Related Experiment Video
Updated: Jun 30, 2026
07:13
Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model
Published on: April 18, 2025
Development and validation of an interpretable machine learning model for predicting central lymph node metastasis in
Li Zhou1, Wei-Ping Lu1, Heng-Lu Zhang1
1Department of Endocrinology and Metabolism, The Affiliated Huaian No.1 People's Hospital of Nanjing Medical University, Huaian, Jiangsu, China.
Frontiers in Oncology
|June 29, 2026
Summary
A new machine learning model accurately predicts central lymph node metastasis (CLNM) in papillary thyroid carcinoma (PTC) using blood markers and clinical data. This tool aids surgical decisions and personalized medicine for PTC patients.
Area of Science:
- Oncology
- Medical Informatics
- Biostatistics
Background:
- Papillary thyroid carcinoma (PTC) is the most common thyroid malignancy.
- Central lymph node metastasis (CLNM) is a key prognostic factor in PTC.
- Accurate prediction of CLNM is crucial for guiding surgical extent and treatment strategies.
Purpose of the Study:
- To develop and validate a machine learning (ML) model for predicting CLNM in PTC patients.
- To utilize routine blood test-derived inflammatory markers and clinical features for prediction.
- To enhance the interpretability and clinical utility of the prediction model.
Main Methods:
- Retrospective analysis of 1,697 PTC patients, divided into training (70%) and validation (30%) sets.
- Collection of clinical variables and inflammatory markers: neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, lymphocyte-to-monocyte ratio (LMR), and systemic immune-inflammation index (SII).
- Feature selection using LASSO regression, followed by development and comparison of eight ML algorithms; optimal model selected based on discrimination, calibration, and clinical utility; SHAP analysis for interpretability.
Main Results:
- Central lymph node metastasis (CLNM) observed in 30.4% of patients.
- LASSO regression identified 29 predictive features.
- The Stacking Ensemble ML model demonstrated superior performance (AUC: 0.988 training, 0.923 validation), significantly outperforming logistic regression (AUC: 0.721).
- SHAP analysis highlighted maximum tumor size, LMR, age, VEGF, and SII as key predictors.
- Decision curve analysis confirmed substantial clinical benefit.
Conclusions:
- The developed ML model effectively predicts CLNM in PTC patients using accessible data.
- The model integrates inflammatory markers and clinical features, offering high predictive accuracy.
- This tool supports informed surgical decision-making and advances precision medicine in PTC management.
