人工智能影像诊断系统在老年人肺结核主动筛查中的应用

Application of artificial intelligence imaging diagnostic system in active screening of pulmonary tuberculosis in the aged

目的初步评估人工智能(artificial intelligence,AI)影像诊断系统(简称“AI系统”)在柳州市老年人中开展肺结核主动筛查的应用价值。方法2021年9月—2022年5月柳州市基层医疗卫生机构应用AI系统对老年人进行肺结核主动筛查。分别收集AI系统与柳州市结核病定点医院影像专家的肺结核诊断结果。采用配对χ^(2)检验考察AI系统和影像专家诊断结果的差异,采用Kappa分析考察诊断结果的一致性。结果共6320名研究对象,男女比例为1∶1.21,年龄65~98岁,中位年龄为70(67,75)岁(D=0.127,P<0.001)。研究对象的AI诊断百分值在0%~99.50%,中位值为0%(0%,1.30%)(D=0.423,P<0.001)。AI系统和影像专家的诊断结果分别为肺结核者占5.44%和4.49%;疑似肺结核者占6.27%和5.70%;非肺结核者占88.29%和89.81%。配对χ^(2)检验结果显示χ^(2)=59.759,P<0.001;AI系统识别(疑似)肺结核者的灵敏度约为95.19%,识别非肺结核者的特异度约为97.76%,约登指数约为0.93;Kappa分析结果显示Kappa=0.861,P<0.001。结论针对柳州市老年人肺结核主动筛查,AI系统的诊断结果与柳州市结核病定点医院影像专家的诊断结果有较高的一致性,AI系统在柳州市老年人肺结核主动筛查中有较高的诊断价值。

Objective To preliminarily evaluate the application value of artificial intelligence(AI)imaging diagnostic system(referred to as AI system)in active tuberculosis screening among the aged in Liuzhou City.Methods From September 2021 to May 2022,the aged in primary healthcare institutions in Liuzhou City who were screened for pulmonary tuberculosis by AI system were enrolled.Tuberculosis diagnosis results were collected from the AI system and imaging specialists in pulmonary tuberculosis designated medical institutions in Liuzhou City.The pairedχ^(2) test was used to investigate the difference between the diagnostic results of the AI system and imaging specialists.Kappa analysis was used to investigate the consistency of the diagnostic results.Results A total of 6320 subjects were included in this study,with a male to female ratio of 1∶1.21,aged 65 to 98 years old,with a median age of 70(67,75)years old(D=0.127,P<0.001).The percentile of AI diagnosis in the subjects ranged from 0%to 99.50%,with a median value of 0%(0%,1.30%)(D=0.423,P<0.001).There were 5.44%and 4.49%of tuberculosis diagnoses by the AI system and imaging specialists,respectively.There were 6.27%and 5.70%suspected pulmonary tuberculosis and 88.29%and 89.81%non-pulmonary tuberculosis.The pairedχ^(2) test showed thatχ^(2)=59.759 and P<0.001.The sensitivity of the AI system to identify(suspected)tuberculosis patients was about 95.19%.The specificity of the AI system to identify non-tuberculosis patients was about 97.76%,and the Youden index was about 0.93.The Kappa analysis showed that Kappa=0.861 and P<0.001.Conclusion In the active screening of pulmonary tuberculosis in the aged in Liuzhou City,the diagnostic results of the AI system has a high consistency with the diagnostic results of imaging specialists in tuberculosis designated hospitals in Liuzhou City.The AI system has a high diagnostic value in the active screening of pulmonary tuberculosis in the aged in Liuzhou City.