QULM-DL: Quantitative Ultrasound Localization Microscopy via Deep Learning

Abstract

Ultrasound localization microscopy (ULM) has been developed in recent years to significantly improve the spatial resolution of ultrasound imaging by localizing the microbubbles (MBs) flowing in microvasculature. Nevertheless, challenges remain in ULM. In our previous work (IEEE Trans. Med. Imaging, 10.1109/TMI.2020.2986781), by using a modified sub-pixel convolutional neural network (CNN), we have implemented ULM with fast data-processing speed, high imaging accuracy, short data-acquisition time, and high flexibility characteristics. However, the signal intensities of MBs are not quantitatively recovered, as similar to many other ULM techniques. To overcome this limitation, in this work, we propose a new deep-learning (DL) method, termed as qULM-DL, to recover the true distributions (positions and intensities) of the MBs in ULM. A series of numerical simulations and in vivo experiment are performed to evaluate the performance of qULM-DL. The results show that qULM-DL can not only super-resolve the complex structures but also quantitatively recover MBs intensities in these structures.