基于r-FOV DWI临床-影像组学模型预测宫颈癌临床分期的研究  

Predicting the Clinical Stage of Cervical Cancer Based on the Clinical-Imaging Model of Reduced Field-of-View Diffusion-Weighted Imaging

作　　者：王艳春 赵建立 黄秋晗 胡道予[1] 邓保娣 WANG Yanchun;ZHAO Jianli;HUANG Qiuhan(Department of Radiology,Tongji Hospital,Tongji Medical College,Huazhong University of Science and Technology,Wuhan,Hubei Province 430030,China)

机构地区：[1]华中科技大学同济医学院附属同济医院放射科,武汉430030 [2]华中科技大学同济医学院附属武汉中心医院急诊科,武汉430014 [3]浙江大学医学院附属邵逸夫医院放射科,杭州310016

出　　处：《临床放射学杂志》2024年第7期1147-1153,共7页Journal of Clinical Radiology

基　　金：国家自然科学基金项目(编号:82102025)。

摘　　要：目的探讨基于小视野(r-FOV)扩散加权成像(DWI)表观扩散系数(ADC)图的临床-影像组学模型预测宫颈癌临床分期的价值。方法回顾性分析89例经手术或活检病理证实为宫颈癌,并于治疗前行MRI检查的患者,进行常规T1WI、T2WI、DWI以及r-FOV DWI扫描。采用3D Slicer软件,在矢状位r-FOV ADC图上手动逐层勾画肿瘤感兴趣区(ROI),并提取病灶的影像组学特征。采用独立样本t检验或Mann-Whitney U检验初步筛选出鉴别早、晚期宫颈癌组间差异有统计学意义的影像组学特征。应用LASSO回归模型及10折交叉验证法筛选出最优影像组学特征,然后采用多元Logistic回归分析,构建包含临床因素和影像组学特征的预测模型。采用受试者工作特征(ROC)曲线评估模型的预测效能,并用决策曲线分析(DCA)来评估该模型的临床应用价值。结果共提取851个影像组学特征,最终筛选出16个最优影像组学特征,在多元Logistic回归分析中,影像组学特征、患者年龄以及患者是否绝经被选择入列线图,并以之构建预测早、晚期宫颈癌的临床-影像组学模型。该预测模型在训练集及测试集中的曲线下面积(AUC)分别为0.998、0.777。DCA显示模型具有鉴别能力的阈值为0.18。结论基于r-FOV DWI ADC图的临床-影像组学模型对预测宫颈癌临床分期具有较高的临床价值。Objective To evaluate the value of clinical-radiomics model based on reduced field-of-view(r-FOV)ADC map in predicting clinical stage of cervical cancer.Methods Eighty-nine patients with pathologically proven cervical cancer who underwent pre-treatment conventional MR sequences,r-FOV DWI and full field-of-view(f-FOV)DWI were retrospectively reviewed.3D Slicer software was used to manually delineate the region of interest of tumor layer by layer on the sagittal r-FOV ADC map,and then radiomics features of the lesions were extracted.Independent sample t test or Mann-Whitney U test was used to preliminarily screen out the radiomics features with statistically significant differences between early and advanced cervical cancer groups.Least absolute shrinkage and selection operator(LASSO)regression analysis and a 10-fold cross-validation method were used to screen the optimal radiomics features.Multivariate Logistic regression analysis was used to construct the prediction model including clinical factors and radiomics features.Receiver operating characteristic(ROC)curve was used to evaluate the predictive efficacy of the model.Decision curve analysis(DCA)was used to evaluate the clinical application value of the model.Results A total of 851 radiomics features were extracted,and 16 optimal radiomics features were finally screened.On multivariate Logistic regression analysis,radiomics features,patient age and menopause were selected into the nomogram,and then they were used to construct clinical-radiomics model for predicting early and advanced cervical cancer.The area under the curve(AUC)of clinical-radiomics model was 0.998 and 0.777 in the training set and test set,respectively.The decision curve analysis(DCA)shows that the threshold value of the model's discrimination ability is 0.18.Conclusion The clinical-radiomics model based on reduced field-of-view(r-FOV)ADC map has high value in predicting clinical stage of cervical cancer.

关 键 词：宫颈癌 磁共振成像 肿瘤分期 影像组学 

分 类 号：R737.33[医药卫生—肿瘤] R445.2[医药卫生—临床医学]