Showing papers in "IEEE Transactions on Ultrasonics Ferroelectrics and Frequency Control in 2020"
TL;DR: This work generalizes the definition of CNR based on the overlap area between two probability density functions and proposes a new metric, gCNR, which is robust against dynamic range alterations and allows us to assess the relevance of new imaging algorithms.
Abstract: In the last 30 years, the contrast-to-noise ratio (CNR) has been used to estimate the contrast and lesion detectability in ultrasound images. Recent studies have shown that the CNR cannot be used with modern beamformers, as dynamic range alterations can produce arbitrarily high CNR values with no real effect on the probability of lesion detection. We generalize the definition of CNR based on the overlap area between two probability density functions. This generalized CNR (gCNR) is robust against dynamic range alterations; it can be applied to all kind of images, units, or scales; it provides a quantitative measure for contrast; and it has a simple statistical interpretation, i.e., the success rate that can be expected from an ideal observer at the task of separating pixels. We test gCNR on several state-of-the-art imaging algorithms and, in addition, on a trivial compression of the dynamic range. We observe that CNR varies greatly between the state-of-the-art methods, with improvements larger than 100%. We observe that trivial compression leads to a CNR improvement of over 200%. The proposed index, however, yields the same value for compressed and uncompressed images. The tested methods showed mismatched performance in terms of lesion detectability, with variations in gCNR ranging from −0.08 to +0.29. This new metric fixes a methodological flaw in the way we study contrast and allows us to assess the relevance of new imaging algorithms.
224 citations
TL;DR: A deep neural network is designed to directly process full or subsampled radio frequency data acquired at various subsampling rates and detector configurations so that it can generate high-quality US images using a single beamformer.
Abstract: In ultrasound (US) imaging, various types of adaptive beamforming techniques have been investigated to improve the resolution and the contrast-to-noise ratio of the delay and sum (DAS) beamformers. Unfortunately, the performance of these adaptive beamforming approaches degrades when the underlying model is not sufficiently accurate and the number of channels decreases. To address this problem, here, we propose a deep-learning-based beamformer to generate significantly improved images over widely varying measurement conditions and channel subsampling patterns. In particular, our deep neural network is designed to directly process full or subsampled radio frequency (RF) data acquired at various subsampling rates and detector configurations so that it can generate high-quality US images using a single beamformer. The origin of such input-dependent adaptivity is also theoretically analyzed. Experimental results using the B-mode focused US confirm the efficacy of the proposed methods.
100 citations
TL;DR: Overall, the DNNs successfully translated feature representations learned from simulated data to phantom and in vivo data, which is promising for this novel approach to simultaneous ultrasound image formation and segmentation.
Abstract: Single plane wave transmissions are promising for automated imaging tasks requiring high ultrasound frame rates over an extended field of view. However, a single plane wave insonification typically produces suboptimal image quality. To address this limitation, we are exploring the use of deep neural networks (DNNs) as an alternative to delay-and-sum (DAS) beamforming. The objectives of this work are to obtain information directly from raw channel data and to simultaneously generate both a segmentation map for automated ultrasound tasks and a corresponding ultrasound B-mode image for interpretable supervision of the automation. We focus on visualizing and segmenting anechoic targets surrounded by tissue and ignoring or deemphasizing less important surrounding structures. DNNs trained with Field II simulations were tested with simulated, experimental phantom, and in vivo data sets that were not included during training. With unfocused input channel data (i.e., prior to the application of receive time delays), simulated, experimental phantom, and in vivo test data sets achieved mean ± standard deviation Dice similarity coefficients of 0.92 ± 0.13, 0.92 ± 0.03, and 0.77 ± 0.07, respectively, and generalized contrast-to-noise ratios (gCNRs) of 0.95 ± 0.08, 0.93 ± 0.08, and 0.75 ± 0.14, respectively. With subaperture beamformed channel data and a modification to the input layer of the DNN architecture to accept these data, the fidelity of image reconstruction increased (e.g., mean gCNR of multiple acquisitions of two in vivo breast cysts ranged 0.89–0.96), but DNN display frame rates were reduced from 395 to 287 Hz. Overall, the DNNs successfully translated feature representations learned from simulated data to phantom and in vivo data, which is promising for this novel approach to simultaneous ultrasound image formation and segmentation.
86 citations
TL;DR: The results indicate that the proposed method substantially improves the robustness of SR-UMI under a clinically relevant imaging scenario where SR- UMI is challenged by a limited MB accumulation time, reduced number of MBs, lowered imaging frame rate, and degraded signal-to-noise ratio.
Abstract: Contrast microbubble (MB)-based super-resolution ultrasound microvessel imaging (SR-UMI) overcomes the compromise in conventional ultrasound imaging between spatial resolution and penetration depth and has been successfully applied to a wide range of clinical applications. However, clinical translation of SR-UMI remains challenging due to the limited number of MBs detected within a given accumulation time. Here, we propose a Kalman filter-based method for robust MB tracking and improved blood flow speed measurement with reduced numbers of MBs. An acceleration constraint and a direction constraint for MB movement were developed to control the quality of the estimated MB trajectory. An adaptive interpolation approach was developed to inpaint the missing microvessel signal based on the estimated local blood flow speed, facilitating more robust depiction of microvasculature with a limited amount of MBs. The proposed method was validated on an ex ovo chorioallantoic membrane and an in vivo rabbit kidney. Results demonstrated improved imaging performance on both microvessel density maps and blood flow speed maps. With the proposed method, the percentage of microvessel filling in a selected blood vessel at a given accumulation period was increased from 28.17% to 74.45%. A similar SR-UMI performance was achieved with MB numbers reduced by 85.96%, compared to that with the original MB number. The results indicate that the proposed method substantially improves the robustness of SR-UMI under a clinically relevant imaging scenario where SR-UMI is challenged by a limited MB accumulation time, reduced number of MBs, lowered imaging frame rate, and degraded signal-to-noise ratio.
