临床、病理、MRI特征及IVIM参数联合模型预测宫颈癌程序性细胞死亡蛋白1及其配体(PD-1/PD-L1)表达

Combined model of clinical,pathological,MRI features and IVIM parameters for predicting programmed cell death protein 1(PD-1)/PD-1 ligand(PD-L1)expression in cervical cancers

目的基于SHAP法观察以临床、病理、MRI特征及体素内不相干运动(IVIM)成像定量参数联合模型预测宫颈癌细胞程序性死亡蛋白1(PD-1)及其配体(PD-L1)表达的价值。方法采集63例治疗前初诊宫颈癌盆腔MRI,并对病理标本行PD-1/PD-L1免疫组织化学染色;比较PD-1表达阳性与阴性组、PD-L1表达阳性与阴性组临床、病理、MRI表现及IVIM参数(真实弥散系数D、灌注相关弥散系数D*及灌注分数f)的差异,并以logistic回归分析筛选宫颈癌PD-1及PD-L1表达阳性的独立影响因素,建立预测宫颈癌PD-1及PD-L1表达阳性联合模型;以受试者工作特征(ROC)曲线评估模型诊断效能,以SHAP法解释其中各变量的贡献价值。结果PD-1阳性组与阴性组、PD-L1阳性组与阴性组之间肿瘤病理分级、宫旁浸润、淋巴结转移及D值差异均有统计学意义(P均<0.05)。FIGO分期、肿瘤病理分级、宫旁浸润、淋巴结转移和D值均为宫颈癌PD-1/PD-L1表达阳性的独立影响因素(P均<0.05);以之建立的联合模型的曲线下面积分别为0.85及0.89。根据SHAP值,联合模型中FIGO分期和肿瘤病理分级的贡献最大。结论以宫颈癌临床、病理、MRI特征及IVIM参数D值构建的联合模型可有效预测其PD-1/PD-L1表达。

Objective To explore the value of combined model of clinical,pathological,MRI features and intravoxel incoherent motion(IVIM)quantitative parameters for predicting expressions of programmed cell death protein 1(PD-1)and its ligand(PD-L1)in cervical cancer based on SHAP analysis.Methods Totally 63 initially diagnosed cervical cancer patients who underwent pelvic MRI before treatment were enrolled.Immunohistochemical staining of PD-1 and PD-L1 were performed to evaluate the expression of PD-1/PD-L1.Clinical,pathological,MRI manifestations and IVIM parameters,including true diffusion coefficient D,perfusion related diffusion coefficient D*and perfusion fraction f were compared between positive and negative groups of PD-1 and PD-L1,and the independent factors of PD-1/PD-L1 positive expression in cervical cancer were screened with logistic regression analysis.Then the combined models were established,and the diagnostic efficacy of the models were evaluated using receiver operating characteristic(ROC)curves,while SHAP method was used to explain the contribution of each parameter to the combined model.Results Significant differences of pathological grade,parametrial invasion,lymph node metastases and D value were found between positive and negative groups of PD-1/PD-L1 cervical cancer(all P<0.05).FIGO staging,pathological grade,parametrial invasion,lymph node metastasis and D value were independent factors of PD-1/PD-L1-positive expression in cervical cancer(all P<0.05).The area under the curve of the combined model in predicting PD-1 and PD-L1-positive expression of cervical cancer was 0.85 and 0.89,respectively.FIGO staging and pathological grade of cervical cancer had the greatest contribution to the combined model.Conclusion Combined model of clinical,pathological,MRI features and D value could effectively predict PD-1/PD-L1 expression in cervical cancer.