人工智能软件在基层医院低剂量胸部CT筛查肺结节中的应用价值  

Application value of artificial intelligence software in low-dose chest CT screening of pulmonary nodules in primary hospitals

作　　者：高雪峰 徐磊 Gao Xuefeng;Xu Lei(Department of Radiology,Yinshanhu Hospital,Wuzhong District,Suzhou 215124,China)

机构地区：[1]苏州市吴中区尹山湖放射科,215124

出　　处：《实用医技杂志》2024年第5期372-374,共3页Journal of Practical Medical Techniques

摘　　要：目的评估人工智能(AI)软件在基层医院辅助放射科医生诊断肺结节的效能和价值。方法回顾性分析2023年7-12月尹山湖医院300例胸部CT资料,由2名具有15年以上阅片经验的放射科医师结合AI诊断肺结节作为筛查金标准。比较人工、AI及人工+AI 3种阅片方式对不同大小(<3 mm、3~5 mm及>5 mm)及不同部位(肺门区、中央区及胸膜下区)肺结节的诊断效能,包括灵敏度、特异度。结果AI阅片对不同大小及不同部位结节的诊断灵敏度均高于人工阅片,其中,二者在对大小3~5 mm及肺门区分布的结节诊断灵敏度差异具有统计学意义;AI阅片对不同大小及不同部位结节的特异度均低于人工阅片,其中,对分布于胸膜下区的结节二者的诊断特异度差异具有统计学意义。人工+AI阅片方式对不同大小及不同部位结节的诊断灵敏度、特异度均显著高于人工阅片或单纯AI阅片(P<0.05)。结论AI有助于提高大小3~5 mm及肺门区的结节检出率及诊断灵敏度,但存在诊断特异度较低的不足,AI+人工阅片方式则能提高肺结节的诊断效能,因此,AI辅助诊断肺结节在基层医院放射科具有较好的临床应用前景。Objective To evaluate the diagnostic efficacy of artificial intelligence(AI)software in assisting radiologists in diagnosing pulmonary nodules in primary hospitals.Methods A retrospective analysis was conducted on 300 low-dose chest CT data from Suzhou Yinshanhu Hospital from March 2023 to December 2023.Two radiologists with over 15 years of radiology working experience combined with AI to diagnose pulmonary nodules as the gold standard for screening.The diagnostic efficacy of three different diagnosis methods(manual,AI,and artificial+AI)for pulmonary nodules of different sizes(<3 mm,3~5 mm,and>5 mm)and different locations(hilar,central,and subpleural)were compared,including sensitivity and specificity.Result The diagnostic sensitivity of AI film reading for nodules of different sizes and locations was higher than that of manual diagnosis alone.Among them,there was a statistically significant difference in the diagnostic sensitivity of nodules with sizes ranging from 3~5 mm and distribution in the pulmonary hilum between the two.The specificity of AI diagnosis for nodules of different sizes and locations is lower than that of manual diagnosis.Among them,there is a statistically significant difference in diagnostic specificity for nodules distributed in the subpleural area between the two.The diagnostic sensitivity,specificity,and accuracy of manual combined with AI diagnosis method for nodules of different sizes and locations are significantly higher than those of manual diagnosis alone or AI diagnosis alone(P<0.05).Conclusion AI can help to improve the detection rate and diagnostic sensitivity of nodules with the size of 3~5 mm and located in pulmonary portal area,but there are shortcomings of relativcely low diagnostic specificity.Therefor,AI assisted diagnosis of pulmonary nodules has a good clinical application prospects in the department of primary hospitals.

关 键 词：人工智能 低剂量胸部CT 基层医院 肺结节 

分 类 号：R734.2[医药卫生—肿瘤] R730.44[医药卫生—临床医学]