Deep Convolutional Neural Networks for Cancer Detection: Faster R-CNN, U-Net and GoogLeNet

Abstract

The emergence of deep learning as one of the key technologies involved in big data analysis has brought unprecedented progress to the medical field that adopts efficient and accurate models achieving good practical results in speech recognition, visual object classifications, cancer detection, drug discovery and many other fields. The main goal of this paper is to examine the popular methods of deep learning currently in the medical field. The major methodology includes Faster region-based convolutional network (Faster R-CNN), U-Net, and GoogLeNet. These three methods show outstanding performance than other popular deep models. Also, the combination of the two of these methods will also achieve decent accuracy like Faster R-CNN combining GoogLeNet. The article expounds on how the medical field takes advantage of these three CNN architectures to detect cancer from medical images. Moreover, incorporating the computer-aided detection (CAD) systems with the three models, doctors and radiologists efficiently and accurately perform the work of diagnosing cancer.