METHODS: A total of 301 eligible patients (165/136 successful/unsuccessful) who underwent SWL were retrospectively evaluated and divided into a training cohort (n = 241) and a test cohort (n = 60) following an 8:2 ratio. Univariate analysis was performed to assess clinical characteristics for constructing a nomogram. Radiomics and conventional radiological characteristics of stones were evaluated. Following feature selection, radiomics and radiological models were constructed using logistic regression (LR), support vector machine (SVM), random forest (RF), K nearest neighbor (KNN), and XGBoost. The models\' performance was compared using metrics such as the area under the receiver operating characteristic curve (AUC), precision, recall, accuracy, and F1 score. Finally, a nomogram was created incorporating the best image model signature and clinical predictors.
RESULTS: The SVM-based radiomics model showed superior predictive performance in both training and test cohorts (AUC: 0.956, 0.891, respectively). The nomogram, which combined SVM-based radiomics signature with proximal ureter diameter (PUD), demonstrated further improved predictive performance in the test cohort (AUC: 0.891 vs. 0.939, P = 0.166).
CONCLUSIONS: Integration of CT-derived radiomics and PUD showed excellent ability to predict SWL treatment success in patients with ureteral stones larger than 10mm, providing a promising approach for clinical decision-making.
方法:对总共301名符合条件的患者(165/136名成功/不成功)进行了回顾性评估,并按照8:2的比例分为训练组(n=241)和测试组(n=60)。进行单变量分析以评估临床特征以构建列线图。评估了结石的影像组学和常规放射学特征。在选择功能之后,使用逻辑回归(LR)构建放射组学和放射学模型,支持向量机(SVM),随机森林(RF),K最近邻(KNN),XGBoost使用诸如接收器工作特征曲线下面积(AUC)、精度,召回,准确度,F1得分。最后,创建了包含最佳图像模型特征和临床预测因子的列线图.
结果:基于SVM的影像组学模型在训练和测试队列中均显示出优异的预测性能(AUC:0.956,0.891)。列线图,将基于SVM的影像组学特征与输尿管近端直径(PUD)相结合,在测试队列中显示出进一步改善的预测性能(AUC:0.891vs.0.939,P=0.166)。
结论:CT衍生的影像组学和PUD的整合显示出在输尿管结石大于10mm的患者中预测SWL治疗成功的良好能力,为临床决策提供了一种有希望的方法。