Search

Algorithmic changes that increase beamforming speed have become increasingly relevant to processing synthetic aperture (SA) ultrasound data. In particular, beamforming SA data in a spatio-temporal frequency domain using the F-k (Stolt) migration have been shown to reduce the beamforming time by up to two orders of magnitude compared with the conventional delay-and-sum (DAS) beamforming, and it has been used in applications where large amounts of raw data make real-time frame rates difficult to attain, such as multistatic SA imaging and plane-wave Doppler imaging with large ensemble lengths. However, beamforming signals in a spatio-temporal Fourier space can require loading large blocks of data at once, making it memory-intensive and less suited for parallel (i.e., multithreaded) processing. As an alternative, we propose beamforming in a range-Doppler (RD) frequency domain using the range-Doppler algorithm (RDA) that has originally been developed for SA radar (SAR) imaging. Through simulation and phantom experiments, we show that RDA achieves similar lateral resolution and contrast compared with DAS and F-k migration. At the same time, higher axial sidelobes in RDA images can be reduced via (temporal) frequency binning. Like the F-k migration, RDA significantly reduces the overall number of computations relative to DAS, and it achieves ten times lower processing time on a single CPU. Because RDA uses only a spatial Fourier transform (FT), it requires two times less memory than the F-k migration to process the simulated multistatic data and can be executed on as many as a thousand parallel threads (compared with eight parallel threads for the F-k migration), making it more suitable for implementation on modern graphics processing units (GPUs). While RDA is not as parallelizable as DAS, it is expected to hold a significant speed advantage on devices with moderate parallel processing capabilities (up to several thousand cores), such as point-of-care and low-cost ultrasound devices. [ABSTRACT FROM AUTHOR]

Phase aberration is widely considered a major source of image degradation in medical pulse-echo ultrasound. Traditionally, near-field phase aberration correction techniques are unable to account for distributed aberrations due to a spatially varying speed of sound in the medium, while most distributed aberration correction techniques require the use of point-like sources and are impractical for clinical applications where diffuse scattering is dominant. Here, we present two distributed aberration correction techniques that utilize sound speed estimates from a tomographic sound speed estimator that builds on our previous work with diffuse scattering in layered media. We first characterize the performance of our sound speed estimator and distributed aberration correction techniques in simulations where the scattering in the media is known a priori. Phantom and in vivo experiments further demonstrate the capabilities of the sound speed estimator and the aberration correction techniques. In phantom experiments, point target resolution improves from 0.58 to 0.26 and 0.27 mm, and lesion contrast improves from 17.7 to 23.5 and 25.9 dB, as a result of distributed aberration correction using the eikonal and wavefield correlation techniques, respectively. [ABSTRACT FROM AUTHOR]

Lesion detectability (LD) quantifies how easily a lesion or target can be distinguished from the background. LD is commonly used to assess the performance of new ultrasound imaging methods. The contrast-to-noise ratio (CNR) is the most popular measure of LD; however, recent work has exposed its vulnerability to manipulations of dynamic range. The generalized CNR (gCNR) has been proposed as a robust histogram-based alternative that is invariant to such manipulations. Here, we identify key shortcomings of CNR and strengths of gCNR as LD metrics for modern beamformers. Using the measure theory, we pose LD as a distance between empirical probability measures (i.e., histograms) and prove that: 1) gCNR is equal to the total variation distance between probability measures and 2) gCNR is one minus the error rate of the ideal observer. We then explore several consequences of measure-theoretic LD in simulation studies. We find that histogram distances depend on bin selection that LD must be considered in the context of spatial resolution and that many histogram distances are invariant under measure-preserving isomorphisms of the sample space (e.g., dynamic range transformations). Finally, we provide a mathematical interpretation for why quantitative values such as contrast ratio (CR), CNR, and signal-to-noise ratio should not be compared between images with different dynamic ranges or underlying units and demonstrate how histogram matching can be used to reenable such quantitative comparisons. [ABSTRACT FROM AUTHOR]

