非小细胞肺癌胸膜侵犯的CT特征评估

Evaluation of CT Features of Pleural Invasion in Non-small Cell Lung Cancer

目的:通过探讨手术切除的非小细胞肺癌胸膜侵犯与其CT特征性表现之间的相关性,为该病预后的判断及临床治疗方案的调整提供参考和依据。方法:回顾性分析南京大学医学院附属鼓楼医院143例非小细胞肺癌患者术前胸部平扫CT图像,共评估17个分类变量及7个连续变量。采用ROC曲线分析CT参数对胸膜侵犯的诊断效能。通过拟合建立多参数模型提高诊断效能。结果:分类变量中病灶边界、分叶征、邻近胸膜增厚、胸膜凹陷征及病灶与胸膜关系在有无胸膜侵犯两组中的差异有统计学意义(P<0.05);连续变量中DLP、病灶长径(最大径)、短径在有无胸膜侵犯两组中的差异有统计学意义(P<0.05);胸膜侵犯多见于病灶与胸膜关系分型Ⅳ型与Ⅴ型中且伴有胸膜凹陷征的肿瘤中。DLP、病灶长径及短径对胸膜侵犯的诊断效能较高。多参数模型预测胸膜侵犯的AUC值(曲线下面积)为0.772,准确度62.25%,较单因素分析诊断效能更高。结论:病灶边界、分叶征、邻近胸膜增厚、胸膜凹陷征及病灶与胸膜关系可术前预测非小细胞肺癌胸膜侵犯,多参数模型预测胸膜侵犯的诊断效能较单因素分析更高,可为临床治疗提供参考价值。

Objective:To investigatetherelationship between thepleural invasion and CT featuresof surgically resected non-small cell lung cancer(NSCLC)and to provide guidance for predicting prognosis and adjusting clinical treatment.Methods:Preoperative non-contrast enhance CT images of 143 patients with NSCLC in Nanjing Drum Tower Hospital were retrospectively analyzed.A total of 17 CT categorical variables and 7 continuous variables were evaluated.ROC curve was used to analyze the diagnostic efficacy of CT parameters for pleural invasion.The multi-parameter model was established to improve the diagnostic efficiency.Results:There were statistically significant differences between the two groups in lesion boundary,lobulation sign,adjacent pleural thickening,pleural pitting,and the relationship between lesion and pleural invasion(all P<0.05).Among the continuous variables,DLP,lesion length diameter(maximum diameter),and short diameter had statistically significant differences between the two groups with or without pleural invasion.Pleural invasion more common in IV and V types(lesions and pleural relationship type)of tumor with pleural sag.DLP,long diameter and short diameter were more effective in the diagnosis of pleural invasion.The area under the curve of multi-parameter model for predicting pleural invasion was 0.772 with an accuracy of 62.25%,which was more effective than that of single-factor analysis.Conclusion:Lesion boundaries,lobulation sign,adjacent pleural thickening,pleural pitting,and the relationship between lesions and pleura can predict pleural invasion of NSCLC before surgery,and the multi-parameter model was more effective than univariate analysis in predicting pleural invasion.It may provide relevant reference value for clinical treatment.