Neural Network Beamformer Research Articles

Improving Image Quality for Single-Angle Plane Wave Ultrasound Imaging With Convolutional Neural Network Beamformer.

Ultrafast ultrasound imaging based on plane wave (PW) compounding has been proposed for use in various clinical and preclinical applications, including shear wave imaging and super resolution blood flow imaging. Because the image quality afforded by PW imaging is highly dependent on the number of PW angles used for compounding, a tradeoff between image quality and frame rate occurs. In the present study, a convolutional neural network (CNN) beamformer based on a combination of the GoogLeNet and U-Net architectures was developed to replace the conventional delay-and-sum (DAS) algorithm to obtain high-quality images at a high frame rate. RF channel data are used as the inputs for the CNN beamformers. The outputs are in-phase and quadrature data. Simulations and phantom experiments revealed that the images predicted by the CNN beamformers had higher resolution and contrast than those predicted by conventional single-angle PW imaging with the DAS approach. In in vivo studies, the contrast-to-noise ratios (CNRs) of carotid artery images predicted by the CNN beamformers using three or five PWs as ground truths were approximately 12 dB in the transverse view, considerably higher than the CNR obtained using the DAS beamformer (3.9 dB). Most tissue speckle information was retained in the in vivo images produced by the CNN beamformers. In conclusion, only a single PW at 0° was fired, but the quality of the output image was proximal to that of an image generated using three or five PW angles. In other words, the quality-frame rate tradeoff of coherence compounding could be mitigated through the use of the proposed CNN for beamforming.

Phase Transmittance RBF Neural Network Beamforming for Static and Dynamic Channels

This letter proposes a novel beamforming technique for self-organizing wireless networks, such as those encountered in fifth generation (5G) and the Internet of Things (IoT). The proposed beamformer is based on a complex radial basis function artificial neural network (ANN), which allows phase transmittance between the input and output nodes. The beamforming algorithm is applied to a six-element uniform circular array, considering the mutual coupling between elements. The proposed technique is evaluated over critical static and dynamic scenarios. The static scenario approaches a multiuser environment, highly polluted with electromagnetic interference and plenty of severe in-band undesired signals. The dynamic scenario emulates a close air support military operation, in which an aircraft moves around the ground station at a high velocity. In both scenarios, nonlinearities at the receiver analog RF front end with a –10 dB signal-to-interference ratio (SIR) are considered. Results show that the proposed technique presents significantly a superior performance when compared to other solutions. For the static scenario, the novel beamformer is able to focus the array radiation pattern in the direction of the desired signal, achieving a zero symbol error rate. For the dynamic scenario, the beamforming is able to track the moving station, achieving a low symbol error rate for a wide angular range.

Fast and robust adaptive beamforming method based on complex‐valued RBF neural network

The adaptive beamforming is one of the key techniques for array signal processing. However, the matrix inversion operation in the existing methods will cost a large amount of computational complexity, which results in poor real-time processing ability. In order to reduce the amount of computational cost, a fast and robust adaptive beamforming method based on complex-valued radial basis function (CRBF) neural network is proposed. In the proposed method, the CRBF neural network is established, thus the direct matrix inversion is avoided by the nonlinear mapping processing from the array covariance matrix to the adaptive weight vector, and therefore the calculation speed of adaptive weight vectors is increased. Based on the simulation results, the proposed method is verified that the speed of adaptive beamforming is increased compared with sample matrix inversion (SMI) algorithm method and an improved performance is achieved compared with that of conventional real-valued RBF neural network beamformer.

Advanced beamforming for improved functional assessment in echocardiography

Echocardiography is one of the most used medical imaging exams. The data from these exams are often used to compute quantitative metrics of cardiac health including measures such as the global longitudinal strain (GLS) or ejection fraction. Metrics related to blood flow are also derived from echocardiography data. These metrics have great potential because ideally, they provide quantitative biomarkers to monitor cardiac function over time and compare patient function to population values. Unfortunately, echocardiography data is often severely corrupted by various forms of acoustic clutter including wavefront aberration, reverberation and off-axis scattering from bright structures like the ribs and lungs. These sources of clutter degrade ultrasound image quality and corrupt the ability to derive reliable quantitative biomarkers of cardiac function. To resolve problems like these, we have been developing non-linear (mathematically) beamformers. These include our aperture domain model image reconstruction (ADMIRE) beamformer and, more recently, our deep neural network beamformer. We have shown that these beamformers are able to reduce errors related to wavefront aberration, reverberation and off-axis scattering, and also that the ADMIRE beamformer can eliminate the underestimation of global longitudinal strain caused by high levels of stationary reverberation caused by the chest wall.

Reduced symmetric self-constructing fuzzy neural network beamforming detectors

Beamforming technology has been widely used in smart antenna systems that can increase the user's capacity and coverage in modern communication products. In this study, a powerful reduced symmetric self-constructing fuzzy neural network (RS-SCFNN) beamforming detector is proposed for multi-antenna-assisted systems. A novel training algorithm for the RS-SCFNN beamformer is proposed based on clustering of array input vectors and an adaptive minimum bit-error rate method. An inherent symmetric property of the array input signal space is exploited to make training procedure of RS-SCFNN more efficient than that of standard SCFNN. In addition, the required amount of fuzzy rules can be greatly reduced in the RS-SCFNN structure. Simulation results demonstrate that RS-SCFNN beamformer provides superior performance to the classical linear ones and the other non-linear ones (including symmetric radial basis function, SCFNN and S-SCFNN), especially when supporting a large amount of users in the rank-deficient multi-antenna-assisted system.

Correction to "Direction Finding in Phased Arrays with a Neural Network Beamformer" [Comments

Correction to "Direction Finding in Phased Arrays with a Neural Network Beamformer" [Comments

Direction finding in phased arrays with a neural network beamformer

Adaptive neural network processing of phased-array antenna received signals promises to decrease antenna manufacturing and maintenance costs while increasing mission uptime and performance between repair actions. We introduce one such neural network which performs aspects of digital beamforming with imperfectly manufactured, degraded, or failed antenna components. This paper presents measured results achieved with an adaptive radial basis function (ARBF) artificial neural network architecture which learned the single source direction finding (DF) function of an eight-element X-band array having multiple, unknown failures and degradations. We compare the single source DF performance of this ARBF neural network, whose internal weights are computed using a modified gradient descent algorithm, with another radial basis function network, Linnet, whose weights are calculated using linear algebra. Both networks are compared to a traditional DF approach using monopulse.