[Texture analysis of diffusion-weighted magnetic resonance imaging to identify atypically enhanced small hepatocellular carcinoma and dysplastic nodules under the background of cirrhosis].

Objective: To investigate the value of texture analysis based on diffusion-weighted magnetic resonance imaging (DWI) in the differential diagnosis of atypically enhanced small hepatocellular carcinoma (sHCC) and dysplastic nodules (DNs) in liver cirrhosis. Methods: Data of 59 cases with atypical enhancement and solitary cirrhotic nodule (≤2 cm) confirmed by dynamic contrast enhanced MRI and surgical pathology specimen were analyzed retrospectively. Among them, 37 cases were of atypically enhanced sHCC and 22 cases of DNS. The DWI signal characteristics of the lesions were analyzed to measure the average apparent diffusion coefficient (ADC) value of the lesions, and the ADC ratio of the lesion to the liver parenchyma. MaZda software was used to manually draw the region of interest to extract the texture parameters of DWI lesions. The three sets (combination of Fisher coefficient, classification of error probability combined with average correlation coefficient and interactive information) were used to select the thirty optimal texture parameters. Raw data analysis (RDA), principal component analysis (PCA), linear discriminant analysis (LDA) and non-linear discriminant analysis (NDA) were performed for texture classification. The difference of ADC value and ADC ratio between sHCC and DNS group was compared by independent sample t-test, and χ2 test was used to compare the count data (or rate). ROC curve analysis was used to evaluate the diagnostic efficiency. Results: The sensitivity, specificity and accuracy of DWI high-signal in the identification of atypically enhanced sHCC and DNs were 94.6% (35/37), 68.2% (15/22), and 84.7% (50/59), respectively. The ADC ratio of atypically enhanced sHCC was significantly lower than DNs, and the difference was statistically significant (t = 2.99, P = 0.002). The sensitivity, specificity, and accuracy for the diagnosis of atypically enhanced sHCC were 73.0% (27/37), 72.7% (16/22) and 72.9% (43/59), respectively. The sensitivity, specificity and accuracy of DWI texture analysis in diagnosing atypically enhanced sHCC were 94.6% (35/37), 95.5% (21/22) and 94.9% (56/59).The diagnostic efficiency of DWI texture analysis (AUC = 0.94) was significantly higher than DWI high-signal (AUC = 0.81) and ADC ratio (AUC = 0.72). Conclusion: The texture analysis based on DWI can identify atypically enhanced sHCC and dysplastic nodules under the background of cirrhosis, and its efficacy is better than qualitative and quantitative DWI.目的： 探讨基于磁共振扩散加权成像（DWI）的纹理分析鉴别诊断肝硬化不典型强化小肝癌（sHCC）和增生结节（DNs）的价值。 方法： 回顾性分析术前MRI动态增强不典型强化、手术标本病理学证实的单发肝硬化结节（≤2 cm）患者59例的资料，其中不典型强化sHCC 37例，DNs 22例。分析病灶的DWI信号特征，测量病灶的平均表观扩散系数（ADC）值、病灶与肝实质ADC比值。采用MaZda软件手工勾画感兴趣区，提取病灶DWI的纹理参数，采用Fisher系数、分类错误概率联合平均相关系数、交互信息三者联合的方法选择30个最佳纹理参数集合。使用原始数据分析、主要成分分析、线性判别分析和非线性判别分析进行纹理分类。采用独立样本t检验比较sHCC与DNs组间ADC值、ADC比值的差异，计数资料（或率）的比较采用χ(2)检验。采用ROC曲线分析评估诊断效能。 结果： DWI高信号鉴别不典型强化sHCC与DNs的灵敏度、特异度及准确度分别为94.6%（35/37）、68.2%（15/22）及84.7%（50/59）。不典型强化sHCC的ADC比值显著低于DNs，差异存在统计学意义（t = 2.99，P = 0.002）；诊断不典型强化sHCC的灵敏度、特异度及准确度分别为73.0%（27/37）、72.7% （16/22）及72.9%（43/59）。DWI纹理分析诊断不典型强化sHCC的灵敏度、特异度及准确度为94.6%（35/37）、95.5%（21/22）及94.9%（56/59）。DWI纹理分析的诊断效能（AUC = 0.94）显著高于DWI高信号（AUC = 0.81）及ADC比值（AUC = 0.72）。 结论： 基于DWI的纹理分析可以鉴别肝硬化背景下不典型强化的小肝癌和增生结节，其效能优于定性及定量DWI。.