Deep Learning-Based Microbubble Localization for Ultrasound Localization Microscopy.

Xi Chen,Aiguo Han,Matthew R Lowerison,Zhijie Dong,Pengfei Song

doi:10.1109/tuffc.2022.3152225

Abstract

Ultrasound localization microscopy (ULM) is an emerging vascular imaging technique that overcomes the resolution-penetration compromise of ultrasound imaging. Accurate and robust microbubble (MB) localization is essential for successful ULM. In this study, we present a deep learning (DL)-based localization technique that uses both Field-II simulation and in vivo chickenembryo chorioallantoic membrane (CAM) data for training. Both radio frequency (RF) and in-phase and quadrature (IQ) data were tested in this study. The simulation experiment shows that the proposed DL-based localization was able to reduce both missing MB localization rate and MB localization error. In general, RF data showed better performance than IQ. For the in vivo CAM study with high MB concentration, DL-based localization was able to reduce the vessel MB saturation time by more than 50% compared to conventional localization. In addition, we propose a DL-based framework for real-time visualization of the high-resolution microvasculature. The findings of this article support the use of DL for more robust and faster MB localization, especially under high MB concentrations. The results indicate that further improvement could be achieved by incorporating temporal information of the MB data.

Highlights

AS a diffraction-limited imaging modality, the performance of ultrasound imaging has long been limited by the classical trade-off between imaging resolution and penetration depth
The performance of the deep learning (DL)-based localization was quantified on a set of Field-II simulation data with low to moderate MB concentration using models trained with group 1 low concentration data
We studied the performance of DL-based method for MB localization for Ultrasound localization microscopy (ULM), using both envelope detected (ENV) and RF data under challenging high MB concentration scenario

Summary

Introduction

AS a diffraction-limited imaging modality, the performance of ultrasound imaging has long been limited by the classical trade-off between imaging resolution and penetration depth. As an acoustic analogy to structured illumination microscopy (SIM), acoustical structured illumination was recently proposed to surpass the resolution limit by generating a series of known patterns with the transducer, which enables encoding of high-resolution information of the observed image [6]. The lateral resolution of ultrasound can be improved by null subtraction imaging (NSI), a technique that applies multiple receive apodizations to reduce sidelobes and enhance the mainlobe [7]. Superresolution of these techniques is achieved contrast-free and based on manipulation of the transmit and/or received point spread function (PSF) of the ultrasound imaging system

Methods

Results

Discussion

Conclusion