本研究旨在建立预测食管癌肝转移的机器学习(ML)模型。我们回顾性分析监测记录的食管癌患者,流行病学,2010年至2020年的最终结果(SEER)数据库。我们通过单因素和多因素logistic回归分析确定了与肝转移风险相关的11项指标。随后,将这些指标纳入6个ML分类器,以构建相应的预测模型.使用接收器工作特征曲线下面积(AUC)评估这些模型的性能,准确度,灵敏度,和特异性。本研究共纳入17,800例诊断为食道癌的患者。年龄,主站点,组织学,肿瘤分级,T级,N级,手术干预,放射治疗,化疗,骨转移,肺转移是食管癌患者肝转移的独立危险因素。在开发的六种模型中,使用GBM算法构建的ML模型在数据集的内部验证过程中表现出最高的性能,AUC,准确度,灵敏度,特异性分别为0.885、0.868、0.667和0.888。基于GBM算法,我们开发了一个可访问的基于Web的预测工具(可访问https://project2-dngisws9d7xkygjcvnue8u。流光。app/)用于预测食管癌肝转移的风险。
This study aimed to establish a machine learning (ML) model for predicting hepatic metastasis in esophageal cancer. We retrospectively analyzed patients with esophageal cancer recorded in the Surveillance, Epidemiology, and End Results (SEER) database from 2010 to 2020. We identified 11 indicators associated with the risk of liver metastasis through univariate and multivariate logistic regression. Subsequently, these indicators were incorporated into six ML classifiers to build corresponding predictive models. The performance of these models was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity. A total of 17,800 patients diagnosed with esophageal cancer were included in this study. Age, primary site, histology, tumor grade, T stage, N stage, surgical intervention, radiotherapy, chemotherapy, bone metastasis, and lung metastasis were independent risk factors for hepatic metastasis in esophageal cancer patients. Among the six models developed, the ML model constructed using the GBM algorithm exhibited the highest performance during internal validation of the dataset, with AUC, accuracy, sensitivity, and specificity of 0.885, 0.868, 0.667, and 0.888, respectively. Based on the GBM algorithm, we developed an accessible web-based prediction tool (accessible at https://project2-dngisws9d7xkygjcvnue8u.streamlit.app/ ) for predicting the risk of hepatic metastasis in esophageal cancer.