Complex-Valued Convolutional Gated Recurrent Neural Network for Ultrasound Beamforming.

Ultrasound detection is a potent tool for the clinical diagnosis of various diseases due to its real-time, convenient, and noninvasive qualities. Yet, existing ultrasound beamforming and related methods face a big challenge to improve both the quality and speed of imaging for the required clinical applications. The most notable characteristic of ultrasound signal data is its spatial and temporal features. Because most signals are complex-valued, directly processing them by using real-valued networks leads to phase distortion and inaccurate output. In this study, for the first time, we propose a complex-valued convolutional gated recurrent (CCGR) neural network to handle ultrasound analytic signals with the aforementioned properties. The complex-valued network operations proposed in this study improve the beamforming accuracy of complex-valued ultrasound signals over traditional real-valued methods. Further, the proposed deep integration of convolution and recurrent neural networks makes a great contribution to extracting rich and informative ultrasound signal features. Our experimental results reveal its outstanding imaging quality over existing state-of-the-art methods. More significantly, its ultrafast processing speed of only 0.07 s per image promises considerable clinical application potential. The code is available at https://github.com/zhangzm0128/CCGR.

Investigation of a method to estimate the average speed of sound using phase variances of element signals for ultrasound compound imaging.

PURPOSE: In the receive beamforming of an ultrasonography system, a B-mode image is reconstructed by assuming an average speed of sound (SoS) as a constant value. In our previous studies, we proposed a method for estimating the average SoS based on the coherence factor (CF) and the reciprocal of phase variances of element signals in delay-and-sum (DAS) beamforming. In this paper, we investigate the accuracy of estimation of the average SoS for compound imaging. METHODS: For this purpose, two numerical simulations were performed with k-Wave software. Also, the estimation methods based on the CF and the reciprocal were applied to in vivo data from the common carotid artery, and B-mode images were reconstructed using the estimated average SoS. RESULTS: In the first numerical simulation using an inhomogeneous phantom, the relationship between the accuracy and the transmission angles for the estimation was investigated, and the root mean squared errors (RMSEs) of estimates obtained based on the CF and the reciprocal of the phase variance were 1.25 ± 0.09, and 0.765 ± 0.17% at the transmission sequence of steering angles of (- 10°, - 5°, 0°, 5°, 10°), respectively. In the second numerical simulation using a cyst phantom, lateral resolutions were improved by reconstructing the image using the estimates obtained using the proposed strategy (reciprocal). By the proposed strategy, improvement of the continuity of the lumen-intima interface in the lateral direction was observed in the in vivo experiment. CONCLUSION: Consequently, the results indicated that the proposed strategy was beneficial for estimation of the average SoS and image reconstruction.

Fully Complex-Valued Gated Recurrent Neural Network for Ultrasound Imaging.

Ultrasound imaging is widely used in medical diagnosis. It has the advantages of being performed in real time, cost-efficient, noninvasive, and nonionizing. The traditional delay-and-sum (DAS) beamformer has low resolution and contrast. Several adaptive beamformers (ABFs) have been proposed to improve them. Although they improve image quality, they incur high computation cost because of the dependence on data at the expense of real-time performance. Deep-learning methods have been successful in many areas. They train an ultrasound imaging model that can be used to quickly handle ultrasound signals and construct images. Real-valued radio-frequency signals are typically used to train a model, whereas complex-valued ultrasound signals with complex weights enable the fine-tuning of time delay for enhancing image quality. This work, for the first time, proposes a fully complex-valued gated recurrent neural network to train an ultrasound imaging model for improving ultrasound image quality. The model considers the time attributes of ultrasound signals and uses complete complex-number calculation. The model parameter and architecture are analyzed to select the best setup. The effectiveness of complex batch normalization is evaluated in training the model. The effect of analytic signals and complex weights is analyzed, and the results verify that analytic signals with complex weights enhance the model performance to reconstruct high-quality ultrasound images. The proposed model is finally compared with seven state-of-the-art methods. Experimental results reveal its great performance.

Effect of Clutter Filter in High-Frame-Rate Ultrasonic Backscatter Coefficient Analysis.

