基于多参数MRI影像组学特征融合模型鉴别高级别胶质瘤与单发性脑转移瘤  

Differentiation of high-grade glioma and solitary brain metastasis based on radiomics features fusion of multiparametric MRI

作　　者：徐向东[1] 梁芳蓉 韦瑞丽 吴嘉良 张婉丽 王琳婧 杨蕊梦[1,2] 甄鑫 赖胜圣[6] XU Xiangdong;LIANG Fangrong;WEI Ruili;WU Jialiang;ZHANG Wanli;WANG Linjing;YANG Ruimeng;ZHEN Xin;LAI Shengsheng(Department of Radiology,Guangzhou First People's Hospital,Guangzhou 510180,China;School of Medicine,South China University of Technology,Guangzhou 510006,China;Department of Radiology,the University of Hong Kong Shenzhen hospital,Shenzhen 518000,China;Radiotherapy Center,Affiliated Cancer Hospital&Institute of Guangzhou Medical University,Guangzhou 510095,China;School of Biomedical Engineering,Southern Medical University,Guangzhou 510515,China;School of Medical Equipment,Guangdong Food and Drug Vocational College,Guangzhou 510520,China)

机构地区：[1]广州市第一人民医院放射科,广州510180 [2]华南理工大学医学院,广州510006 [3]香港大学深圳医院放射科,深圳518000 [4]广州医科大学附属肿瘤医院放疗中心,广州510095 [5]南方医科大学生物医学工程学院,广州510515 [6]广东食品药品职业学院医疗器械学院,广州510520

出　　处：《磁共振成像》2022年第11期53-59,65,共8页Chinese Journal of Magnetic Resonance Imaging

基　　金：国家自然科学基金(编号:81971574)。

摘　　要：目的 探索基于多参数MRI影像组学特征融合的新型预测模型在高级别胶质瘤(high-grade glioma, HGG)和单发性脑转移瘤(solitary brain metastasis, SBM)中的鉴别价值。材料与方法 收集121名(61名HGG和60名SBM)患者的多参数MRI扫描图像,在常规轴位MRI图像[T1WI、T2WI、T2加权液体衰减反转恢复(T2-weighted fluid attenuated inversion recovery, T2_FLAIR)和T1WI增强图像(post-contrast enhancement T1WI, CE_T1WI)]上勾画了肿瘤实性强化部分的体积(tumor volume of enhancement region, VOI)。通过合并HGG和SBM的类别信息,对不同MRI序列提取的影像组学特征进行融合,并定量比较了不同MRI序列及其组合的性能。结果 从T1WI和T2_FLAIR序列中提取的图像特征的融合比来自其他单一序列或组合的特征具有更显著的预测性能,实现了受试者工作特征曲线下面积、准确率、敏感度和特异度分别为0.946、86.4%、84.1%和88.7%的良好鉴别性能。结论 基于多参数MRI影像组学特征的融合模型通过整合肿瘤的多序列MR图像信息,可以实现对HGG和SBM的无创、高效鉴别。Objective: To explore the value of a new prediction model based on the fusion of multiparametric MRI imaging features in the differential diagnosis of high-grade glioma(HGG) and solitary brain metastasis(SBM). Materials and Methods: We collected multiparametric MRI images of 121(61 HGG and 60 SBM) patients in this study, and delineated the tumor volume of solid enhancement region(VOI) on the conventional axial MRI images [T1WI, T2WI, T2-weighted fluid attenuated inversion recovery(T2_FLAIR) and post-contrast enhancement T1WI(CE_T1WI)]. The radiomics features extracted from different MRI sequences were fused by merging the class information of HGG and SBM, and the performance of different MRI sequences and their combinations were compared quantitatively. Results: Fusion of image features extracted from the T1WI and T2_FLAIR sequences had dominant predictive performances over features from other single sequence or combinations, achieving a discrimination accuracy of area under the ROC curve(AUC), accuracy, sensitivity and specificity of 0.946, 86.4%, 84.1% and 88.7%, respectively. Conclusions: The fusion model based on radiomics features from multiparameter MRI could noninvasively and efficiently identify HGG and SBM via integrating multi-sequence image information of the tumor.

关 键 词：脑肿瘤 高级别胶质瘤 单发性脑转移瘤 影像组学 磁共振成像 鉴别诊断 

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] R739.41[医药卫生—临床医学]