83 citations
TL;DR: This work demonstrates the feasibility of 3-D super-resolution imaging and super-resolved velocity mapping using a customized 2-D sparse array transducer.
Abstract: High-frame-rate 3-D ultrasound imaging technology combined with super-resolution processing method can visualize 3-D microvascular structures by overcoming the diffraction-limited resolution in every spatial direction. However, 3-D super-resolution ultrasound imaging using a full 2-D array requires a system with a large number of independent channels, the design of which might be impractical due to the high cost, complexity, and volume of data produced. In this study, a 2-D sparse array was designed and fabricated with 512 elements chosen from a density-tapered 2-D spiral layout. High-frame-rate volumetric imaging was performed using two synchronized ULA-OP 256 research scanners. Volumetric images were constructed by coherently compounding nine-angle plane waves acquired at a pulse repetition frequency of 4500 Hz. Localization-based 3-D super-resolution images of two touching subwavelength tubes were generated from 6000 volumes acquired in 12 s. Finally, this work demonstrates the feasibility of 3-D super-resolution imaging and super-resolved velocity mapping using a customized 2-D sparse array transducer.
80 citations
TL;DR: Overall, the DL algorithm performed well and could be integrated into an ultrasound system in order to help diagnose and track B-line severity and could decrease variability and provide a standardized method for improved diagnosis and outcome.
Abstract: Shortness of breath is a major reason that patients present to the emergency department (ED) and point-of-care ultrasound (POCUS) has been shown to aid in diagnosis, particularly through evaluation for artifacts known as B-lines. B-line identification and quantification can be a challenging skill for novice ultrasound users, and experienced users could benefit from a more objective measure of quantification. We sought to develop and test a deep learning (DL) algorithm to quantify the assessment of B-lines in lung ultrasound. We utilized ultrasound clips ( ${n} =400$ ) from an existing database of ED patients to provide training and test sets to develop and test the DL algorithm based on deep convolutional neural networks. Interpretations of the images by algorithm were compared to expert human interpretations on binary and severity (a scale of 0–4) classifications. Our model yielded a sensitivity of 93% (95% confidence interval (CI) 81%–98%) and a specificity of 96% (95% CI 84%–99%) for the presence or absence of B-lines compared to expert read, with a kappa of 0.88 (95% CI 0.79–0.97). Model to expert agreement for severity classification yielded a weighted kappa of 0.65 (95% CI 0.56–074). Overall, the DL algorithm performed well and could be integrated into an ultrasound system in order to help diagnose and track B-line severity. The algorithm is better at distinguishing the presence from the absence of B-lines but can also be successfully used to distinguish between B-line severity. Such methods could decrease variability and provide a standardized method for improved diagnosis and outcome.
77 citations
TL;DR: A plausible genesis of some important lung artifacts is suggested and a convincing physical explanation of the genesis of important ultrasound lung artifacts does not exist yet.
Abstract: In standard B mode imaging, a set of ultrasound pulses is used to reconstruct a 2-D image even though some of the assumptions needed to do this are not fully satisfied. For this reason, ultrasound medical images show numerous artifacts which physicians recognize and evaluate as part of their diagnosis since even one artifact can provide clinical information. Understanding the physical mechanisms at the basis of the formation of an artifact is important to identify the physiopathological state of the biological medium which generated the artifact. Ultrasound lung images are a significant example of this challenge since everything that is represented beyond the thickness of the chest wall ( $\approx ~2$ cm) is artifactual information. A convincing physical explanation of the genesis of important ultrasound lung artifacts does not exist yet. Physicians simply base their diagnosis on a correlation observed over the years between the manifestation of some artifacts and the occurrence of particular lung pathologies. In this article, a plausible genesis of some important lung artifacts is suggested.
70 citations
TL;DR: A deep neural network-based model with loss function being scaled root-mean-squared error was proposed for super-resolution, denoising, as well as BW enhancement of the PA signals collected at the boundary of the domain and is shown to improve the collected boundary data, in turn, providing superior quality reconstructed PA image.
Abstract: Photoacoustic tomography (PAT) is a noninvasive imaging modality combining the benefits of optical contrast at ultrasonic resolution. Analytical reconstruction algorithms for photoacoustic (PA) signals require a large number of data points for accurate image reconstruction. However, in practical scenarios, data are collected using the limited number of transducers along with data being often corrupted with noise resulting in only qualitative images. Furthermore, the collected boundary data are band-limited due to limited bandwidth (BW) of the transducer, making the PA imaging with limited data being qualitative. In this work, a deep neural network-based model with loss function being scaled root-mean-squared error was proposed for super-resolution, denoising, as well as BW enhancement of the PA signals collected at the boundary of the domain. The proposed network has been compared with traditional as well as other popular deep-learning methods in numerical as well as experimental cases and is shown to improve the collected boundary data, in turn, providing superior quality reconstructed PA image. The improvement obtained in the Pearson correlation, structural similarity index metric, and root-mean-square error was as high as 35.62%, 33.81%, and 41.07%, respectively, for phantom cases and signal-to-noise ratio improvement in the reconstructed PA images was as high as 11.65 dB for in vivo cases compared with reconstructed image obtained using original limited BW data. Code is available at https://sites.google.com/site/sercmig/home/dnnpat .
