Automated assessment of breast density on mammogram images based on convolutional neural networks

Inception_ResNet_V2 model is one of the most effective deep learning methods and one of the outputs of artificial intelligence in image classification as accurately as possible, as it can be relied upon to create highly accurate classification models in the context of medical images analysis. In this work, we will test the effectiveness of this model by applying it to the assessment of breast density in radiographs, based on the Breast imaging-reporting and data system (BI-RADS) standard. This method solved the problem of lack of data and training and gave us impressive results, with accuracy reaching 99% in the two-class classification and 97% in the fourth classification. This can be considered remarkable progress in the process of early detection of breast cancer.