A comparative ensemble approach of deep learning models for binary and multiclass classification of histopathological images for breast cancer detection

Abstract Breast cancer (BC) is the most frequently occurring cancer in women after lung cancer. There are different stages of breast cancer. Among them, Invasive ductal BC causes the maximum number of deaths in women. Different radio imaging techniques are available to diagnose this fatal disease. With the development of different radiographic imaging techniques, it is now possible to identify this fatal condition early on. However, qualified radiologists and pathologists must analyze the radiographic and Histopathological images. The procedure is expensive and prone to mistakes, as radiologists and pathologists are human beings. In this paper, three deep learning models such as Vision Transformer (ViT), Convmixer and Visual Geometry Group-19 (VGG-19), are proposed for the detection and classification of different breast cancer tumours using Breast cancer histopathological (Break His) image database. The performance of each of these models is evaluated using an 80:20 training scheme and measured in terms of accuracy, precision, recall, loss, F1 score and area under the curve. From the simulation result, we found that ViT performs best for binary classification of breast cancer tumours with accuracy, precision, recall and F1-score of 99.89%,98.29%,98.29% and 98.29%, respectively. Also, ViT gives the best performance in terms of accuracy, Precision, recall and F1-score 98.21%, 89.84% and 89.97%, respectively, for eight class classifications of breast histopathological images. Then, we have an ensemble ViT-Convmixer model for detecting breast cancer and observe that the ensemble model's performance degrades compared to the ViT model with an accuracy of 95 and 85 percent. We have also compared the performance of the proposed best model with the other existing models. The proposed model can also detect other diseases with improved accuracy.