MRI图像纹理分析在高级别和低级别脑膜瘤鉴别诊断中的价值分析

Value analysis of texture analysis of MRI images in differential diagnosis for high-grade and low-grade meningiomas

目的:分析磁共振成像(MRI)图像纹理分析在高级别和低级别脑膜瘤鉴别诊断中的价值。方法:回顾性收集90例术前行常规MRI检查的脑膜瘤患者临床资料,根据术前病理分级情况将其分为低级别组(Ⅰ级,73例)和高级别组(Ⅱ~Ⅲ级,17例)。对两组MRI图像肿瘤最大径感兴趣区(ROI)进行勾画,测量其直方图参数和灰度共生矩阵参数,比较两组MRI图像中直方图与灰度共生矩阵参数,对有统计学意义的指标绘制受试者工作特征(ROC)曲线,计算ROC曲线下面积(AUC)。结果:高级别组T2加权像(T2WI)和液体衰减反转恢复(FLAIR)序列的最大值高于低级别组,差异有统计学意义(t=2.45,t=3.38;P<0.05);T2WI、表观弥散系数(ADC)和FLAIR序列标准差及偏度值明显升高,差异有统计学意义(t=1.64,t=1.32;P<0.05);而其他直方图参数比较无明显差异。高级别组T1加权像(T1WI)、T2WI、ADC及FLAIR序列的熵值高于低级别组,差异有统计学意义(t=4.62,t=3.00,t=3.42,t=2.93;P<0.05);高级别组T2WI序列的ASM能量值明显下降,与低级别组比较差异有统计学意义(t=2.35,P<0.05),而其他序列灰度共生矩阵参数比较无明显差异。有统计学意义的指标ROC曲线下面积(AUC)高于0.70的参数包括ADC序列的偏度值、T2WI和ADC序列的熵值,三者联合检测的AUC值为0.84;高级别组脑膜瘤的灵敏度为75.34%,特异度为88.24%。结论:MRI图像纹理分析可为临床诊断提供更多定量信息,在术前鉴别和诊断脑膜瘤患者病理分级具有较高的效能。

Objective:To analyze the value of texture analysis of magnetic resonance imaging(MRI)images in the differential diagnosis for high-grade and low-grade meningiomas.Methods:The clinical data of 90 patients with meningioma who underwent routine MRI examination pre operation were retrospective collected.All patients were divided into low-grade group(73 patients with pathological grading I)and high-grade group(17 patients with pathological grading II and III)according to the preoperative pathological grading.On the MRI image,the region of interest(ROI)of the maximum diameter level of the tumor was delineated,and histogram parameters and gray level co-occurrence matrix parameters were measured,and the differences of histogram and gray level co-occurrence matrix parameters of MRI imaging between the two groups were compared.The indicators with statistical significance were adopted to draw receiver operating characteristic(ROC)curve and their area under the curve(AUC)of ROC were analyzed.Results:The maximum values of T2-weighted image(T2WI)and fluid-attenuated inversion recovery(FLAIR)sequence of high-level group were significantly higher than those of low-level group(t=2.45,t=3.38,P<0.05).And the standard deviation and SKEW of T2WI,apparent diffusion coefficient(ADC)and FLAIR sequence of observation group were significantly higher than those of control group(t=1.64,t=1.32,P<0.05),while there was no significant difference in other histogram parameters between the two groups.The entropy values of T1WI,T2WI,ADC and FLAIR sequences of high-grade group were significantly higher than those of low-grade group(t=4.62,t=3.00,t=3.42,t=2.93,P<0.05).While the ASM energy values of T2WI sequences decreased significantly,and the difference of that between the two groups was significant(t=2.35,P<0.05).And there were no significant difference in the parameters of gray level co-occurrence matrix of other sequences between the two groups.In these indicators with significance,the parameters whose AUC of ROC were higher than 0.70 included SKEW of ADC sequence and the entropy value of T2WI and ADC sequence,and the AUC value of combined detection with three indicators was 0.84.And the sensitivity and specificity of high-grade group in diagnosing meningiomas were 75.34%and 88.24%,respectively.Conclusion:The texture analysis of MRI images can provide more quantitative information for clinical diagnosis,and it has higher efficiency in preoperative identification and diagnosis for pathological grading of patients with meningioma.