人工智能(AI)辅助低剂量CT联合肿瘤标志物对社区体检人群肺癌机会性筛查的应用研究

Research on the Application of Artificial Intelligence (AI) Assisted Low Dose CT Combined with Tumor Markers for the Chance Screening of Lung Cancer in Community Physical Examination Population

摘要: 目的：研究人工智能(AI)辅助低剂量CT联合多种肿瘤标志物在社区体检人群中肺癌机会性筛查的应用价值。方法：回顾性分析2023年1月~2024年12月间在本院经LDCT检查表现为肺结节的患者80例，且所有患者均进行过4种肺癌肿瘤标记物(SCC、CEA、Cyfra21-1及NSE)检测，以病理结果为金标准，通过对比分析各组对肺结节良恶性诊断的准确性、敏感性、特异性等，分析不同诊断方式的价值。结果：80例经病理诊断显示恶性结节63个，良性结节17个，人工智能系统(AI)辅助 + 医师诊断组 + 联合肿瘤标志物诊断组诊断肺癌的准确率为97.5%，灵敏度为98.4%，特异度为94.1%，高于其他两组，而漏诊率为1.6%，误诊率为5.9%，均低于其他两组，差异有统计学意义(P < 0.05)，此方法优于传统方法，能够在早期发现更多的潜在病例。AI辅助 + 医师诊断组的平均阅片时间为(167 ± 45) s短于其他两组，差异有统计学意义(P < 0.05)。肺癌组的血清肿瘤标志物水平高于良性病变组，差异有统计学意义(P < 0.05)。结论：人工智能(AI)辅助低剂量CT联合肿瘤标志物在社区体检人群肺癌机会性筛查中检出率、灵敏度、特异度及准确率均较高，且诊断准确率与病理结果一致性较高，可作为临床筛查肺结节的常规手段。

Abstract: Objective: To investigate the application value of artificial intelligence (AI) assisted low-dose CT combined with multiple tumor markers in screening for lung cancer in community physical examination population. Method: A retrospective analysis was conducted on 80 patients who presented with pulmonary nodules on LDCT examination in our hospital from January 2023 to December 2024. All patients underwent testing for four lung cancer tumor markers (SCC, CEA, Cyfra21-1, and NSE), with pathological results as the gold standard. The accuracy, sensitivity, and specificity of each group in the diagnosis of benign and malignant pulmonary nodules were compared and analyzed, and the value of different diagnostic methods was analyzed. Result: Pathological diagnosis showed 63 malignant nodules and 17 benign nodules in 80 cases. The accuracy, sensitivity, and specificity of the artificial intelligence system (AI) assisted physician diagnosis group combined with tumor marker diagnosis group for diagnosing lung cancer were 97.5%, 98.4%, and 94.1%, higher than the other two groups. However, the missed diagnosis rate was 1.6% and the misdiagnosis rate was 5.9%, both lower than the other two groups, with statistical significance (P < 0.05). This method is superior to traditional methods and can detect more potential diseases in the early stage. The average reading time of the AI assisted + physician diagnosis group was (167 ± 45) seconds, which was shorter than the other two groups, and the difference was statistically significant (P < 0.05). The serum tumor marker levels in the lung cancer group were higher than those in the benign lesion group, and the difference was statistically significant (P < 0.05). Conclusion: Artificial intelligence (AI) assisted low-dose CT combined with tumor markers has a high detection rate, sensitivity, specificity, and accuracy in the screening of lung cancer opportunities in community physical examination populations. The diagnostic accuracy is consistent with pathological results, and can be used as a routine method for clinical screening of lung nodules.