基于CE-T1WI图像影像组学和病理参数模型预测胶质瘤术后复发的研究  

Prediction based on CE-T1WI omics and pathological parameter models research on postoperative recurrence of glioma

作　　者：金一萱 吕瑞瑞 杨治花 孙萌 牛芳 吕鸿洁 马蓉 王晓东 JIN Yixuan;LÜRuirui;YANG Zhihua;SUN Meng;NIU Fang;LÜHongjie;MA Rong;WANG Xiaodong(Ningxia Medical University School of Clinical Medicine,Yinchuan 750004,China;Department of Radiology,General Hospital of Ningxia Medical University,Yinchuan 750004,China;Department of Radiotherapy,Cancer Hospital of General Hospital of Ningxia Medical University,Yinchuan 750004,China)

机构地区：[1]宁夏医科大学临床医学院,银川750004 [2]宁夏医科大学总医院放射科,银川750004 [3]宁夏医科大学总医院肿瘤医院放疗科,银川750004

出　　处：《磁共振成像》2024年第10期103-108,共6页Chinese Journal of Magnetic Resonance Imaging

基　　金：宁夏回族自治区自然科学基金项目(编号:2022AAC03487);宁夏回族自治区科技重点研发计划项目(编号:2019BEG03037)。

摘　　要：目的初步探讨基于术前对比增强T1WI(contrast enhancement T1WI,CE-T1WI)影像组学联合病理参数的列线图预测脑胶质瘤患者术后复发的应用价值。材料与方法回顾性分析2020年4月至2023年4月于宁夏医科大学总医院经术后病理确诊为脑胶质瘤患者115例,按照7∶3随机分为训练集(n=81)和验证集(n=34)。于术前CE-T1WI图像上进行容积感兴趣区(volume of interest,VOI)的勾画并提取影像组学特征。采用U检验及最小绝对收缩和选择算子(least absolute shrinkage and selection operator,LASSO)进行影像组学特征的筛选,将最终筛选出的特征纳入影像组学标签并建立影像组学模型。根据筛选所得组学特征的相应系数计算影像组学评分(radiomics score,Radscore)。通过logistic回归筛选与复发存在相关性的病理预测因子并建立病理参数模型。二者结合为联合模型,绘制列线图将联合模型可视化。采用受试者工作特征曲线下面积(area under the curve,AUC)评估各模型的预测性能,使用DeLong检验比较不同模型间AUC值差异,采用决策曲线(decision curve analysis,DCA)分析观察各模型的临床价值。结果基于术前CE-T1WI图像勾画的VOI中共提取出200个影像组学特征,筛选出与复发相关的组学特征6个。通过logistic回归分析纳入异柠檬酸脱氢酶-1(isocitric dehydrogenase-1,IDH-1)基因型(OR=2.070,P=0.041)、Ki-67表达水平(OR=1.065,P<0.001)为与胶质瘤复发相关的病理参数。相较于单独的病理参数模型和影像组学模型,联合模型在预测效能上表现最佳(训练组AUC:0.875 vs.0.835、0.769,Z=-1.585、-2.458,P=0.013、0.014)。DCA分析示风险阈值概率大于0.32时,应用联合模型的临床获益水平高于另外两种模型。结论基于术前CE-T1WI图像影像组学和病理参数构建的联合模型在预测脑胶质瘤复发中具有较好的临床应用价值,为脑胶质瘤患者治疗决策及预后提供重要预测信息。Objective:To explore the application value of using a column chart based on preoperative contrast enhancement T1WI(CE-T1WI)omics combined with pathological parameters to predict postoperative recurrence in patients with gliomas.Materials and Methods:A retrospective analysis was conducted on 115 patients diagnosed with glioma after surgery at the General Hospital of Ningxia Medical University from April 2020 to April 2023.They were randomly divided into a training set(n=81)and a validation set(n=34)at a ratio of 7∶3.Draw the volume of interest(VOI)on preoperative enhanced T1WI(CE-T1WI)and extract imaging omics features.U test and least absolute shrinkage and selection operator(LASSO)algorithm were used to screen imaging omics features.The final selected features were included in imaging omics labels and an imaging omics model was established.Calculate the radiomics score(Radscore)based on the corresponding coefficients of the selected omics features.Screening pathological predictive factors that are correlated with recurrence through logistic regression and establishing a pathological parameter model.The combination of the two forms a joint model,and a column chart is drawn to visualize the joint model.Evaluate the predictive performance of each model using the area under curve(AUC)of the subject's working characteristic curve.Using the DeLong test to compare the differences in AUC values between different models,and observe the clinical value of each model using decision curve analysis(DCA).Results:Based on preoperative CE-T1WI image delineation,a total of 200 imaging omics features were extracted from VOI,and 6 omics features related to recurrence were selected.Logistic regression analysis was used to include isocitrate dehydrogenase 1(IDH-1)genotype(OR=2.070,P=0.041)and Ki-67 expression level(OR=1.065,P<0.001)as pathological parameters associated with glioma recurrence.Compared to individual pathological parameter models and radiomics models,the combined model showed the best predictive performance(AUC:0.875

关 键 词：胶质瘤 复发 影像组学 磁共振成像 列线图表 预测模型 

分 类 号：R445.2[医药卫生—影像医学与核医学] R730.264[医药卫生—诊断学]