Development of a supervised machine learning model to predict recurrence of oral tongue squamous cell carcinoma

Objective : Despite diagnostic advancements, development of reliable prognostic systems for assessing risk of cancer recurrence still remains a challenge. In this study, we developed a novel framework to leverage the expansive Surveillance, Epidemiology, and End Results (SEER) database to generate highly representative machine learning prediction models for oral tongue squamous cell carcinoma (OTSCC) cancer recurrence. Materials and Methods : Using our framework, we identified cases of 5- and 10-year OTSCC recurrence from the SEER database. Cases were split into training (80%) and test (20%) sets for model development and testing. Four classification models were trained by using the H2O artificial intelligence platform, whose performances were assessed according to their accuracy, recall, precision, and the area under the curve (AUC) of their receiver operating characteristic (ROC) curves. By evaluating Shapley additive explanations contribution plots, feature importance was studied. Results: Of 130,979 patients, 36,042 (27.5%) were female and the mean (SD) age was 58.2 (13.7) years. The Gradient Boosting Machine model performed the best, achieving 81.8% accuracy, 0.75 AUC, 83.0% recall, and 97.7% precision for 5-year prediction. Moreover, 10-year predictions demonstrated 80.0% accuracy, and 94.0% precision. The number of prior tumors, patient age, site of cancer recurrence, and tumor histology were the most significant predictors. Conclusion : Implementation of our novel SEER framework enabled successful identification of patients with OTSCC recurrence, with which highly accurate and sensitive prediction models were generated. Thus, we demonstrated our framework’s potential to be applied to various cancers for building generalizable screening tools to predict tumor recurrence.