Deep Learning for Image Segmentation: A Focus on Medical Imaging

Abstract

Image segmentation is crucial for various research areas. Many computer vision applications depend on segmenting images to understand the scene, such as autonomous driving, surveillance systems, robotics, and medical imaging. With the recent advances in deep learning (DL) and its confounding results in image segmentation, more attention has been drawn to its use in medical image segmentation. This article introduces a survey of the state-of-the-art deep convolution neural network (CNN) models and mechanisms utilized in image segmentation. First, segmentation models are categorized based on their model architecture and primary working principle. Then, CNN categories are described, and various models are discussed within each category. Compared with other existing surveys, several applications with multiple architectural adaptations are discussed within each category. A comparative summary is included to give the reader insights into utilized architectures in different applications and datasets. This study focuses on medical image segmentation applications, where the most widely used architectures are illustrated, and other promising models are suggested that have proven their success in different domains. Finally, the present work discusses current limitations and solutions along with future trends in the field.