High-frame-rate imaging with a clutter filter can clearly visualize blood flow signals and provide more efficient discrimination with tissue signals. In vitro studies using clutter-less phantom and high-frequency ultrasound suggested a possibility of evaluating the red blood cell (RBC) aggregation by analyzing the frequency dependence of the backscatter coefficient (BSC). However, in in vivo applications, clutter filtering is required to visualize echoes from the RBC. This study initially evaluated the effect of the clutter filter for ultrasonic BSC analysis for in vitro and preliminary in vivo data to characterize hemorheology. Coherently compounded plane wave imaging at a frame rate of 2 kHz was carried out in high-frame-rate imaging. Two samples of RBCs suspended by saline and autologous plasma for in vitro data were circulated in two types of flow phantoms without or with clutter signals. The singular value decomposition was applied to suppress the clutter signal in the flow phantom. The BSC was calculated using the reference phantom method, and it was parametrized by spectral slope and mid-band fit (MBF) between 4-12 MHz. The velocity distribution was estimated by the block matching method, and the shear rate was estimated by the least squares approximation of the slope near the wall. Consequently, the spectral slope of the saline sample was always around four (Rayleigh scattering), independently of the shear rate, because the RBCs did not aggregate in the solution. Conversely, the spectral slope of the plasma sample was lower than four at low shear rates but approached four by increasing the shear rate, because the aggregations were presumably dissolved by the high shear rate. Moreover, the MBF of the plasma sample decreased from -36 to -49 dB in both flow phantoms with increasing shear rates, from approximately 10 to 100 s-1. The variation in the spectral slope and MBF in the saline sample was comparable to the results of in vivo cases in healthy human jugular veins when the tissue and blood flow signals could be separated.

Contrast analysis in ultrafast ultrasound blood flow imaging of jugular vein.

PURPOSE: The contrasts of flowing blood in in vitro experiments using porcine blood and in vivo measurements of human jugular veins were analyzed to demonstrate that the hemorheological property was dependent on the shear rate. METHODS: Blood samples (45% hematocrit) suspended in saline or plasma were compared with examine the difference in viscoelasticity. Ultrafast plane-wave imaging at an ultrasonic center frequency of 7.5 MHz was performed on different steady flows in a graphite-agar phantom. Also, in vivo measurement was performed in young, healthy subjects and patients with diabetes. A spatiotemporal matrix of beamformed radio-frequency data was used for the singular value decomposition (SVD) clutter filter. The clutter-filtered B-mode image was calculated as the amplitude envelope normalized at the first frame in the diastolic phase to evaluate contrast. The shear rate was estimated as the velocity gradient perpendicular to the lateral axis. RESULTS: Although nonaggregated erythrocytes at a high shear rate exhibited a low echogenicity, the echogenicity in the plasma sample overall increased due to erythrocyte aggregation at a low shear rate. In addition, the frequency of detection of specular components, defined as components beyond twice the standard deviation of a contrast map obtained from a clutter-filtered B-mode image, increased in the porcine blood at a high shear rate and the venous blood in healthy subjects versus patients with diabetes. CONCLUSION: The possibility of characterizing hemorheological properties dependent on the shear rate and diabetes condition was indicated using ultrafast plane-wave imaging with an SVD-based clutter filter.

Two-Dimensional Wavenumber Analysis Implemented in Ultrasonic Vector Doppler Method with Focused Transmit Beams.

The multi-angle Doppler method was introduced for the estimation of velocity vectors by measuring axial velocities from multiple directions. We have recently reported that the autocorrelation-based velocity vector estimation could be ameliorated significantly by estimating the wavenumbers in two dimensions. Since two-dimensional wavenumber estimation requires a snapshot of an ultrasonic field, the method was first implemented in plane wave imaging. Although plane wave imaging is predominantly useful for examining blood flows at an extremely high temporal resolution, it was reported that the contrast in a B-mode image obtained with a few plane wave emissions was lower than that obtained with focused beams. In this study, the two-dimensional wavenumber analysis was first implemented in a framework with focused transmit beams. The simulations showed that the proposed method achieved an accuracy in velocity estimation comparable to that of the method with plane wave imaging. Furthermore, the performances of the methods implemented in focused beam and plane wave imaging were compared by measuring human common carotid arteries in vivo. Image contrasts were analyzed in normal and clutter-filtered B-mode images. The method with focused beam imaging achieved a better contrast in normal B-mode imaging, and similar velocity magnitudes and angles were obtained by both the methods with focused beam and plane wave imaging. In contrast, the method with plane wave imaging gave a better contrast in a clutter-filtered B-mode image and smaller variances in velocity magnitudes than those with focused beams.

