A Fast and Efficient Deep Learning Aided Diagnosis of Breast Cancer Using Histopathological Images

Breast Cancer (BC) is a type of intrusive illness and considered as the most prevalent disease for women that is responsible for huge number of deaths in the world. An early prognosis could boost the treatment’s efficiency and depress the fatality rate largely. Computer-Aided Diagnosis-(CAD) relies heavily on the automatic classification of BC utilizing histopathology images; however, the accuracy of the feature-based classification method is usually highly based on the accurate cell segmentation and feature extractions. Because of impurities, overlapping cells, and uneven radiation; feature extraction and accurate segmentation are still difficult. Contrarily, the existing system possesses some significant drawbacks namely; poor choice of features, and classification efficiency, computational complicatedness, imbalanced datasets, and absence of data augmentation. To overcome these difficulties, an efficient BC classification model which includes Double-Shot Transfer Learning (DSTL) and Cross-over Tanh-based Extreme Learning Machine was proposed. The proposed model was simulated tested using the BreaKHis dataset. The accuracy of about 96.67% was obtained which demonstrates the efficiency of the model.