【摘要】 目的 探讨基于磁共振成像（magnetic resonance imaging,MRI）的纹理分析（texture analysis, TA）在上皮性卵巢癌（epithelial ovarian cancer,EOC）分型中的价值。方法 回顾性分析我院2016年7月至2020年12月经手术病理证实的45例EOC患者的病例资料，所有患者术前均行盆腔MRI常规序列扫描及弥散加权成像（diffusion-weighted imaging,DWI）。根据EOC的二元论模型^[1]，将所有患者分为Ⅰ型EOC和Ⅱ型EOC两组。利用后处理软件于轴位T₂加权成像（T₂-weighted imaging,T₂WI）和表观扩散系数（diffusion-weighted imaging,ADC）图上提取病灶的相关纹理参数，包括灰度直方图纹理参数（平均值、标准差、峰度、偏度、均匀性）和灰度共生矩阵纹理参数（能量、惯性、熵、相关性、逆差距）。采用t检验、Mann-Whitney U检验比较Ⅰ型和Ⅱ型EOC纹理参数的差异。对于差异有统计学意义纹理参数（P<0.05），绘制受试者工作特征曲线（receiver operating characteristic, ROC）曲线，得到曲线下面积（area under the curve,AUC），评估它们对Ⅰ型EOC和Ⅱ型EOC的鉴别诊断效能。进一步将有统计学意义的纹理参数进行logistic回归分析，确定鉴别Ⅰ型EOC和Ⅱ型EOC的独立影响因素。结果 各纹理参数中，基于T₂WI的标准差、熵、相关、逆差距和基于ADC图的标准差、偏度、熵、逆差距在两组之间差异具有统计学意义（P<0.05），其中基于T₂WI的熵对Ⅰ型EOC和Ⅱ型EOC的鉴别诊断价值最大；多因素分析显示，基于T₂WI的熵和基于ADC图的标准差、偏度、熵是鉴别Ⅰ型EOC和Ⅱ型EOC的独立影响因素。结论 基于MRI图像的纹理参数有助于区分Ⅰ型EOC和Ⅱ型EOC，尤其是T₂WI的熵和ADC图的标准差、偏度、熵，TA有望成为术前无创性评估EOC分型的重要工具。更多还原

【Abstract】 Objective To investigate the value of texture parameters based on MRI images in the classification of epithelial ovarian cancer. Methods The cases of 45 patients with epithelial ovarian cancer confirmed by surgery and pathology in our hospital from July 2016 to December 2020 were retrospectively analyzed. All patients underwent routine pelvic MRI sequence scan and diffusion weighted imaging before surgery. According to the binary classification theory of epithelial ovarian cancer^[1], patients were divided into two groups: type Ⅰ EOC and type Ⅱ EOC.Post-processing software was used to extract the relevant texture parameters of the lesions from axial T₂WI images and ADC images, including grayscale histogram texture parameters（mean, standard deviation, kurtosis, skewness, uniformity） and grayscale co-occurrence matrix texture parameters（energy, inertial, entropy, correlation, inverse difference moment）. The t-test and Mann-Whitney U test were used to compare the texture parameters of type Ⅰ and type Ⅱ epithelial ovarian cancer. For texture parameters with statistically significant differences（P<0.05）, draw the curve of receiver operating characteristic（ROC）, and the area under the curve（AUC） was obtained to evaluate their efficacy in the differential diagnosis of type Ⅰ EOC and type Ⅱ EOC. Furthermore, Logistic regression analysis was performed on statistically significant texture parameters to determine the independent influencing factors for the differentiation of type Ⅰ EOC and type Ⅱ EOC.Results For texture parameters, the difference between T₂WI based standard deviation, entropy, correlation, inverse difference moment and ADC based standard deviation, skewness, entropy, inverse difference moment between the two groups was statistically significant（all P<0.05）, Among them, Entropy based on T₂WI showed the greatest value in the differential diagnosis of type Ⅰ EOC and type ⅡEOC.Multivariate analysis showed that entropy based on T₂WI and standard deviation, skewness and entropy based on ADC map were independent influencing factors for differentiating type Ⅰ EOC from type ⅡEOC. Conclusions The texture parameters based on MRI images are helpful to distinguish type Ⅰ EOC from type Ⅱ EOC, especially the entropy of T₂WI and the standard deviation, skewness, and entropy of ADC images. TA is expected to become an important tool for preoperative non-invasive evaluation of EOC classification.更多还原