基于MRI增强绝对梯度纹理分析对垂体瘤质地术前评估的可行性分析

Feasibility Analysis of Preoperative Assessment of Pituitary Tumor Texture Based on MRI-enhanced Absolute Gradient Texture Analysis

目的采用MRI绝对梯度纹理分析方法对垂体瘤的质地进行分析,并探讨该技术的可行性,为垂体瘤的术前评估及术式选择提供影像学参照。方法回顾性分析2018年1月~2020年6月本院经手术病理证实的55例垂体瘤患者的病例资料,其中质地硬者20例,质地软者35例,选取病变矢状位增强T1WI图像最大层面,采用Mazda纹理分析软件勾画感兴趣区,比较软、硬两种质地垂体瘤绝对梯度参数,采用二元Logistic回归分析法及ROC曲线分析法联合预测因子及各参数对软、硬两种质地垂体瘤的诊断效能。结果不同质地垂体瘤梯度均值及变异度比较,差异有统计学意义(P<0.05);不同质地垂体瘤偏度、峰度、非零梯度百分比比较,差异无统计学意义(P>0.05)。二元Logistic回归分析显示,联合预测因子取值0.810时,其对应的敏感度和特异度分别为100.00%、80.00%。ROC曲线分析显示,梯度均值、变异度及联合预测因子的AUC值均大于0.7,同时联合预测因子的AUC值达0.920,较均值及梯度具有更好的诊断效能。结论MRI绝对梯度分析方法可提供量化型特征参数,为垂体瘤质地的术前评估提供了新的切入点。

Objective To analyze the texture of pituitary tumors using MRI absolute gradient texture analysis method,and to explore the feasibility of this technique,so as to provide an imaging reference for the preoperative evaluation and surgical selection of pituitary tumors.Methods A retrospective analysis of the case data of 55 patients with pituitary tumors confirmed by surgery and pathology in our hospital from January 2018 to June 2020.Among them,20 cases had hard texture and 35 cases had soft texture.The maximum level of the T1WI image was enhanced in the sagittal view of the lesion.Using Mazda texture analysis software to delineate the area of interest,compare the absolute gray gradient parameters of pituitary tumors with soft and hard textures.The dual logistic regression analysis method and ROC curve analysis method were used to combine predictive factors and various parameters in the diagnosis of soft and hard pituitary tumors.Results There was a statistically significant difference in the gradient mean and variability of pituitary tumors of different textures(P<0.05);There was no statistically significant difference in the skewness,kurtosis,and percentage of non-zero gradient of pituitary tumors of different textures(P>0.05).Binary Logistic regression analysis showed that when the combined predictor value was 0.810,the corresponding sensitivity and specificity are 100.00%and 80.00%,respectively.ROC curve analysis showed that the average value of the gradient,the degree of variability and the AUC value of the combined predictor were all greater than 0.7,and the AUC value of the combined predictor reached 0.920,which has better diagnostic performance than the average and gradient.Conclusion The MRI gray-scale absolute gradient analysis method can provide quantitative characteristic parameters and provide a new entry point for the preoperative evaluation of pituitary tumor texture.