Deep learning for ultrasound image formation is rapidly garnering research support and attention, quickly rising as the latest frontier in ultrasound image formation, with much promise to balance both image quality and display speed. Despite this promise, one challenge with identifying optimal solutions is the absence of unified evaluation methods and datasets that are not specific to a single research group. This article introduces the largest known international database of ultrasound channel data and describes the associated evaluation methods that were initially developed for the challenge on ultrasound beamforming with deep learning (CUBDL), which was offered as a component of the 2020 IEEE International Ultrasonics Symposium. We summarize the challenge results and present qualitative and quantitative assessments using both the initially closed CUBDL evaluation test dataset (which was crowd-sourced from multiple groups around the world) and additional in vivo breast ultrasound data contributed after the challenge was completed. As an example quantitative assessment, single plane wave images from the CUBDL Task 1 dataset produced a mean generalized contrast-to-noise ratio (gCNR) of 0.67 and a mean lateral resolution of 0.42 mm when formed with delay-and-sum beamforming, compared with a mean gCNR as high as 0.81 and a mean lateral resolution as low as 0.32 mm when formed with networks submitted by the challenge winners. We also describe contributed CUBDL data that may be used for training of future networks. The compiled database includes a total of 576 image acquisition sequences. We additionally introduce a neural-network-based global sound speed estimator implementation that was necessary to fairly evaluate the results obtained with this international database. The integration of CUBDL evaluation methods, evaluation code, network weights from the challenge winners, and all datasets described herein are publicly available (visit https://cubdl.jhu.edu for details). [ABSTRACT FROM AUTHOR]

Power Doppler (PD) is a commonly used technique for flow detection and vessel visualization in radiology clinics. Despite its broad set of applications, PD suffers from multiple noise sources and artifacts, such as thermal noise, clutter, and flash artifacts. In addition, a tradeoff exists between acquisition time and Doppler image quality. These limit the ability of clinical PD imaging in deep-lying and small-vessel detection and visualization, particularly among patients with high body mass indices (BMIs). To improve the Doppler vessel detection, we have previously proposed coherent flow PD (CFPD) imaging and demonstrated its performance on porcine vasculature. In this article, we report on a pilot clinical study of CFPD imaging on healthy human volunteers and patients with high BMI to assess the clinical feasibility of the technique in liver imaging. In this study, we built a real-time CFPD imaging system using a graphical processing unit (GPU)-based software beamformer and a CFPD processing module. Using the real-time CFPD imaging system, the liver vasculature of 15 healthy volunteers with normal BMI below 25 and 15 patients with BMI greater than 25 was imaged. Both PD and CFPD image streams were produced simultaneously. The generalized contrast-to-noise ratio (gCNR) of the PD and CFPD images was measured to provide the quantitative evaluation of image quality and vessel detectability. Comparison of PD and CFPD image shows that gCNR is improved by 35% in healthy volunteers and 28% in high BMI patients with CFPD compared to PD. Example images are provided to show that the improvement in the Doppler image gCNR leads to greater detection of small vessels in the liver. In addition, we show that CFPD can suppress in vivo reverberation clutter in clinical imaging. [ABSTRACT FROM AUTHOR]

Photoacoustic (PA) imaging can revolutionize medical ultrasound by augmenting it with molecular information. However, clinical translation of PA imaging remains a challenge due to the limited viewing angles and imaging depth. Described here is a new robust algorithm called Superiorized Photo-Acoustic Non-NEgative Reconstruction (SPANNER), designed to reconstruct PA images in real-time and to address the artifacts associated with limited viewing angles and imaging depth. The method utilizes precise forward modeling of the PA propagation and reception of signals while accounting for the effects of acoustic absorption, element size, shape, and sensitivity, as well as the transducer’s impulse response and directivity pattern. A fast superiorized conjugate gradient algorithm is used for inversion. SPANNER is compared to three reconstruction algorithms: delay-and-sum (DAS), universal back-projection (UBP), and model-based reconstruction (MBR). All four algorithms are applied to both simulations and experimental data acquired from tissue-mimicking phantoms, ex vivo tissue samples, and in vivo imaging of the prostates in patients. Simulations and phantom experiments highlight the ability of SPANNER to improve contrast to background ratio by up to 20 dB compared to all other algorithms, as well as a 3-fold increase in axial resolution compared to DAS and UBP. Applying SPANNER on contrast-enhanced PA images acquired from prostate cancer patients yielded a statistically significant difference before and after contrast agent administration, while the other three image reconstruction methods did not, thus highlighting SPANNER’s performance in differentiating intrinsic from extrinsic PA signals and its ability to quantify PA signals from the contrast agent more accurately. [ABSTRACT FROM AUTHOR]

The widespread development of new ultrasound image formation techniques has created a need for a standardized methodology for comparing the resulting images. Traditional methods of evaluation use quantitative metrics to assess the imaging performance in specific tasks, such as point resolution or lesion detection. Quantitative evaluation is complicated by unconventional new methods and nonlinear transformations of the dynamic range of data and images. Transformation-independent image metrics have been proposed for quantifying task performance. However, clinical ultrasound still relies heavily on visualization and qualitative assessment by expert observers. We propose the use of histogram matching to better assess differences across image formation methods. We briefly demonstrate the technique using a set of sample beamforming methods and discuss the implications of such image processing. We present variations of histogram matching and provide code to encourage the application of this method within the imaging community. [ABSTRACT FROM AUTHOR]

