Comparative Analysis of CNN Architectures for Enhancing Oral Cancer Detection Using Advanced Image Processing Techniques

To effectively diagnose oral cancer, sophisticated technologies like deep learning must be investigated. This is still a significant difficulty in medical diagnostics. To assess Convolutional Neural Network (CNN) architectures' efficacy in oral cancer detection, we compare and contrast them in this research. Specifically, we compare the performance of VGG19, ResNet50, ResNet101, and a standard CNN model using a dataset of Oral Cancer (Lips and Tongue) images (OCI). We use sophisticated image processing methods including data augmentation, contrast improvement, and noise reduction to improve the quality of the supplied data. We use accuracy, precision rate, recall rate, and F1-score as our assessment measures. Our findings show that the suggested CNN design outperforms VGG19 (72.97%), ResNet50 (79.73%), and ResNet101 (71.62%) in terms of accuracy, with a maximum of 87.10%. Moreover, our approach shows excellent recall, F1score, and accuracy, highlighting its efficacy in correctly detecting malignant areas in oral pictures. Through this comparative analysis, we provide valuable insights into the suitability of different CNN architectures for oral cancer detection, without the integration of specific optimization algorithms. Our findings highlight the potential of deep learning techniques to improve treatment outcomes and early diagnosis for patients with oral cancer, ultimately leading to the advancement of computer-aided cancer detection (CAD) systems.