On the Investigation of Autocorrelation-Based Vector Doppler Method With Plane Wave Imaging.

Although color flow imaging is one of the representative applications of the Doppler method, it can estimate only the velocity component in the direction of ultrasonic propagation, that is, the axial velocity component. The vector Doppler method with high-frame-rate plane wave imaging overcomes such a limitation by estimating the blood flow velocity vectors using the axial velocities obtained by emitting plane waves in multiple directions. The autocorrelation technique can be used for the estimation of the axial velocity using the phase shift of an ultrasonic echo signal between two transmit-receive events. The technique also requires the frequency of the received echo signal. Although the center frequency of the emitted ultrasonic signal is commonly used in the estimation of axial velocities, the center frequency should be estimated from the received signals. In this study, a method for the estimation of the center frequency designed particularly for the high-frame-rate plane wave imaging was developed. The proposed method estimates the wavenumbers of the received signal in lateral and vertical directions to estimate the wavenumber in the axial direction, from which the center frequency was estimated. The beam steering angle was also estimated from the wavenumbers in the two directions. The effect of the proposed method was validated in simulations. The absolute bias error (ABE) and root-mean squared error in estimated velocity vectors obtained by plane wave imaging with three beam steering angles (-15°, 0°, and 15°) were reduced from 9.27% and 14.80% to 1.15% and 8.75%, respectively, by the proposed method. The applicability of the proposed method to in vivo measurements was also demonstrated using the in vivo recordings of human common carotid arteries. Physiologically consistent blood flow velocity distributions were obtained with respect to three subjects using the proposed method.

Improving image contrast and accuracy in velocity estimation by convolution filters for intracardiac blood flow imaging.

In this study, the point spread function (PSF) of an ultrasound imaging system was estimated and used as a reference signal in a filtering method for improvement of image quality. The PSF of the imaging system was estimated from measured echo signals from an imaging target. Convolution filters (including deconvolution) were used for improvement of image contrast and spatial resolution. Furthermore, the accuracy in estimation of velocity vectors was evaluated for investigation of the impact of the proposed filters on velocity estimation. In the phantom experiment, contrast of the B-mode image was improved from 76.4 dB to 81.1 dB and 77.8 dB using the convolution and deconvolution filters, respectively. Also, the two-dimensional (2D) velocity distribution in the phantom was estimated by the block matching method, and the bias error (BE) in the estimated lateral velocity was reduced from -19.7% to 2.16% and 2.29% using the convolution and the deconvolution filters, respectively.

Suppression of reflected signals from substrate as clutters for cell measurements using acoustic impedance microscopy.

Recently, a method for estimating three-dimensional acoustic impedance profiles in cultured cells and human dermal organs was proposed by interpreting the reflected ultrasonic signal based on a 1-D transmission line model for acoustic impedance microscopy (AIM). However, AIM has a disadvantage that reflected signals from cells overlap with that from a reference substrate. Additionally, the amplitudes of the reflected signals from the specimens are significantly weaker than that from the substrate. In this paper, we proposed a new method for separation of those signals based on a concept of clutter filter, which had been developed for a color Doppler method in medical ultrasonic imaging. The proposed filter using singular value decomposition (SVD) could separate original signals into desired signals such as those from the substrate and cells. Additionally, an effect from a tilt of the substrate was investigated in this study. Separability of the proposed filter was evaluated by two investigations. First one was to evaluate the separability by estimating a correlation coefficient between the filtered signal and signal reflected from a position only with the substrate. Second one was to compare a slope of the substrate estimated from the original signal with that estimated from the filtered signals from the substrate. The experimental results showed that the proposed filter could separate signals from the substrate, and the compensation of the tilt of the substrate could improve the performance of the proposed filter.

Advances in ultrasonography: image formation and quality assessment.

