Oral Lesions Classification using EfficientNet Transfer Learning Model

Due to their wide variety of diseases, oral lesions present a substantial diagnostic problem. This research uses deep learning techniques, particularly the EfficientNetB7 model, to present an automated categorisation analysis of oral lesions. The study divides lesions into benign and malignant categories using the Oral Lesions: Cancer Detection Dataset, comprising 2270 high-resolution pictures. Known for its effectiveness in processing large-scale image collections, the EfficientNetB7 architecture is employed in this work. The model successfully distinguishes between benign and malignant tumors with an exceptional accuracy rate of 99.12%. The study highlights the diagnostic dependability of the model by analyzing its performance, including metrics for sensitivity, specificity, and accuracy. Moreover, the study investigates how interpretable the model's predictions are, emphasizing essential aspects that support its decision-making process.