Tumor Detection and Classification of Oral Squamous Cell Carcinoma on Histopathological Images utilizing Personalized Transfer Learning Approaches
Singaraju Ramya, R. I. Minu · 2025
In the healthcare sector, automated systems utilizing artificial intelligence have emerged as the leading solutions for diagnosing various disorders in human beings. The emergence of automated systems is exerting a significant influence in the field of pathology, particularly in the precise treatment and classification of oral squamous cell carcinoma (OSCC) tumours. The current manual microscopic tissue screening procedures present several challenges, including time consumption and the complexity of tumour classification. The presence of stain artefacts in biopsy tissues is a significant obstacle to accurately classifying OSCC and normal tumours, as it results in reduced morphological variations within biopsy samples using different transfer learning (TL) models. This book presents a comparative analysis of different TL approaches utilizing feature extraction models, including ResNet50, DenseNet121, Xception, VGG16, InceptionV3, VGG19, MobileNetV2, and InceptionResNetV2. The study utilizes pre-trained models combined with a custom dense layer within a CNN framework to extract distinctive features from OSCC H&E stained images for training on a new task. Based on the simulation data, the proposed DenseNet121 model attained a classification ACC of 99% on the training set and 97.55% on the testing set.