Yang Ran, Zhao Dan, Ye Chunxue, Hu Ming, Qi Xiao, Li Zhichao
Department of Radiology, Chongqing Western Hospital, No. 301, Huafu Avenue North, Jiulongpo District, Chongqing, 400050, China.
Department of Radiology, Second People's Hospital of Jiu Long Po District, No. 318 Huayu Road, Jiulongpo District, Chongqing, 400052, China.
BMC Med Imaging. 2025 Jul 4;25(1):268. doi: 10.1186/s12880-025-01817-8.
This study aimed to develop and validate a machine learning (ML) model that integrates radiomics and conventional radiological features to predict the success of single-session extracorporeal shock wave lithotripsy (ESWL) for ureteral stones.
This retrospective study included 329 patients with ureteral stones who underwent ESWL between October 2022 and June 2024. Patients were randomly divided into a training set (n = 230) and a test set (n = 99) in a 7:3 ratio. Preoperative clinical data and noncontrast CT images were collected, and radiomic features were extracted by outlining the stone's region of interest (ROI). Univariate analysis was used to identify clinical and conventional radiological features related to the success of single-session ESWL. Radiomic features were selected using the least absolute shrinkage and selection operator (LASSO) algorithm to calculate a radiomic score (Rad-score). Five machine learning models (RF, KNN, LR, SVM, AdaBoost) were developed using 10-fold cross-validation. Model performance was assessed using AUC, accuracy, sensitivity, specificity, and F1 score. Calibration and decision curve analyses were used to evaluate model calibration and clinical value. SHAP analysis was conducted to interpret feature importance, and a nomogram was built to improve model interpretability.
Ureteral diameter proximal to the stone (UDPS), stone-to-skin distance (SSD), and renal pelvic width (RPW) were identified as significant predictors. Six radiomic features were selected from 1,595 to calculate the Rad-score. The LR model showed the best performance on the test set, with an accuracy of 83.8%, sensitivity of 84.9%, specificity of 82.6%, F1 score of 84.9%, and AUC of 0.888 (95% CI: 0.822-0.949). SHAP analysis indicated that the Rad-score and UDPS were the most influential features. Calibration and decision curve analyses confirmed the model's good calibration and clinical utility.
The LR model, integrating radiomics and conventional radiological features, demonstrated strong performance in predicting the success of single-session ESWL for ureteral stones. This approach may assist clinicians in making more accurate treatment decisions.
Retrospectively.
Not applicable.
本研究旨在开发并验证一种机器学习（ML）模型，该模型整合了放射组学和传统放射学特征，以预测输尿管结石单次体外冲击波碎石术（ESWL）的成功率。
这项回顾性研究纳入了2022年10月至2024年6月期间接受ESWL治疗的329例输尿管结石患者。患者按7:3的比例随机分为训练集（n = 230）和测试集（n = 99）。收集术前临床数据和非增强CT图像，并通过勾勒结石的感兴趣区域（ROI）提取放射组学特征。采用单因素分析确定与单次ESWL成功相关的临床和传统放射学特征。使用最小绝对收缩和选择算子（LASSO）算法选择放射组学特征，以计算放射组学评分（Rad-score）。使用10折交叉验证开发了五个机器学习模型（随机森林（RF）、K近邻（KNN）、逻辑回归（LR）、支持向量机（SVM）、自适应增强（AdaBoost））。使用曲线下面积（AUC）、准确率、灵敏度、特异度和F1分数评估模型性能。采用校准分析和决策曲线分析评估模型校准和临床价值。进行SHAP分析以解释特征重要性，并构建列线图以提高模型的可解释性。
结石近端输尿管直径（UDPS）、结石至皮肤距离（SSD）和肾盂宽度（RPW）被确定为显著预测因素。从1595个放射组学特征中选择了6个来计算Rad-score。LR模型在测试集上表现最佳，准确率为83.8%，灵敏度为84.9%，特异度为82.6%，F1分数为84.9%，AUC为0.888（95%CI：0.822 - 0.949）。SHAP分析表明Rad-score和UDPS是最具影响力的特征。校准分析和决策曲线分析证实了该模型具有良好的校准和临床实用性。
整合放射组学和传统放射学特征的LR模型在预测输尿管结石单次ESWL的成功率方面表现出强大性能。这种方法可能有助于临床医生做出更准确的治疗决策。
回顾性研究。
不适用。
