聚类和判别分析法在肺癌六种肿瘤标志物诊断中的应用

目的 探讨聚类和判别分析法在肺癌6种肿瘤标志物诊断中的应用。方法 收集2012年5月至2013年5月新疆医科大学附属肿瘤医院初次入院就诊且最终确诊的肺癌患者342例，在未进行任何治疗前检测其血清中Pro-GRP、CEA、CA125、SCC、CYFRA21-1、NSE的含量。比较六种肿瘤标志物在小细胞肺癌、肺腺癌、肺鳞癌中的差别，应用聚类分析的方法对六个指标进行指标聚类，并对342例样本进行判别分析。结果 NSE和Pro-GRP在小细胞肺癌中的含量明显高于肺鳞癌和肺腺癌；SCC和CYFRA21-1在肺鳞癌中的含量明显高于小细胞肺癌和肺腺癌；CEA和CA125在肺腺癌中的含量明显高于小细胞肺癌和肺鳞癌。聚类分析表明NSE和Pro-GRP是诊断小细胞肺癌的良好指标，而CEA、CA125、SCC、CYFRA21-1却是对诊断非小细胞肺癌具有帮助。利用6种指标建立的判别函数对于小细胞肺癌的诊断符合率为93.3%；对于非小细胞肺癌的诊断符合率为83.0%。结论 利用聚类分析和判别分析的统计方法可以证明NSE、Pro-GRP、CEA、CA125、SCC、CYFRA21-1对肺癌不同病理分型诊断具有不同的应用价值。

Application of Cluster Analysis and Discriminant Analysis in Six Kinds of Lung Cancer Tumor Markers

Objective To investigate the application of cluster analysis and discriminant analysis in diagnosing different pathological types of lung cancer by six kinds of tumor markers. Methods We collected 342 patients who received the first hospitalization and were finally diagnosed as lung cancer in Tumor Hospital Affiliated to Xinjiang Medical University from May 2012 to May 2013. Serum concentrations of SCC, CYFRA 21-1, CEA, CA125, Pro-GRP and NSE were assayed for every patient before any treatment. We compared the differences of the six tumor markers among SCLC, lung adenocarcinoma and lung squamous carcinoma. We clustered the six tumor marker indexes by cluster analysis, and samples of 342 cases of samples were analyzed by discriminant analysis. Results NSE and Pro-GRP levels were significantly higher in SCLC tissues than those in squamous carcinoma and lung adenocarcinoma tissues. SCC and CYFRA21-1 were obviously higher in lung squamous carcinoma tissues than those in SCLC and lung adenocarcinoma tissues; CEA and CA125 levels were significantly higher in lung adenocarcinoma tissues than those in SCLC and lung squamous carcinoma tissues. Cluster analysis showed that NSE and Pro-GRP was helpful for diagnosing SCLC, and CEA, CA125, SCC, CYFRA21-1 were beneficial in the diagnosis of NSCLC. The diagnosis coincidence rate for SCLC was 93.3% and for NSCLC was 83.0% by the discrimination function established on six tumor markers. Conclusion Cluster analysis and discriminant analysis indicate that NSE, Pro-GRP, CEA, CA125, SCC and CYFRA21-1 have certain diagnostic value in diagnosing the different pathological types of lung cancer.