Technique for Breast Cancer Classification Using Semi-supervised Deep Convolutional Neural Networks with Transfer Learning Models

Breast cancer affects several women worldwide every year. The survival rate of breast cancer depends on numerous factors. Early diagnosis and treatment are the most practical approaches to managing this disease. Deep learning-assisted cancer diagnosis is an effective technology for doctors to detect breast cancer quickly. Here, we propose a novel deep convolutional neural network-based transfer learning model for the accurate and most effective classification of breast cancer among women. This model is built using the pre-trained model. Inception-V3Net. First, the model is built with binary classification and then utilized to classify breast cancer histopathological images on a multi-class basis. The highest average accuracy attained by the proposed model is 94.8% when assessed under various magnifying factors. The final outcome of the proposed approach proves it to be more reliable and robust than the existing models. The proposed semi-supervised model of breast cancer classification is validated using the BreakHis public dataset. The result of the proposed CNN-built deep transfer learning approach is found to be better than the existing methods.