No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Medical ultrasound imaging: to GPU or not to GPU?
Abstract
Medical ultrasound imaging stands out from other modalities in providing real-time diagnostic capability at an affordable price while being physically portable. This article explores the suitability of using GPUs as the primary signal and image processors for future medical ultrasound imaging systems. A case study on synthetic aperture (SA) imaging illustrates the promise of using high-performance GPUs in such systems.
... When fired individually, an increase in the number of elements leads to more signal processing and a higher computational complexity [112]. However, recent advances in embedded systems open the door to the use of advanced algorithms for real-time imaging. ...
... Processing Units (GPU) coupled with Central Processing Units (CPU) shows great promises in many domains [112]. For US imaging, many efforts are already oriented toward the use of heterogeneous computing to handle higher signal processing complexity for improved image quality [78,113,114]. ...
... Delay-and-Sum (DAS) beamforming is routinely adopted to produce image output in medical ultrasound imaging by compensating the time delay of the received echo according to the geometric path of propagation before coherent summation [1]. However, it suffers from intrinsic limitations such as insufficient image resolution and noticeable off-axis clutter. ...
... Note that the summation in Equation (2) is to produce the high-resolution CPWC image. After substituting Equation (2) into Equation (1), the conventional power Doppler detection of DAS beamforming in CPWC imaging is calculated as ...
Conventional ultrasonic coherent plane-wave (PW) compounding corresponds to Delay-and-Sum (DAS) beamforming of low-resolution images from distinct PW transmit angles. Nonetheless, the trade-off between the level of clutter artifacts and the number of PW transmit angle may compromise the image quality in ultrafast acquisition. Delay-Multiply-and-Sum (DMAS) beamforming in the dimension of PW transmit angle is capable of suppressing clutter interference and is readily compatible with the conventional method. In DMAS, a tunable p value is used to modulate the signal coherence estimated from the low-resolution images to produce the final high-resolution output and does not require huge memory allocation to record all the received channel data in multi-angle PW imaging. In this study, DMAS beamforming is used to construct a novel coherence-based power Doppler detection together with the complementary subset transmit (CST) technique to further reduce the noise level. For p = 2.0 as an example, simulation results indicate that the DMAS beamforming alone can improve the Doppler SNR by 8.2 dB compared to DAS counterpart. Another 6-dB increase in Doppler SNR can be further obtained when the CST technique is combined with DMAS beamforming with sufficient ensemble averaging. The CST technique can also be performed with DAS beamforming, though the improvement in Doppler SNR and CNR is relatively minor. Experimental results also agree with the simulations. Nonetheless, since the DMAS beamforming involves multiplicative operation, clutter filtering in the ensemble direction has to be performed on the low-resolution images before DMAS to remove the stationary tissue without coupling from the flow signal.
... Typically, the systems first use B-Mode real-time imaging to set the best probe position, then they switch to HFR acquisition to save raw data. Finally, the raw data are processed off-line in a PC, which may use a graphical processing unit (GPU) [4] to accomplish the heaviest elaborations. In this approach, after the acquisition, several seconds may be needed to view the results. ...
... In this paper we describe the new Virtual real-time (VRT) approach, which has been implemented on the ultrasound research scanner ULA-OP 256 [1] [4]. Raw data is continuously stored on RAM during the real-time investigation and processed according to the on-board calculation power. ...
... This is especially true in the case of MVS as unlike the single beamformer of MVT, every sub-band in the MVS method requires its own dedicated beamformer. To overcome this obstacle, the greatest research effort in this area is currently concentrated on implementing the MV beamformer using Graphics Processing Units (GPUs) [126,127]. In this way the MV method can be possibly used for real-time cardiac ultrasound imaging [128,129]. ...
... The corresponding number for the extraction of an MVS weight was 28 times higher imposing limitations to its direct use. The calculation load in the MVS case can only be managed by parallel receive processing which is lately linked with medical ultrasound adaptive beamforming [126,127]. ...
This thesis is concerned with developing algorithms for the precise localization
of ultrasound point scatterers with an eye to super-resolution ultrasound contrast
imaging. In medical ultrasound, the conventional resolution is limited by diffraction
and, in contrast to other sensing fields, point source imaging has not been extensively
investigated. Here, two independent methods were proposed aiming to increase the
lateral and the axial resolution respectively, by improving the localization accuracy
of a single scatterer. The methods were examined with simulated and experimental
data by using standard transmission protocols. Where a technique is applicable to
imaging of more complicated structures than point sources, this was also examined.
