表观扩散系数全域灰度直方图鉴别腮腺肿瘤的价值

Value of whole-tumor histogram profiling of apparent diffusion coefficient maps in differential diagnosis of parotid tumors

目的探讨表观扩散系数(ADC)全域直方图对鉴别腮腺多形性腺瘤和恶性肿瘤的价值。方法回顾性分析2016年11月至2018年1月于郑州大学第一附属医院经手术病理证实的41例腮腺多形性腺瘤和22例恶性肿瘤治疗前MRI平扫、DWI及ADC图像。在包含肿瘤实质的每一层ADC图像上用Mazda软件勾画感兴趣区并进行灰度全域直方图分析,对两组肿瘤直方图参数特征进行统计学分析,并应用ROC曲线评价直方图参数鉴别腮腺多形性腺瘤和恶性肿瘤的价值。结果通过灰度直方图提取出的12个参数中,最小值、最大值、均值、偏度、第10百分位数、第50百分位数、第90百分位数及第99百分位数这8个参数在两组间的差异有统计学意义(P<0.05),其余4个参数面积、峰度、变异度和第1百分位数在两组间差异无统计学意义(P>0.05)。其中,第10百分位数诊断效能最高,曲线下面积为0.834,灵敏度和特异度分别为85.4%和75.7%。灰度直方图多个参数联合进行诊断的曲线下面积为0.888,灵敏度和特异度分别为85.6%和77.3%。结论磁共振全域灰度直方图分析可作为鉴别腮腺多形性腺瘤和恶性肿瘤的重要手段,第10百分位数具有较高的诊断效能。灰度直方图多个参数联合能够进一步提高诊断效能。

Objective To determine the value whole-tumor histogram profiling of apparent diffusion coefficient(ADC)maps in differentiating pleomorphic adenoma and malignant tumor of the parotid gland.Methods A retrospective analysis was performed on plain scan,DWI and ADC MRI images prior to treatments from 41 cases of parotid polymorphic adenoma and 22 of malignant parotid tumors which were confirmed by post-surgical pathology between November 2016 and January 2018 at the First Affiliated Hospital of Zhengzhou University.For each section of ADC image containing tumor parenchyma,Mazda software was used to delineate the region of interest(ROI)and analyze the gray-scale whole-tumor histogram of ADC map.The characteristics of tumor histogram profiling in the two groups were statistically analyzed.ROC curve was used to evaluate the value of histogram profiling in differentiating pleomorphic adenoma and malignant tumor of the parotid gland.Results Among the 12 parameters retrieved from grayscale histograms,there were statistically significant differences between the two groups in the minimum,maximum,mean,skewness,10th,50th,90th and 99th percentiles(P<0.05),but not in the other four parameters(area,kurtosis,variability and 1st percentile(P>0.05).Specifically,the 10th percentile offered the highest diagnostic efficiency,with an area under the curve(AUC)of 0.834,a sensitivity of 85.4%and a specificity of 75.7%,respectively.The AUC with a combination of multiple gray-scale histogram in the diagnosis was 0.888,with a sensitivity of 85.6%and a specificity of 77.3%,respectively.Conclusion Whole-tumor grayscale histogram profiling can be used as an important tool to distinguish between pleomorphic adenoma and malignant tumor of the parotid gland.High diagnostic efficiency may be achieved by using the 10th percentile,and can be further improved by using combination of multiple parameters on the grayscale histogram.