Enhancing Oral Cancer Diagnosis through Attention Mechanisms and Explainable AI: A VGG-19 with CBAM Integration

Oral cancer remains a significant global health concern, particularly due to challenges in early detection and diagnosis. This research introduces a new approach that leverages deep learning techniques that integrates the VGG-19 and convolutional neural network with the CBAM (Convolutional Block Attention Module) to enhance feature representation and emphasize diagnostically salient regions in oral imagery. To further ensure model transparency and clinical interpretability, we incorporate advanced Explainable Artificial Intelligence (XAI) techniques, including LIME, SHAP, and Grad-CAM, which generate both global and localized visual explanations for the model’s predictions. The proposed model was trained and rigorously evaluated on the benchmark Oral Cancer (Lips and Tongue) Images Dataset, publicly available on Kaggle. Our attention-augmented and explainability-enhanced model achieved a notable classification accuracy of 96.30%, precisely identifying and differentiating between cancerous and non-cancerous oral images. The integration of CBAM and XAI not only improves diagnostic accuracy but also supports transparency and trust, aligning with clinical demands for interpretable AI in medical diagnostics.