基于增强T2*加权血管成像序列R2*图肿瘤全域纹理分析预测子宫内膜癌微卫星不稳定性

目的评估基于增强T2^*加权血管成像(ESWAN)序列R2^*图的肿瘤全域纹理分析(TA)预测子宫内膜癌(EC)微卫星不稳定性(MSI)的价值。方法回顾性分析38例经术后病理证实、术前接受ESWAN序列盆腔MR扫描的EC患者,其中12例MSI(MSI组)、26例微卫星稳定(MSS,MSS组),经后处理获得R2^*图。于R2^*图像上逐层手动勾画肿瘤ROI,融合后获得全域感兴趣容积(VOI);采用A.K.分析软件提取其纹理特征,以Spearman相关性分析和梯度提升决策树(GBDT)方法筛选最优纹理特征,构建多元logistic回归模型,用于预测EC MSI状态;以受试者工作特征(ROC)曲线评价模型的诊断效能。结果共提取74个纹理特征,最终筛选出6个最优纹理特征,以之构建预测EC MSI的回归模型。ROC曲线显示,模型的曲线下面积(AUC)、准确率、敏感度及特异度分别为0.95、89.50%、83.30%及92.30%。结论基于ESWAN序列R2^*图的肿瘤全域TA有助于术前预测EC MSI。

Tumor global texture analysis based on enhanced T2*weighted angiography sequence R2* map for predicting microsatellite instability of endometrial carcinoma

Objective To explore the value of tumor global texture analysis (TA) based on enhanced T2 star weighted angiography (ESWAN) sequence R2^* map for predicting microsatellite instability (MSI) of endometrial carcinoma (EC). Methods Preoperative pelvic MRI data of 38 patients with EC confirmed by surgery and pathology, including 12 cases of MSI (MSI group) and 26 of microsatellite stability (MSS, MSS group) were retrospectively analyzed. All patients underwent ESWAN sequence scanning. R2^* map was obtained after post-processing. Tumor ROI was manually delineated layer by layer on R2^* map, then the whole region volume of interest (VOI) was fused. Texture features of VOI were extracted using A.K. analysis software, and the optimal texture features were selected using Spearman correlation analysis and gradient boosting decision tree (GBDT) methods. Then, a multivariate logistic regression model was constructed to predict EC MSI, and relative operating characteristic (ROC) curve was used to evaluate the diagnostic efficiency of this model. Results A total of 74 texture features were extracted, and finally 6 optimal texture features were selected to build the regression model for predicting EC MSI. ROC curve showed that the area under the curve (AUC), accuracy, sensitivity and specificity of this model was 0.95, 89.50%, 83.30% and 92.30%, respectively. Conclusion Tumor global TA based on ESWAN R2^* map was helpful to preoperative predicting of EC MSI.