基于影像特征构建列线图模型预测浸润性肺腺癌发生气腔播散概率的研究

Study on constructing a nomogram model based on radiomic features to predict the probability of spread through air spaces in invasive lung adenocarcinoma

    摘要
  • 摘要:
    目的  探讨基于影像特征构建列线图模型预测浸润性肺腺癌发生气腔播散（STAS）概率的价值。
    方法  回顾性分析2020年5月至2023年4月于南方医科大学顺德医院经手术组织病理学检查结果证实的391例浸润性肺腺癌患者的临床资料，其中男性168例、女性223例，年龄（61.0±11.8）岁。将患者分为STAS阳性组（166例）和STAS阴性组（225例），比较2组患者性别、年龄、肿瘤最大径、CT值、肿瘤位置、分叶征、毛刺征、胸膜牵拉征、肿瘤密度、空气支气管征、微血管穿行征、月牙征、空泡征、肿瘤-肺界面的差异。计量资料的组间比较采用Mann-Whitney U检验；无序计数资料的组间比较采用卡方检验；有序计数资料的组间比较采用Mann-Whitney U检验。通过单因素分析和多因素Logistic回归分析筛选出STAS的独立预测因素，构建临床预测模型的风险列线图。采用受试者工作特征曲线评估预测模型的预测性能，采用校准曲线和决策曲线评估预测模型的综合临床价值。
    结果 单因素分析结果表明，STAS阳性组与STAS阴性组患者性别、年龄、肿瘤最大径、CT值、分叶征、毛刺征、胸膜牵拉征、肿瘤密度、空气支气管征、微血管穿行征、月牙征、空泡征的差异均有统计学意义（χ2=3.933~161.518、t=−14.508~−2.710，均P<0.05）。多因素Logistic回归分析结果表明，肿瘤的CT值、分叶征、胸膜牵拉征、空泡征、月牙征是浸润性肺腺癌STAS的独立预测因素（Z=−3.490~5.447，均P<0.05），预测模型的曲线下面积为0.892（95% CI：0.862~0.923），灵敏度为81.9%，特异度为78.2%。校准曲线和决策曲线表明预测模型具有较高的可靠性。
    结论  基于影像特征构建到线图模型对浸润性肺腺癌STAS发生概率的预测具有较高的临床价值。

     

    Abstract:
    Objective To discuss the value of constructing a nomogram model based on imaging features in predicting the probability of spread through air spaces (STAS) in invasive lung adenocarcinoma.
    Methods The clinical data of 391 patients (including 168 males and 223 females, with an average age of (61.0±11.8) years) with invasive lung adenocarcinoma confirmed by surgery and histopathology in Shunde Hospital of Southern Medical University from May 2020 to April 2023 were retrospectively analyzed. The patients were divided into STAS-positive group (166 cases) and STAS-negative group (225 cases). The differences in gender, age, tumor maximum diameter, CT value, tumor location, lobulation sign, spiculation sign, pleural traction sign, tumor density, air bronchogram sign, microvascular crossing sign, crescent sign, vacuole sign, and tumor-lung interface were compared between the two groups. Chi square test was used for comparison of unordered count data between groups, while Mann-Whitney U test was employed for comparison of measurement data and ordered count data between groups. The independent predictors of STAS were screened by univariate analysis and multivariate Logistic regression analysis, and the risk nomogram of the clinical prediction model was constructed. A receiver operating characteristic curve was used to evaluate the predictive performance of the prediction model, and calibration and decision curves were adopted to assess its comprehensive clinical value.
    Results The univariate analysis showed significant differences between STAS-positive and STAS-negative groups in gender, age, tumor maximum diameter, CT value, lobulation sign, spiculation sign, pleural traction sign, tumor density, air bronchogram sign, microvascular crossing sign, crescent sign, and vacuole sign (χ2=3.933–161.518, t=−14.508–−2.710; all P<0.05). The multivariate Logistic regression analysis indicated that the CT value of tumor, lobulation sign, pleural traction sign, vacuole sign, and crescent sign were independent predictors of STAS in invasive lung adenocarcinoma (Z=−3.490–5.447; all P<0.05). The area under the curve of the prediction model was 0.892 (95%CI: 0.862–0.923), the sensitivity was 81.9%, and the specificity was 78.2%. The calibration and decision curves showed that the prediction model had high reliability.
    Conclusion Constructing a nomogram model based on image features has high clinical value in predicting the probability of STAS in invasive lung adenocarcinoma.

     

    • HTML全文
    • 参考文献(23)
    • 相关文章
    • 施引文献
  • 资源附件(0)

/

返回文章
返回