Development and validation of machine learning models for predicting lung metastasis risk in differentiated thyroid...

Area of Science:

• Oncology
• Machine Learning
• Medical Informatics

Background:

• Differentiated thyroid cancer (DTC) typically progresses slowly.
• Lung metastasis (LM) in DTC patients is associated with a poor prognosis.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for predicting LM risk in DTC patients.
• To identify independent risk factors for LM, considering age and gender subgroups.

Main Methods:

• Utilized demographic and clinicopathological data from SEER and Zhangzhou Municipal Hospital databases.
• Developed and compared five ML models: logistic regression, random forest, decision tree, XGBoost, and GBM.
• Validated models using accuracy, precision, recall, F1 score, Brier score, ROC-AUC, PR-AUC, calibration curves, and DCA.

Main Results:

• Identified age, gender, tumor size, T stage, N stage, and histologic type as significant risk factors for LM.
• Observed variations in the impact of gender, T stage, and histology on LM risk across age subgroups.
• GBM model achieved high performance (AUROC 0.982, accuracy 0.818, F1 score 0.818) in the validation set.

Conclusions:

• The GBM model is effective for identifying high-risk DTC patients with LM.
• Findings can guide clinical practice and personalized treatment planning.
• Further validation in diverse populations and settings is recommended.