Delay-and-sum (DAS) beamforming is widely used for generation of B-mode images from echo signals obtained with an array probe composed of transducer elements. However, the resolution and contrast achieved with DAS beamforming are determined by the physical specifications of the array, e.g., size and pitch of elements. To overcome this limitation, adaptive imaging methods have recently been explored extensively thanks to the dissemination of digital and programmable ultrasound systems. On the other hand, it is also important to evaluate the performance of such adaptive imaging methods quantitatively to validate whether the modification of the image characteristics resulting from the developed method is appropriate. Since many adaptive imaging methods have been developed and they often alter image characteristics, attempts have also been made to update the methods for quantitative assessment of image quality. This article provides a review of recent developments in adaptive imaging and image quality assessment.

Statistical Analysis of Ultrasonic Scattered Echoes Enables the Non-invasive Measurement of Temperature Elevations inside Tumor Tissue during Oncological Hyperthermia.

Non-invasive monitoring of temperature elevations inside tumor tissue is imperative for the oncological thermotherapy known as hyperthermia. In the present study, two cancer patients, one with a developing right renal cell carcinoma and the other with pseudomyxoma peritonei, underwent hyperthermia. The two patients were irradiated with radiofrequency current for 40 min during hyperthermia. We report the results of our clinical trial study in which the temperature increases inside the tumor tissues of patients with right renal cell carcinoma and pseudomyxoma peritonei induced by radiofrequency current irradiation for 40 min could be detected by statistical analysis of ultrasonic scattered echoes. The Nakagami shape parameter m varies depending on the temperature of the medium. We calculated the Nakagami shape parameter m by statistical analysis of the ultrasonic echoes scattered from the tumor tissues. The temperature elevations inside the tumor tissues were expressed as increases in brightness on 2-D hot-scale maps of the specific parameter αmod, indicating the absolute values of the percentage changes in m values. In the αmod map for each tumor tissue, the brightness clearly increased with treatment time. In quantitative analysis, the mean values of αmod were calculated. The mean value of αmod for the right renal cell carcinoma increased to 1.35 dB with increasing treatment time, and the mean value of αmod for pseudomyxoma peritonei increased to 1.74 with treatment time. The increase in both αmod brightness and the mean value of αmod implied temperature elevations inside the tumor tissues induced by the radiofrequency current; thus, the acoustic method is promising for monitoring temperature elevations inside tumor tissues during hyperthermia.

Measurement of flow velocity vectors in carotid artery using plane wave imaging with repeated transmit sequence.

PURPOSE: Doppler-based methods are widely used for blood flow imaging in clinical settings. However, they inherently estimate the velocity component only in the axial direction. Therefore, various studies of angle-independent methods have been conducted. The multi-angle Doppler method is one such angle-independent method, in which the velocity vector is estimated using axial velocities obtained from multiple directions by steering an ultrasonic beam. Recently, plane wave imaging, which realizes a very high frame rate of several thousand frames per second, was applied to the multi-angle Doppler method. However, the maximum detectable velocity, i.e., the aliasing limit, was reduced depending on the number of steering angles. In the present study, the feasibility of a specific transmit sequence, namely, the repeated transmit sequence, was examined using the plane-wave multi-angle Doppler method. METHOD: In the repeated transmit sequence, plane waves were emitted to the same direction twice, after which the steering angle was changed. By repeating the same procedure, a pair of beamformed radio-frequency (RF) signals could be obtained under each beam steering angle. By applying the autocorrelation method to each pair of RF signals, the time interval between the RF signals could be kept as the pulse repetition interval (PRI). The feasibility of such a transmit sequence was examined by numerical simulation and in vivo measurement of a human carotid artery. RESULTS: The simulation results showed that the maximum steering angles of over 10 degrees were not feasible with the linear array used in the present study. The feasible maximum steering angle would depend on the element pitch of the probe relative to the ultrasonic wavelength. By limiting the maximum steering angles to 5 and 10 degrees, bias errors were 9.2% and 11.3%, respectively, and root mean squared errors were 21.5% and 16.9%, respectively. Also, flow velocity vectors in a human carotid artery could be visualized with the proposed method. CONCLUSION: The multi-angle Doppler method was implemented in plane wave imaging with the repeated transmit sequence, and the proposed method was shown to be feasible through numerical simulation and in vivo measurement of a carotid artery.

