Abstract: Oral cancer is a severe public health problem with a high fatality rate, and early identification is critical for increasing survival rates. We propose a deep learning-based technique for detecting oral cancer utilizing two types of oral cavity pictures: normal oral images and histopathologic images in this work. We trained our models on a huge dataset of oral cavity pictures using two cutting-edge convolutional neural network (CNN) models, ResNet50 and VGG16, using transfer learning. According to our findings, ResNet50 had an accuracy of 95% and VGG16 had an accuracy of 94% in identifying oral cavity pictures as cancerous or non-cancerous.

We have integrated our model into OralScreen, an online tool that can be utilized by both patients and medical experts such as physicians and histopathologists. Our findings show that deep learning-based techniques have the potential to increase the accuracy and efficiency of oral cancer diagnosis dramatically.

Keywords: Deep Learning, CNN, ResNet50, VGG16.

PDF | DOI: 10.17148/IJARCCE.2023.125167

Open chat
Chat with IJARCCE