Histopathological Image Based Oral Squamous Cell Carcinoma Classification Using Deep Network Fusion

Oral cancer represents a pressing public health concern, posing a substantial threat to individuals due to its impact on critical anatomical structures, including the oral cavity, tongue, and lips. The continual evolution of technology has played a pivotal role in advancing disease diagnostics, and this holds true for oral cancer detection. Utilizing histopathology images as a diagnostic aid enhances the accuracy of disease identification. When coupled with deep learning methodologies, the potential for significantly improved disease prediction becomes evident, promising practical benefits in real-world scenarios. Our research dives deep into these cutting-edge techniques, systematically exploring a range of possibilities to enhance the predictive capacity for this disease. In this regard, we propose a deep fusion-based approach that combines three different models, namely Resnet50, Efficientnet-b0, and Convnext-tiny, exhibit noteworthy advancements in predictive accuracy. Our findings strongly indicate that the fused model outperforms the individual models and other methods in prediction accuracy, thus having substantial potential to advance the field of its analysis in the future.