Multimodal AI in Oral Cancer: A PRISMA-Guided Systematic Review.
Authors
Affiliations (4)
Affiliations (4)
- Maharaja Sriram Chandra Bhanja Deo University Baripada Odisha India.
- Department of Information Technology School of Engineering and Technology Nagaland University Kohima Nagaland India.
- Centre For Intelligent Healthcare Coventry University Coventry UK.
- National Medical Research Association Leicester UK.
Abstract
Oral cancer presents significant diagnostic and prognostic challenges, often due to late-stage detection and tumour heterogeneity. Recent advancements in artificial intelligence (AI), particularly deep learning (DL) and machine learning (ML), offer promising solutions by integrating multimodal biomedical data. This PRISMA-guided systematic review evaluates over 82 studies from 2015 to 2025 that apply convolutional neural networks, transformers and graph neural networks to oral cancer, using combinations of histopathology, radiology, genomics and clinical data. Results indicate that multimodal AI models consistently outperform unimodal approaches in diagnostic accuracy, AUC and survival prediction. Architectures employing attention mechanisms, graph embedding and fusion transformers demonstrate superior performance in tasks such as early lesion detection and treatment outcome modelling. Despite these advances, challenges persist in generalisability, data standardisation and clinical validation. Multimodal AI thus represents a paradigm shift toward precision oncology in oral cancer, warranting further research focused on interpretability, prospective validation and ethical clinical deployment.