| 赵阳,于秀蕾,曹波,何林阳,王永强.应用人工智能辅助胸片筛查在肺结核感染高危人群中的应用价值[J].浙江中西医结合杂志,2024,34(11): |
| 应用人工智能辅助胸片筛查在肺结核感染高危人群中的应用价值 |
| To evaluate the clinical application value of artificial intelligence assisted screening for high-risk population of mycobacterium tuberculosis infection based on chest X-ray |
| 投稿时间：2024-07-05 修订日期：2024-09-20 |
| DOI： |
| 中文关键词: 结核，肺 人工智能 诊断, 计算机辅助 放射摄影术,胸部 |
| 英文关键词:Tuberculosis pulmonary Artificial intelligence Diagnosis, Computer-Assisted Radiography thoracic,Chest X-ray |
| 基金项目:杭州市科技发展计划（20201231Y035）；杭州市生物医药和健康产业发展扶持专项（2023WJC068） |
|
| 摘要点击次数: 271 |
| 全文下载次数: 1 |
| 中文摘要: |
| 目的 评价结合人工智能(AI)辅助胸片主动筛查肺结核感染高危人群的诊断效能,为完善结核病重点人群防控策略提供科学依据。方法 选取2021年5月某劳动密集型工厂2178例工人作为研究对象,应用搭载胸部X线的移动医疗车进行胸片筛查及实验室检查,低年资诊断医师现场阅片；现场筛查结束后两名高年资诊断医师重新阅片,阳性结果作为参照标准。结合AI的计算机辅助诊断技术(CAD)分析胸片获得评分,14天后同一低年资诊断医师结合CAD评分重新阅片。采用ROC曲线评价CAD肺结核诊断效能；采用DeLong检验对比有无结合CAD评分低年资医生诊断效能。结果 2178例工人胸片经两位高年资诊断医师阅片,23例(1.06%)为结核阳性,其中实验室检查证实5例(0.23%)为病原学阳性结核。CAD诊断的敏感度、特异度分别为100%、65.3%,AUC值为0.911。低年资医师现场诊断阳性为15例,敏感度为30.4%,特异度为99.6%,AUC值为0.650；低年资医师结合CAD评分诊断阳性为71例,敏感度为91.3%,特异度为97.4%,AUC值为0.944。两次诊断AUC之差为0.293,Z统计量为5.633,P<0.0001。结论 结合AI模型的CAD诊断可以提高低年资医师的肺结核胸片诊断效能,有利于结核病重点人群的主动筛查。 |
| 英文摘要: |
| 【Objective】 To evaluate the effect of chest X-ray active tuberculosis screening combined with artificial intelligence (AI) assisted diagnosis in high-risk population of mycobacterium tuberculosis infection, and to provide scientific basis for improving the prevention and control strategy of tuberculosis among key population.【Methods】 In May 2021, 2178 cases of a labour-intensive factory used mobile digital radiography for screening active chest X-ray and laboratory examination, and a report was issued by a junior radiologist on site.At the end of the screening, all images were re-read by two senior radiologist, with positive tuberculosis results as a reference.All images were analyzed with Computer-Aided Detection (CAD) of AI, and scores were obtained. After 14 days, the junior radiologist reviewed the images again with CAD results. ROC curve was used to evaluate the diagnostic efficiency of AI diagnostic model,DeLong test was used to compare the diagnostic efficacy of CAD assisted with or without junior radiologist.【Results】 Among 2178 chest radiographs reviewed by two senior junior radiologist, 23 cases (1.06%) were positive for tuberculosis, among which 5 cases (0.23%) were confirmed to be etiological positive for tuberculosis.The sensitivity and specificity of AI system were 100% and 63.2%,the AUC value is 0.911. Without CAD, 15 cases were diagnosed with positive results, with sensitivity of 30.4%, specificity of 99.6% and AUC value of 0.650. With CAD, 71 cases were diagnosed with positive results, with sensitivity of 91.3%, specificity of 97.4% and AUC value of 0.944. The difference between areas of two diagnoses was 0.293 and Z statistic was 5.633, P<0.0001.【Conclusion】 CAD diagnosis combined with AI can improve the sensitivity of tuberculosis diagnosis in junior radiologist.It demonstrates the feasibility of CAD diagnostics for tuberculosis screening in high-risk populations on a mobile digital radiography. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
