摘要
目的:探讨MRI影像组学在颅内胶质瘤术前分级评估中的应用价值,提高MRI对颅内胶质瘤的诊断效能。方法:回顾分析58例经病理证实的颅内胶质瘤患者的MRI资料,使用Matlab软件提取病灶特征并使用Lasso软件(R语言3.4.0版)进行降维处理,降维后的特征再进行Logistic多元回归相关性分析,从而筛选出具有诊断价值且与肿瘤级别显著相关的影像组学特征,运用受试者工作特征(ROC)曲线对特征参数的诊断效能进行分析。结果:DWI序列降维后经Logistic多元回归分析筛选出影像组学相关特征为:均值(P=0.04)、RMS(P=0.04)、Percentile-65(P=0.023)、Percentile-70(P=0.03)和Percentile-75(P=0.009)。DWI提取特征中Percentile-75对高、低级别胶质瘤的诊断效能最高,ROC曲线下面积(AUC)为0.793。T2WI序列及T1WI增强序列未筛选出与肿瘤分级相关的参数。在热力图上DWI序列Percentile-75特征与肿瘤分级的相关性较明显。结论:在T2WI、DWI和T1WI增强三个序列中,DWI序列对高低级别胶质瘤的鉴别效能较高,该序列上所筛选的特征参数中以Percentile-75的诊断价值最高。
Objective：Discussing the application of radiomics was used in MRI image by diagnosis cerebral glioma and improving the diagnosis efficacy. Methods：The research findings 58 cases of cerebra/glioma by pathology were analyzed ret- rospectively. Characteristics were picked up from tumor in two-dimensional planes by Matlab software,and the key factors were retrieved by Lasso package in R-software with 3.4. 0 generation. Stepwise multiple logistic regression analysis was used to deterrnin the independent factor that MRI image can diagnosis grading of cerebral glioma. Using box-type and scat- ter to describe the factor and at the same time using the ROC curve to analysis the key factor diagnosis efficacy. Results：The key factors of the lesions on MR-DWI were analyzed by stepwise multiple logistic regression as the best select examination. The correlative factor as following：mean value （P=0.04） ,RMS （P=0. 04） ,percentile-65 （P=0. 023） ,percentile-70 （P=0.03） and Percentile-75 （P=0. 009）. The factor of percentile-75 as the dependent correlative factor and it is the highest diagnosis efficacy with the AUC was 0. 793. Characteristics of Percentile-75 have high correlation with pathology grading where filling in more area red colore in hotmap of DWI. TI WI with contrast-enhanced and T2 WI had not correlation factors. Conclusions： DWI was the highest diagnosis efficacy sequence among T1 WI contrast-enhanced with T2 WI,and the factor that percentile-75 was the independent correlation factor with the best diagnosis efficacy.
出处
《放射学实践》
北大核心
2017年第12期1229-1233,共5页
Radiologic Practice
基金
国家自然科学基金(8126048
81760316
81703155)
关键词
胶质瘤
脑肿瘤
影像组学
磁共振成像
扩散加权成像
肿瘤分级
Glioma
Cerebral neoplasm
Radiomics
Magnetic resonance imaging
Diffusion weighted imaging
Tumor grading
