An Interpretable Machine-Learning Model for Predicting Occult Central Lymph Node Metastasis in Papillary Thyroid
Zhongyu Wang1, Shangman Yang2,3, Yin Li4
1Department of Head and Neck Surgery, State Key Laboratory of Oncology in South China, Guangdong Provincial Clinical Research Center for Cancer, Sun Yat-sen University Cancer Center, Guangzhou 510060, P. R. China.
The Journal of Clinical Endocrinology and Metabolism
|December 11, 2025
Summary
This study identifies RET fusion positivity as a key risk factor for occult lymph node metastasis in papillary thyroid cancer. A machine learning model aids in predicting this risk for better treatment planning.
Area of Science:
- Oncology
- Medical Imaging
- Genetics
Background:
- Accurate prediction of occult lymph node metastasis (OLNM) in clinically lymph node negative (cN0) papillary thyroid carcinoma (PTC) is crucial for treatment optimization.
- Strategies like thermal ablation and active surveillance require precise OLNM risk assessment.
Purpose of the Study:
- To identify independent risk factors for OLNM in cN0 PTC.
- To develop and evaluate machine learning models for predicting OLNM using clinical, ultrasonographic, and molecular features.
Main Methods:
- Retrospective analysis of 961 cN0 PTC patients.
- Multivariate logistic regression to identify risk factors.
- Development and validation of eight machine learning models, including random forest, with SHAP for interpretability.
Main Results:
- RET fusion and BRAF mutations identified as independent molecular risk factors for OLNM.
- The random forest model showed optimal performance (AUC 0.906 training, 0.733 test).
- Tumor size, age, and echogenic foci were top predictors identified by SHAP analysis.
Conclusions:
- RET fusion positivity is a novel independent risk factor for OLNM in cN0 PTC.
- The developed random forest model offers a framework for integrating diverse data for OLNM risk prediction.
- A web calculator is available for practical application of the predictive model.
