| 刘亚斌,李 庆,周 围,郑常俊,舒 亚.胸部CT结合AI诊断系统对疑似肺结节患者的诊断及对结节类型的评估价值[J].,2022,(5):955-959 |
| 胸部CT结合AI诊断系统对疑似肺结节患者的诊断及对结节类型的评估价值 |
| Chest CT Combined with AI Diagnosis System in the Diagnosis of Patients with Suspected Pulmonary Nodules and the Evaluation Value of Nodule Types |
| 投稿时间：2021-08-03 修订日期：2021-08-27 |
| DOI：10.13241/j.cnki.pmb.2022.05.032 |
| 中文关键词: 胸部CT 人工智能诊断 肺结节 诊断价值 |
| 英文关键词: Chest CT AI diagnosis Pulmonary nodules Diagnostic value |
| 基金项目:四川省医学科研课题计划项目（Q18005） |
|
| 摘要点击次数: 478 |
| 全文下载次数: 282 |
| 中文摘要: |
| 摘要 目的：探讨胸部CT结合AI诊断系统对疑似肺结节患者的诊断及对结节类型的评估价值。方法：选取2019年12月-2020年12月在我院进行CT检查的358例疑似肺结节患者，将其按照随机数字表法分为两组：对照组（放射科医生根据CT扫描结果，通过人工阅片分析记录检出结节数量和影像特征），观察组（将CT扫描结果导入AI辅助诊断系统，经AI运算得到结节检出数量和影像特征）。AI辅助系统IMsight用于肺结节的图像分析和自动检测。通过组织病理学确定结节的良恶性。绘制受试者工作特征曲线（ROC）曲线以评估AI和CT结合图像的诊断价值。结果：病理结果最后确诊结节数量736个，恶性结节139个（18.89 %），良性结节597个（81.11 %）。观察组诊断结节数量717个，检出率97.42%，对照组诊断出结节数量603个，检出率81.93 %。观察组较对照组的结节检出率、阳性检出率升高（P<0.05），漏检率和假阴性率均显著降低（P<0.05）。当结节小于10 mm时，观察组较对照组的检出率升高（P<0.05），观察组较对照组对磨玻璃密度结节和实性结节检出率升高（P<0.05），观察组较对照组位于胸膜结节检出率升高（P<0.05）。观察组较对照组AUC（P<0.05），表明AI系统下的结节检出准确率高。ROC曲线显示观察组的敏感性和特异性分别为88.39%和89.68 %，对照组的敏感性和特异性分别为75.24 %和82.34 %，观察组较对照组的ROC曲线敏感性和特异性升高（P<0.05）。结论：AI辅助诊断系统可有效提高肺结节的检出率,减少误检率及漏检率,值得在肺结节CT检测中应用推广。 |
| 英文摘要: |
| ABSTRACT Objective: To investigate the diagnostic value of chest CT combined with AI diagnostic system in patients with suspected pulmonary nodules and the evaluation of nodule types. Methods: From December 2019 to December 2020, 358 patients with suspected pulmonary nodules who underwent CT examinations in our hospital were selected, and they were divided into two groups according to the random number table method: control group (radiologists used CT scan results to record the number of nodules and image characteristics through manual reading analysis) and observation group (import the CT scan results into the AI-assisted diagnosis system, and get the number of nodules detected and the image characteristics through AI calculations). All patients were examined by Siemens photonic dual-source CT. The AI assist system IMsight is used for image analysis and automatic detection of lung nodules. Determine the cancerousness of the nodule by histopathology. Draw a receiver-operator characteristic(ROC) curve to evaluate the diagnostic value of AI and CT combined images. Results: The pathological results finally confirmed 736 nodules, 139 malignant nodules (18.89 %), and 597 benign nodules(81.11 %). The number of nodules diagnosed in the observation group was 717, with a diagnosis rate of 97.42 %, and the number of nodules diagnosed in the control group was 603, with a diagnosis rate of 81.93 %. Compared with the control group, the nodule detection rate and positive detection rate of the observation group were higher(P<0.05), and both the missed detection rate and false negative rate are significantly reduced(P<0.05). When the nodule is smaller than 10mm, the detection rate of the observation group is higher than that of the control group(P<0.05), and compared with the control group, the observation group has higher detection rate of ground glass density nodules and solid nodules(P<0.05). Compared with the control group, the detection rate of nodules located in the middle of the pleura was higher (P<0.05). Compared with the control group, the observation group has a higher AUC (P<0.05), indicating that the detection accuracy of nodules under the AI system is higher. The ROC curve showed that the sensitivity and specificity of the observation group were 88.39% and 89.68%, and the sensitivity and specificity of the control group were 75.24% and 82.34%, respectively. Compared with the control group, the sensitivity and specificity of the observation group were higher(P<0.05). Conclusion: The AI-assisted diagnosis system can effectively improve the detection rate of lung nodules, reduce the false detection rate and the missed detection rate, and it is worthy of application and promotion in CT detection of lung nodules. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
