Comparison of two Deep Learning Methods for Classification of Dataset of Breast Ultrasound Images

Breast cancer is the prominent cancer types which account to high mortality. Early detection could serve to improve clinical outcomes. Ultrasonography is a digital imaging technique which is used for differentiating benign and malignant tumors. Recently a dataset of breast ultrasound image has been released which gives opportunity to apply machine learning methods over it and enable one to automatically classify them in the respective tumor class. In the present study, I compared the performance of two deep learning methods: AlexNet, and MobileNet for the classification of breast ultrasound images using an augmented data set of 12000 images (6000 each of benign, and malignant) and received an F1-score of 0.914 and 0.974 respectively so herewith concluded that the later deep learning approach performs better then the former in classifying the ultrasound images of breast tissues in benign and malignant cancer types.