Transfer Learning-Based Model for Oral Cancer Detection

This electronic Only 29% of oral cancer cases are currently diagnosed in the early stages, even though that oral cancer has an 83% survival probability if found early. Deep learning methods can recognize patterns in the cells that make up oral cancer and help in early detection. In this paper, we provide preliminary findings from neural networks for the identification of oral cancer in microscopic pictures. We gave a thorough analysis of several models’ performance with this kind of data. In comparison to a baseline simple convolution neural network, we achieved an overall improvement of 10-15% by utilizing transfer learning approach. The enhanced value of data preparation methods with a transfer learning model for this purpose is demonstrated by ablation studies.