Hybrid Deep Learning Framework for Real-Time Oral Cancer Detection and Prevention Using Multi-Model CNN Integration

Early identification is essential for increasing treatment outcomes and survival rates for oral cancer, a serious worldwide health concern. Conventional diagnostic techniques are inaccessible in environments with low resources because they are frequently costly, time-consuming, and dependent on expert opinion. Using convolutional neural networks, or CNNs, built on a dataset of pictures of malignant and non-cancerous lips and mouth areas, we offer a deep learning-based method for real-time cancer detection in order to address this issue. Our model leverages the distinct properties of several cutting-edge architectures, such as VGG-19, AlexNet, LeNet-5, ZFNet and InceptionV3, to create a pipeline for improved feature extraction and classification. The VGG-19 component ensures deep recursive feature learning by using layered Conv2D layers prior to MaxPool2D. AlexNet optimizes spatial hierarchy by contributing consecutive Conv2D and Max-Pooling layers. The resulting model showed strong generalization with 96.55% training accuracy and 94.55% test accuracy. Furthermore, a Streamlit-Based graphical interface is integrated into the system, allowing for the real-time identification of cancers in the mouth from uploaded photographs. The portal incorporates YouTube API videos to further promote awareness and prevention by giving users access to instructional materials about the different forms of oral cancer, early warning signs, and preventative strategies.