Super resolution ultrasound imaging using deep learning based micro-bubbles localization

Abstract

Super resolution ultrasound imaging has shown its potential to detect minor structures of tissues beyond the limit of diffraction and achieve sub-wavelength resolution through localizing and tracking the ultrasound contrast agents, such as micro-bubbles. Normally, one important step of super resolution ultrasound imaging, micro-bubbles localization is implemented through conventional computer vision techniques, such as local maxima detection etc. However, these classical techniques are generally time consuming and need fine-tuning multiple parameters to achieve the optimal results. Hence, in the manuscript, a deep learning based micro-bubbles localization is proposed, trying to replace or simplify the complex operations of classical methods. The efficiency of our proposed models is preliminarily proved through 2022 ultra-SR challenge.