Optimized Hybrid Cascaded Approach for Accurate OSCC Detection in Histopathology Images Using Deep CNNs

Oral Squamous Cell Carcinoma (OSCC) is a common and fatal kind of oral cancer, accounting for around 3% of all cancer cases worldwide and over 330,000 deaths every year. Early detection and timely intervention are essential to managing OSCC and preventing it. Traditionally, the examination is performed by a pathologist manually examining the histopathological images to check for signs of OSCC. This approach is helpful but subjective, time-consuming, and tedious. Artificial intelligence (AI)-guided computer vision has recently become very compelling and practical for image analysis and diagnosis. Existing AI-based methods achieve sufficient accuracy at the cost of high computing resources and large datasets. This paper proposes a cascaded network incorporating deep learning and traditional machine learning approaches with a principle components analysis (PCA) algorithm for OSCC detection from histopathology images. The proposed method achieved perfect scores in accuracy, precision, recall, and F1-score (100%), with faster image processing time. The AI models for the cascaded networks were selected through an exhaustive search in which five popular CNN models were used for extracting features; the PCA was used to determine optimal features, and seven traditional machine learning models were used to detect the OSCC. The MobileNetV2-PCA-LR cascaded network was found to be most suitable in this study. The proposed cascaded network brings efficiency, accuracy, scalability, and robustness to OSCC screening.