FS-GAN: Fuzzy Self-guided structure retention generative adversarial network for medical image enhancement

Improving the quality of medical images is helpful for doctors to perform clinical diagnosis and treatment. Many medical image enhancement methods can achieve good performance when training on paired data. However, acquiring paired high/low-quality medical images for training image enhancement models is challenging, and these methods are not applicable to unpaired images while also lacking structural information preservation. To address these problems, this paper proposes a fuzzy self-guided structure retention generative adversarial network (FS-GAN), which can perform unpaired learning. Specifically, we develop a fuzzy discriminator to distinguish real images from generated images in the fuzzy domain, which is able to improve the enhanced performance of the model. Moreover, we design the self-guided structure retention module (SSRM) and illumination distribution correction module (IDCM) to capture the structure information of nerve fibers in a self-guided manner and correct the illumination distribution of the image to improve the visual effect. Compared with the existing methods, the experimental results show that the enhanced images generated by the FS-GAN exhibit the most complete texture structure and uniform illumination distribution, and FS-GAN performs well in downstream application tasks.