MRI多指标空间相关性在弥漫性胶质瘤分级及预后中的效能  

MRI multi-index spatial correlation in the grading and prognosis of diffuse glioma

作　　者：周娴 许强 李建瑞 詹天亮 张志强 ZHOU Xian;XU Qiang;LI Jianrui;ZHAN Tianliang;ZHANG Zhiqiang(Department of Radiology,Jinling Hospital,Nanjing Medical University/General Hospital of Eastern Theater Com-mand,PLA,Nanjing 210002,Jiangsu,China;Department of Radiology,Jinling Clinical Medical College,Nanjing University of Traditional Chinese Medicine/General Hospital of Eastern Theater Command,PLA,Nanjing 210002,Jiangsu,China)

机构地区：[1]南京医科大学金陵临床医学院(东部战区总医院)放射诊断科,南京210002 [2]南京中医药大学金陵临床医学院(东部战区总医院)放射诊断科,南京210002

出　　处：《医学研究与战创伤救治》2025年第1期28-34,共7页Journal of Medical Research & Combat Trauma Care

基　　金：国家重点研发计划(2018YFA0701703);国家自然科学基金(81530054)。

摘　　要：目的探讨弥漫性脑胶质瘤磁共振成像(MRI)多指标空间相关性特征,并观察其在肿瘤分级及生存期预测中的效能。方法回顾性分析2017年1月至2020年1月东部战区总医院经病理证实的弥漫性胶质瘤患者198例(Ⅱ级71例、Ⅲ级42例、Ⅳ级85例)的多序列[T_(1)加权成像(T_(1)WI)、T_(2)加权成像(T_(2)WI)、T_(1)WI增强扫描(T_(1)CE)、扩散加权成像(DWI)及动脉自旋标记(ASL)]MRI数据。定量分析肿瘤实性部分多指标:T_(1)WI及T_(2)WI相对信号强度、T_(1)CE相对强化值、表观扩散系数(ADC)及脑血流量(CBF)间的空间相关性特征。观察多指标间空间相关性特征与病理分级及生存预后的关系。并利用支持向量机(SVM)分类器建立MRI多指标空间相关性特征模型,用于胶质瘤的病理分级及生存期的预测。结果弥漫性胶质瘤高、低级别间MRI多指标的空间相关性有显著差异,且T_(1)CE_T_(2)、T_(1)CE_T_(1)、T_(1)CE_CBF、T_(1)CE_ADC Z值ROC曲线下面积(AUC)>0.7,基于指标间空间相关性构建的SVM预测模型显示有较高的病理分级效能(AUC=0.9222),模型中T_(1)CE_T_(2)Z值特征权重最高;长、短生存期中T_(1)CE_ADC、T_(1)CE_T_(2)、T_(1)CE_T_(1)指标间的空间相关性有显著差异(P<0.05),生存期预测模型中AUC为0.6318且T_(1)CE_T_(1)Z值特征权重最高。结论MRI多指标间空间相关特征分析为胶质瘤异质性病理生理机制提供证据,并在肿瘤病理分级和生存期预测方面具有应用价值。Objective This study aims at investigating the spatial correlation characteristics of MRI multi-index in diffuse glioma and evaluating its efficacy in predicting tumor grade and survival time.Methods A retrospective analysis was conducted on the multi-sequence MRI data of 198 patients with pathologically confirmed diffuse glioma(71 cases of gradeⅡ,42 cases of gradeⅢ,85 cases of grade IV)from Jinling Hospital between January 2017 and January 2020.T_(1)weighted imaging(T_(1)WI),T_(2)weighted imaging(T_(2)WI),contrast-enhanced MRI data(T_(1)CE),diffusion-weighted imaging(DWI),and arterial spin labeling(ASL)were obtained.Quantitative analysis was performed on the spatial correlation between multiple indicators of the solid part of the tumor,including relative signal intensity on T_(1)WI and T_(2)WI,relative enhancement value on T_(1)WI,apparent diffusion coefficient(ADC),and cerebral blood flow(CBF).The relationship between the spatial correlation of multiple indicators and pathological grade as well as survival time was observed.A support vector machine(SVM)classifier was used to establish an MRI multi-index spatial correlation feature model for pathological grading and survival prediction of glioma.Results There was significant difference in the spatial correlation of MRI multiple indicators between high and low grade diffuse gliomas,and the area under the ROC curve(AUC)of T_(1)CE_T_(2),T_(1)CE_T_(1),T_(1)CE_CBF,T_(1)CE_ADC Z values was>0.7.The SVM prediction model based on the spatial correlation between indicators showed a high pathological grading efficiency(AUC=0.9222),and the T_(1)CE_T_(2)Z score feature weight in the model was the highest.The spatial correlations among T_(1)CE_ADC,T_(1)CE_T_(2)and T_(1)CE_T_(1)indicators in the long and short survival periods were significantly different(P<0.05).The AUC ofthe survivalprediction modelwas 0.6318 and T_(1)CE_T_(1)Z score feature weightwasthe highest.Conclusion The spatial correlation feature analysis of multiple MRI indicators provides evidence for the heter

关 键 词：弥漫性脑胶质瘤 多指标MR 空间相关性 组织学分级 生存期 支持向量机 

分 类 号：R445.2[医药卫生—影像医学与核医学] R739.41[医药卫生—诊断学]