Effective Breast Cancer Detection using Novel Deep Learning Algorithm

Abstract

Ultrasound is one of the most common screening tools for breast cancer detection. However, the lack of qualified radiologists causes the diagnosis process to become a challenging task. Deep learning's promising achievement in various computer vision problems inspires us to apply the technology to medical image recognition problems. We propose a detection model based on the Rapid-CNN to detect breast cancer quickly and accurately. We conduct this experiment by collecting breast cancer datasets, pre-processing, training models, and evaluating the model performance. This model can detect breast cancer with bounding boxes based on the experiment result. In this model, it is possible to detect the bounding box that is more than what it should be, so we applied NMS  to eliminate the prediction of the bounding box that is less precise to increase accuracy.