Generative Adversarial Networks in the Field of Medical Image Segmentation

This paper provides a comprehensive review of image segmentation techniques based on generative adversarial networks (GANs). For completeness, the traditional image segmentation techniques are also discussed briefly. In addition, the paper delves into the principles of GANs and highlights representative models. Various methods of image segmentation utilizing GAN models are thoroughly examined and categorized into three main groups: segmentation regions, imaging modalities, and classification methods. The essential GAN-based segmentation techniques are elucidated, shedding light on their core concepts. This three-tiered classification system not only facilitates a comprehensive understanding of the diverse approaches within GAN-based medical image segmentation but also serves as a valuable framework for researchers and practitioners in navigating the evolving landscape of this specialized field. To assess the effectiveness of these segmentation techniques, the paper relies on reported results from researchers using benchmark datasets such as Brain Tumor Segmentation (BraTS), Structured Analysis of the Retina (STARE), Digital Retinal Images for Vessel Extraction (DRIVE), Child Heart and Health Study in England (CHASE), Medical Image Computing and Computer-Assisted Intervention (MICCAI)-2017, PROMISE12, INBreast, Alzheimer's Disease Neuroimaging Initiative (ADNI), Lung Image Database Consortium (LIDC), PDDCA, and CHAOS2019. The analysis of these results indicates a discernible potential for the development of novel GAN-based architectures that can enhance the performance of medical image segmentation. The paper thus underscores the need for further exploration and innovation in this domain to address existing limitations and optimize the outcomes of image segmentation in medical contexts.