Further, a preliminary study was included with algorithm application to microbubbles
that are currently used in contrast enhanced ultrasound. It was demonstrated
that it is feasible to translate to ultrasonics, adaptive processes or techniques from
optical imaging/astronomy. This way, it was possible to overcome the diffraction
limit and achieve sub-wavelength localization. The accuracy gains are subject to
many parameters but may reach up to two orders of magnitude, and are based
exclusively on array signal processing. The latter is an important advantage since
current attempts for super-resolution ultrasound are image-based which is generally
undesired.
... In subsequent studies, acquisition parameters such as number of compounded angles [44], number of pulse cycles, pulse frequency modulation (chirp) [57], SNR, probe elements geometry, MB size distribution, MB nonlinear response, could be interesting to better understand the underlying limitations of ULM and progress towards finding optimal imaging parameters. ...
The resolution of 3D Ultrasound Localization Microscopy (ULM) is determined by acquisition parameters such as frequency and transducer geometry but also by microbubble (MB) concentration, which is linked to the total acquisition time needed to sample the vascular tree at different scales. In this study, we introduce a novel 3D anatomically-realistic ULM simulation framework based on two-photon microscopy (2PM) and in-vivo MB perfusion dynamics. As a proof of concept, using metrics such as MB localization error, MB count and network filling, we quantify the effect of MB concentration and PSF volume by varying probe transmit frequency (3-15 MHz). We found that while low frequencies can achieve sub-wavelength resolution as predicted by theory, they are also associated with prolonged acquisition times to map smaller vessels, thus limiting effective resolution (i.e., the smallest vessel that can be reconstructed). A linear relationship was found between the maximal MB concentration and the inverse of the point spread function (PSF) volume. Since inverse PSF volume roughly scales cubically with frequency, the reconstruction of the equivalent of 10 minutes at 15 MHz would require hours at 3 MHz. We expect that these findings can be leveraged to achieve effective reconstruction and serve as a guide for choosing optimal MB concentrations in ULM.
... Plane-wave (PW) imaging depends on the unfocused transmit wave to illuminate a wide field of view [1], and the corresponding backscattered echoes are time-compensated for the geometric path of propagation from each receiving channel to the image pixel. When the delayed channel echoes are summed to produce the image output of that pixel, it is referred to as Delay-and-Sum (DAS) beamforming [2]. ...
Coherent plane wave compounding (CPWC) reconstructs transmit focusing by coherently summing several low-resolution plane-wave (PW) images from different transmit angles to improve its image resolution and quality. The high frame rate of CPWC imaging enables a much larger number of Doppler ensembles such that the Doppler estimation of blood flow becomes more reliable. Due to the unfocused PW transmission, however, one major limitation of the Doppler estimation in CPWC imaging is the relatively low signal-to-noise ratio (SNR). Conventionally, the Doppler power is estimated by a zero-lag autocorrelation which reduces the noise variance, but not the noise level. A higher-lag autocorrelation method such as the first-lag (R(1)) power Doppler image has been developed to take advantage of the signal coherence in the temporal direction for suppressing uncorrelated random noises. In this paper, we propose a novel Temporal Multiply-and-Sum (TMAS) power Doppler detection method to further improve the noise suppression of the higher-lag method by modulating the signal coherence among the temporal correlation pairs in the higher-lag autocorrelation with a tunable pt value. Unlike the adaptive beamforming methods which demand for either receive–channel–domain or transmit–domain processing to exploit the spatial coherence of the blood flow signal, the proposed TMAS power Doppler can share the routine beamforming architecture with CPWC imaging. The simulated results show that when it is compared to the original R(1) counterpart, the TMAS power Doppler image with the pt value of 2.5 significantly improves the SNR by 8 dB for the cross-view flow velocity within the Nyquist rate. The TMAS power Doppler, however, suffers from the signal decorrelation of the blood flow, and thus, it relies on not only the pt value and the flow velocity, but also the flow direction relative to the geometry of acoustic beam. The experimental results in the flow phantom and in vivo dataset also agree with the simulations.
... Over the past two decades, the emergence of graphics processing units (GPUs) has facilitated high performance computing [45][46][47]. Medical ultrasound imaging also took advantage of this technology [48][49][50]. Different beamforming techniques such as Capon [49,[51][52][53], short-lag spatial coherence [50,[54][55][56], synthetic aperture sequential beamforming [57,58], delay-multiply-and-sum [59] and double-stage delay-multiply-and-sum [60,61], and volumetric imaging systems [62][63][64] were accelerated by GPU. GPU-based beamforming softwares were developed [65,66], and ultrasound vector flow imaging systems were accelerated [67][68][69]. ...
