Detection of Oral cancer using Image analysis

Early detection and accurate classification of oral cancer are crucial for improving patient prognosis and guiding effective treatment strategies. This study proposes a comprehensive framework combining image processing techniques and deep learning models for oral cancer classification and localization. ResNet-101, a deep convolutional neural network, is employed to classify oral images into five clinically relevant categories: No lesion, no referral needed, refer for other reasons, refer – low risk OPMD, and Refer – cancer/high risk OPMD. In parallel, YOLOv2, a real-time object detection model, is utilized to accurately detect and localize cancerous regions within oral images, facilitating precise visualization of affected areas. The system’s performance is evaluated using standard metrics, including accuracy, precision, recall, and F1-score, to ensure both classification reliability and detection robustness. Experimental results demonstrate that integrating deep feature extraction through ResNet-101 with YOLOv2-based localization improves diagnostic efficiency and reduces misclassification risk. The proposed approach not only aids clinicians in early identification of potentially malignant oral disorders (OPMDs) but also provides a scalable framework for automated oral cancer screening. By combining classification and detection in a unified deep learning pipeline, this study highlights the potential of AI-driven techniques to enhance clinical decision-making and support timely, targeted interventions in oral oncology.

Index Terms—Composite annotation, deep learning, image classification, object detection, oral cancer, oral potentially malignant disorders