Machine learning predicts lymph node metastasis in early-stage oral tongue squamous cell carcinoma

Abstract: Purpose Early-stage oral tongue squamous cell cancer (OTSCC) has a rate of 20-50% metastasis to cervical lymph nodes. This study aimed to build and validate four machine learning (ML) models to predict the occurrence of lymph node metastasis before and after surgery for early-stage (cT1-cT2N0) OTSCC. Materials and Methods We designed a retrospective cross-sectional study and reviewed the clinical and pathological records of early-stage OTSCC patients. The sample was comprised of two groups with different node status (‘positive’ or ‘negative’) and randomly split into training (70%) and testing (30%) sets. Four common ML algorithms, including Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM), and Naive Bayes (NBs), were employed to predict pathologic nodal metastasis of early-stage OTSCC. Area under the receiver operating characteristic (ROC) curve (AUC), sensitivity and specificity were used to assess the performance of these models and conventional methods including depth of invasion (DOI), neutrophil-to-lymphocyte ratio (NLR) and tumor budding. Results 145 patients (56 with positive lymph node and 89 negative) were included in the current study. The performance of ML models was significantly superior to conventional prediction methods. The RF model performed best (AUC=0.786, sensitivity=85%, specificity=75%), and exceeded that of NLR (AUC=0.539, sensitivity=53.6%, specificity=53.9%, P = 0.003). When DOI, worst pattern of invasion (WPOI), lymphocytic host response (LHR) and tumor budding were added to models analysis according to patients’ postoperative pathological records, the SVM model performed best (AUC=0.956, sensitivity=100%, specificity=87.5%), and was superior to univariate assessment of tumor budding (AUC=0.830, sensitivity=80.9%, specificity=87.5%, P=0.002), DOI (AUC=0.613, sensitivity=91.1%, specificity=31.5%, P Conclusions ML shows better performance in predicting lymph node metastasis of early-stage OTSCC than conventional prediction methods of DOI, NLR or tumor budding.