A study on understanding the physical mechanism of change in ultrasonic envelope statistical property during temperature elevation.

PURPOSE: Our previous studies demonstrate that the variation in ultrasonic envelope statistics is correlated with the temperature change inside scattering media. This variation is identified as the change in the scatterer structure during thermal expansion or contraction. However, no specific evidence has been verified to date. This study numerically reproduces the change in the scatterer distribution during thermal expansion or contraction using finite element simulations and also investigates how the situation is altered by different material properties. METHODS: The material properties of a linear elastic solid depend on the thermal expansion coefficient, thermal conductivity, specific heat, and initial scatterer number density. Three-dimensional displacements, calculated in the simulation, were sequentially used to update the positions of the randomly distributed scatterers. Ultrasound signals from the scatterer distribution were generated by simulating a 7.5-MHz linear array transducer whose specifications were the same as those in the experimental measurements of several phantoms and excised porcine livers. To represent the change in the envelope statistical feature, the absolute value of the ratio change in the logarithmic Nakagami (NA) parameter, Δ m , at each time was calculated as a value normalized with the initial NA parameter. RESULTS: The change in the scatterer number density relates to the volume change during temperature elevation. The magnitude of the Δ m shift against the temperature change increases depending on the higher thermal expansion coefficient. In contrast, the relationship between Δ m and the scatterer number density is similar with any material property. Additionally, the changes in Δ m obtained by several experimental phantoms with low to high scatterer number densities are comparable with the numerical simulation results. CONCLUSIONS: The change in Δ m is indirectly related to the change in the scatterer number density owing to the volume change during thermal expansion or contraction.

Genetic response to low­intensity ultrasound on mouse ST2 bone marrow stromal cells.

Although low­intensity ultrasound (LIUS) is a clinically established procedure, the early cellular effect of LIUS on a genetic level has not yet been studied. The current study investigated the early response genes elicited by LIUS in bone marrow stromal cells (BMSCs) using global­scale microarrays and computational gene expression analysis tools. Mouse ST2 BMSCs were treated with LIUS [ISATA, 25 mW/cm2 for 20 min with a frequency of 1.11 MHz in a pulsed­wave mode (0.2­s burst sine waves repeated at 1 kHz)], then cultured for 0.5, 1 and 3 h at 37˚C. The time course of changes in gene expression was evaluated using GeneChip® high­density oligonucleotide microarrays and Ingenuity® Pathway Analysis tools. The results were verified by reverse transcription­quantitative polymerase chain reaction (RT­qPCR). A single exposure of LIUS did not affect cell morphology, cell growth or alkaline phosphatase activity. However, 61 upregulated and 103 downregulated genes were identified from 0.5 to 3 h after LIUS treatment. Two significant gene networks, labeled E and H, were identified from the upregulated genes, while a third network, labeled T, was identified from the downregulated genes. Gene network E or H containing the immediate­early genes FBJ osteosarcoma oncogene and early growth response 1 or the heat shock proteins heat shock protein 1a/b was associated mainly with the biological functions of bone physiology and protein folding or apoptosis, respectively. Gene network T containing transcription factors fos­like antigen 1 and serum response factor was also associated with the biological functions of the gene expression. RT­qPCR indicated that the expression of several genes in the gene networks E and H were elevated in LIUS­treated cells. LIUS was demonstrated to induce gene expression after short application in mouse ST2 BMSCs. The results of the present study provide a basis for the elucidation of the detailed molecular mechanisms underlying the cellular effects of LIUS.

Investigation of feasibility of singular value decomposition clutter filter in plane wave imaging with packet transmission sequence.