64 citations
TL;DR: An automatic and unsupervised method for the detection and localization of the pleural line in LUS data based on the hidden Markov model and Viterbi Algorithm is proposed and provides a significant support to medical staff for further evaluating the patient condition.
Abstract: Recent works highlighted the significant potential of lung ultrasound (LUS) imaging in the management of subjects affected by COVID-19. In general, the development of objective, fast, and accurate automatic methods for LUS data evaluation is still at an early stage. This is particularly true for COVID-19 diagnostic. In this article, we propose an automatic and unsupervised method for the detection and localization of the pleural line in LUS data based on the hidden Markov model and Viterbi Algorithm. The pleural line localization step is followed by a supervised classification procedure based on the support vector machine (SVM). The classifier evaluates the healthiness level of a patient and, if present, the severity of the pathology, i.e., the score value for each image of a given LUS acquisition. The experiments performed on a variety of LUS data acquired in Italian hospitals with both linear and convex probes highlight the effectiveness of the proposed method. The average overall accuracy in detecting the pleura is 84% and 94% for convex and linear probes, respectively. The accuracy of the SVM classification in correctly evaluating the severity of COVID-19 related pleural line alterations is about 88% and 94% for convex and linear probes, respectively. The results as well as the visualization of the detected pleural line and the predicted score chart, provide a significant support to medical staff for further evaluating the patient condition.
63 citations
TL;DR: This article proposes a channel attention module with multiscale grid average pooling (MSGRAP) for the precise segmentation of breast cancer regions in ultrasound images and develops a novel semantic segmentation network with the proposed attention module.
Abstract: Breast cancer accounts for the second-largest number of deaths in women around the world, and more than 8% of women will suffer from the disease in their lifetime. Mortality due to breast cancer can be reduced by its early and precise diagnosis. Many studies have investigated methods for segmentation, and computer-aided diagnosis based on deep learning techniques, in particular, has recently gained attention. However, recently proposed methods such as fully convolutional network (FCN), SegNet, and U-Net still need to be further improved to provide better semantic segmentation when diagnosing breast cancer by ultrasound imaging, because of their low performance. In this article, we propose a channel attention module with multiscale grid average pooling (MSGRAP) for the precise segmentation of breast cancer regions in ultrasound images. We demonstrate the effectiveness of the channel attention module with MSGRAP for semantic segmentation and develop a novel semantic segmentation network with the proposed attention module for the precise segmentation of breast cancer regions in ultrasound images. While a conventional convolutional operation cannot use global spatial information on input images and only use the small local information in a kernel of a convolution filter, the proposed attention module allows using both global and local spatial information. In addition, through ablation studies, we come up with a network architecture for precise breast cancer segmentation in an ultrasound image. The proposed network was constructed with an open-source breast cancer ultrasound image data set, and its performance was compared with those of other state-of-the-art deep-learning models for the segmentation of breast cancer. The experimental results showed that our network outperformed other segmentation methods, and the proposed channel attention module improved the performance of the network for breast cancer segmentation in ultrasound images.
59 citations
TL;DR: The authors' two proposed networks substantially outperform current deep learning methods in terms of contrast-to-noise ratio (CNR) and strain ratio (SR) and have better SR by substantially reducing the underestimation bias.
Abstract: In this article, two novel deep learning methods are proposed for displacement estimation in ultrasound elastography (USE). Although convolutional neural networks (CNNs) have been very successful for displacement estimation in computer vision, they have been rarely used for USE. One of the main limitations is that the radio frequency (RF) ultrasound data, which is crucial for precise displacement estimation, has vastly different frequency characteristics compared with images in computer vision. Top-rank CNN methods used in computer vision applications are mostly based on a multilevel strategy, which estimates finer resolution based on coarser ones. This strategy does not work well for RF data due to its large high-frequency content. To mitigate the problem, we propose modified pyramid warping and cost volume network (MPWC-Net) and RFMPWC-Net, both based on PWC-Net, to exploit information in RF data by employing two different strategies. We obtained promising results using networks trained only on computer vision images. In the next step, we constructed a large ultrasound simulation database and proposed a new loss function to fine-tune the network to improve its performance. The proposed networks and well-known optical flow networks as well as state-of-the-art elastography methods are evaluated using simulation, phantom, and in vivo data. Our two proposed networks substantially outperform current deep learning methods in terms of contrast-to-noise ratio (CNR) and strain ratio (SR). Also, the proposed methods perform similar to the state-of-the-art elastography methods in terms of CNR and have better SR by substantially reducing the underestimation bias.
TL;DR: The results showed that choosing which layers to fine-tune is a critical task and demonstrated that fine-tuning the last layers of the network, which is the common practice for classification networks, is often the worst strategy.
