Title: Redefining oral cancer detection: The role of AI-based biotechnological methods beyond conventional technologies—a systematic review
Abstract:
Oral cancer, particularly oral squamous cell carcinoma (OSCC), remains a major global health concern, with delayed diagnosis contributing substantially to morbidity and mortality. Conventional clinical examination, imaging and histopathological assessment remain the cornerstone of diagnosis but may have limitations related to accessibility, sampling, subjectivity and recognition of early molecular changes. Recent advances in biotechnology have enabled the identification of molecular, biochemical, metabolic and spectral signatures associated with oral carcinogenesis. Integration of these technologies with artificial intelligence (AI) may facilitate automated and non-invasive detection of oral cancer.
Objective: This systematic review evaluated the diagnostic performance of AI-based biotechnological methods for oral cancer detection beyond conventional diagnostic technologies and examined differences in performance according to the underlying biotechnology and AI methodology.
Materials and Methods: A systematic review was designed according to PRISMA 2020 principles. PubMed/MEDLINE, Scopus, Web of Science, Embase, Google Scholar and IEEE Xplore were searched from January 2010 to June 2026. Studies evaluating AI or machine-learning algorithms applied to molecular, salivary, genomic, epigenomic, metabolomic, proteomic, spectroscopic, biosensor or multimodal biotechnological data for oral cancer or oral potentially malignant disorder detection were considered. Studies involving only conventional photographs, radiographs or histopathological images were excluded. Risk of bias was assessed using QUADAS-2 with additional AI-specific domains. Diagnostic accuracy was synthesized using hierarchical random-effects models where appropriate.
Results: The search resulted in the identification of 1,284 records and 38 studies were eligible and 24 studies were included in the research. Included studies involved molecular and salivary biomarker evaluation, genomic and epigenomic signatures, metabolomics, proteomics, spectroscopy, biosensors and multimodal AI based approaches. Pooled sensitivity of AIbased biotechnological methods used for the detection of oral cancer was 0.91 (95% CI: 0.87–0.94), and pooled specificity was 0.89 (95% CI: 0.85–0.92). There was a significant between-study variability. Molecular biomarkers and multimodal approaches were the most successful categories of biotechnology in their diagnostic performance. A small number of studies reported external validation, hinting at the lack of generalizability of the existing AI models.
Conclusion: There is significant potential in the use of AI approaches to biotechnology for extending oral cancer diagnosis beyond imaging-based technologies. The molecular biomarkers, the salivary diagnostics, the metabolomics, the spectroscopy and biosensing seem to be the most promising. However, larger multicentre studies, standardised biological sampling, transparency of AI reporting and independent external validation are needed for clinical translation.
Keywords: Artificial Intelligence, Oral Cancer, Oral Squamous Cell Carcinoma, Biotechnology, Molecular Biomarker, Saliva, Metabolomics, Spectroscopy, Biosensors.