In a recent study, we proposed a technique to correct aberration caused by the skull and reconstruct a transcranial B-mode image with a refraction-corrected synthetic aperture imaging (SAI) scheme. Given a sound speed map, the arrival times were calculated using a fast marching technique (FMT), which solves the Eikonal equation and, therefore, is computationally expensive for real-time imaging. In this article, we introduce a two-point ray tracing method, based on Fermat’s principle, for fast calculation of the travel times in the presence of a layered aberrator in front of the ultrasound probe. The ray tracing method along with the reconstruction technique is implemented on a graphical processing unite (GPU). The point spread function (PSF) in a wire phantom image reconstructed with the FMT and the GPU implementation was studied with numerical synthetic data and experiments with a bone-mimicking plate and a sagittally cut human skull. The numerical analysis showed that the error on travel times is less than 10% of the ultrasound temporal period at 2.5 MHz. As a result, the lateral resolution was not significantly degraded compared with images reconstructed with FMT-calculated travel times. The results using the synthetic, bone-mimicking plate, and skull dataset showed that the GPU implementation causes a lateral/axial localization error of 0.10/0.20, 0.15/0.13, and 0.26/0.32 mm compared with a reference measurement (no aberrator in front of the ultrasound probe), respectively. For an imaging depth of 70 mm, the proposed GPU implementation allows reconstructing 19 frames/s with full synthetic aperture (96 transmission events) and 32 frames/s with multiangle plane wave imaging schemes (with 11 steering angles) for a pixel size of
$200~\mu \text{m}$
. Finally, refraction-corrected power Doppler imaging is demonstrated with a string phantom and a bone-mimicking plate placed between the probe and the moving string. The proposed approach achieves a suitable frame rate for clinical scanning while maintaining the image quality.
... There are two main approaches to implementing all the above RX functionalities. In "software-based" ultrasound scanners, each function is implemented in programmable devices such as the graphic processor units (GPUs) [7][8][9]. In the so-called hardware-based scanners, specific devices such as field-programmable gate arrays (FPGAs), digital signal processors (DSPs), and GPUs are used [10][11][12]. ...
The spread of high frame rate and 3-D imaging techniques has raised pressing requirements for ultrasound systems. In particular, the processing power and data transfer rate requirements may be so demanding to hinder the real-time (RT) implementation of such techniques. This paper first analyzes the general requirements involved in RT ultrasound systems. Then, it identifies the main bottlenecks in the receiving section of a specific RT scanner, the ULA-OP 256, which is one of the most powerful available open scanners and may therefore be assumed as a reference. This case study has evidenced that the “star” topology, used to digitally interconnect the system’s boards, may easily saturate the data transfer bandwidth, thus impacting the achievable frame/volume rates in RT. The architecture of the digital scanner was exploited to tackle the bottlenecks by enabling a new “ring“ communication topology. Experimental 2-D and 3-D high-frame-rate imaging tests were conducted to evaluate the frame rates achievable with both interconnection modalities. It is shown that the ring topology enables up to 4400 frames/s and 510 volumes/s, with mean increments of +230% (up to +620%) compared to the star topology.
... Esaote, Philips) are equipped with GPU cards [1, 2] Furthermore, some recent denoising methods for ultrasound images are developed on GPUs [9,22], which are also used for denoising. Furthermore, the application of GPUs to image processing for future medical ultrasound imaging systems [59] presents the advantages of GPUs over CPUs in terms of performance, power consumption, and cost. ...
Ultrasound images are widespread in medical diagnosis for muscle-skeletal, cardiac, and obstetrical diseases, due to the efficiency and non-invasiveness of the acquisition methodology. However, ultrasound acquisition introduces noise in the signal, which corrupts the resulting image and affects further processing steps, e.g. segmentation and quantitative analysis. We define a novel deep learning framework for the real-time denoising of ultrasound images. Firstly, we compare state-of-the-art methods for denoising (e.g. spectral, low-rank methods) and select WNNM ( Weighted Nuclear Norm Minimisation ) as the best denoising in terms of accuracy, preservation of anatomical features, and edge enhancement. Then, we propose a tuned version of WNNM ( tuned-WNNM ) that improves the quality of the denoised images and extends its applicability to ultrasound images. Through a deep learning framework, the tuned-WNNM qualitatively and quantitatively replicates WNNM results in real-time. Finally, our approach is general in terms of its building blocks and parameters of the deep learning and high-performance computing framework; in fact, we can select different denoising algorithms and deep learning architectures.