Abstract: One way of resolving the problem of scarce and expensive data in deep learning for medical applications is using transfer learning and fine-tuning a network which has been trained on a large data set. The common practice in transfer learning is to keep the shallow layers unchanged and to modify deeper layers according to the new data set. This approach may not work when using a U-Net and when moving from a different domain to ultrasound (US) images due to their drastically different appearance. In this study, we investigated the effect of fine-tuning different sets of layers of a pretrained U-Net for US image segmentation. Two different schemes were analyzed, based on two different definitions of shallow and deep layers. We studied simulated US images, as well as two human US data sets. We also included a chest X-ray data set. The results showed that choosing which layers to fine-tune is a critical task. In particular, they demonstrated that fine-tuning the last layers of the network, which is the common practice for classification networks, is often the worst strategy. It may therefore be more appropriate to fine-tune the shallow layers rather than deep layers in US image segmentation when using a U-Net. Shallow layers learn lower level features which are critical in automatic segmentation of medical images. Even when a large US data set is available, we observed that fine-tuning shallow layers is a faster approach compared to fine-tuning the whole network.
TL;DR: Linear scanning of a single element of a PMUT array was performed on different tissue phantoms embedded with light-absorbing targets to successfully demonstrate B-mode PAI using PMUTs.
Abstract: A linear piezoelectric micromachined ultrasound transducer (PMUT) array was fabricated and integrated into a device for photoacoustic imaging (PAI) of tissue phantoms. The PMUT contained 65 array elements, with each element having 60 diaphragms of $60~\mu \text{m}$ diameter and $75~\mu \text{m}$ pitch. A lead zirconate titanate (PZT) thin film was used as the piezoelectric layer. The in-air vibration response of the PMUT array elements showed a first mode resonance between 6 and 8 MHz. Hydrophone measurements showed 16.2 kPa average peak ultrasound pressure output at 7.5 mm from one element excited with 5 Vpp input. A receive sensitivity of ~0.48 mV/kPa was observed for a PMUT array element with 0 dB gain. The PMUT array was bonded to a custom-printed circuit board to enable compact integration with an optical fiber bundle for PAI. A broad photoacoustic bandwidth of ~89% was observed for the photoacoustic response captured from absorbing pencil lead targets. Linear scanning of a single element of a PMUT array was performed on different tissue phantoms embedded with light-absorbing targets to successfully demonstrate B-mode PAI using PMUTs.
TL;DR: Fundamental mechanisms underlying the interaction between ultrasound and cancellous bone, including anisotropy, the effect of the fluid filler medium, phantoms, computational modeling of ultrasound propagation, acoustic microscopy, and nonlinear properties in cancellousBone are discussed.
Abstract: Ultrasound is now a clinically accepted modality in the management of osteoporosis. The most common commercial clinical devices assess fracture risk from measurements of attenuation and sound speed in cancellous bone. This review discusses fundamental mechanisms underlying the interaction between ultrasound and cancellous bone. Because of its two-phase structure (mineralized trabecular network embedded in soft tissue—marrow), its anisotropy, and its inhomogeneity, cancellous bone is more difficult to characterize than most soft tissues. Experimental data for the dependencies of attenuation, sound speed, dispersion, and scattering on ultrasound frequency, bone mineral density, composition, microstructure, and mechanical properties are presented. The relative roles of absorption, scattering, and phase cancellation in determining attenuation measurements in vitro and in vivo are delineated. Common speed of sound metrics, which entail measurements of transit times of pulse leading edges (to avoid multipath interference), are greatly influenced by attenuation, dispersion, and system properties, including center frequency and bandwidth. However, a theoretical model has been shown to be effective for correction for these confounding factors in vitro and in vivo . Theoretical and phantom models are presented to elucidate why cancellous bone exhibits negative dispersion, unlike soft tissue, which exhibits positive dispersion. Signal processing methods are presented for separating “fast” and “slow” waves (predicted by poroelasticity theory and supported in cancellous bone) even when the two waves overlap in time and frequency domains. Models to explain dependencies of scattering on frequency and mean trabecular thickness are presented and compared with measurements. Anisotropy, the effect of the fluid filler medium (marrow in vivo or water in vitro ), phantoms, computational modeling of ultrasound propagation, acoustic microscopy, and nonlinear properties in cancellous bone are also discussed.
TL;DR: The design space of the SH0 mode ADLs at GHz is first theoretically investigated, showing that the large coupling and sufficient spectral clearance to adjacent modes collectively enable the broadband performance of SH0 delay lines.
Abstract: We present the first group of GHz broadband SH0 mode acoustic delay lines (ADLs). The implemented ADLs adopt unidirectional transducer designs in a suspended X-cut lithium niobate thin film. The design space of the SH0 mode ADLs at GHz is first theoretically investigated, showing that the large coupling and sufficient spectral clearance to adjacent modes collectively enable the broadband performance of SH0 delay lines. The fabricated devices show 3-dB fractional bandwidth ranging from 4% to 34.3% insertion loss between 3.4 and 11.3 dB. Multiple delay lines have been demonstrated with center frequencies from 0.7 to 1.2 GHz, showing great frequency scalability. The propagation characteristics of SH0 in lithium niobate thin film are experimentally extracted. The demonstrated ADLs can potentially facilitate broadband signal processing applications.
TL;DR: A 3-D super-resolution (SR) pipeline based on data from a row–column (RC) array that uses the volumetric envelope data to find the bubble’s positions from their interpolated maximum signal and yields a high resolution in all three coordinates is presented.
