Computational Super-Resolution for Ultrasound Localization Microscopy Through Solving an Inverse Problem.

Ultrasound Localization Microscopy (ULM) enables to accurately characterize micro-vascular structures by means of ultrasound imaging. In ULM, the backscattering signals of injected microbubbles (MBs) are utilized to sub-wavelength localize and track MBs flowing in the circulatory system. To ensure precise MB localizations, ULM is constrained to low MB concentrations, leading to prolonged acquisition times. Owing to patient comfort, motion, and computational cost to elaborate data, the constraint on low MB concentrations is a bottleneck for ULM clinical translation. To mitigate the need of low MB concentration, we propose to uncouple a bi-disperse MB population. The bi-disperse population is composed of two monodisperse MB populations having diameter (and resonance frequency) of 2.5 μm (3 MHz) and 3.8 μm (5.5 MHz), respectively. The different diameter determines a different resonance response. Experiments are performed injecting the populations singularly and simultaneously in a 3-D printed phantom. Uncoupling is performed by means of a signal processing pipeline, which exploits the strong nonlinear response of MBs having resonance frequency tuned with the transmission frequency. After uncoupling, super-resolved density and velocity flow maps are generated for each MB population. The results demonstrate the capability of uncoupling the selected pair of MB populations, thus potentially permitting increased MB concentrations.

Ultrasound Localization Microscopy (ULM) is a milestone in the medical vascular imaging context, enabling the precise characterization of micro-vascular structures using ultrasound imaging. By accurately localizing contrast microbubbles (MBs) flowing in the circulatory system, ULM generates micro-resolved vascular images, overcoming the ultrasonic diffraction limit. However, as ULM relies on precise localization and tracking of individual MBs, high MB concentrations yield to increased localization errors and, ultimately, ULM failure. This constraint limits ULM to low MB concentrations, resulting in long acquisition times that pose challenges in clinical settings. Methods: Here, we show the feasibility of uncoupling a bi-disperse MB population, composed of two monodisperse MB populations. The uncoupling is performed through a signal processing pipeline that exploits the strong nonlinear response of MBs having resonance frequency tuned with the transmission frequency. After uncoupling, ULM density and velocity flow maps are generated. Results: Density and velocity maps are generated after uncoupling, when injecting the bi-disperse population individually and simultaneously in a vascular 3D-printed phantom. Furthermore, density maps generated after uncoupling are compared with the one obtained using standard ULM. Results demonstrate the capability of the proposed uncoupling pipeline to separate the bi-disperse population. Conclusion: This work presents a signal processing pipeline to uncouple a bi-disperse MB population, formed by two monodisperse MB populations. Results are validated in a 3D-printed phantom and demonstrate the feasibility of the uncoupling which, in turn, would enable higher concentrations and reduce acquisition times for micro-vascular imaging.

Ultrasound localization microscopy (ULM) based on microbubble (MB) localization was recently introduced to overcome the resolution limit of conventional ultrasound. However, ULM is currently challenged by the requirement for long data acquisition times to accumulate adequate MB events to fully reconstruct vasculature. In this study, we present a curvelet transform-based sparsity promoting (CTSP) algorithm that improves ULM imaging speed by recovering missing MB localization signal from data with very short acquisition times. CTSP was first validated in a simulated microvessel model, followed by the chicken embryo chorioallantoic membrane (CAM), and finally, in the mouse brain. In the simulated microvessel study, CTSP robustly recovered the vessel model to achieve an 86.94% vessel filling percentage from a corrupted image with only 4.78% of the true vessel pixels. In the chicken embryo CAM study, CTSP effectively recovered the missing MB signal within the vasculature, leading to marked improvement in ULM imaging quality with a very short data acquisition. Taking the optical image as reference, the vessel filling percentage increased from 2.7% to 42.2% using 50ms of data acquisition after applying CTSP. CTSP used 80% less time to achieve the same 90% maximum saturation level as compared with conventional MB localization. We also applied CTSP on the microvessel flow speed maps and found that CTSP was able to use only 1.6s of microbubble data to recover flow speed images that have similar qualities as those constructed using 33.6s of data. In the mouse brain study, CTSP was able to reconstruct the majority of the cerebral vasculature using 1–2s of data acquisition. Additionally, CTSP only needed 3.2s of microbubble data to generate flow velocity maps that are comparable to those using 129.6s of data. These results suggest that CTSP can facilitate fast and robust ULM imaging especially under the circumstances of inadequate microbubble localizations.

