摘要
目的:探讨CT直方图参数分析对肺部浸润前病变和浸润性肺癌的鉴别诊断价值.方法:回顾性收集经手术病理证实的浸润性肺癌和肺部浸润前病变共199例,其中浸润性肺癌111例,肺部浸润前病变88例.所有患者均完成术前CT检查.CT图像经MATLAB软件处理后得到肿瘤区域的直方图参数,包括CT最大强度值、CT最小强度值、CT直方图偏度等共12个参数.对CT直方图参数的分析分为2个部分:①Mann-Whitney检验和ROC曲线分析被用来评价直方图参数在两类病变中的统计学差异;②将回归分析用于所有的CT直方图参数,尝试对两类肺部病变进行分类鉴别.结果:在单参数分析中,除直方图峰度外,余11个CT直方图参数对两类肺部病变的分类鉴别差异有统计学意义(均P>0.05),其中直方图熵值表现出最好的诊断能力(P<0.01,ROC曲线下面积0.912,敏感度0.901,特异度0.784).Logistic回归分析对两类样本的鉴别准确率为93.0%,敏感度0.909,特异度0.946;Fisher系数回归分析对两类样本的鉴别准确率为90.5%,敏感度0.852,特异度0.946.结论:CT直方图参数分析及回归分析可鉴别肺部浸润前病变与浸润性肺癌.
Objective:To investigate on the differential diagnosis of preinvasive lesions and invasive cancer of lung by CT histogram analysis. Methods: Data of 199 cases with invasive lung cancer and lung preinvasive lesions confirmed by surgery and pathology were collected systematically, including 111 cases of invasive lung cancer and 88 cases of preinvasive lesions. All patients completed preoperative CT examination. Twelve histogram parameters in the lesion area were obtained by MATLAB software ,including the maximum and minimum intensity of CT value, the skewness of CT histogram and so on. The analysis of histogram parameters was divided into two parts. Mann-Whitney test and ROC analysis were firstly used to evaluate the significance of histogram parameters in the two types of lesions and the sensitivity and specificity of the parameters with significant differences in the differential diagnosis of the two types of lesions. Secondly, regression analysis was applied to CT histogram parameters to identify two types of lung lesions. Results: In the single parameter analysis, except histogram kurtosis, the other 11 parameters showed statistical differences in the classification and differentiation of two types of lung lesions , and histogram entropy showed the best diagnostic ability (P<0.01, with AUC of 0.912, sensitivity of 0.901, specificity of 0.784). In regression analysis, the accuracy, sensitivity and specificity of Logistic regression were 93.0%, 0.909,0.946,respectively, and those of Fisher coefficient regression were 90.5%, 0.852,0.946, respectively. Conclusion: CT histogram parameters and regression analysis are useful in differential diagnosis of preinvasive lesions and invasive cancer of lung.
作者
王文斐
王玉强
苏伟
张凡涛
杨新国
Wang Wenfei;Wang Yuqiang;Su Wei;Zhang Fantao;Yang Xinguo(Department of Imaging, Shengli Oilfield Centre Hospital, Dongying, 257034, China)
出处
《中国中西医结合影像学杂志》
2019年第4期356-359,共4页
Chinese Imaging Journal of Integrated Traditional and Western Medicine