Abstract: A 3-D super-resolution (SR) pipeline based on data from a row–column (RC) array is presented. The 3-MHz RC array contains 62 rows and 62 columns with a half wavelength pitch. A synthetic aperture (SA) pulse inversion sequence with 32 positive and 32 negative row emissions is used for acquiring volumetric data using the SARUS research ultrasound scanner. Data received on the 62 columns are beamformed on a GPU for a maximum volume rate of 156 Hz when the pulse repetition frequency is 10 kHz. Simulated and 3-D printed point and flow microphantoms are used for investigating the approach. The flow microphantom contains a 100- $\mu \text{m}$ radius tube injected with the contrast agent SonoVue. The 3-D processing pipeline uses the volumetric envelope data to find the bubble’s positions from their interpolated maximum signal and yields a high resolution in all three coordinates. For the point microphantom, the standard deviation on the position is (20.7, 19.8, 9.1) ${\mu }\text{m}~(x,y,z)$ . The precision estimated for the flow phantom is below $23~\mu \text{m}$ in all three coordinates, making it possible to locate structures on the order of a capillary in all three dimensions. The RC imaging sequence’s point spread function has a size of 0.58 $\times $ 1.05 $\times $ 0.31 mm3 ( $1.17\lambda \times 2.12\lambda \times 0.63\lambda $ ), so the possible volume resolution is 28900 times smaller than for SA RC B-mode imaging.
TL;DR: The challenges encountered in these measurements have led to different technological strategies, which are presented and critically discussed, with a focus on the US transducer geometry, Doppler signal processing, and fHR extraction methods.
Abstract: Fetal well-being is commonly assessed by monitoring the fetal heart rate (fHR). In clinical practice, the de facto standard technology for fHR monitoring is based on the Doppler ultrasound (US). Continuous monitoring of the fHR before and during labor is performed using a US transducer fixed on the maternal abdomen. The continuous fHR monitoring, together with simultaneous monitoring of the uterine activity, is referred to as cardiotocography (CTG). In contrast, for intermittent measurements of the fHR, a handheld Doppler US transducer is typically used. In this article, the technology of Doppler US for continuous fHR monitoring and intermittent fHR measurements is described, with emphasis on fHR monitoring for CTG. Special attention is dedicated to the measurement environment, which includes the clinical setting in which fHR monitoring is commonly performed. In addition, to understand the signal content of acquired Doppler US signals, the anatomy and physiology of the fetal heart and the surrounding maternal abdomen are described. The challenges encountered in these measurements have led to different technological strategies, which are presented and critically discussed, with a focus on the US transducer geometry, Doppler signal processing, and fHR extraction methods.
TL;DR: An encoder–decoder convolutional neural network architecture was constructed with custom modules and trained to identify the origin of the PA wavefronts inside an optically scattering deep-tissue medium and has broad applications such as diffused optical wavefront shaping, circulating melanoma cell detection, and real-time vascular surgeries.
Abstract: Optical photons undergo strong scattering when propagating beyond 1-mm deep inside biological tissue. Finding the origin of these diffused optical wavefronts is a challenging task. Breaking through the optical diffusion limit, photoacoustic (PA) imaging (PAI) provides high-resolution and label-free images of human vasculature with high contrast due to the optical absorption of hemoglobin. In real-time PAI, an ultrasound transducer array detects PA signals, and B-mode images are formed by delay-and-sum or frequency-domain beamforming. Fundamentally, the strength of a PA signal is proportional to the local optical fluence, which decreases with the increasing depth due to depth-dependent optical attenuation. This limits the visibility of deep-tissue vasculature or other light-absorbing PA targets. To address this practical challenge, an encoder–decoder convolutional neural network architecture was constructed with custom modules and trained to identify the origin of the PA wavefronts inside an optically scattering deep-tissue medium. A comprehensive ablation study provides strong evidence that each module improves the localization accuracy. The network was trained on model-based simulated PA signals produced by 16 240 blood-vessel targets subjected to both optical scattering and Gaussian noise. Test results on 4600 simulated and five experimental PA signals collected under various scattering conditions show that the network can localize the targets with a mean error less than 30 microns (standard deviation 20.9 microns) for targets below 40-mm imaging depth and 1.06 mm (standard deviation 2.68 mm) for targets at a depth between 40 and 60 mm. The proposed work has broad applications such as diffused optical wavefront shaping, circulating melanoma cell detection, and real-time vascular surgeries (e.g., deep-vein thrombosis).
TL;DR: The proposed real-time pipeline allows for a continuous acquisition and measurement workflow without user interaction, and has the potential to significantly reduce the time spent on the analysis and measurement error due to foreshortening, while providing quantitative volume measurements in the everyday echo lab.
Abstract: Volume and ejection fraction (EF) measurements of the left ventricle (LV) in 2-D echocardiography are associated with a high uncertainty not only due to interobserver variability of the manual measurement, but also due to ultrasound acquisition errors such as apical foreshortening. In this work, a real-time and fully automated EF measurement and foreshortening detection method is proposed. The method uses several deep learning components, such as view classification, cardiac cycle timing, segmentation and landmark extraction, to measure the amount of foreshortening, LV volume, and EF. A data set of 500 patients from an outpatient clinic was used to train the deep neural networks, while a separate data set of 100 patients from another clinic was used for evaluation, where LV volume and EF were measured by an expert using clinical protocols and software. A quantitative analysis using 3-D ultrasound showed that EF is considerably affected by apical foreshortening, and that the proposed method can detect and quantify the amount of apical foreshortening. The bias and standard deviation of the automatic EF measurements were −3.6 ± 8.1%, while the mean absolute difference was measured at 7.2% which are all within the interobserver variability and comparable with related studies. The proposed real-time pipeline allows for a continuous acquisition and measurement workflow without user interaction, and has the potential to significantly reduce the time spent on the analysis and measurement error due to foreshortening, while providing quantitative volume measurements in the everyday echo lab.