Diffuse reverberation is ultrasound image noise caused by multiple reflections of the transmitted pulse before returning to the transducer, which degrades image quality and impedes the estimation of displacement or flow in techniques such as elastography and Doppler imaging. Diffuse reverberation appears as spatially incoherent noise in the channel signals, where it also degrades the performance of adaptive beamforming methods, sound speed estimation, and methods that require measurements from channel signals. In this paper, we propose a custom 3D fully convolutional neural network (3DCNN) to reduce diffuse reverberation noise in the channel signals. The 3DCNN was trained with channel signals from simulations of random targets that include models of reverberation and thermal noise. It was then evaluated both on phantom and in-vivo experimental data. The 3DCNN showed improvements in image quality metrics such as generalized contrast to noise ratio (GCNR), lag one coherence (LOC) contrast-to-noise ratio (CNR) and contrast for anechoic regions in both phantom and in-vivo experiments. Visually, the contrast of anechoic regions was greatly improved. The CNR was improved in some cases, however the 3DCNN appears to strongly remove uncorrelated and low amplitude signal. In images of in-vivo carotid artery and thyroid, the 3DCNN was compared to short-lag spatial coherence (SLSC) imaging and spatial prediction filtering (FXPF) and demonstrated improved contrast, GCNR, and LOC, while FXPF only improved contrast and SLSC only improved CNR. [ABSTRACT FROM AUTHOR]

Passive cavitation mapping (PCM) techniques typically utilize a time-exposure acoustic (TEA) approach, where the received radio frequency data are beamformed, squared, and integrated over time. Such PCM-TEA cavitation maps typically suffer from long-tail artifacts and poor axial resolution with pulse–echo diagnostic arrays. Here, we utilize a recently developed PCM technique based on cavitation source localization (CSL), which fits a hyperbolic function to the received cavitation wavefront. A filtering method based on the root-mean-square error (rmse) of the hyperbolic fit is utilized to filter out spurious signals. We apply a wavefront correction technique to the signals with poor fit quality to recover additional cavitation signals and improve cavitation localization. Validation of the PCM-CSL technique with rmse filtering and wavefront correction was conducted in experiments with a tissue-mimicking flow phantom and an in vivo mouse model of cancer. It is shown that the quality of the hyperbolic fit, necessary for the PCM-CSL, requires an rmse < 0.05 mm2 in order to accurately localize the cavitation sources. A detailed study of the wavefront correction technique was carried out, and it was shown that, when applied to experiments with high noise and interference from multiple cavitating microbubbles, it was capable of effectively correcting noisy wavefronts without introducing spurious cavitation sources, thereby improving the quality of the PCM-CSL images. In phantom experiments, the PCM-CSL was capable of precisely localizing sources on the therapy beam axis and off-axis sources. In vivo cavitation experiments showed that PMC-CSL showed a significant improvement over PCM-TEA and yielded acceptable localization of cavitation signals in mice. [ABSTRACT FROM AUTHOR]

Ultrasound molecular imaging (UMI) is enabled by targeted microbubbles (MBs), which are highly reflective ultrasound contrast agents that bind to specific biomarkers. Distinguishing between adherent MBs and background signals can be challenging in vivo. The preferred preclinical technique is differential targeted enhancement (DTE), wherein a strong acoustic pulse is used to destroy MBs to verify their locations. However, DTE intrinsically cannot be used for real-time imaging and may cause undesirable bioeffects. In this work, we propose a simple 4-layer convolutional neural network to nondestructively detect adherent MB signatures. We investigated several types of input data to the network: “anatomy-mode” (fundamental frequency), “contrast-mode” (pulse-inversion harmonic frequency), or both, i.e., “dual-mode”, using IQ channel signals, the channel sum, or the channel sum magnitude. Training and evaluation were performed on in vivo mouse tumor data and microvessel phantoms. The dual-mode channel signals yielded optimal performance, achieving a soft Dice coefficient of 0.45 and AUC of 0.91 in two test images. In a volumetric acquisition, the network best detected a breast cancer tumor, resulting in a generalized contrast-to-noise ratio (GCNR) of 0.93 and Kolmogorov-Smirnov statistic (KSS) of 0.86, outperforming both regular contrast mode imaging (GCNR = 0.76, KSS = 0.53) and DTE imaging (GCNR = 0.81, KSS = 0.62). Further development of the methodology is necessary to distinguish free from adherent MBs. These results demonstrate that neural networks can be trained to detect targeted MBs with DTE-like quality using nondestructive dual-mode data, and can be used to facilitate the safe and real-time translation of UMI to clinical applications. [ABSTRACT FROM AUTHOR]

