Artificial Intelligence and Digital Epidemiology for Population-Based Oral Cancer Risk Prediction

Abstract

Oral cancer remains a major public health problem because many patients are diagnosed after substantial disease progression. Oral squamous cell carcinoma represents the major histological form of oral cancer. Conventional oral examination remains essential for early recognition. However, screening programmes are difficult to implement uniformly across populations because specialist resources are unevenly distributed. Digital epidemiology provides an opportunity to collect population-level information through electronic health records. Artificial intelligence can analyse these heterogeneous datasets and identify patterns associated with disease risk. Recent studies have demonstrated the potential of machine learning for oral cancer risk prediction. Deep learning has also produced high diagnostic performance using clinical photographs and histopathological images. Smartphone-based systems are beginning to extend AI-assisted screening into community settings. However, current evidence remains dominated by retrospective studies. Many datasets are derived from single institutions and contain limited demographic diversity. External validation remains insufficient. Differences in image acquisition, disease prevalence, and population characteristics can influence model performance. This review discusses how artificial intelligence and digital epidemiology can be integrated into a population-based oral cancer prediction framework.