... In subsequent studies, acquisition parameters such as number of compound angles [34], number of pulse cycles, pulse frequency modulation (chirp) [41], SNR, probe elements geometry, MB size distribution just to name a few could be interesting to better understand the underlying limitations of ULM and progress towards finding optimal imaging parameters. ...
The resolution of 3D Ultrasound Localization Microscopy (ULM) is determined by acquisition parameters such as frequency and transducer geometry but also by microbubble (MB) concentration, which is also linked to the total acquisition time needed to sample the vascular tree at different scales. In this study, we introduce a novel 3D anatomically- and physiologically-realistic ULM simulation framework based on two-photon microscopy (2PM) and in-vivo MB perfusion dynamics. As a proof of concept, using metrics such as MB localization error, MB count and network filling, we could quantify the effect of MB concentration and PSF volume by varying probe transmit frequency (3-15 MHz). We find that while low frequencies can achieve sub-wavelength resolution as predicted by theory, they are also associated with prolonged acquisition times to map smaller vessels, thus limiting effective resolution. A linear relationship was found between maximal MB concentration and inverse point spread function (PSF) volume. Since inverse PSF volume roughly scales cubically with frequency, the reconstruction of the equivalent of 10 minutes at 15 MHz would require hours at 3 MHz. We expect that these findings can be leveraged to achieve effective reconstruction and serve as a guide for choosing optimal MB concentrations in ULM.
... H EALTH professionals extensively use 2D Ultrasound (US) videos and images to visualize and measure internal organs for various purposes including evaluation of muscle architectural and functional changes for diagnostic purposes, and as an outcome measure to evaluate rehabilitation effects. Compared to Computerized Tomography (CT) scans, US imaging is low-cost [1], more sensitive [2], and does not expose patients to ionizing radiation [3]. ...
Health professionals extensively use Two- Dimensional (2D) Ultrasound (US) videos and images to visualize and measure internal organs for various purposes including evaluation of muscle architectural changes. US images can be used to measure abdominal muscles dimensions for the diagnosis and creation of customized treatment plans for patients with Low Back Pain (LBP), however, they are difficult to interpret. Due to high variability, skilled professionals with specialized training are required to take measurements to avoid low intra-observer reliability. This variability stems from the challenging nature of accurately finding the correct spatial location of measurement endpoints in abdominal US images. In this paper, we use a Deep Learning (DL) approach to automate the measurement of the abdominal muscle thickness in 2D US images. By treating the problem as a localization task, we develop a modified Fully Convolutional Network (FCN) architecture to generate blobs of coordinate locations of measurement endpoints, similar to what a human operator does. We demonstrate that using the TrA400 US image dataset, our network achieves a Mean Absolute Error (MAE) of 0.3125 on the test set, which almost matches the performance of skilled ultrasound technicians. Our approach can facilitate next steps for automating the process of measurements in 2D US images, while reducing inter-observer as well as intra-observer variability for more effective clinical outcomes.
... Since multi-core CPUs can also be used in parallel computing, the hybrid application of multi-core CPUs and GPUs should theoretically be a better solution. So et al. [32] discussed the GPU and CPU used in ultrasound imaging and drew a collusion that hybrid systems, e.g., CPUs-GPUs, were better than GPUs alone. Therefore, a cheap PC where a multi-core CPU and a GPU are installed is used in our work. ...
A novel beamforming algorithm named Delay Multiply and Sum (DMAS), which excels at enhancing the resolution and contrast of ultrasonic image, has recently been proposed. However, there are nested loops in this algorithm, so the calculation complexity is higher compared to the Delay and Sum (DAS) beamformer which is widely used in industry. Thus, we proposed a simple vector-based method to lower its complexity. The key point is to transform the nested loops into several vector operations, which can be efficiently implemented on many parallel platforms, such as Graphics Processing Units (GPUs), and multi-core Central Processing Units (CPUs). Consequently, we considered to implement this algorithm on such a platform. In order to maximize the use of computing power, we use the GPUs and multi-core CPUs in mixture. The platform used in our test is a low cost Personal Computer (PC), where a GPU and a multi-core CPU are installed. The results show that the hybrid use of a CPU and a GPU can get a significant performance improvement in comparison with using a GPU or using a multi-core CPU alone. The performance of the hybrid system is increased by about 47%–63% compared to a single GPU. When 32 elements are used in receiving, the fame rate basically can reach 30 fps. In the best case, the frame rate can be increased to 40 fps.
