Description:

Cancer is still considered as the most common source of human illness and death globally. Lately, the number of patients classified with cancer was 18.1 million, amongst these cases breast cancer was found to be the most common of cancer. Breast cancer is a threatening malignant tumor that affects women all over the world. Recently, it has become the second leading reason of death among women after bronchus and lung cancer. Primary discovery and accurate diagnosis are the means to lower breast cancer death rates. Frequent observations and medical imaging such as; ultrasound, MRI, and mammography are essential in order to differentiate between begin lesions and malignant. However, Pathological diagnosis with histopathological imaging is preferred as it has a higher capability to offer straightforward evidence for classification assessment and experimental treatment. But pathological analysis is complex, needs much time, and is tiring. Expert pathologists may also have deviations in their diagnosis. Recently, the deep learning model is preferred for quantitative image analysis. Deep learning diagnoses the disease and prepares suitable prediction models to assist doctors in developing effective treatment plans. Therefore, the automatic and quantitative analysis of images can be done through deep learning-based approaches. One of these approaches is transfer learning, which is a sub-branch of deep learning. Transfer learning improves learning in a new task through the transfer of knowledge from a related task that has already been learned. To this end, a variety of model compression and acceleration techniques have been developed. As a representative type of model compression and acceleration, knowledge distillation effectively learns a small student model from a large teacher model. This paper presents an automatic system to classify breast cancer as benign and malignant based on knowledge distillation and transfer learning models.