TL;DR: Results are promising for the application of deep learning to estimate correlation functions derived from ultrasound data in multiple areas of ultrasound imaging and beamforming, possibly replacing GPU-based approaches in low-power, remote, and synchronization-dependent applications.
Abstract: Deep fully connected networks are often considered “universal approximators” that are capable of learning any function. In this article, we utilize this particular property of deep neural networks (DNNs) to estimate normalized cross correlation as a function of spatial lag (i.e., spatial coherence functions) for applications in coherence-based beamforming, specifically short-lag spatial coherence (SLSC) beamforming. We detail the composition, assess the performance, and evaluate the computational efficiency of CohereNet, our custom fully connected DNN, which was trained to estimate the spatial coherence functions of in vivo breast data from 18 unique patients. CohereNet performance was evaluated on in vivo breast data from three additional patients who were not included during training, as well as data from in vivo liver and tissue mimicking phantoms scanned with a variety of ultrasound transducer array geometries and two different ultrasound systems. The mean correlation between the SLSC images computed on a central processing unit (CPU) and the corresponding DNN SLSC images created with CohereNet was 0.93 across the entire test set. The DNN SLSC approach was up to 3.4 times faster than the CPU SLSC approach, with similar computational speed, less variability in computational times, and improved image quality compared with a graphical processing unit (GPU)-based SLSC approach. These results are promising for the application of deep learning to estimate correlation functions derived from ultrasound data in multiple areas of ultrasound imaging and beamforming (e.g., speckle tracking, elastography, and blood flow estimation), possibly replacing GPU-based approaches in low-power, remote, and synchronization-dependent applications.
TL;DR: A novel perfect edge reflector is proposed with elimination of longitudinal spurious ripples in the design of the SH0 plate acoustic wave (PAW) resonators, applicable to the ultra-wideband filters and next-generation tunable filters.
Abstract: This study lays a foundation for the design of the SH0 plate acoustic wave (PAW) resonators based on LiNbO3 plate, and are applicable to the ultra-wideband filters and next-generation tunable filters. The coupling coefficient ( $k^{2}$ ) of SH0 mode is optimized to as high as 55% and wideband spurious are well controlled by analyzing the propagation characteristics of plate modes in LiNbO3. The SH0 mode is demonstrated to be slowly dispersive and features the superiority of interdigital transducer (IDT)—defining frequency over most other plate modes. On the device level, the transducer types and electrode materials are investigated and compared. In addition, for edge-reflected SH0 resonators, stopband dispersion analysis is provided and the longitudinal ripples are suppressed. For edge-reflected SH0 resonators, a novel perfect edge reflector is proposed with elimination of longitudinal spurious ripples.
TL;DR: The goal of this study was to evaluate the effectiveness of deep learning to realize a spatiotemporal filter in the context of SR-US processing and the performance of the 3DCNN was encouraging for real-time SR- US imaging.
Abstract: Super-resolution ultrasound (SR-US) imaging is a new technique that breaks the diffraction limit and allows visualization of microvascular structures down to tens of micrometers. The image processing methods for the spatiotemporal filtering needed in SR-US, such as singular value decomposition (SVD), are computationally burdensome and performed offline. Deep learning has been applied to many biomedical imaging problems, and trained neural networks have been shown to process an image in milliseconds. The goal of this study was to evaluate the effectiveness of deep learning to realize a spatiotemporal filter in the context of SR-US processing. A 3-D convolutional neural network (3DCNN) was trained on in vitro and in vivo data sets using SVD as ground truth in tissue clutter reduction. In vitro data were obtained from a tissue-mimicking flow phantom, and in vivo data were collected from murine tumors of breast cancer. Three training techniques were studied: training with in vitro data sets, training with in vivo data sets, and transfer learning with initial training on in vitro data sets followed by fine-tuning with in vivo data sets. The neural network trained with in vitro data sets followed by fine-tuning with in vivo data sets had the highest accuracy at 88.0%. The SR-US images produced with deep learning allowed visualization of vessels as small as $25~\mu \text{m}$ in diameter, which is below the diffraction limit (wavelength of $110~\mu \text{m}$ at 14 MHz). The performance of the 3DCNN was encouraging for real-time SR-US imaging with an average processing frame rate for in vivo data of 51 Hz with GPU acceleration.
TL;DR: In this article, a geometrical focusing approach was proposed to efficiently concentrate energy at the region of interest, and a phase correction focusing approach that takes the skull morphology into account. But the results were limited to the case of a single micro tube.
