MRI瘤周影像组学术前预测ⅠB~ⅡB期宫颈癌宫旁浸润

Preoperative prediction of parametrial invasion of stage IB-IIB cervical cancer by intratumoural and peritumoural MRI radiomics

目的建立一个基于矢状位T2加权图像(T2WI)中瘤内结合不同瘤周组学特征和临床危险因素联合模型来预测ⅠB~ⅡB期宫颈癌宫旁浸润。方法回顾性分析2018年1月~2024年4月在陕西中医药大学第二附属医院西咸院区术前接受MRI检查和根治性子宫切除术加盆腔淋巴清扫并经术后病理证实的ⅠB~ⅡB期宫颈癌患者180例,对其矢状位T2WI图像瘤内感兴趣区(ROI)和1~6 mm的瘤周环(ROI-1、ROI-2、ROI-3、ROI-4、ROI-5和ROI-6)分别提取影像组学特征,采用Pearson分析和LASSO回归进行特征选择,构建不同瘤周组学模型并筛选出最佳组学模型,基于最佳组学特征与临床独立危险因素构建联合模型,利用ROC、校准曲线和决策曲线分析(DCA)评估模型的预测性能、校准度和应用价值。结果由瘤内结合瘤周3 mm区域得到的4个有效特征建立的组学模型预测效能最佳,在训练组和内部验证组,AUC分别为0.980和0.770;肿瘤最大径、血小板计数为临床独立危险因素,由肿瘤最大径、血小板计数建立的临床模型预测效能次之,AUC分别为0.860和0.673;基于瘤内结合瘤周3 mm的4个有效特征、肿瘤最大径和血小板计数构建的联合模型预测效能更稳定,AUC分别为0.952和0.939,经校准曲线和决策曲线分析,瘤内结合瘤周3 mm建立的组学模型校准度较高,临床净收益较大。结论基于MRI瘤内瘤周影像组学特征和临床参数的联合模型可以更好地术前预测ⅠB~ⅡB期宫颈癌宫旁浸润,对指导患者个体化治疗有着重要临床意义。

Objective To establish a comprehensive model based on sagittal T2-weighted imaging(T2WI)combined with different peritumoral characteristics and clinical risk factors for the prediction of parametrial invasion in stage IB-IIB cervical cancer.Methods Confirmed by postoperative pathology,a total of 180 patients with stage IB-IIB cervical cancer were enrolled from Xixian Campus of the Second Affiliated Hospital of Shaanxi University of Chinese Medicine from January 2018 to April 2024.All the patients they received preoperative MRI examination and radical hysterectomy with systematic pelvic lymph node dissection and retrospectively analyzed.The radiomics features were extracted from the volumetric region of interest of the tumor(ROI)and 1 mm-,2 mm-,3 mm-,4 mm-,5 mm-,6 mm-peritumoural rings(ROI-1,ROI-2,ROI-3,ROI-4,ROI-5,ROI-6)of the sagittal T2WI,respectively,and were selected by Pearson analysis and LASSO regression.Different feature-based radiomics models were independently built and their predictive performances were compared to select the optimal ones.Finally,the comprehensive model was developed based on optimal radiomics characteristics and clinical independent risk factors.And the predictive performance,calibration degree and application value of the models were evaluated by the ROC curve,calibration curves and the decision curve analysis(DCA).Results Four effective radiomics features,obtained from the peritumoral regions with 3 mm distances,had the best predictive performance,achieving an AUC of 0.980 and 0.770 in the training and internal validation cohorts,respectively.The maximum tumor diameter and platelet count were identified as independent clinical risk factors.The clinical model established by maximum tumor diameter and platelet count had the second predictive performance,with AUC of 0.860 and 0.673,respectively.The combined model constructed by integrating independent risk factors and four effective radiomics features from the peritumoral regions with 3 mm distances had more stable predictive performance,with an AUC of 0.952 and 0.939,respectively.After calibration curve and decision curve analysis,the intratumoral binding 3 mm around the tumor omics model had higher calibration degree and greater clinical net benefit.Conclusion The combined model based on intratumoral peritumoral radiomics and clinical parameters of MRI can better predict the preoperation of stage IB-IIB cervical cancer,which has important clinical significance for guiding the individualized treatment of patients.