A Proposed Heterogeneous Ensemble Algorithm Model for Predicting Central Lymph Node Metastasis in Papillary Thyroid...

Area of Science:

• Oncology
• Medical Informatics
• Machine Learning

Background:

• Papillary thyroid cancer (PTC) management involves controversial decisions regarding prophylactic central lymph node dissection.
• Accurate prediction of central lymph node metastasis (CLNM) is crucial for personalized treatment strategies.

Purpose of the Study:

• To develop and validate a heterogeneous ensemble algorithm for precise prediction of CLNM in PTC patients.
• To provide a data-driven reference for the clinical dilemma of prophylactic central lymph node dissection.

Main Methods:

• A retrospective study included PTC patients undergoing initial thyroid resection (2014-2018).
• Eighteen clinical and ultrasound (US) variables were analyzed to develop a heterogeneous ensemble model using extreme gradient boosting, k-nearest neighbors, random forest, gradient boosting, and AdaBoost.
• Partial dependent plots were employed for model interpretability.

Main Results:

• The heterogeneous ensemble model achieved an area under the receiver operating characteristic curve of 0.67, significantly outperforming individual machine learning models and US.
• Key predictors for CLNM included younger age (≤33 years), larger tumor size (≥0.8 cm), US-suspected CLNM, and microcalcification.
• Favorable factors for CLNM were identified as positive anti-thyroid peroxidase antibody and serum thyroglobulin levels.

Conclusions:

• The developed heterogeneous ensemble algorithm demonstrates potential as an optimal tool for predicting CLNM in PTC.
• Integrating clinical and US features enhances predictive accuracy, supporting informed clinical decision-making regarding central lymph node dissection.