Abstract: High-resolution transcranial ultrasound imaging in humans has been a persistent challenge for ultrasound due to the imaging degradation effects from aberration and reverberation. These mechanisms depend strongly on skull morphology and have high variability across individuals. Here, we demonstrate the feasibility of human transcranial super-resolution imaging using a geometrical focusing approach to efficiently concentrate energy at the region of interest, and a phase correction focusing approach that takes the skull morphology into account. It is shown that using the proposed focused super-resolution method, we can image a 208- $\mu \text{m}$ microtube behind a human skull phantom in both an out-of-plane and an in-plane configuration. Individual phase correction profiles for the temporal region of the human skull were calculated and subsequently applied to transmit–receive a custom focused super-resolution imaging sequence through a human skull phantom, targeting the 208- $\mu \text{m}$ diameter microtube at 68.5 mm in depth and at 2.5 MHz. Microbubble contrast agents were diluted to a concentration of $1.6\times 10^{6}$ bubbles/mL and perfused through the microtube. It is shown that by correcting for the skull aberration, the RF signal amplitude from the tube improved by a factor of 1.6 in the out-of-plane focused emission case. The lateral registration error of the tube’s position, which in the uncorrected case was 990 $\mu \text{m}$ , was reduced to as low as 50 $\mu \text{m}$ in the corrected case as measured in the B-mode images. Sensitivity in microbubble detection for the phase-corrected case increased by a factor of 1.48 in the out-of-plane imaging case, while, in the in-plane target case, it improved by a factor of 1.31 while achieving an axial registration correction from an initial 1885- $\mu \text{m}$ error for the uncorrected emission, to a 284- $\mu \text{m}$ error for the corrected counterpart. These findings suggest that super-resolution imaging may be used far more generally as a clinical imaging modality in the brain.
TL;DR: The development of an optically transparent high-frequency ultrasonic transducer using lithium niobate single-crystal and indium–tin–oxide electrodes with up to 90% optical transmission in the visible-to-near-infrared spectrum is reported.
Abstract: We report the development of an optically transparent high-frequency ultrasonic transducer using lithium niobate single-crystal and indium–tin–oxide electrodes with up to 90% optical transmission in the visible-to-near-infrared spectrum. The center frequency of the transducer was at 36.9 MHz with 33.9%, at −6 dB fractional bandwidth. The photoacoustic imaging capability of the fabricated transducer was also demonstrated by successfully imaging a resolution target and mouse-ear vasculatures in vivo , which were irradiated by a 532 nm pulse laser transmitted through the transducer.
TL;DR: A protocol using teleultrasound, which is supported by the 5G network, is applied to explore the feasibility of solving the problem of early imaging assessment of COVID-19 to make early diagnosis and repeated assessment available in the isolation ward.
Abstract: Early diagnosis is critical for the prevention and control of the coronavirus disease 2019 (COVID-19). We attempted to apply a protocol using teleultrasound, which is supported by the 5G network, to explore the feasibility of solving the problem of early imaging assessment of COVID-19. Four male patients with confirmed or suspected COVID-19 were hospitalized in isolation wards in two different cities. Ultrasound specialists, located in two other different cities, carried out the robot-assisted teleultrasound and remote consultation in order to settle the problem of early cardiopulmonary evaluation. Lung ultrasound, brief echocardiography, and blood volume assessment were performed. Whenever difficulties of remote manipulation and diagnosis occurred, the alternative examination was repeated by a specialist from another city, and in sequence, remote consultation was conducted immediately to meet the consensus. The ultrasound specialists successfully completed the telerobotic ultrasound. Lung ultrasound indicated signs of pneumonia with varying degrees in all cases and mild pleural effusion in one case. No abnormalities of cardiac structure and function and blood volume were detected. Remote consultation on the issue of manipulation practice, and the diagnosis in one case was conducted. The cardiopulmonary information was delivered to the frontline clinicians immediately for further treatment. The practice of teleultrasound protocol makes early diagnosis and repeated assessment available in the isolation ward. Ultrasound specialists can be protected from infection, and personal protective equipment can be spared. Quality control can be ensured by remote consultations among doctors. This protocol is worth consideration as a feasible strategy for early imaging assessment in the COVID-19 pandemic.
TL;DR: Simulations performed using the open-source k-Wave acoustics toolbox were quantitatively validated against measurements of acoustic pressure in water and layered absorbing fluid media and demonstrate that when the acoustic medium properties and geometry are well known, accurate quantitative predictions of the acoustic field can be made using k- Wave.
Abstract: Models of ultrasound propagation in biologically relevant media have applications in planning and verification of ultrasound therapies and computational dosimetry. To be effective, the models must be able to accurately predict both the spatial distribution and amplitude of the acoustic pressure. This requires that the models are validated in absolute terms, which for arbitrarily heterogeneous media should be performed by comparison with measurements of the acoustic field. In this article, simulations performed using the open-source k-Wave acoustics toolbox, with a measurement-based source definition, were quantitatively validated against measurements of acoustic pressure in water and layered absorbing fluid media. In water, the measured and simulated spatial-peak pressures agreed to within 3% under linear conditions and 7% under nonlinear conditions. After propagation through a planar or wedge-shaped glycerol-filled phantom, the difference in spatial-peak pressure was 8.5% and 10.7%, respectively. These differences are within or close to the expected uncertainty of the acoustic pressure measurement. The −6 dB width and length of the focus agreed to within 4% in all cases, and the focal positions were within 0.7 mm for the planar phantom and 1.2 mm for the wedge-shaped phantom. These results demonstrate that when the acoustic medium properties and geometry are well known, accurate quantitative predictions of the acoustic field can be made using k-Wave.
TL;DR: This new temperature compensation procedure was applied to a set of guided wave signals collected in a blind trial of a guided wave pipe monitoring system, yielding residuals de-coupled from temperature and reduced by at least 50% as compared with those obtained using the standard approach.
