人工智能辅助诊疗系统在新型冠状病毒肺炎诊疗中的应用初探  

Application of Artificial Intelligence-Assisted Diagnostic System in Diagnosis and Treatment of Coronavirus Disease 2019

作　　者：王娜娜[1] 王大为 李大胜[1] 于卫永[3,4] 田凇 于巍伟 WANG Nana;WANG Dawei;LI Dasheng;YU Weiyong;TIAN Song;YU Weiwei(Department of Radiology,Beijing Haidian Hospital(Haidian Section of Peking University Third Hospital),Beijing 100080,China;Institute of Advanced Research,Beijing Infervision Technology Co.,Ltd,Beijing 100025,China;School of Rehabilitation Medicine,Capital Medical University,Beijing 100068,China;Department of Radiology,Beijing Bo’ai Hospital,China Rehabilitation Research Center,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]北京推想科技有限公司先进研究院,北京100025 [3]首都医科大学康复医学院,北京100068 [4]中国康复研究中心北京博爱医院放射科,北京100068

出　　处：《中国医疗设备》2020年第6期75-79,共5页China Medical Devices

摘　　要：目的探索基于CT影像的人工智能(Artificial Intelligence,AI)辅助诊疗系统在新型冠状病毒肺炎(Coronavirus Disease2019,COVID-19)诊疗中的应用价值。方法回顾性收集与分析29例COVID-19确诊患者的44次CT检查数据。2名资深影像诊断医师采用常规阅片方式和AI辅助阅片方式对CT图像进行诊断评价,对比两种方法对COVID-19受累肺叶的检出率和诊断用时,并应用AI辅助诊疗系统对感染区域进行定量分析,探讨定量分析结果与病情进展之间的相关性。结果AI辅助诊疗系统阅片方式肺叶检出数高于常规诊断方法(128个vs. 102个),且诊断用时明显减少[(0.67±0.56)min vs.(3.18±2.03)min],均具有统计学意义(P<0.05)。COVID-19早期、进展期及放射学转归期在感染区域(全肺、右肺上叶、右肺下叶、左肺上叶、左肺下叶)体积占比及感染区域内CT值分布(>60 HU及-270~30 HU)体积占比组间对比具有统计学差异(P<0.05)。结论 AI辅助诊疗系统可以高效辅助COVID-19诊断及量化评估,为COVID-19的病程分析提供帮助。Objective To explore the application value of artificial intelligence(AI)-assisted diagnostic system based on CT images in the diagnosis and treatment of coronavirus diseases 2019(COVID-19).Methods A total of 29 confirmed COVID-19 patients with 44 chest CT scans were retrospectively collected and analyzed in this study.Two senior radiologists evaluated CT images with or without the assistant of AI diagnostic system.The detection efficiency of infected lung lobes and the reading time were compared between the two methods.The AI diagnostic system was then used to quantitatively measure the infected volume in both lungs,and the correlation between the quantitative analysis results and the disease progress was explored.Results The detection rate of pulmonary lobes in AI diagnosis system was higher than that in the conventional diagnosis method(128 vs.102),and the diagnosis time was significantly reduced[(0.67±0.56)min vs.(3.18±2.03)min](P<0.05).There were significant differences in the proportion of the volume of the infected area(whole lung,right upper lobe,right lower lobe,left upper lobe,left lower lobe)and the distribution of the CT value(>60 HU,-270~30 HU)between the groups in the early stage,progressive stage and radiologic transition stage of COVID-19(P<0.05).Conclusion The AI-assisted diagnostic system can efficiently assist the diagnosis and quantitative assessment of COVID-19 lesions,which provides a great help in assessing the disease process.

关 键 词：新型冠状病毒肺炎 计算机断层扫描 人工智能 定量评估 

分 类 号：R563.1[医药卫生—呼吸系统] R318[医药卫生—内科学]