Machine Learning: A Multicenter Study on Predicting Lateral Lymph Node Metastasis in cN0 Papillary Thyroid Carcinoma.

Area of Science:

• Oncology
• Surgical Oncology
• Medical Informatics

Background:

• The need for prophylactic lateral neck dissection in clinically node-negative (cN0) papillary thyroid carcinoma (PTC) is debated.
• Accurate prediction of lymph node metastasis (LNM) is crucial for treatment planning.

Purpose of the Study:

• To compare the efficacy of traditional nomograms with machine learning (ML) models in predicting LNM in cN0 PTC.
• To evaluate ML models for predicting ipsilateral lateral neck LNM, specifically in levels II, III, and IV.

Main Methods:

• 1616 PTC patient records from Hospital A (training/testing) and 243 from Hospital B (validation) were analyzed.
• Eight ML models, including Random Forest (RF), were developed and validated using 10-fold cross-validation.
• Model performance was assessed using metrics like accuracy, AUC, specificity, and sensitivity, with the best ML model compared against traditional nomograms.

Main Results:

• The RF model demonstrated superior performance, achieving high accuracy and AUC in predicting ipsilateral lateral LNM.
• A streamlined RF model, utilizing key features like central LNM and extrathyroidal extension, maintained strong predictive power.
• RF models consistently outperformed traditional nomograms across various prediction tasks and were implemented as a web-based calculator.

Conclusions:

• Machine learning, especially RF, offers a reliable method for predicting lateral LNM in cN0 PTC patients.
• ML models surpass traditional nomograms in accuracy and clinical utility for guiding neck dissection decisions.
• The developed ML tools can aid clinicians in optimizing treatment strategies for PTC.