论文部分内容阅读
目的:探讨基于腹腔镜超声的影像组学方法在肾脏肿块良、恶性鉴别中的价值。方法:回顾性纳入2012年12月至2018年12月在华中科技大学附属同济医院就诊并经手术病理证实的肾脏肿块患者119例，所有患者术前均行腹腔镜超声检查。根据病理将纳入患者分为良性组和恶性组，所有患者随机分为训练组(84例)和验证组(35例)。采用pyradiomics包提取良性组和恶性组病灶的107个影像组学特征。采用两独立样本n t检验或Mann-Whitney n U检验、Lasso回归分析对训练组中的特征进行筛选和建模，采用ROC曲线评估模型诊断效能，并通过验证组数据验证模型的稳定性。n 结果:119例患者的119个病灶纳入研究(良性54个，恶性65个)，最终筛选出5个差异有统计学意义的影像组学特征建模，在训练组和验证组中模型预测肾脏肿块病理良性和恶性ROC曲线下面积、敏感性、特异性、准确性分别为0.92、0.96、0.79、0.88和0.91、0.89、0.88、0.89。结论:基于腹腔镜超声影像组学特征构建的预测模型对肾肿块良、恶性的诊断为经皮超声诊断探索了新途径，奠定了应用基础。“,”Objective:To investigate the value of radiomics in discriminating benign and malignant renal masses based on laparoscopic ultrasound imaging.Methods:One hundred and nineteen patients undergoing laparoscopic surgical resection and histopathological analysis of renal masses in Tongji Hospital of Huazhong University of Science and Technology were retrospectively included from December 2012 to December 2018. All patients underwent preoperative laparoscopic ultrasound examinations. According to pathology, the patients were divided into benign group and malignant group, and all patients were randomly divided into training group (n n=84) and verification group (n n=35). One hundred seven radiomics features were extracted respectively in benign group and malignant group using pyradiomics. The independent sample n t-test or Mann-Whitney n U test and Lasso regression were used for features selection and model establishment in the training group. The performance of the model was evaluated by ROC curve, and the stability of the model was verified in the validation group.n Results:A total of 119 patients met inclusion criteria (benign: n n=54; malignant: n n=65), and 5 radiomics features with statistical significance were selected to establish predictive model for discrimination of malignant and benign renal masses. The area under the ROC curve, sensitive, specific and accurate were 0.92, 0.96, 0.79, 0.88 in the training group and 0.91, 0.89, 0.88, 0.89 in the validation group, respectively.n Conclusions:The predictive model based on radiomic features of laparoscopic ultrasonographic imaging provides a new approach for the discrimination of malignant and benign renal masses and lays a foundation for its application.