AI driven oral cancer diagnosis from clinical imaging and computational pathology to precision therapeutics

This is the authors' abstract. We don't add key points for this paper.

Artificial intelligence (AI) is emerging as a computational approach for analysing the complex and heterogeneous data generated across oral cancer diagnosis and management. This narrative review provides an integrated overview of AI applications from early oral lesion detection, computational pathology, prognostic assessment, and precision therapeutics. It describes how machine learning and deep learning process clinical photographs, digitized histopathological images, whole-slide images, radiological examinations, and molecular or clinical data to generate task-specific outputs. The review covers AI-assisted identification of suspicious oral lesions, histopathological detection of oral epithelial dysplasia and oral squamous cell carcinoma, assessment of tumor differentiation, characterization of disease progression, and prediction of treatment response for therapeutic stratification. Particular attention is given to the relationship between the input data, computational task, and clinically relevant output rather than to model architecture alone. Although several studies report high diagnostic performance, much of the current evidence remains based on retrospective datasets and internal validation, while external, multicentre, and prospective evaluations are comparatively limited. Major challenges include dataset quality and representativeness, domain shift, annotation variability, interpretability, and integration into clinical workflows. Future progress will require patient-level evaluation, multicentre validation, multimodal integration, uncertainty-aware prediction, and prospective assessment of clinical utility. This review outlines the evolving role of AI as a potential decision-support framework across the oral cancer care pathway.