Magnification Dependent Histopathology Breast Cancer Image Binary Classification Using Different Levels Of EfficientNet
Saman Saadi Zadeh · 2021
Breast cancer kills millions of women annually around the world; early diagnosis of this perilous disease can result in efficient treatment. Computer Aided Detection systems by using advanced screening tools can automatically help the physician all over the world to detect the cancer. In our approach we are dealing with histopathological Images and state-of-art deep learning network named EfficientNet, it has been applied to BreakHis images to classify them into benign or malignant. For breast cancer classification in histopathological images using deep learning we applied eight different stages of EfficientNet that has different amounts of parameters on four diverse resolutions from our image dataset. This pre-trained Network with ImageNet weights represents high accuracy in training as well as in testing part. The best result among 32 findings in our proposed work gives the accuracy of 94.41 % for training, 99.28% while predictingand 99%Sensitivity, Recall and F1-score.