Nomogram在脑胶质瘤MR术前病理分级中的应用  

Application of Nomogram in Preoperative Pathological Grading of Brain Gliomas via MR

作　　者：林坤[1] 王晓明[1] LIN Kun;WANG Xiaoming(Department of Radiology,Shengjing Hospital,China Medical University,Shenyang 110004,China)

机构地区：[1]中国医科大学附属盛京医院放射科,辽宁沈阳110004

出　　处：《中国医学影像学杂志》2023年第5期442-447,共6页Chinese Journal of Medical Imaging

基　　金：国家自然科学基金项目(81871408);辽宁省临床能力建设项目(LNCCC-B06-2014);盛京自由研究者基金项目(2014-02)。

摘　　要：目的 探讨基于传统MR影像特征的定性分析、半定量分析以及可视化Nomogram在脑胶质瘤术前病理分级诊断中的应用价值。资料与方法 回顾性分析中国医科大学附属盛京医院2015年7月—2019年10月经手术病理证实为脑胶质瘤68例患者的增强MR图像(Ⅱ级39例、Ⅲ级13例、Ⅳ级16例),对比不同级别胶质瘤的临床-影像特征,筛选差异有统计学意义的影像特征建立分级诊断的Logistic回归模型,并勾画诺曼图。结果 囊变、出血、瘤周水肿和强化方式在Ⅱ、Ⅲ、Ⅳ级3组间以及高、低级别组间比较,差异均有统计学意义(χ~2=9.770~38.510,P<0.05;χ~2=8.043~37.704,P<0.05)。定性分析对高、低级别胶质瘤分级诊断的曲线下面积(AUC)为0.803。Logistic回归分析对以上影像特征行逐步回归分析,回归模型对高、低级别胶质瘤分级诊断试验的AUC为0.918。回归模型分级诊断Ⅱ、Ⅲ级胶质瘤的AUC为0.870;分级诊断Ⅱ、Ⅳ级胶质瘤的AUC为0.956。勾画的可视化诺曼图的校准曲线显示模型的诊断效能较高。结论 传统MR检查能够提供对胶质瘤分级诊断有价值的影像特征。Logistic回归模型分级诊断高、低级别胶质瘤的诊断效能高于传统定性分析。基于模型勾画的诺曼图有望为临床提供直观的评估工具。Purpose To investigate the value of qualitative analysis,semi-quantitative analysis,and nomogram of conventional MR sequence in preoperative pathological grading of brain glioma.Materials and Methods Conventional enhanced MR images of 68 patients(grade 39,ⅡgradeⅢ13,gradeⅣ16)with pathologicallyconfirmed glioma,from July2015 toOctober2019 in Shengjing Hospital of ChinaMedical University,were retrospectively studied.The clinical-imaging features of different grades of glioma were compared.The features with significant differences in different grades of gliomas were used for Logistic stepwise regression analysis to fit the grading model.To visualize the model,the nomogram was developed.Results Cyst change,hemorrhage,peri-tumoral edema and enhancement pattern were significantly different among grade,andⅡⅢⅣgliomas,and between high and low grade gliomas(χ2=9.770-38.510,P<0.05;χ2=8.043-37.704,P<0.05).The area under the curve(AUC)for grading of glioma based on qualitative analysis was 0.803.Stepwise regression analysis was performed for the above features,and the AUC of the regression model was 0.918.The AUC of the model classifying grade and gliomas was 0.870.The AUC of tⅡⅢhe model classifying gradeⅡandⅣgliomas was 0.956.The visualized nomogram showed high diagnostic efficacy.Conclusion Conventional MR sequences can provide valuable imaging features for glioma grading.The Logistic stepwise regression model can effectively differentiate high from low grade gliomas,with high accuracy and superiority to qualitative analysis.The nomogram is one of the potential tool for glioma clinical assessment.

关 键 词：胶质瘤 诺曼图 回归分析 磁共振成像 

分 类 号：R739.41[医药卫生—肿瘤] R445.2[医药卫生—临床医学]