Automated Oral Cancer Detection using Convolutional Neural Networks and Support Vector Machines

Abstract

Oral cancer is a significant public health concern, demanding early detection and intervention for improved patient outcomes. In this study, we propose an automated method for oral cancer detection leveraging state-of-the-art deep learning techniques. Convolutional Neural Networks (CNNs), specifically the ResNet18 architecture [3, 12], are employed for feature extraction from oral images, followed by classification using Support Vector Machines (SVMs). The dataset comprises a collection of oral images encompassing various stages and types of oral cancer [17]. Our methodology involves preprocessing steps to standardize image dimensions and augment the dataset. The ResNet18 model is utilized to extract discriminative features from the images, which are subsequently fed into an SVM classifier for binary classification distinguishing between cancerous and non-cancerous oral images [2, 20]. The evaluation of our proposed approach demonstrates promising results in automated oral cancer detection. Performance metrics, including accuracy, sensitivity [15], and specificity, exhibit commendable levels, suggesting the effectiveness of the combined ResNet18-SVM methodology. Comparative analyses against existing methods underscore the potential of our approach in facilitating early and accurate oral cancer diagnosis [7, 9]. The implications of automated oral cancer detection are far-reaching, with the potential to revolutionize clinical practices by enabling prompt interventions and improving patient prognosis. Future research directions encompass exploring diverse CNN architectures, integrating multi-modal data sources, and refining the proposed methodology for enhanced diagnostic precision [8,14]. This study signifies a significant stride towards automated oral cancer detection, laying the groundwork for leveraging advanced deep learning techniques in the realm of medical image analysis for improved healthcare outcomes [1, 5].