Machine learning to predict occult metastatic lymph nodes along the recurrent laryngeal nerves in thoracic esophageal squamous cell carcinoma

Abstract

Purpose Esophageal squamous cell carcinoma (ESCC) metastasizes in an unpredictable fashion to adjacent lymph nodes, including those along the recurrent laryngeal nerves (RLNs). This study is to apply machine learning (ML) for prediction of RLN node metastasis in ESCC. Methods The dataset contained 3352 surgically treated ESCC patients whose RLN lymph nodes were removed and pathologically evaluated. Using their baseline and pathological features, ML models were established to predict RLN node metastasis on each side with or without the node status of the contralateral side. Models were trained to achieve at least 90% negative predictive value (NPV) in fivefold cross-validation. The importance of each feature was measured by the permutation score. Results Tumor metastases were found in 17.0% RLN lymph nodes on the right and 10.8% on the left. In both tasks, the performance of each model was comparable, with a mean area under the curve ranging from 0.731 to 0.739 (without contralateral RLN node status) and from 0.744 to 0.748 (with contralateral status). All models showed approximately 90% NPV scores, suggesting proper generalizability. The pathology status of chest paraesophgeal nodes and tumor depth had the highest impacts on the risk of RLN node metastasis in both models. Conclusion This study demonstrated the feasibility of ML in predicting RLN node metastasis in ESCC. These models may potentially be used intraoperatively to spare RLN node dissection in low-risk patients, thereby minimizing adverse events associated with RLN injuries.

Machine learning to predict the occurrence of thyroid nodules: towards a quantitative approach for judicious utilization of thyroid ultrasonography

“…While previous studies have yielded inconclusive findings regarding the advantage of ML over traditional statistics in performing different clinical tasks ( 34 , 35 , 59 , 60 ), it is generally believed that complex ML models often require big data to achieve optimal performance ( 61 ). This study involved a substantial dataset of more than 4,000 subjects with a balanced class distribution, where all three ML models showed significant improvements over the standard logistic regression, albeit to a modest extent.…”

“…This finding reinforces the potential of ML in predicting medical or epidemiological outcomes when large datasets are available. Nonetheless, classic regression approaches may continue to play a pivotal role in these tasks by virtue of the model simplicity, which can mitigate bias and overfitting in scenarios with smaller or imbalanced datasets ( 35 , 62 ). Moreover, the superior explainability of simple regression models over complex ML algorithms may make them more suitable for predicting clinical outcomes, as explainable equations may facilitate clinical adaptation.…”

“…The LightGBM has a similar structure to the XGBoost but uses a different strategy to split the data. These ML models represent the state-of-the-art ML techniques that show remarkable outcomes in a variety of tasks (32)(33)(34)(35). Logistic regression served as a baseline model to allow unbiased performance assessment for these ML models.…”