磁共振DWI纹理分析预测高血压脑出血扩大的价值

Investigate the value of diffusion-weighted imaging magnetic resonance texture analysis for prediction enlargement of hypertensive intracerebral hemorrhage

目的探讨基于磁共振弥散加权成像(DWI)纹理分析预测高血压脑出血(ICH)扩大的价值。方法搜集56例于发病12 h内行CT和MRI扫描的高血压脑出血患者,根据24 h内CT复查结果,分为血肿扩大组和未扩大组。采用Mazda软件提取两组患者DWI图像血肿区域纹理参数,分别运用Fisher法、分类错误率联合平均相关系数法(POE+ACC)、交互信息(MI)对提取的纹理参数进行筛选降维,然后在B11分析模块中使用原始数据分析(RDA)、主要成分分析(PCA)、线性(LDA)和非线性(NDA)分析法分别对血肿扩大进行判别,将误判率最低的特征集纳入二元Logistic回归,把进入回归模型的纹理参数绘制受试者工作特性曲线(ROC曲线)。结果POE+ACC结合NDA获得的误判率最小,为5.36%(3/56)。S(2,2)SumEntrp、S(3,-3)SumAverg、Perc.10%三个纹理参数及三者联合预测ICH扩大的曲线下面积分别为0.744、0.947、0.930和0.989。结论高血压脑出血DWI图像纹理分析能有效预测血肿扩大,S(2,2)SumEntrp、S(3,-3)SumAverg、Perc.10%三个纹理参数联合预测的效能最佳,可为高血压脑出血预测及辅助诊疗提供更多参考。

Objective To investigate the value of texture analysis based on diffusion-weighted magnetic resonance imaging(DWI)in predicting hematoma enlargement of hypertensive intracerebral hemorrhage(ICH).Methods Fifty-six patients with hypertensive intracerebral hemorrhage who underwent CT and MRI within 12 hours of onset were collected.According to the results of CT reexamination within 24 hours,the patients were divided into enlarged hematoma group(positive group)and non-enlarged hematoma group(negative group).MaZda texture analysis software was used to extract texture features of the hematoma area in DWI images.The Fisher coefficient,POE+ACC and MI were used to reduce the dimensionality of the texture features,respectively,and the B11 analysis module embedded was used to distinguish the expansion of hematoma by the raw data analysis(RDA),principal component analysis(PCA),linear discriminant analysis(LDA)and non-linear discriminant analysis(NDA).The texture features with the lowest false positive rate were included in the binary logistic regression modeling analysis,and the receiver operating characteristic curve(ROC curve)was drawn for the texture parameters entering the regression model.Results POE+ACC with NDA obtained the smallest false positive rate with 5.36%(3/56).The area under curve(AUC)predictedvalue for the three texture features of S(2,2)SumEntrp,S(3,-3)SumAverg,Perc.10%and their combination were 0.744,0.947,0.930 and 0.989,respectively.Conclusion The texture analysis based on DWI image can effectively predict hematoma enlargement in hypertensive intracerebral hemorrhage.The combination of S(2,2)SumEntrp,S(3,-3)SumAverg and Perc.10%is the best index in prediction of hematoma enlargement,which could provide a reference for diagnosis and treatment of ICH.