基于深度学习的主动脉根部计算机断层扫描血管造影图像快速分割的质量评估与临床价值  

Quality assessment and clinical value of rapid segmentation of computed tomography angiography images of aortic root based on deep learning

作　　者：黄晓斌 吴正参 黄莹 陈谦 柴浩 叶鹏 谢光辉[1] 徐辉[1] 马琛明 杨剑 张海波[7] 陈国中 HUANG Xiaobin;WU Zhengcan;HUANG Ying;CHEN Qian;CHAI Hao;YE Peng;XIE Guanghui;XU Hui;MA Chenming;YANG Jian;ZHANG Haibo;CHEN Guozhong(Department of Medical Imaging,Nanjing First Hospital,Nanjing 210012,China;Department of Ultrasound,Nanjing Youan Hospital,Nanjing 211122,China;Department of Cardiothoracic Surgery,Nanjing First Hospital,Nanjing 210012,China;Department of Cardiology,Nanjing First Hospital,Nanjing 210012,China;Tavi Mercy Data Technology(Nanjing)Co.,Ltd.,Nanjing 210035,China;Department of Cardiovascular Surgery,Xijing Hospital,Air Force Medical University,Xi'an 710032,China;Department of Cardiovascular Surgery,Beijing Anzhen Hospital,Capital Medical University,Beijing 100029,China)

机构地区：[1]南京市第一医院医学影像科,江苏南京210012 [2]南京市佑安医院超声科,江苏南京211122 [3]南京市第一医院心胸外科,江苏南京210012 [4]南京市第一医院心脏内科,江苏南京210012 [5]拓微摹心数据科技(江苏南京)有限公司,江苏南京210035 [6]空军军医大学西京医院心血管外科,陕西西安710032 [7]首都医科大学附属北京安贞医院心外科,北京100029

出　　处：《空军军医大学学报》2025年第2期187-191,共5页Journal of Air Force Medical University

基　　金：国家自然科学基金青年科学基金(82001811);国家重点研发计划项目(2020YFC2008105);南京市生命健康科技专项(202205047)。

摘　　要：目的通过与计算机断层扫描血管造影(CTA)图像分析相比较,探讨基于深度学习方法实现快速、准确的主动脉根自动分割与评估的可靠性和临床价值。方法收集2021年8月至2022年10月在南京市第一医院接受经导管主动脉瓣置换术(TAVR)的68例患者的主动脉根部CTA图像,分别使用3mensio工作站和Cvpilot人工智能系统工作站进行分析。对分析结果进行回顾性统计分析与专家主观评价。结果在专家主观评价的68例中,63例(92.6%)患者达到Ⅲ级,4例(5.9%)达到Ⅱ级,1例(1.5%)达到Ⅰ级。Cvpilot人工智能系统工作站的测量结果与3mensio工作站的测量结果无显著的统计学差异(P>0.05)。相较于低年资医生,Cvpilot人工智能系统工作站的测量结果更加接近于专家的评估结果。此外,Cvpilot人工智能系统工作站显著减少了测量的总耗时与鼠标总移动距离(P<0.01)。结论对于接受TAVR的患者而言,Cvpilot人工智能系统工作站具有与常规计算机断层扫描测量结果良好的相关性(r=0.883)。该平台可靠的评估质量使其在未来具有广阔的临床应用前景。Objective To explore the reliability and clinical value of fast and accurate automatic segmentation and evaluation of aortic root based on deep learning compared with computed tomography angiography(CTA) image analysis.Methods CTA images of 68 patients undergoing transcatheter aortic valve replacement(TAVR) in Nanjing First Hospital from August 2021 to October 2022 were collected and analyzed using 3mensio and Cvpilot platform,respectively.The analysis results were analyzed retrospectively and subjectively by experts.Results Of the 68 patients subjectively evaluated by experts,63(92.6%) reached Grade Ⅲ,4(5.9%) reached Grade Ⅱ,and 1(1.5%) reached Grade Ⅰ.There were no statistically significant differences between the measurement results of Cvpilot platform and 3mensio(P>0.05).Compared with junior doctors,the measurement results of the Cvpilot platform were closer to the evaluation results of experts.In addition,the Cvpilot platform significantly reduced the total measured time and mouse travel distance(P<0.01).Conclusion For patients undergoing TAVR,the Cvpilot platform has a good correlation with conventional computed tomography measurements(r=0.883).The reliable evaluation quality of the platform makes it have broad clinical application prospects in the future.

关 键 词：主动脉瓣 经导管主动脉瓣置换术 自动分割 深度学习 计算机断层扫描 

分 类 号：R654.2[医药卫生—外科学] R816.2[医药卫生—临床医学]