Neovascularisation of carotid plaques contributes to their vulnerability. Current imaging methods such as contrast-enhanced ultrasound (CEUS) usually lack the required spatial resolution and quantification capability for precise neovessels identification. We aimed at quantifying plaque vascularisation with ultrasound localization microscopy (ULM) and compared the results to histological analysis. We conducted a prospective, monocentric, study involving patients who were undergoing carotid endarterectomy (CEA) for carotid artery stenosis. The day before CEA ultrasound examination coupled with the injection of microbubbles (MB) as a contrast agent (CEUS) to image the MB circulating within and around the carotid plaque was performed. CEUS images analysis classified patients into 2 groups: absence of neovascularisation (group A) or presence of neovascularisation (group B). ULM was performed by localising and tracking individual MB centres to reconstruct the neovessels structure with a resolution of around 60μm. Plaques were manually segmented on the images to quantify the number of neovessels and various haemodynamic metrics inside the plaques. Histological analysis of the excised carotid plaque specimens classified patients into 2 groups: absence of neovascularisation (group I) or presence of neovascularisation (group II). Among the 26 patients included, classification was as follows: group I: n=8 and group II: n=18, 18 patients had analysable CEUS images and were classified as follows: group A: n=10, group B: n=8. The median (Q1-Q3) number of MB tracked per second inside the plaque was 0.03 (0-0.37) for patients in group I and 0.51 (0-3) for patients in group A versus (vs.) 3.55 (1.26-17.68) for patients in group II and 9.69 (5.83-34.68) for patients in group B (p=0.00049; p=0.010 respectively). The length of the MB tracks was 0.02mm (0-0.16) in group I vs. 0.29mm (0.22-0.45) in group II (p=0.0069). The study also showed that flow in the neovessels was greater during systole than during diastole period: 9.38 (1.67-19.17) MB tracked per second vs. 1.35 (0.28-6.56) (p=0.021). ULM allows the detection of neovessels within the carotid atherosclerotic plaque. Thus, ULM provides a precise picture of plaque neovascularisation in patients and could be used as a non-invasive imaging technique to assess carotid plaque vulnerability. The study was sponsored and funded by Assistance Publique-Hôpitaux de Paris (CRC 1806 APHP INNOVATION 2018). Co-funding by ART (Technological Research Accelerator) biomedical ultrasound program of INSERM, France.

PurposeThree-dimensional (3D) ultrasound localization microscopy (ULM) using a 2-D matrix probe and microbubbles (MBs) has recently been proposed to visualize microvasculature in three spatial dimensions beyond the ultrasound diffraction limit. However, 3D ULM has several limitations, including: (1) high system complexity, (2) complex MB flow dynamics in 3D, and (3) extremely long acquisition time that had to be addressed.MethodTo reduce the system complexity while maintaining high image quality, we used a sub-aperture process to reduce received channel counts. To address the second issue, a 3D bipartite graph-based method with Kalman filtering-based tracking was used in this study for MB tracking. An MB separation approach was incorporated to separate high concentration MB data into multiple, sparser MB datasets, allowing better MB localization and tracking for a limited acquisition time.ResultsThe proposed method was first validated in a flow channel phantom, showing improved spatial resolutions compared with the contrasted enhanced power Doppler image. Then the proposed method was evaluated with an in vivo chicken embryo brain dataset. Results showed that the reconstructed 3D super-resolution image achieved a spatial resolution of around 52 μm (smaller than the wavelength of around 200 μm).ConclusionA lower system complexity of 3D ULM has been proposed. In addition, our proposed 3D ULM provided the capability of 3D motion compensation and MB tracking. Microvessels that cannot be resolved clearly using localization only, can be well identified with the proposed method.

Ultrasound localization microscopy (ULM) has revolutionized microvascular imaging by breaking the acoustic diffraction limit. However, different ULM workflows depend heavily on distinct prior knowledge, such as the impulse response and empirical selection of parameters (e.g., the number of microbubbles (MBs) per frame M), or the consistency of training-test dataset in deep learning (DL)-based studies. We hereby propose a general ULM pipeline that reduces priors. Our approach leverages a DL model that simultaneously distills MB signals and reduces speckles from every frame without estimating the impulse response and M. Our method features an efficient channel attention Vision Transformer (ViT) and a progressive learning strategy, enabling it to learn global information through training on progressively increasing patch sizes. Ample synthetic data were generated using the k-Wave toolbox to simulate various MB patterns, thus overcoming the deficiency of labeled data. The ViT output was further processed by a standard radial symmetry (RS) method for subpixel localization. Our method performed well on model-unseen public datasets: one in silico dataset with ground truth (GT) and four in vivo datasets of mouse tumor, rat brain, rat brain bolus, and rat kidney. Our pipeline outperformed conventional ULM, achieving higher positive predictive values (precision in DL, 0.88-0.41 versus 0.83-0.16) and improved accuracy (root-mean-square errors (RMSEs): 0.25- $0.14~\lambda $ versus 0.31- $0.13~\lambda $ ) across a range of signal-to-noise ratios (SNRs) from 60 to 10 dB. Our model could detect more vessels in diverse in vivo datasets while achieving comparable resolutions to the standard method. The proposed ViT-based model, seamlessly integrated with state-of-the-art downstream ULM steps, improved the overall ULM performance with no priors.

