810例周边型肺阴影的人工智能辅助胸部CT诊断与术后病理诊断对比分析  

A comparative analysis of peripheral lung shadow diagnosed by artificial intelligence-assisted chest CT versus postoperative pathology in 810 patients

作　　者：杜林 张洪[1] 罗翔凤 吕军 孙大强 DU Lin;ZHANG Hong;LUO Xiangfeng;LV Jun;SUN Daqiang(Department of Thoracic Surgery,Tianjin Chest Hospital Affiliated to Tianjin University,Tianjin,300222,P.R.China;LinkDoc Science and Technology Co.Ltd.,Beijing,100080,P.R.China)

机构地区：[1]天津大学附属天津市胸科医院胸外科,天津300222 [2]零氪科技有限公司,北京100080

出　　处：《中国胸心血管外科临床杂志》2022年第7期854-858,共5页Chinese Journal of Clinical Thoracic and Cardiovascular Surgery

基　　金：2018年天津市新一代人工智能科技重大专项(18ZXZNSY00400)。

摘　　要：目的探讨人工智能(artificial intelligence,AI)辅助胸部电子计算机断层扫描(computed tomography,CT)在周围型肺阴影诊断中的临床应用价值。方法回顾性分析2018年1月—2019年7月在天津大学附属天津市胸科医院胸外科手术治疗的810例周围型肺阴影患者CT影像资料,其中男339例、女471例,中位年龄63岁。将术前AI辅助诊断肺周围型肺阴影的恶性概率与术后病理诊断结果进行对比分析。结果810例周围型肺阴影患者术后病理诊断分别为肺癌627例(77.4%)、癌前病变30例(3.7%)、肺良性病变153例(18.9%)。上述三组病例术前AI诊断的中位恶性概率分别为86.0%(肺癌)、90.0%(癌前病变)和37.0%(肺良性病变)。对本组病例进行AI恶性概率分布的工作特征曲线分析,结果显示,曲线下面积为0.882,诊断肺癌的恶性概率临界值为75.0%,敏感度为0.856,特异性为0.814。本组患者中,AI诊断恶性概率≥75.0%者有571例,其中537例术后病理诊断为肺癌,阳性预测值为94.0%(537/571)。结论AI辅助胸部CT诊断系统以恶性概率≥75.0%为诊断阈值,对周围型肺癌诊断有很高的准确率。Objective To investigate the clinical value of artificial intelligence(AI)-assisted chest computed tomography(CT)in the diagnosis of peripheral lung shadow.Methods The CT image data of 810 patients with peripheral pulmonary shadow treated by thoracic surgery in Tianjin Chest Hospital Affiliated to Tianjin University from January 2018 to July 2019 were retrospectively analyzed using AI-assisted chest CT imaging diagnosis system.There were339 males and 471 females with a median age of 63 years.The malignant probability of preoperative AI-assisted diagnosis of peripheral pulmonary shadow was compared with the results of postoperative pathology.Results The pathological diagnosis of 810 patients with peripheral pulmonary shadow was lung cancer in 627(77.4%)patients,precancerous lesion in 30(3.7%)patients and benign lesion in 153(18.9%)patients.The median probability of malignant AI diagnosis before operation was 86.0%(lung cancer),90.0%(precancerous lesion)and 37.0%(benign lesion),respectively.According to the analysis of receiver operating characteristic(ROC)curve of AI malignant probability distribution in this group of patients,the area under the ROC curve was 0.882.The critical value of malignant probability for diagnosis of lung cancer was 75.0%with a sensitivity of 0.856 and specificity of 0.814.A total of 571 patients were diagnosed with AI malignancy probability≥75.0%,among whom 537 patients were pathologically diagnosed as lung cancer with a positive predictive value of 94.0%(537/571).Conclusion The AI-assisted chest CT diagnosis system has a high accuracy in the diagnosis of peripheral lung cancer with malignant probability≥75.0%as the diagnostic threshold.

关 键 词：肺癌 人工智能 辅助诊断 周围型肺阴影 

分 类 号：R734.2[医药卫生—肿瘤] TP18[医药卫生—临床医学]