstruction algorithm on improving the quality of aortic CT angiography(CTA)images in patients with Takayasu arteritis.Methods Retrospectively collected 53 patients with large artery inflammation underwent aortic CTA in our hospital were included.The arterial phase and delayed phase images were reconstructed with the advanced intelligent Clear-IQ engine(AiCE)based on deep learning and three-dimensional adaptive iterative dose reduction(AIDR 3D),resulting in four groups of images.The differences in subjective and objective parameters[CT values and standard deviation(SD)of the ascending aorta,descending aorta abdominal aorta,left iliac artery,right iliac artery,and paraspinal muscles at the same level,as well as the calculation of signal-to-noise ratio(SNR)and contrast-to-noise ratio(CNR)]were compared.Results CT values of all regions in the arterial phase of AiCE group were slightly higher than those of the AIDR 3D group,and the differences were statistically significant(P＜0.001).In the delayed phase,there was no significant difference in the CT values of left iliac artery between the two groups(P＞0.05),but the CT values of the other vessels were significantly different(P＜0.05).The SD values of the AiCE group were lower than those of the AIDR 3D group,and the differences were statistically significant(P＜0.001).The SNR and CNR of the images in the AiCE group were higher than those in the AIDR 3D.In the arterial phase,the SNR increased by an average of 55.7％and the CNR increased by an average of 81.9％with the AiCE reconstruction algorithm.In the delayed phase,SNR increased by an average of 56.1％and CNR increased by an average of 75.7％,and all the differences were statistically significant(P＜O.001).The consistency of the subjective scores between two doctors was good,with Kappa values ranging from 0.64 to 0.88.The overall subjective image quality score of the AiCE group was better than that of the AIDR 3D group,and difference in subjective scores was statistically significant(P＜0.001).Conclusion Compared with AiCE algorithm and AIDR 3D algorithm,AiCE algorithm can effectively improve the image quality of aortic CTA and improve the application value of aortic CTA in the clinical diagnosis of Takayasu arteritis.
关键词
体层摄影术,X线计算机/血管造影/深度学习重建/图像质量/主动脉/大动脉炎Key words
Tomography,X-ray computed/Computed tomography angiography/Deep learning reconstruction/Image quality/Aorta/Takayasu arteritis引用本文复制引用
出版年
2024
NETL
NSTL
万方数据