摘要
目的探讨人工智能肺结节筛查在肺癌高危人群中的应用效果及价值。方法选择2018年12月—2020年2月肺癌高危人群583例作为研究对象,所有患者均经手术组织检查最终确诊(金标准),确诊前患者均给予拟行CT检查,分析人工智能肺结节筛查在肺癌高危人群中的诊断效能(敏感性、特异性)。结果583例肺癌高危人群均经手术最终确诊肺癌患者56例,确诊率为9.61%。所有患者均完成CT检查,最终确诊55例,诊断符合率为98.12%;人工读片和人工智能肺结节筛查肺结节<1 mm、1~5 mm及>5 mm检出率均差异无统计学意义(χ^(2)=1.291、1.042、0.793,P=0.743、0.891、0.598);人工智能检测方法对1 mm层厚肺结节CT片2次读片Kappa值为0.938,接近1,具有良好的一致性。ROC曲线结果表明,人工智能检测方法用于肺癌高危人群中AUC值为0.867,诊断敏感性为0.814,特异性为0.894。结论将人工智能肺结节筛查用于肺癌高危人群中能达到早期肺癌识别的敏感性、特异性,能辅助医生诊断,为临床诊疗提供影像学依据和参考,值得推广应用。
Objective To explore the application effect and value of artificial intelligence lung nodule screening in high-risk groups of lung cancer.Methods 583 high-risk patients with lung cancer from December 2018 to February 2020 were selected as the target.All patients were finally diagnosed by surgical tissue examination(gold standard).Before the diagnosis,the patients were given CT examinations to analyze the diagnostic power(sensitivity,specificity)of artificial intelligence lung nodules screening in high-risk groups of lung cancer.Results All 583 high-risk patients with lung cancer were finally diagnosed with 56 cases of lung cancer after surgery,the diagnosis rate was 9.61%.All patients completed CT examinations,and 55 cases were finally diagnosed,with a diagnosis coincidence rate of 98.12%;the detection rates of pulmonary nodules<1 mm,1-5 mm and>5 mm by manual reading and artificial intelligence pulmonary nodule screening were not statistically significant(χ^(2)=1.291,1.042,0.793,P=0.743,0.891,0.598);the Kappa value of 2 readings of 1mm thick pulmonary nodule CT films with artificial intelligence detection method was 0.938,which was close to 1,which has good consistency.The ROC curve results show that the AUC value of the artificial intelligence detection method for the high-risk population of lung cancer was 0.867,the diagnostic sensitivity was 0.814,and the specificity was 0.894.Conclusion The use of artificial intelligence pulmonary nodule screening in high-risk groups of lung cancer can achieve the sensitivity and specificity of early lung cancer recognition,assist doctors in diagnosis,provide imaging evidence and reference for clinical diagnosis and treatment,and is worthy of popularization and application.
作者
张宝忠
ZHANG Bao-zhong(Department of Respiratory Medicine,People's Hospital of Weifang High-tech Industrial Development Zone,Weifang,Shandong Province,262700 China)
出处
《系统医学》
2021年第2期59-61,80,共4页
Systems Medicine
关键词
人工智能肺结节筛查
肺癌高危人群
肺癌识别
应用效果
ROC曲线
Artificial intelligence pulmonary nodule screening
High-risk population of lung cancer
Lung cancer recognition
Application effect
ROC curve