HYBRID MACHINE LEARNING APPROACH FOR ORAL CANCER DIAGNOSIS AND CLASSIFICATION USING HISTOPATHOLOGICAL IMAGES
HAN THI NGOC PHAN, ARJINA AKTER · International Journal of Medical Science and Dental Health · 2025
Oral cancer remains a significant global health challenge, with early diagnosis crucial for improving patient outcomes. This study explores the integration of machine learning (ML) techniques in the detection and classification of oral cancer using histopathological images. A hybrid approach combining deep learning-based feature extraction (via pre-trained convolutional neural networks) and traditional handcrafted methods is proposed. The study uses a dataset of 10,000 annotated histopathological images, carefully preprocessed to enhance consistency and mitigate quality variations. Multiple ML models, including ResNet50 and traditional algorithms like SVM and random forests, were trained, evaluated, and validated across several performance metrics such as accuracy, precision, recall, and AUC-ROC. The models demonstrated high performance, with deep learning models showing superior classification ability. Explainability techniques, such as Grad-CAM and SHAP, were incorporated to enhance model transparency and trust. External validation and real-world simulation testing confirmed the robustness and generalizability of the system. The deployment of the models within a user-friendly software application offers a potential pathway for clinical integration, streamlining the diagnostic process for oral cancer detection.