CT影像学特征、血清肺癌自身抗体的肺炎型肺癌诊断模型构建

Construction of a diagnostic model for pneumonic-type lung cancer based on CT imaging features and seven serum tumor-associated autoantibodies

摘要:

目的 探讨肺炎型肺癌患者CT影像学特征、血清7项肺癌自身抗体（TAAbs）表达水平，据此构建肺炎型肺癌诊断模型。方法 回顾性纳入2021年2月—2024年8月安徽医科大学附属亳州医院收治的224例疑似肺癌患者为研究对象，分为建模组184例和内部验证组40例。另取同期该院40例疑似肺癌患者的临床资料作为外部验证组。根据组织病理诊断结果将建模组120例肺炎型肺癌患者纳入肺癌组，64例大叶性肺炎患者纳入肺炎组，比较两组一般临床资料、CT影像特征，酶联免疫吸附试验检测血清7项TAAbs［肿瘤-睾丸抗原（CAGE）、G抗原7（GAGE7）等］水平；采用多因素一般Logistic回归分析筛选肺炎型肺癌的危险因素并建立诊断模型，Homser-Lemeshow检验拟合优度，受试者工作特征（ROC）曲线对模型的有效性进行评价，并对模型进行内部和外部验证。结果 多因素一般Logistic回归分析结果显示，纵隔淋巴结肿大［O^R =1.550（95% CI：1.298，1.851）］、支气管充气征［O^R =7.147（95% CI：1.641，31.129）］、毛刺征［O^R =7.486（95% CI：2.182，25.681）］、空泡征［O^R =13.077（95% CI：2.912，58.715）］、7项TAAbs阳性［O^R =6.746（95% CI：1.933，23.549）］是诊断肺炎型肺癌的危险因素（P <0.05）。诊断模型为Logit（P） =-7.691+0.438×纵隔淋巴结肿大+1.967×支气管充气征+2.013×毛刺征+2.571×空泡征+1.909×7项TAAbs阳性；ROC曲线分析结果显示，模型在建模组、内部验证组、外部验证组的曲线下面积分别为0.943（95% CI：0.899，0.972）、0.870（95% CI：0.752，0.989）、0.842（95% CI：0.710，0.974），敏感性分别为94.17%（95% CI：0.884，0.976）、92.31%（95% CI：0.640，0.998）、84.62%（95% CI：0.546，0.981），特异性分别为84.37%（95% CI：0.731，0.922）、81.48%（95% CI：0.619，0.937）、81.48%（95% CI：0.619，0.937），Hosmer-Lemeshow拟合优度检验显示拟合良好（P >0.05）。结论 肺癌患者CT图像具有典型特征，7项TAAbs阳性率较高；基于CT影像特征和血清7项TAAbs构建的诊断模型或可用于肺炎型肺癌的临床辅助诊断。

Abstract:

Objective To construct a diagnostic model for pneumonic-type lung cancer (PTLC) based on CT imaging features and seven serum tumor-associated autoantibodies (TAAbs).Methods This retrospective study included 224 suspected lung cancer patients (February 2021 to August 2024), divided into modeling (n = 184) and internal validation (n = 40) cohorts. An external validation cohort (n = 40) was independently enrolled. PTLC (n = 120) and lobar pneumonia (n = 64) groups were compared for CT features (mediastinal lymphadenopathy, air bronchogram, spiculation, vacuolar sign) and TAAbs. Multivariate logistic regression and ROC analysis were performed.Results Multivariate general logistic regression analysis showed that enlarged mediastinal lymph node[O^R = 1.550 (95% CI: 1.298, 1.851) ], air bronchogram sign [O^R = 7.147 (95% CI: 1.641, 31.129) ], spicule sign [O^R = 7.486 (95% CI: 2.182, 25.681) ], vacuolar sign [O^R = 13.077 (95% CI: 2.912, 58.715) ] and positivity of 7 TAAbs [O^R = 6.746 (95% CI: 1.933, 23.549) ] were risk factors for diagnosing pneumonic-type lung cancer (P < 0.05). The diagnostic model was Logit (P) = -7.691 + 0.438 × enlarged mediastinal lymph node + 1.967 × air bronchogram sign + 2.013 × spicule sign + 2.571 × vacuolar sign + 1.909 × positivity of 7 TAAbs. ROC curves indicated that the areas under the curve (AUCs) of the model in the modeling group, the internal validation group and the external validation group were 0.943 (95% CI: 0.899, 0.972), 0.870(95% CI: 0.752, 0.989), and 0.842 (95% CI: 0.710, 0.974), respectively. The sensitivities were 94.2% (95% CI: 0.884, 0.976), 92.3% (95% CI: 0.640, 0. 998), and 84.62% (95% CI: 0.546, 0.981). The specificities were 84.4% (95% CI: 0.731, 0.922), 81.5% (95% CI: 0.619, 0.937), and 81.5% (95% CI: 0.619, 0.937). Homser-Lemeshow goodness of fit test test showed good fit (P > 0.05).Conclusion CT images of lung cancer patients have typical features, and the positive rates of 7 TAAbs are relatively high. The diagnostic model constructed based on CT image features and 7 TAAbs in serum may be used in clinical auxiliary diagnosis of pneumonic-type lung cancer.