Robust recovery of multistatic synthetic aperture data from conventional ultrasound sequences can enable complete transmit-and-receive focusing at all points in the field of view without the drawbacks of virtual-source synthetic aperture and further enables more advanced imaging applications, such as backscatter coherence, sound speed estimation, and phase aberration correction. Recovery of the multistatic data set has previously been demonstrated on a steered transmit sequence for phased arrays using an adjoint-based method. We introduce two methods to improve the accuracy of the multistatic data set. We first modify the original technique used for steered transmit sequences by applying a ramp filter to compensate for the nonuniform frequency scaling introduced by the adjoint-based method. Then, we present a regularized inversion technique that allows additional aperture specification and is intended to work for both steered transmit and walking aperture sequences. The ramp-filtered adjoint and regularized inversion techniques, respectively, improve the correlation of the recovered signal with the ground truth from 0.9404 to 0.9774 and 0.9894 in steered transmit sequences and 0.4610 to 0.4733 and 0.9936 in walking aperture sequences. [ABSTRACT FROM AUTHOR]

Short-lag spatial coherence (SLSC) imaging has demonstrated improved performance over conventional B-mode ultrasound imaging. Previous work has evaluated the performance of SLSC using 2-D matrix arrays in simulation and in vivo studies across various levels of subaperture beamforming, demonstrating improved contrast-to-noise ratio (CNR) and speckle signal-to-noise ratio (SNR) over 1-D arrays. This work explores the application of SLSC imaging in 1.5-D and 1.75-D arrays to quantify the impacts of elevation element count, mirroring, and Fresnel element spacing on SLSC image quality. Through simulation and in vivo studies, increased elevation element count was shown to improve CNR and speckle SNR relative to 1-D SLSC and B-mode images. Elevation mirroring (1.5-D) was shown to force the inclusion of long lags into the SLSC calculation, introducing additional decorrelation and reducing image quality relative to 1.75-D arrays with individually connected elements. These results demonstrate the effectiveness of SLSC imaging in 1.5-D and 1.75-D arrays. [ABSTRACT FROM AUTHOR]

With traditional beamforming methods, ultrasound B-mode images contain speckle noise caused by the random interference of subresolution scatterers. In this paper, we present a framework for using neural networks to beamform ultrasound channel signals into speckle-reduced B-mode images. We introduce log-domain normalization-independent loss functions that are appropriate for ultrasound imaging. A fully convolutional neural network was trained with the simulated channel signals that were coregistered spatially to ground-truth maps of echogenicity. Networks were designed to accept 16 beamformed subaperture radio frequency (RF) signals. Training performance was compared as a function of training objective, network depth, and network width. The networks were then evaluated on the simulation, phantom, and in vivo data and compared against the existing speckle reduction techniques. The most effective configuration was found to be the deepest (16 layer) and widest (32 filter) networks, trained to minimize a normalization-independent mixture of the $\ell _{1}$ and multiscale structural similarity (MS-SSIM) losses. The neural network significantly outperformed delay-and-sum (DAS) and receive-only spatial compounding in speckle reduction while preserving resolution and exhibited improved detail preservation over a nonlocal means method. This work demonstrates that ultrasound B-mode image reconstruction using machine-learned neural networks is feasible and establishes that networks trained solely in silico can be generalized to real-world imaging in vivo to produce images with significantly reduced speckle. [ABSTRACT FROM AUTHOR]

Stress echocardiography is used to detect myocardial ischemia by evaluating cardiovascular function both at rest and at elevated heart rates. Stress echocardiography requires excellent visualization of the left ventricle (LV) throughout the cardiac cycle. However, LV endocardial border visualization is often negatively impacted by high levels of clutter associated with patient obesity, which has risen dramatically worldwide in recent decades. Short-lag spatial coherence (SLSC) imaging has demonstrated reduced clutter in several applications. In this work, a computationally efficient formulation of SLSC was implemented into an object-oriented graphics processing unit-based software beamformer, enabling real-time (>30 frames per second) SLSC echocardiography on a research ultrasound scanner. The system was then used to image 15 difficult-to-image stress echocardiography patients in a comparison study of tissue harmonic imaging (THI) and harmonic spatial coherence imaging (HSCI). Video clips of four standard stress echocardiography views acquired with either THI or HSCI were provided in random shuffled order to three experienced readers. Each reader rated the visibility of 17 LV segments as “invisible,” “suboptimally visualized,” or “well visualized,” with the first two categories indicating a need for contrast agent. In a symmetry test unadjusted for patientwise clustering, HSCI demonstrated a clear superiority over THI ($p< 0.0001$). When measured on a per-patient basis, the median total score significantly favored HSCI with $p< 0.001$. When collapsing the ratings to a two-level scale (“needs contrast” versus “well visualized”), HSCI once again showed an overall superiority over THI, with $p< 0.0001$ by McNemar test adjusted for clustering. [ABSTRACT FROM AUTHOR]

