Predicting oral cancer risk in patients with oral leukoplakia and oral lichenoid mucositis using machine learning

Abstract Oral cancer may arise from oral leukoplakia and lichenoid mucositis (oral lichen planus lesions) subtypes of potentially malignant disorders. As not all patients will develop in their lifetime, the availability transformation predictive platforms would assist individualized treatment planning formulation optimal follow-up regimens for these patients. Therefore, this study aims to compare select machine learning (ML)-based models stratifying status with mucositis. One thousand one hundred eighty-seven treated at three tertiary health institutions Hong Kong, Newcastle UK, Lagos Nigeria were included study. Demographic, clinical, pathological, treatment-based factors obtained diagnosis during used populate forty-six learning-based models. These implemented as a set twenty-six predictors centers substantial data quantity fifteen insufficient data. Two best selected according number variables. We found that ML-based risk achieved an accuracy 97% 94% respectively following model testing. Upon external validation, both sensitivity, specificity, F1-score 1, 0.88, 0.67 on consecutive after construction Furthermore, 15-predictor ML reduced higher sensitivity identifying developed malignancies other settings compared binary epithelial dysplasia system stratification (0.96 vs 0.82). findings suggest could be useful stratify different settings.