摘要
目的:探讨深度学习图像重建(DLIR)算法在提升头颈CT血管成像(CTA)颅底血管图像质量及诊断信心中的价值。方法:回顾性分析我院行头颈CTA检查的40例患者,对原始数据进行不同算法及水平的重建,包括基于多模型的自适应统计迭代重建-Veo (60%ASIR-V、90%ASIR-V)及DLIR (DLIR-L、DLIR-M、DLIR-H);比较不同算法及水平重建图像的主观及客观评价。结果:椎动脉及颈内动脉ASIR-V重建图像CT值与DLIR重建图像CT值间的差异均无统计学意义(P>0.05)。随着ASIR-V及DLIR重建水平的增加,图像噪声均减低,DLIR-H的噪声最低,与60%ASIR-V比较,差异有统计学意义(P<0.05)。随着ASIR-V及DLIR水平的增加,椎动脉及颈内动脉SNR、CNR均升高,椎动脉DLIR-H重建图像信噪比(SNR)、对比度噪声比(CNR)最高,与60%ASIR-V比较,差异有统计学意义(P<0.05)。90%ASIR-V主观评分与60%ASIR-V间的差异无统计学意义(P>0.05);DLIR主观评分在不同重建水平间的差异无统计学意义(P>0.05),DLIR-H及DLIR-M的主观评分均明显高于60%ASIR-V,差异有统计学意义(P<0.05)。结论:在头颈CTA颅底血管成像中,与ASIR-V相比,DLIR可以进一步降低图像噪声,提升图像质量和诊断信心。
Purpose: To explore the value of deep learning image reconstruction(DLIR) algorithm in improving the image quality and diagnostic confidence of skull base vessels in head and neck CTA. Methods: Retrospective analysis was performed on 40 patients who underwent head and neck CTA examination in our hospital. The original data were reconstructed with different algorithms and levels, including adaptive statistical iterative reconstructionVeo(60%ASIR-V and 90%ASIR-V) and DLIR(DLIR-L, DLIR-M, DLIR-H);The subjective and objective evaluation of different algorithms and levels reconstructed images were compared. Results: CT values of ASIR-V and DLIR reconstruction images of vertebral artery and internal carotid artery were with no statistically significant difference(P>0.05). With the increase of ASIR-V and DLIR reconstruction levels, the image noise decreased, and DLIR-H was with the lowest noise, which was with statistically significant difference when compared with 60%ASIR-V(P<0.05). With the increase of ASIR-V and DLIR levels, both signal-to-noise ratio(SNR) and contrastto-noise ratio(CNR) of vertebral artery and internal carotid artery increased, DLIR-H reconstruction images of vertebral artery was with the highest SNR and CNR, which were with statistically significant difference when compared with 60%ASIR-V(P<0.05). The subjective score of 90%ASIR-V was lower than that of 60%ASIR-V, with no statistically significant difference(P>0.05). The subjective score of DLIR increased with the increase of reconstruction level, and the subjective score of DLIR-H and DLIR-M was significantly higher than that of 60%ASIR-V, with statistical significance(P<0.05). Conclusion: Compared with ASIR-V, DLIR can be used to further reduce image noise and improve image quality and diagnostic confidence of skull base vessels in head and neck CTA.
作者
杨彦兵
田艳鑫
谢志远
于梓婷
阮小伟
汪芳
杨利莉
YANG Yanbing;TIAN Yanxin;XIE Zhiyuan;YU Ziting;RUAN Xiaowei;WANG Fang;YANG Lili(Department of Radiology,The People's Hospital of Ningxia Hui Autonomous Region(The First Afiliated Hospital of Northwest University for Nationalities))
出处
《中国医学计算机成像杂志》
CSCD
北大核心
2022年第6期579-583,共5页
Chinese Computed Medical Imaging
基金
宁夏回族自治区重点研发(一般)项目(2019BEG03046,2021BEG03092)
西北民族大学2019年中央高校基本科研项目(31920190184)。
关键词
头颈部
CT血管成像
图像质量
深度学习图像重建
自适应统计迭代重建V
Head and neck
Computed tomography angiography
Image quality
Deep learning image reconstruction
Adaptive statistical iterative reconstruction-Veo
