GAN-Based Ultrasound Localization Microscopy

Ultrasound localization microscopy (ULM) breaks the acoustics diffraction limit and allows the imaging of the microvasculature within organs and tumors at sub-wavelength resolution while maintaining the imaging penetration. However, the reconstruction quality of localization-based methods highly depends on sufficient ultrasound data frames with sparsely distributed microbubbles (MBs) in each frame which results in great data storage burden and low processing speed. Here, we proposed a novel method based on generative adversarial networks (GANs) to implement the MB localization in ULM imaging to accelerate the data processing speed. The synthetic results indicate that the proposed method performs well in MB localization task with great robustness to overlapping MBs and achieves higher localization speed once the network be well trained.