Prediction model of ipsilateral level II lymph node metastasis in papillary thyroid carcinoma.

Objectives: This study aimed to develop a predictive model for ipsilateral level II lymph node metastasis (LNM) in patients with papillary thyroid carcinoma (PTC) using machine learning techniques. The necessity of level II dissection in lateral neck dissection (LND) remains debated, and accurate prediction of metastasis at this level could help refine surgical decision-making and minimize unnecessary dissection.

Methods: A retrospective review of 138 patients with PTC who underwent initial LND with curative intent was performed. Preoperative patient background and imaging findings were analysed to identify factors associated with ipsilateral level II LNM. Decision trees (DT), random forests (RF) and support vector machines (SVM) were trained using a 70:30 data split and 10-fold cross-validation. Model performance was assessed using area under the receiver operating characteristic curve (AUC) and Brier score.

Results: Ipsilateral level II LNM was present in 55 patients (39.9 %); the DT model identified significant predictors: level II LNM ≥15 mm, multiple level III lymph nodes suspicious for metastases preoperatively (LNSM), superior pole extension, level III/IV LNSM <18 mm (AUC: 0.831, Brier score: 0.140). RF and SVM showed improved predictive performance (RF: AUC 0.901, Brier score 0.124; SVM: AUC 0.929, Brier score 0.110). Features of high importance in RF and SVM were similar to those in DT.

Conclusions: This study highlights the potential of machine learning-based models in predicting ipsilateral level II LNM in PTC patients and contributes to a more personalized approach to LND. The findings support the selective omission of ipsilateral level II dissection in carefully evaluated cases, which may reduce surgical morbidity without compromising oncologic outcomes.