Abstract: In guided wave structural health monitoring, defects are typically detected by identifying high residuals obtained through the baseline subtraction method, where an earlier measurement is subtracted from the “current” signal. Unfortunately, varying environmental and operational conditions (EOCs), such as temperature, also produce signal changes and hence, potentially, high residuals. While the majority of the temperature compensation methods that have been developed target the changed wave speed induced by varying temperature, a number of other effects are not addressed, such as the changes in attenuation, the relative amplitudes of different modes excited by the transducer, and the transducer frequency response. A temperature compensation procedure is developed, whose goal is to correct any spatially dependent signal change that is a systematic function of temperature. At each structural position, a calibration function that models the signal variation with temperature is computed and is used to correct the measurements, so that in the absence of a defect the residual is reduced to close to zero. This new method was applied to a set of guided wave signals collected in a blind trial of a guided wave pipe monitoring system using the T(0, 1) mode, yielding residuals de-coupled from temperature and reduced by at least 50% as compared with those obtained using the standard approach at positions away from structural features, and by more than 90% at features such as the pipe end. The method, therefore, promises a substantial improvement in the detectability of small defects, particularly at the existing pipe features.
TL;DR: First clinical results obtained with quantitative LUS spectroscopy when applied to the differentiation of pulmonary fibrosis are presented, with promising results strongly motivate toward the extension of the clinical study, aiming at analyzing a larger cohort of patients and including a broader range of pathologies.
Abstract: The application of ultrasound imaging to the diagnosis of lung diseases is nowadays receiving growing interest. However, lung ultrasound (LUS) is mainly limited to the analysis of imaging artifacts, such as B-lines, which correlate with a wide variety of diseases. Therefore, the results of LUS investigations remain qualitative and subjective, and specificity is obviously suboptimal. Focusing on the development of a quantitative method dedicated to the lung, in this work, we present the first clinical results obtained with quantitative LUS spectroscopy when applied to the differentiation of pulmonary fibrosis. A previously developed specific multifrequency ultrasound imaging technique was utilized to acquire ultrasound images from 26 selected patients. The multifrequency imaging technique was implemented on the ULtrasound Advanced Open Platform (ULA-OP) platform and an LA533 (Esaote, Florence, Italy) linear-array probe was utilized. RF data obtained at different imaging frequencies (3, 4, 5, and 6 MHz) were acquired and processed in order to characterize B-lines based on their frequency content. In particular, B-line native frequencies (the frequency at which a B-line exhibits the highest intensity) and bandwidth (the range of frequencies over which a B-line shows intensities within −6 dB from its highest intensity), as well as B-line intensity, were analyzed. The results show how the analysis of these features allows (in this group of patients) the differentiation of fibrosis with a sensitivity and specificity equal to 92% and 92%, respectively. These promising results strongly motivate toward the extension of the clinical study, aiming at analyzing a larger cohort of patients and including a broader range of pathologies.
TL;DR: This work investigates the possibility to effectively suppress crosstalk artifacts in MLT while improving image quality by applying beamforming techniques based on backscattered signals spatial coherence, and demonstrates that this class of algorithms can provide significant benefits compared with delay and sum.
Abstract: One of the current challenges in ultrasound imaging is achieving higher frame rates, particularly in cardiac applications, where tracking the heart motion and other rapid events can provide potential valuable diagnostic information. The main drawback of ultrasound high-frame-rate strategies is that usually they partly sacrifice image quality in order to speed up the acquisition time. In particular, multi-line transmission (MLT), which consists in transmitting multiple ultrasound beams simultaneously in different directions, has been proven able to improve frame rates in echocardiography, unfortunately generating artifacts due to inter-beam crosstalk interferences. This work investigates the possibility to effectively suppress crosstalk artifacts in MLT while improving image quality by applying beamforming techniques based on backscattered signals spatial coherence. Several coherence-based algorithms (i.e., short-lag filtered-delay multiply and sum beamforming, coherence and generalized coherence factor, phase and sign coherence, and nonlinear beamforming with $p$ th root compression) are implemented and compared, and their performance trends are evaluated when varying their design parameters. Indeed, experimental results of phantom and in vivo cardiac acquisitions demonstrate that this class of algorithms can provide significant benefits compared with delay and sum, well-suppressing artifacts (up to 48.5-dB lower crosstalk), and increasing image resolution (by up to 46.3%) and contrast (by up to 30 dB in terms of contrast ratio and 12.6% for generalized contrast-to-noise ratio) at the same time.
TL;DR: This article reports on the modeling, design, fabrication, and testing of high-performance X-cut lithium niobate (LN) laterally vibrating resonators (LVRs) operating around 50 MHz to exploit the high figure of merit (FoM) to provide for large passive voltage amplification in the front end of emerging radio frequency (RF) applications.
Abstract: This article reports on the modeling, design, fabrication, and testing of high-performance X-cut lithium niobate (LN) laterally vibrating resonators (LVRs) operating around 50 MHz. The objective of this work is to exploit the high figure of merit (FoM)—product of quality factor at series resonance ( ${Q}_{s}$ ) and electromechanical coupling ( ${k}_{t}^{{2}}$ )—to provide for large passive voltage amplification in the front end of emerging radio frequency (RF) applications, i.e., wake-up radio receivers (WuRx). Finite-element analysis (FEA) is performed to optimize the devices’ geometry and ensure simultaneous high ${Q}_{s}$ and ${k}_{t}^{{2}}$ . Resonators exhibiting ${Q}_{s} >5300$ and ${k}_{t}^{{2}} >27$ % are demonstrated, with FoM >1650—the highest value recorded for resonators in the megahertz range to the best of our knowledge. Finally, passive voltage gains between 35 and 57 V/V are showcased for capacitive loads ranging from 400 fF to 1 pF.