Resolution Enhancement in Ultrasound Localization Microscopy using a Super-Resolution Generative Adversarial Network

Ultrasound localization microscopy (ULM), as an advanced imaging technique that breaks through the acoustic diffraction limit, has demonstrated great potential for visualizing microvascular structures. However, its imaging quality can be compromised under challenging conditions such as sparse microbubble distribution, insufficient frame rates, or noise interference, often resulting in image blurring and loss of fine details. To address these issues, this study introduces the application of a super-resolution generative adversarial network (SRGAN) for post-processing primary ULM images, aiming to further enhance image resolution and vascular structure clarity. SRGAN is capable of reconstructing high-frequency texture information, thereby improving the visibility and continuity of microvessels without increasing additional data acquisition costs. This study experimentally validated the method based on the PALA dataset, and the results showed that SRGAN improved the resolution and microvascular detail representation of ULM images, providing a new approach for achieving efficient and low-cost ULM imaging.