Scholarlink

<p>Hepatoblastoma (HB) is the most common primary malignant liver tumor in children. Although the incidence is low, it is a serious threat to children’s health. Traditional methods have limitations in prognosis prediction. In the era of precision medicine, machine learning (ML) shows great potential in medical fields, especially in prognosis prediction. This study aims to develop ML-based predictive models for the overall survival (OS) of children with HB, to improve prognosis assessment and quality of life. Data from the Surveillance, Epidemiology, and End Results (SEER) database from 2000 to 2021 were used. A total of 525 pediatric HB patients meeting the inclusion criteria were included. The data were randomly divided into a training cohort (<i>n = </i>420) and a validation cohort (<i>n = </i>105) in an 8:2 ratio. Four ML algorithms, including Decision Tree, Random Survival Forest (RSF), Gradient Boosting Survival Analysis (GBSA), and Support Vector Machine (SVM), were used to evaluate the OS of HB patients. The performance of the models was assessed using the area under the receiver-operating characteristic curve (AUC) and the consistency index (C-Index). Shapley Additive Explanations (SHAP) plots were used to interpret the contribution of each variable to the model prediction. The basic characteristics of the patients were analyzed. Through feature variable selection, six variables (age, tumor size, lymph-node invasion, metastatic status, surgical therapy, and chemotherapy) were identified as significant prognostic factors. Among the four ML models, the RSF model showed the best predictive performance. In the training cohort, the AUC for predicting 1-year, 3-year, and 5-year OS was 0.822, 0.810, and 0.809, respectively, and the C-index was 0.791 (95% CI 0.667–0.813). In the validation cohort, the AUC was 0.740, 0.765, and 0.765, and the C-index was 0.764 (95% CI 0.571–0.909). The SHAP summary plot showed that surgical therapy, metastatic status, and tumor size were the most important variables. This study successfully developed and validated four ML-based predictive models for the prognosis of HB patients. The RSF model had superior predictive performance and has broad application prospects in assisting clinicians in making individualized treatment decisions, improving survival prediction accuracy, and optimizing the prognosis of HB patients.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Prognostic prediction of hepatoblastoma in children: development and validation of machine learning models—an SEER-based study

  • 期刊论文

  • 发表日期：

    2025年10月9日

查看全文
  • Jianfeng Luo,
  • Kaikun Huang,
  • Le Zheng,
  • Long Li

摘要

Hepatoblastoma (HB) is the most common primary malignant liver tumor in children. Although the incidence is low, it is a serious threat to children’s health. Traditional methods have limitations in prognosis prediction. In the era of precision medicine, machine learning (ML) shows great potential in medical fields, especially in prognosis prediction. This study aims to develop ML-based predictive models for the overall survival (OS) of children with HB, to improve prognosis assessment and quality of life. Data from the Surveillance, Epidemiology, and End Results (SEER) database from 2000 to 2021 were used. A total of 525 pediatric HB patients meeting the inclusion criteria were included. The data were randomly divided into a training cohort (n = 420) and a validation cohort (n = 105) in an 8:2 ratio. Four ML algorithms, including Decision Tree, Random Survival Forest (RSF), Gradient Boosting Survival Analysis (GBSA), and Support Vector Machine (SVM), were used to evaluate the OS of HB patients. The performance of the models was assessed using the area under the receiver-operating characteristic curve (AUC) and the consistency index (C-Index). Shapley Additive Explanations (SHAP) plots were used to interpret the contribution of each variable to the model prediction. The basic characteristics of the patients were analyzed. Through feature variable selection, six variables (age, tumor size, lymph-node invasion, metastatic status, surgical therapy, and chemotherapy) were identified as significant prognostic factors. Among the four ML models, the RSF model showed the best predictive performance. In the training cohort, the AUC for predicting 1-year, 3-year, and 5-year OS was 0.822, 0.810, and 0.809, respectively, and the C-index was 0.791 (95% CI 0.667–0.813). In the validation cohort, the AUC was 0.740, 0.765, and 0.765, and the C-index was 0.764 (95% CI 0.571–0.909). The SHAP summary plot showed that surgical therapy, metastatic status, and tumor size were the most important variables. This study successfully developed and validated four ML-based predictive models for the prognosis of HB patients. The RSF model had superior predictive performance and has broad application prospects in assisting clinicians in making individualized treatment decisions, improving survival prediction accuracy, and optimizing the prognosis of HB patients.