Image Based Cancer Classification Using Metaheuristic Algorithms and Deep Transfer Learning

Computer-Aided classification of histopathological images is one of the best approaches used to achieve early and precise breast cancer diagnosis. In recent years, various researchers have been using transfer learning techniques for feature extraction. This research employs optimized deep features to address breast cancer image classification problems. This integration ensures more efficient feature selection and enhances the accuracy of the classification process. In this study, a two-stage image classification is proposed. Deep-Transfer learning (DTL) methods such as ResNet, ZF-Net, and GoogleNet extract the data from the image dataset. Next a Dragonfly Algorithm (DA) is used to select the optimal features. SVM, DT, and KNN are three distinguished classifiers implemented to evaluate the performance of the proposed model. The highest accuracy achieved with GooGLeNet-Svm is 0.991, while ZFNetSVM reaches 0.978, and ResNet-Svm attains 0.987. The experimental study results prove that the proposed method integrating DTL and the Dragonfly Algorithm, is more efficient than the existing one.