| 摘要: |
| 目的 评价基于腹腔镜超声的纹理分析在鉴别小肾脏占位中肾透明细胞癌和血管平滑肌脂肪瘤的诊断价值。方法 回顾性收集我院2012年12月至2020年6月间经病理证实的66例肾透明细胞癌和58例血管平滑肌脂肪瘤患者的腹腔镜超声检查资料。所有患者随机分为训练组（87例）和验证组（37例）。通过ITK-SNAPE软件手工勾画感兴趣区，采用pyradiomics包提取纹理特征。使用组内和组间相关系数来选择具有良好稳定性和可重复性的特征。然后使用最大相关最小冗余(mRMR)和最小绝对收缩和选择算子(LASSO)算法进行特征选择和建模，采用ROC曲线对模型诊断效能进行评估。结果 共纳入124例小肾脏占位病灶（肾透明细胞癌组66例，血管平滑肌脂肪瘤组58例）。预测模型由两个纹理特征构成，在训练组和验证组中鉴别肾透明细胞癌和血管平滑肌脂肪瘤的ROC 曲线下面积，敏感度，特异度分别为0.911、0.891、0.854和0.900、0.950、0.647。结论 基于腹腔镜超声的纹理分析可以准确鉴别小肾脏占位中的肾透明细胞癌和血管平滑肌脂肪瘤。 |
| 关键词: 纹理分析 腹腔镜超声 肾透明细胞癌 肾血管平滑肌脂肪瘤 小肾脏占位 |
| DOI： |
| 投稿时间：2020-11-28修订日期：2020-12-22 |
| 基金项目:人工智能技术对肾脏肿块性疾病的二维超声诊断研究 |
|
| Texture analysis of laparoscopic ultrasound images of small renal masses: Differentiation of angiomyolipoma from clear cell renal cell carcinoma |
| Yang Yu,Wei Guan,Fan Li,Ting Wang,You-bin Deng |
| (Department of Medical Ultrasound,Tongji Hospital,Tongji Medical College,Huazhong University of Science and Technology) |
| Abstract: |
| Objective To investigate the value of texture analysis based on laparoscopic ultrasound images of small renal masses in discrimination of clear cell renal cell carcinoma and angiomyolipoma. Methods 66 patients with clear cell renal cell carcinoma and 58 patients with angiomyolipoma confirmed by pathology in Tongji Hospital of Huazhong University of Science and Technology from December 2012 to June 2020 were analyzed retrospectively. The patients were randomly divided into training group (n=87) and verification group (n=37). ROI was drawn manually by ITK‐SNAP software. The texture features were extracted by python‐based pyradiomics package. Intra- and interclass correlation coefficients were used to select the features with excellent stability and repeatability. The minimum-redundancy maximum-relevance (mRMR) and the least absolute shrinkage and selection operator (LASSO) algorithms were used for feature selection and model development for discriminating clear cell renal cell carcinoma from angiomyolipoma. Receiver operating characteristic curve analysis was used to evaluate the diagnostic performance of the model. Result 124 patients with small renal masses confirmed by pathology after resection were analyzed retrospectively, including 66patients with clear cell renal cell carcinoma and 58 patients with angiomyolipoma. The area under the receiver-operating( AUC), sensitive and specific were 0.911、0.891、0.854in the training group and 0.900、0.950、0.647in the validation group. Conclusion Texture analysis of laparoscopic ultrasound images of small renal masses could be used to discriminate clear cell renal cell carcinoma from angiomyolipoma. |
| Key words: Texture analysis Laparoscopic ultrasonography Renal cell carcinoma Renal angiomyolipoma Small renal masses |
