人工智能辅助诊断系统在COVID-19患者病程变化中的诊疗作用  

Role of Artificial Intelligence-assisted Diagnosis and Treatment System in the Change of Course of COVID-19 Patients

作　　者：王娜娜[1] 李大胜[1] 王大为 李珺[2] 于巍伟 于卫永[4,5] 陈振波[4,5] WANG Nana;LI Dasheng;WANG Dawei;LI Jun;YU Weiwei;YU Weiyong;CHEN Zhenbo(Department of Radiology,Beijing Haidian Hospital(Haidian Section of Peking University Third Hospital),Beijing 100080,China;Department of Infection,Beijing Haidian Hospital(Haidian Section of Peking University Third Hospital),Beijing 100080,China;Institute of Advanced Research,Infervision Medical Technology Co.,Ltd.,Beijing 100025,China;School of Rehabilitation Medicine,Capital Medical University,Beijing 100068,China;Department of Radiology,Beijing Boai Hospital,Beijing 100068,China)

机构地区：[1]北京市海淀医院(北京大学第三医院海淀院区)放射科,北京100080 [2]北京市海淀医院(北京大学第三医院海淀院区)感染科,北京100080 [3]北京推想科技有限公司先进研究院,北京100025 [4]首都医科大学康复医学院,北京100068 [5]北京博爱医院放射科,北京100068

出　　处：《中国医疗设备》2022年第8期37-41,65,共6页China Medical Devices

基　　金：北京市科技计划项目(Z211100003521011)。

摘　　要：目的使用基于电子计算机断层扫描(Computed Tomography,CT)影像学资料的人工智能(Artificial Intelligence,AI)辅助诊断系统动态评估新型冠状病毒肺炎(Corona Virus Disease 2019,COVID-19)治疗前后的影像表现,为COVID-19的精准诊疗提供影像学定量依据。方法本研究回顾性收集28例COVID-19患者的共78次CT影像学资料和临床资料,通过对比分析患者治疗前后临床实验室检查结果,多次CT影像结果并应用AI技术进行辅助诊断,分析CT影像学特征及AI定量特征结果。结果对于28例COVID-19患者,临床治疗平均(10.50±3.60)d后,实验室检查指标比较,差异无统计学意义(P>0.05);AI辅助诊断与常规阅片诊断一致性较高(Kappa=0.708,P<0.05),诊断用时显著减少[(2.77±0.90)min vs.(0.78±0.40)min,P<0.05],并能够评估肺部病情定量变化(肺内平均CT值、全肺感染体积占比及全肺感染体积),在早期、进展期和影像转归期三期两两比较有显著差异(P<0.05)。结论CT+AI联合应用可作为现期筛查COVID-19的高效模式,可以在临床疗效中提供重要的参考价值。Objective To dynamically evaluate the imaging performance of Corona Virus Disease(COVID-19)before and after treatment by applicating CT image-based artificial intelligence(AI)-assisted diagnosis system,in order to provide reliable imaging quantification for the precise diagnosis and treatment of COVID-19.Methods A total of 78 CT scan data and clinical data of 28 patients with confirmed COVID-19 retrospectively were collected and analyzed,contrasted the data including multiple CT scan images and results of AI technology for auxiliary diagnosis system before and after treatments.Results For 28 patients with COVID-19,there was no significant change in index of laboratory tests after an average of(10.50±3.60)d of clinical treatment(P>0.05);the consistency between AI-assisted diagnosis and routine reading diagnosis was high(Kappa=0.708,P<0.05),the diagnostic time was significantly reduced(2.77±0.90)min vs.(0.78±0.40)min(P<0.05),and its quantitative changes could be evaluated(the mean CT value in the lung,the proportion of the whole lung infection volume and the whole lung infection volume).There were statistically significant differences in early stage,advanced stage,and outcome stage(P<0.05).Conclusion The combined application of CT+AI can be used as an efficient mode for current screening of COVID-19.CT+AI can provide important information in clinical diagnosis and treatment.

关 键 词：新型冠状病毒肺炎 电子计算机断层扫描 辅助诊断 人工智能 定量评估 

分 类 号：R563.1[医药卫生—呼吸系统]