Medical Image Segmentation Using Deep Learning: A Survey

During the last few years, medical image segmentation using deep learning has become the most active research area in computer vision. Effectively, researchers become more and more interested in this accurate technique that has a direct impact on the decisions made in different medical fields. The deep learning image segmentation success in different areas, including the medical area, enable us to have the best results. The aim of this paper is two folds, firstly, it presents a study about the most important deep learning architectures used in the medical image segmentation such as the Fully Convolutional Network (FCN), the DeepLab Family and the Convolutional networks for biomedical image segmentation (U-Net) and Generative Adversarial Networks (GANs). Secondly, it provides an analysis for each implemented model in these architectures, which allows highlighting the various common challenges between those models and their adopted approaches.