PURPOSE: Assessment of blood flow is an important function in diagnostic ultrasound imaging. Color flow imaging is one such method widely used in the clinical setting. Since autocorrelation suffers from aliasing, the time interval between successive transmissions of ultrasonic pulses should be as short as possible. For this purpose, a specific transmit-receive sequence, namely, packet transmission, is widely used in color flow imaging. Also, plane wave imaging recently introduced to ultrasound imaging significantly contributes to improvement of the temporal resolution. Furthermore, a singular value decomposition (SVD) clutter filter reportedly outperforms a conventional clutter filter. In the present study, the feasibility of the SVD clutter filter in plane wave imaging with the packet transmission sequence was investigated. METHOD: In the present study, the packet transmission sequence was implemented in plane wave imaging by sending plane waves multiple times in the same direction before changing the steering angle. In the first strategy, like conventional color flow imaging with line-by-line acquisition using a focused transmit beam, a clutter filter was applied to ultrasonic radio-frequency (RF) signals in each packet. In the second strategy, the number of transmissions per packet was set at two, and a clutter filter was applied to RF signals obtained from the first or second transmission in different packets. RESULTS: The in vivo experimental results on a human carotid artery showed that the second strategy with an SVD filter realized significantly better performance than the first strategy with a polynomial regression filter used as a conventional filter. CONCLUSION: An SVD clutter filter was feasible in plane wave imaging with the packet transmission sequence, and the performance was improved by limiting the number of transmissions per packet to two.

Novel Evidence Concerning Lacrimal Sac Movement Using Ultra-High-Frequency Ultrasound Examinations of Lacrimal Drainage Systems.

PURPOSE: Current hypothesis regarding the mechanism of active tear drainage is based on studies performed ex vivo or under nonphysiological conditions. Novel ultra-high-frequency ultrasound has the advantage of generating images with superior resolution, enabling measurements of low flow in small vessels, and the tracking of tissue motion in real time. The purpose of this study was to investigate the lacrimal drainage system and active drainage using this modality. METHODS: The upper lacrimal drainage system was investigated with 40-70 MHz ultrasound in 22 eyes in 13 patients. Irrigation confirmed a lacrimal obstruction in 10 eyes. Motion tracking was used to map movement of the lateral lacrimal sac wall and to measure flow when possible. RESULTS: The anatomy of the upper lacrimal drainage system was mapped in vivo, including the proximal canaliculi, which have not previously been imaged. The lacrimal sac lumen is slit shaped in its resting state but is distended when irrigated or if a nasolacrimal duct obstruction is present. Thus, the healthy lacrimal sac is not a cavity, and the medial retinaculum does not act against a stretched structure. Motion tracking visualized the "lacrimal pump," showing that the direction of motion of the lateral lacrimal sac wall is mainly in the sagittal plane during blinking. CONCLUSIONS: Ultra-high-frequency ultrasound allows detailed physiological monitoring of the upper lacrimal drainage system in vivo. Our findings suggest that current theories of active tear drainage need to be reappraised.

Ultrasound measurement of fetal arterial pulse pressure using phased-tracking methods: A phantom study and clinical experience with antenatal corticosteroid therapy.

AIM: This study aimed to compare the accuracy of fetal pulse pressure estimated with a vascular simulator with that obtained by a manometer (reference) and evaluate the pulse pressure in normal human fetuses and fetuses whose mothers received corticosteroids. METHODS: Fetal pulse pressure was estimated as the product of blood flow velocity and pulse wave velocity, based on the water hammer equation. Ultrasonic raw radiofrequency signals for blood flow velocity were captured from the fetal descending aortas at the diaphragm level, and pulse wave velocity was simultaneously measured from different directions using the phased-tracking method. First, the precision and accuracy of pulse pressure in the estimated method were verified by a circulatory phantom simulator, which reproduced fetal blood flow using a pulsating pump. Then, the pulse pressure of 98 normal human fetuses after 17 weeks of gestation and the fetal pulse pressure in 21 mothers who received antenatal corticosteroids for fetal maturation were measured. RESULTS: A significant correlation between the estimated pulse pressure values and the actual values was found in the phantom simulation (r = 0.99, P < 0.01). The estimated pulse pressure was significantly correlated with gestational age in normal fetuses (r = 0.74, P < 0.01). In steroid-treated pregnant women, fetal pulse pressure was observed to increase significantly on the second day of administration (P < 0.01). CONCLUSION: A noninvasive and accurate estimation model of fetal pulse pressure could be established using phased-tracking method, and this method has the potential to improve the assessment of human fetal hemodynamics.