... The PC installed in the system contains a total of 10 PCIe interfaces (Supermicro X10DRG-O+-CPU, X9DRG-O-PCIE) and integrates a dual Intel Xeon processor (2x E5-2697AV4, 32 cores). In addition to the four PCIe interfaces of the 256-channel ultrasound sub-systems, further PCIe interfaces are used to connect four additional Nvidia graphics cards, which are intended for further signal processing on the GPUs [22]. In total, the system uses 128 Gbytes of DDR4 memory, which is used for the DMA transfer of the raw ultrasound data, allowing the user to access the single-element channel data of all 1024 channels in parallel. ...
Volumetric ultrasound imaging is of great importance in many medical fields, especially in cardiology, but also in therapy monitoring applications. For development of new imaging technologies and scanning strategies, it is crucial to be able to use a hardware platform that is as free and flexible as possible and does not restrict the user in his research in any way. For this purpose, multi-channel ultrasound systems are particularly suitable, as they are able to control each individual element of a matrix array without the use of a multiplexer. We set out to develop a fully integrated, compact 1024-channel ultrasound system that provides full access to all transmission parameters and all digitized raw data of each transducer element. For this purpose, we synchronize four research scanners of our latest “DiPhAS” ultrasound research system generation, each with 256 parallel channels, all connected to a single PC on whose GPUs the entire signal processing is performed. All components of the system are housed in a compact, movable 19-inch rack. The system is designed as a general-purpose platform for research in volumetric imaging; however, the first-use case will be therapy monitoring by tracking radiation-sensitive ultrasound contrast agents.
... H EALTH professionals extensively use 2D Ultrasound (US) videos and images to visualize and measure internal organs for various purposes including evaluation of muscle architectural and functional changes for diagnostic purposes, and as an outcome measure to evaluate rehabilitation effects. Compared to Computerized Tomography (CT) scans, US imaging is low-cost [1], more sensitive [2], and does not expose patients to ionizing radiation [3]. ...
Health professionals extensively use 2D US videos and images to visualize and measure internal organs for various purposes including evaluation of muscle architectural changes. US images can be used to measure abdominal muscles dimensions for the diagnosis and creation of customized treatment plans for patients with LBP, however, they are difficult to interpret. Due to high variability, skilled professionals with specialized training are required to take measurements to avoid low intra-observer reliability. This variability stems from the challenging nature of accurately finding the correct spatial location of measurement endpoints in abdominal US images. In this paper, we use a DL approach to automate the measurement of the abdominal muscle thickness in 2D US images. By treating the problem as a localization task, we develop a modified FCN architecture to generate blobs of coordinate locations of measurement endpoints, similar to what a human operator does. We demonstrate that using the TrA400 US image dataset, our network achieves a MAE of 0.3125 on the test set, which almost matches the performance of skilled ultrasound technicians. Our approach can facilitate next steps for automating the process of measurements in 2D US images, while reducing inter-observer as well as intra-observer variability for more effective clinical outcomes.
... The so-called software-based open platforms [6,11,12] contain limited front-end electronics, and raw echo data are immediately digitized and streamed toward a PC where they are usually processed off-line by GPU boards such as, e.g., [18][19][20]. The main advantage of this approach is that the high computing power of GPUs is available to researchers having average knowledge of C language coding [21][22][23][24]. ...
Methods of increasing complexity are currently being proposed for ultrasound (US) echographic signal processing. Graphics Processing Unit (GPU) resources allowing massive exploitation of parallel computing are ideal candidates for these tasks. Many high-performance US instruments, including open scanners like ULA-OP 256, have an architecture based only on Field-Programmable Gate Arrays (FPGAs) and/or Digital Signal Processors (DSPs). This paper proposes the implementation of the embedded NVIDIA Jetson Xavier AGX module on board ULA-OP 256. The system architecture was revised to allow the introduction of a new Peripheral Component Interconnect Express (PCIe) communication channel, while maintaining backward compatibility with all other embedded computing resources already on board. Moreover, the Input/Output (I/O) peripherals of the module make the ultrasound system independent, freeing the user from the need to use an external controlling PC.
... Recently, efficient architectures for SA-VS imaging have been proposed and implemented in prototype systems [21,24]. In addition, recent advances in graphic processor unit (GPU) computing in medical ultrasound imaging may facilitate more rapid commercialization of SA techniques [25,26]. ...
