A Hybrid Transfer Learning-based Model for Histopathologic Image Classification

As the global frequency of cancer rises, developing a computer-aided diagnosis system has become crucial for improving clinical diagnosis. In this research, a hybrid transfer learning-based model named Inception-v۳-AdaBoost is proposed. Features are automatically extracted from histopathology images using Inception-v۳. The obtained feature map is used as input for the AdaBoost classifier to improve cancer classification further. Experimental studies are conducted on the proposed model using the BreakHis dataset, and its performance is compared to state-of-the-art research. The suggested hybrid model achieved the highest accuracy, precision, recall and F۱-score at ۴۰X, ۱۰۰X, ۲۰۰X and ۴۰۰X magnification levels.