MRI纹理分析对移行带前列腺癌与基质型增生结节的鉴别诊断价值  

Differential diagnosis of transition zone prostate cancer and stromal prostate hyperplasia by MRI-derived texture analysis

作　　者：张宏江[1] 吴昆华[1] 董彪[2] 罗梓维 杨竣宇 王少彧 王波[1] ZHANG Hongjiang;WU Kunhua;DONG Biao;LUO Ziwei;YANG Junyu;WANG Shaoyu;WANG Bo(Department of MRI,the First People＇s Hospital of Yunnan,Kunming 650031,China;Department of Andrology,the Second Affiliated Hospital of Kunming Medical University,Kunming 650101,China;Siemens Healthineers,Shanghai 201318,China)

机构地区：[1]云南省第一人民医院磁共振科,云南昆明650031 [2]昆明医科大学第二附属医院男性科,云南昆明650101 [3]西门子医疗系统有限公司磁共振事业部,上海201318

出　　处：《实用放射学杂志》2018年第10期1560-1563,共4页Journal of Practical Radiology

摘　　要：目的 探讨采用MRI纹理分析对移行带前列腺癌（PCa）与基质型增生结节鉴别诊断的可行性。方法 回顾性分析本院经病理证实的67例前列腺疾病患者，移行带PCa 29例，基质型增生结节38例；采用MaZda软件手动勾画感兴趣区（ROI），利用灰度直方图、灰度共生矩阵（GLGM）、游程矩阵（RLM）、绝对梯度（GRA）、自回归模型（ARM）及小波变换（WAV）中提取纹理特征值，通过费希尔系数法（Fisher）、分类误差概率联合平均相关系数（POE＋ACC）、交互信息（MI）及3种方法联合（MI＋PA＋F）分别提取最佳纹理特征，再采用原始数据分析（RDA）、主成分分析（PCA）、线性判别分析（LDA）、非线性判别分析（NDA）4种分类统计方法对2种病变进行判别，结果以判错率形式表示。结果 表观扩散系数（ADC）纹理分析的判错率（0.00%~23.88%）总体上低于T2WI纹理分析的判错率（1.49%~37.31%）。ADC纹理分析中，MI＋PA＋F联合NDA的判错率最低（0.00%）。结论 MRI纹理分析可用于移行带PCa和基质型增生结节的鉴别诊断。Objective To determine feasibility of MRI-derived texture analysis for differential diagnosis of transition zone prostate cancer {PCa）and stromal prostate hyperplasia. Methods 67 patients with transition zone PCa and stromal prostate hyperplasia confirmed by pathology were enrolled in this retrospective study including 29 cases of transitional zone PCa and 38 cases of stromal hyperplasia nodules.ROI manually drew by using MaZda software,and texture features derived from the gray-level histogram, the gray-level co occurrence matrix（GLGM）, run-length matrix （RLM）, absolute gradient （GRA）, auto-regressive model { ARM） and wavelet transform （ WAV ） were calculated, and the optimal texture features were extracted by Fisher, probability of classification error combined with average correlation coefficients （POE＋ACC）,mutual information （MI） coefficients and the combination of the above three methods （MI＋ PA＋ F）, then the classification statistical methods including raw data analysis （RDA）, principal component analysis （PCA）, linear discriminam analysis （LDA） and nonlinear discriminant analysis （NDA） were used to distinguish transition zone prostate cancer from stromal prostate hyperplasia.The results were shown by miselassification rate.Results The misclassification rate of ADC texture analysis （0.00%-23.88%） generally lower than that of T2 WI texture analysis （1.49 %- 37.31%）. In the AIX2 texture analysis, the misclassification rate of MI ＋ PA＋F combined with NDA was lowest （0.00%）.Conclusion MRI-derived texture analysis can be used to differentiate transition zone PCa from stromal prostate hyperplasia.

关 键 词：良性前列腺增生 前列腺癌 磁共振成像 纹理分析 

分 类 号：R697.3[医药卫生—泌尿科学] R737.25[医药卫生—外科学]