Wu Shujian, Fan Lifang, Wu Yimin, Xu Jingya, Guo Yong, Zhang Hu, Xu Zhengyuan
Department of Radiology, The First Affiliated Hospital of Wannan Medical College, Wuhu, Anhui 241001, China.
School of Medical Imageology, Wannan Medical College, Wuhu, Anhui 241002, China.
Eur J Radiol. 2025 Sep;190:112266. doi: 10.1016/j.ejrad.2025.112266. Epub 2025 Jun 26.
To develop and validate a computerized tomography (CT)‑based deep transfer learning radiomics model combined with explainable machine learning for preoperative risk prediction of thymoma.
This retrospective study included 173 pathologically confirmed thymoma patients from our institution in the training group and 93 patients from two external centers in the external validation group. Tumors were classified according to the World Health Organization simplified criteria as low‑risk types (A, AB, and B1) or high‑risk types (B2 and B3). Radiomics features and deep transfer learning features were extracted from venous‑phase contrast‑enhanced CT images by using a modified Inception V3 network. Principal component analysis and least absolute shrinkage and selection operator regression identified 20 key predictors. Six classifiers-decision tree, gradient boosting machine, k‑nearest neighbors, naïve Bayes, random forest (RF), and support vector machine-were trained on five feature sets: CT imaging model, radiomics feature model, deep transfer learning feature model, combined feature model, and combined model. Interpretability was assessed with SHapley Additive exPlanations (SHAP), and an interactive web application was developed for real‑time individualized risk prediction and visualization.
In the external validation group, the RF classifier achieved the highest area under the receiver operating characteristic curve (AUC) value of 0.956. In the training group, the AUC values for the CT imaging model, radiomics feature model, deep transfer learning feature model, combined feature model, and combined model were 0.684, 0.831, 0.815, 0.893, and 0.910, respectively. The corresponding AUC values in the external validation group were 0.604, 0.865, 0.880, 0.934, and 0.956, respectively. SHAP visualizations revealed the relative contribution of each feature, while the web application provided real‑time individual prediction probabilities with interpretative outputs.
We developed a CT‑based deep transfer learning radiomics model combined with explainable machine learning and an interactive web application; this model achieved high accuracy and transparency for preoperative thymoma risk stratification, facilitating personalized clinical decision‑making.
开发并验证一种基于计算机断层扫描（CT）的深度迁移学习放射组学模型，并结合可解释的机器学习方法用于胸腺瘤术前风险预测。
这项回顾性研究纳入了训练组中来自本机构的173例经病理证实的胸腺瘤患者以及外部验证组中来自两个外部中心的93例患者。根据世界卫生组织简化标准将肿瘤分为低风险类型（A、AB和B1）或高风险类型（B2和B3）。使用改良的Inception V3网络从静脉期对比增强CT图像中提取放射组学特征和深度迁移学习特征。主成分分析和最小绝对收缩和选择算子回归确定了20个关键预测因子。在五个特征集上训练了六个分类器——决策树、梯度提升机、k近邻、朴素贝叶斯、随机森林（RF）和支持向量机：CT成像模型、放射组学特征模型、深度迁移学习特征模型、联合特征模型和联合模型。使用SHapley加性解释（SHAP）评估可解释性，并开发了一个交互式网络应用程序用于实时个性化风险预测和可视化。
在外部验证组中，RF分类器在受试者工作特征曲线（AUC）下的面积达到最高值0.956。在训练组中，CT成像模型、放射组学特征模型、深度迁移学习特征模型、联合特征模型和联合模型的AUC值分别为0.684、0.831、0.815、0.893和0.910。在外部验证组中的相应AUC值分别为0.604、0.865、0.880、0.934和0.956。SHAP可视化揭示了每个特征的相对贡献，而网络应用程序提供了具有解释性输出的实时个体预测概率。
我们开发了一种基于CT的深度迁移学习放射组学模型，并结合可解释的机器学习方法以及一个交互式网络应用程序；该模型在胸腺瘤术前风险分层方面实现了高精度和透明度，有助于个性化临床决策。