Ultrasound localization microscopy (ULM) has been developed to significantly improve the spatial resolution of ultrasound imaging by localizing the microbubbles (MBs). However, owing to the unsatisfactory performance of conventional localization methods in localizing overlapped MBs, low MB concentration is typically required, which leads to a long data-acquisition time. Recently, deep learning (DL)-based localization methods have been proposed to improve the precision and processing speed of localization at high MB concentrations. Because the ground truth of in vivo MB locations is difficult to obtain, labeled training datasets are usually generated by simulations. Considering the differences between the simulated and experimental data, the performance of the trained convolutional neural network (CNN) may not generalize well on experimental data. In this study, a self-supervised learning scheme is proposed to train the CNN directly using experimental data and is evaluated on ex vivo chicken embryo chorioallantoic membrane (CAM) data at high MB concentration. Conventional cross-correlation (CC)-based localization method is used for comparison, and the results show that the proposed method performs much better in localization of high-concentration MBs than the CC-based method. In addition, the proposed method forms a closed loop for neural network training through experimental data and bypasses the requirement for labeled training dataset, which is typically generated using simulations.

Accurate signal detection of ultrasound contrast agents, such as microbubble (MB) and gas vesicle (GV), in the presence of clutter and noise is essential to increase image quality in super-resolution ultrasound imaging (SRUS) and achieve precise GV localization. We developed and evaluated an eigen-image based signal detection method using singular value decomposition (SVD) and changepoint detection to automatically segment the data that are closely related to physical events such as MB flow and GV collapse. Eigen-image based method was compared with the elbow point and hard thresholding method when selecting MB signals after SVD of raw data, acquired from phantom and in vivo experiments. Image reconstructed by eigen-image based method was also compared with unregistered difference image for GV localization when moving GVs in a phantom were collapsed by ultrafast plane waves. The eigen-image based MB signal detection method resulted in higher vessel density (VD) visualization in both the phantom and in vivo mouse tumor. It also achieved increased signal-to-noise ratio (SNR) in both cases. Moreover, this method localized moving GVs more efficiently than the difference imaging method, without requiring pixel registration based on landmarks. The eigen-image based method offers a reliable and automated approach to MB and GV signal detection for both SRUS and point target localization. This approach is a valuable tool for medical imaging providing high-quality vessel images along with accurate locations of moving ultrasound contrast agents, which can be potentially translatable to clinical diagnosis and pre-clinical research.

Ultrasound localization microscopy (ULM) enabled high-accuracy measurements of microvessel flow beyond the resolution limit of conventional ultrasound imaging by utilizing contrast microbubbles (MBs) as point targets. Robust tracking of MBs is an essential task for fast and high-quality ULM image reconstruction. Existing MB tracking methods suffer from challenging imaging scenarios such as high-density MB distributions, fast blood flow, and complex flow dynamics. Here we present a deep learning-based MB pairing and tracking method based on a bi-directional long short-term memory neural network for ULM. The proposed method integrates multiparametric MB characteristics to facilitate more robust and accurate MB pairing and tracking. The method was validated on a simulation data set, a tissue-mimicking flow phantom, and in vivo on a mouse and rat brain.

Assessment of coronary microcirculation alterations in a porcine model of no-reflow using ultrasound localization microscopy: a proof of concept study.

Ultrasound Localisation Microscopy (ULM) is an imaging framework which consists of following ultrasound contrast agents, microbubbles, in time, on ultrasound images. The three main steps of ULM are: detecting microbubbles by reducing tissue signal, localizing them with subwavelength precision and tracking their trajectories. ULM performances were evaluated in different studies throughout metrics such as localisation accuracy or capacity to filter the tissues. In parallel, adaptive beamforming offers narrower Point Spread Function (PSF) and/or better tissue filtering than delay-and-sum method classically used within ULM. In this paper, the ability of adaptive beamformers to enhance ULM performances is evaluated, with a particular focus on the trade-off between acquisition time and bubble concentration to achieve super-resolution results.

