摘要
目的探讨人工智能(AI)影像系统在肺部结节诊断中的真实世界数据。方法收集我院2019年2月至2020年1月单治疗组患有肺部结节并行手术治疗的患者222例。运用基于自适应3D-CNN技术的AI影像系统分析其肺部结节影像,得出关于肺结节大小、密度、位置、良恶性的信息。查询患者术后病理诊断结果,比较AI影像系统分析结果与术后病理诊断结果符合情况,探究AI影像系统在真实世界肺结节诊断能力。结果 222例肺部结节患者AI影像系统分析结果灵敏度、特异度、总符合率、漏诊率分别为67.0%、34.5%、58.6%和33.0%(Kappa=0.0143,P>0.05)。结论 AI影像系统在肺部结节诊断中可信度仍较低。
Objective Explore the real-world data of AI imaging system in the diagnosis of lung nodules.Methods Collected 222 cases of pulmonary nodules in the single treatment group in our hospital from February,2019 to January,2020.Analyze the images of lung nodules with the Deep Wise Medical AI imaging system based on adaptive 3D-CNN technology to obtain information about the size,density,location,benign and malignant of lung nodules.Query the postoperative pathological diagnosis results of patients,compare the analysis results of the AI imaging system with the postoperative pathological diagnosis results,and explore the ability of the AI imaging system to diagnose lung nodules in the real world.Results The sensitivity,specificity,total coincidence rate,and false negative rate of the AI imaging system analysis results of 222 patients with pulmonary nodules was 67. 0%,34. 5%,58. 6%,and 33. 0%,respectively(Kappa = 0. 0143,P>0. 05).Conclusion The credibility of this AI imaging system in the diagnosis of pulmonary nodules is still low.
作者
张涛
张登国
李建
蒲江涛
戴天阳
Zhang Tao;Zhang Dengguo;Li Jian(Department of Thoracic Surgery,Affiliated Hospital of Southwest Medical University,Luzhou,Sichuan 646000;Department 3 of Surgery,Hejiang County People′s Hospital,Luzhou,Sichuan 646000,China)
出处
《四川医学》
CAS
2021年第2期193-196,共4页
Sichuan Medical Journal
关键词
人工智能
肺部结节
肺癌
诊断
artificial intelligence
pulmonary nodules
lung cancer
diagnosis
