高分辨率CT图像纹理分析对孤立性肺实性结节的诊断价值研究

The diagnosis value of high-resolution computed tomography image texture analysis for solitary pulmonary solid nodule

目的:探索高分辨率CT图像纹理分析对孤立性肺实性结节的诊断价值。方法:应用Mazda软件分析98例孤立性肺实性结节高分辨率CT图像纹理参数,并Fisher系数、交互信息、分类错误概率组合平均相关系数法以及3种方法联合(FPM)筛选出具有代表意义纹理参数。然后,应用线性判别分析、非线性判别分析、原始数据分析、主要成分分析4种分类统计方法计算纹理参数诊断孤立性肺实性结节效能,以错判率表示。最后,将纹理参数诊断肺结节效能与胸部影像医师进行对比分析。结果:应用FPM法提取特征性纹理参数并采用非线性分类判别方法孤立性肺实性结节良恶性错判率为6.1%(6/98),胸部影像医师错判率为16.3%(16/98),两者差异具有统计学意义(P<0.05)。结论:高分辨率CT图像纹理分析可作为孤立性肺实性结节鉴别诊断的新手段之一。

Objective:To explore the diagnosis value of high-resolution computed tomography image texture analysis for solitary pulmonary solid nodule.Methods:A total of 98 patients with solitary pulmonary solid nodule were reviewed.Texture analysis was performed on high-resolution computed tomography image by software of Mazda to obtain the misclassification rate between benign and malignant pulmonary solid nodules.The procedure of texture analysis included analysis of texture parameters,selection of typical texture parameters(fisher coefficient,mutual information,classification error probability combined average correlation coefficients,and combined the three above(FPM)),and classification of the typical texture parameters(linear discriminant analysis,nonlinear discriminant analysis,raw data analysis).Then,the misclassification rate was compared between texture analysis and chest-radiologists.Results:The misclassification rate was the lowest with 6.1%(6/98)through FPM selection and nonlinear discriminant analysis statistical method,and was comparable with chest-radiologists(P<0.05).Conclusion:High-resolution computed tomography image texture analysis is a new important way for differential diagnosis of solitary pulmonary solid nodule.