Optimizing Oral Cancer Diagnosis with Advanced Deep Learning Approaches

One of the most common malignancies worldwide, oral cancer mostly affects the neck and mouth. Tobacco use is one of the elements driving its rising prevalence in many different cultures. The high death and morbidity rates linked to this malignancy are exacerbated by obstacles like delayed identification and insufficient treatment planning. If we want better prognoses, better treatment results, and higher survival rates, we need to diagnose more quickly. There is hope that machine learning (ML) approaches can improve diagnostic accuracy, which will decrease cancer-related deaths. Despite their impressive classification accuracy, current ML systems struggle with feature extraction. To overcome these shortcomings and enhance oral cancer diagnosis, this study suggests a system based on deep learning. Prior to classification, the framework does preprocessing and extracts features. Preprocessing steps include using contrast limited adaptive histogram equalisation (CLAHE) to boost contrast and image resolution. For feature extraction, which captures statistical data associated to tactile features, the Local Binary Pattern (LBP) method is utilised. This technique is well-known for its resilience to fluctuations in lighting. When it comes time for classification, a Convolutional Autoencoder Network (CAEN) is used. By surpassing current state-of-the-art methods, the suggested approach achieved a classification accuracy of 96.19% when tested on a Kaggle dataset pertaining to oral cancer. This deep learning method shows great promise for enhancing oral cancer patients' access to early detection and treatment.