Ultrasound molecular imaging (USMI) is accomplished by detecting microbubble (MB) contrast agents that have bound to specific biomarkers, and can be used for a variety of imaging applications, such as the early detection of cancer. USMI has been widely utilized in preclinical imaging in mice; however, USMI in humans can be challenging because of the low concentration of bound MBs and the signal degradation caused by the presence of heterogenous soft tissue between the transducer and the lesion. Short-lag spatial coherence (SLSC) beamforming has been proposed as a robust technique that is less affected by poor signal quality than standard delay-and-sum (DAS) beamforming. In this paper, USMI performance was assessed using contrast-enhanced ultrasound imaging combined with DAS (conventional CEUS) and with SLSC (SLSC-CEUS). Each method was characterized by flow channel phantom experiments. In a USMI-mimicking phantom, SLSC-CEUS was found to be more robust to high levels of additive thermal noise than DAS, with a 6dB SNR improvement when the thermal noise level was +6dB or higher. However, SLSC-CEUS was also found to be insensitive to increases in MB concentration, making it a poor choice for perfusion imaging. USMI performance was also measured in vivo using VEGFR2-targeted MBs in mice with subcutaneous human hepatocellular carcinoma tumors, with clinical imaging conditions mimicked using a porcine tissue layer between the tumor and the transducer. SLSC-CEUS improved the SNR in each of ten tumors by an average of 41%, corresponding to 3.0dB SNR. These results indicate that the SLSC beamformer is well-suited for USMI applications because of its high sensitivity and robust properties under challenging imaging conditions. [ABSTRACT FROM AUTHOR]

The spatial coherence of ultrasound backscatter has been proposed to reduce clutter in medical imaging, to measure the anisotropy of the scattering source, and to improve the detection of blood flow. These techniques rely on correlation estimates that are obtained using computationally expensive strategies. In this paper, we assess the existing spatial coherence estimation methods and propose three computationally efficient modifications: a reduced kernel, a downsampled receive aperture, and the use of an ensemble correlation coefficient. The proposed methods are implemented in simulation and in vivo studies. Reducing the kernel to a single sample improved computational throughput and improved axial resolution. Downsampling the receive aperture was found to have negligible effect on estimator variance, and improved computational throughput by an order of magnitude for a downsample factor of 4. The ensemble correlation estimator demonstrated lower variance than the currently used average correlation. Combining the three methods, the throughput was improved 105-fold in simulation with a downsample factor of 4- and 20-fold in vivo with a downsample factor of 2. [ABSTRACT FROM AUTHOR]

Power Doppler (PD) imaging is a widely used technique for flow detection. Despite the wide use of Doppler ultrasound, limitations exist in the ability of Doppler ultrasound to assess slow flow in the small-diameter vasculature, such as the maternal spiral arteries and fetal villous arteries of the placenta and focal liver lesions. The sensitivity of PD in small vessel detection is limited by the low signal produced by slow flow and the noise associated with small vessels. The noise sources include electronic noise, stationary or slowly moving tissue clutter, reverberation clutter, and off-axis scattering from tissue, among others. In order to provide more sensitive detection of slow flow in small diameter vessels, a coherent flow imaging technique, termed coherent flow PD (CFPD), is characterized and evaluated with simulation, flow phantom experiment studies, and an in vivo animal small vessel detection study. CFPD imaging was introduced as a technique to detect slow blood flow. It has been demonstrated to detect slow flow below the detection threshold of conventional PD imaging using identical pulse sequences and filter parameters. In this paper, we compare CFPD with PD in the detection of blood flow in small-diameter vessels. The results from the study suggest that CFPD is able to provide a 7.5–12.5-dB increase in the signal-to-noise ratio (SNR) over PD images for the same physiological conditions and is less susceptible to reverberation clutter and thermal noise. Due to the increase in SNR, CFPD is able to detect small vessels in high channel noise cases, for which PD was unable to generate enough contrast to observe the vessel. [ABSTRACT FROM AUTHOR]