Convolutional Neural Network Models for Throat Cancer Classification Using Histopathological Images

Cancer is a condition where the abnormal growth of cells is observed, initially identified as tumors. Cancer can happen to any part of the body, early detection of cancer can increase the life expectancy of patients in this paper throat cancer is considered for the study, cancer can start at any part of the throat and later stage can spread to the lungs and liver. Samples are extracted through the fine needle biopsy method and by using Hematoxylin, Eosin (H&E) stain images are obtained with three magnitudes original, 10 $$\times$$ , 20 $$\times$$ from an electron microscope. Convolution Neural Networks (CNN) with Support Vector Machine (SVM) and Transfer Learning (TL) are used to classify the images as either cancer or normal both techniques have shown promising results with an average accuracy of 93%. Transfer learning with 10 $$\times$$ data has provided better results with an average accuracy of 94.18% as TL slightly outperformed SVM and itself with other magnitudes.