Super-resolution ultrasound (SRUS) has become a tool for in vivo microvascular imaging. Most of the SRUS methods are based on microbubble localization: namely, ultrasound localization microscopy (ULM). The aim of this study was to develop a nonlocalization SRUS method and verify its feasibility in microvascular imaging. We introduce a new super-resolution strategy based on the postprocessing of contrast-enhanced ultrasound. The proposed method, which is termed ultrasound diffraction attenuation microscopy (UDAM), uses super-resolution radial fluctuations instead of microbubble localization to overcome acoustic diffraction limits. Biceps of Japanese long-ear white rabbits were adopted to validate its feasibility on muscle vascular imaging, using a clinical accessible ultrasound system at a frame rate of 30 Hz under a single bolus injection of SonoVue (Bracco SpA, Milan, Italy). The super-resolution image was compared with the maximum-intensity projection and ULM. The animal study illustrates that the proposed UDAM can obtain super-resolution microvascular images of rabbits' muscles under a single bolus injection of SonoVue with a 150-second contrast-enhanced ultrasound video. Both ULM and UDAM can achieve a very similar vascular structure with the maximum-intensity projection but much higher spatial resolution. The measurement of 1-dimensional signals shows that UDAM can distinguish the subwavelength structures and substantial reduce the full width at half-maximum of microvessels. We conclude UDAM provides a noninvasive tool for in vivo super-resolution microvascular imaging.

Evaluation of tumor microvascular morphology is of great significance in tumor diagnosis, therapeutic effect prediction, and surgical planning. Recently, two-dimensional ultrasound localization microscopy (2DULM) has demonstrated its superiority in the field of microvascular imaging. However, it suffers from planar dependence and is unintuitive. We propose a novel three-dimensional ultrasound localization microscopy (3DULM) to avoid these limitations. We investigated 3DULM based on a 2D array for tumor microvascular imaging. After intravenous injection of contrast agents, all elements of the 2D array transmit and receive signals to ensure a high and stable frame rate. Microbubble signal extraction, filtering, positioning, tracking, and other processing were used to obtain a 3D vascular map, flow velocity, and flow direction. To verify the effectiveness of 3DULM, it was validated on double helix tubes and rabbit VX2 tumors. Cisplatin was used to verify the ability of 3DULM to detect microvascular changes during tumor treatment. In vitro, the sizes measured by 3DULM at 3mm and 13mm were 178 and 182 , respectively. In the rabbit tumors, we acquired 9000 volumes to reveal vessels about 30 in diameter, which surpasses the diffraction limit of ultrasound in traditional ultrasound imaging, and the results matched with micro-angiography. In addition, there were significant changes in vascular density and curvature between the treatment and control groups. The effectiveness of 3DULM was verified in vitro and in vivo. Hence, 3DULM may have potential applications in tumor diagnosis, tumor treatment evaluation, surgical protocol guidance, and cardiovascular disease. 3D ultrasound localization microscopy is highly sensitive to microvascular changes; thus, it has clinical potential for tumor diagnosis and treatment evaluation. • 3D ultrasound localization microscopy is demonstrated on double helix tubes and rabbit VX2 tumors. • 3D ultrasound localization microscopy can reveal vessels about 30 in diameter-far smaller than traditional ultrasound. • This form of imaging has potential applications in tumor diagnosis, tumor treatment evaluation, surgical protocol guidance, and cardiovascular disease.

Photoacoustic (PA) imaging can map the physiological conditions of tissue and track the biodistribution of contrast agents; ultrasound localization microscopy (ULM) with microbubbles provides deep-tissue super-resolution blood vessel images and blood velocity maps. When combined, the dual imaging technique will become a powerful tool with diverse applications. However, longer acquisition time of ULM poses a challenge. To address this, we propose sparsity constraint optimization to accelerate ULM and develop an interleaved PA/fast ULM imaging sequences for the dual imaging. We demonstrate 3D and time-lapse dual imaging with a commonly used linear array imaging system in a mouse model.

Super-resolution (SR) methods enable to obtain high resolution image (HR) from multiple low resolution input images (LR) of the same scene. Development of SR algorithms for real applications require reliable image quality metrics for performance comparison of various SR methods. In this paper, six standard SR methods were implemented and subjective image quality assessment was performed to evaluate the visual quality of the enhanced images. Nine image quality metrics were examined with respect to their ability to characterize the observed subjective image quality of SR enhancement.