Sign in Subscribe
Pediatrics

Machine Learning Models Show Promise in Predicting Hepatoblastoma Prognosis

Strategic Insights

— 2 min read

A research study has developed and validated four machine learning-based predictive models for the overall survival of children with hepatoblastoma, a type of liver cancer. The study, conducted by researchers from the Graduate School of Peking Union Medical College, used data from the Surveillance, Epidemiology, and End Results (SEER) database from 2000 to 2021 to train and validate the models.

The research found that the Random Survival Forest (RSF) model showed the best predictive performance, with an area under the receiver-operating characteristic curve (AUC) of 0.822, 0.810, and 0.809 for predicting 1-year, 3-year, and 5-year overall survival, respectively. The consistency index (C-index) was 0.791 (95% CI 0.667-0.813) in the training cohort, and 0.764 (95% CI 0.571-0.909) in the validation cohort.

Key Takeaways:

  • The study developed and validated four machine learning-based predictive models for hepatoblastoma prognosis using SEER database data from 2000 to 2021.
  • The RSF model showed the best predictive performance, with an AUC of 0.822, 0.810, and 0.809 for predicting 1-year, 3-year, and 5-year overall survival, respectively.
  • The C-index was 0.791 (95% CI 0.667-0.813) in the training cohort, and 0.764 (95% CI 0.571-0.909) in the validation cohort.
  • The study identified six significant prognostic factors: age, tumor size, lymph-node invasion, metastatic status, surgical therapy, and chemotherapy.
  • Shapley Additive Explanations (SHAP) plots were used to interpret the contribution of each variable to the model prediction.
  • The RSF model has broad application prospects in assisting clinicians in making individualized treatment decisions and improving survival prediction accuracy.

Statistics:

  • A total of 525 pediatric HB patients meeting the inclusion criteria were included in the study.
  • The data were randomly divided into a training cohort (n = 420) and a validation cohort (n = 105) in an 8:2 ratio.
  • The AUC for the RSF model in the training cohort was 0.822, 0.810, and 0.809 for predicting 1-year, 3-year, and 5-year overall survival, respectively.
  • The C-index was 0.791 (95% CI 0.667-0.813) in the training cohort, and 0.764 (95% CI 0.571-0.909) in the validation cohort.

Sources:

  • Prognostic prediction of hepatoblastoma in children: development and validation of machine learning models-an SEER-based study. Updates in Surgery, 2025.
  • NewsRx. Findings from Graduate School of Peking Union Medical College Broaden Understanding of Hepatoblastomas (Prognostic prediction of hepatoblastoma in children: development and validation of machine learning models-an SEER-based study). Pediatrics Week. October 25, 2025; p 521.

Read more