High-frequency ultrasound (HFUS) imaging has emerged as an essential tool for pre-clinical studies and clinical applications such as ophthalmic and dermatologic imaging. HFUS imaging systems based on array transducers capable of dynamic receive focusing have considerably improved the image quality in terms of spatial resolution and signal-to-noise ratio (SNR) compared to those by the single-element transducer-based one. However, the array system still suffers from low spatial resolution and SNR in out-of-focus regions, resulting in a blurred image and a limited penetration depth. In this paper, we present synthetic aperture imaging with a virtual source (SA-VS) for an ophthalmic application using a high-frequency convex array transducer. The performances of the SA-VS were evaluated with phantom and ex vivo experiments in comparison with the conventional dynamic receive focusing method. Pre-beamformed radio-frequency (RF) data from phantoms and excised bovine eye were acquired using a custom-built 64-channel imaging system. In the phantom experiments, the SA-VS method showed improved lateral resolution (>10%) and sidelobe level (>4.4 dB) compared to those by the conventional method. The SNR was also improved, resulting in an increased penetration depth: 16 mm and 23 mm for the conventional and SA-VS methods, respectively. Ex vivo images with the SA-VS showed improved image quality at the entire depth and visualized structures that were obscured by noise in conventional imaging.
... Some open-architecture scanner platforms have been developed by academic laboratories [42][43][44], while a few commercial prototyping systems have also been marketed. To facilitate fast processing of raw prebeamform data, high-speed software-based signal processing frameworks are concurrently being devised using parallel computing hardware such as graphics processing units [45]. Recent achievements on this topic include beamforming [46], speckle tracking [47], adaptive apodization [48], and eigen-processing [49]. ...
In this book chapter, a pedagogical overview on the technical foundations
of ultrasound imaging will be provided to guide readers who are new to the
biomedical ultrasound field. The intent here is to equip readers with fundamental
principles that would serve well as background knowledge to understand the broad
range of technical concepts covered in this book. Not only will the general physics
of ultrasound and the key imaging considerations be outlined, engineering aspects
such as system hardware will also be covered. In addition, commentary on the
emerging trend toward high-frame-rate imaging will be included to highlight latest
innovation thrusts in ultrasound imaging.
... 54 Software-based open scanners [13], [14], which imple-55 ment the signal processing chain through high-level program-56 ming [1], typically approach this problem by transmitting 57 plane or diverging waves and acquiring raw echo-data in local 58 random access memories (RAMs). The data are then sent, 59 usually through PCI-express connection, to a PC containing 60 powerful graphic processing units (GPUs) [15], [16]. In this 61 approach, the results are typically viewed with some delay 62 after the acquisition. ...
The recent development of high-frame-rate (HFR) imaging/Doppler methods based on the transmission of plane or diverging waves, has proposed new challenges to echographic data management and display. Due to the huge amount of data that need to be processed at very high speed, the pulse repetition frequency (PRF) is typically limited to hundreds Hz or few kHz. In Doppler applications, a PRF limitation may result unacceptable since it inherently translates to a corresponding limitation in the maximum detectable velocity. In this paper, the ULA-OP 256 implementation of a novel ultrasound modality, called virtual real-time (VRT), is described. First, for a given HFR real-time modality, the scanner displays the processed results while saving channel data into an internal buffer. Then, ULA-OP 256 switches to VRT mode, according to which the raw data stored in the buffer are immediately re-processed by the same hardware used in real-time. In the two phases, the ULA-OP 256 calculation power can be differently distributed to increase the acquisition frame rate or the quality of processing results. VRT was here used to extend the PRF limit in a multi-line vector Doppler application. In real-time, the PRF was maximized at the expense of the display quality; in VRT, data were reprocessed at a lower rate in a high-quality display format, which provides more detailed flow information. Experiments are reported in which the multi-line vector Doppler technique is shown capable of working at 16 kHz PRF, so that flow jet velocities higher up to 3 m/s can be detected.
... Also, the GPU has many cores, which can be most efficiently utilized for high throughput applications that are common to many classes of database queries. Furthermore, GPUs have been noted for having a greater energy efficiency than the CPU for many applications [42], and have been employed for their relatively low monetary cost per unit metric (e.g., floating point operations per second) [30,44]. Due to the characteristics outlined above, the most powerful supercomputers rely on GPUs. ...
The similarity self-join finds all objects in a dataset that are within a search distance, ∈, of each other. As such, the self-join is a building block of many algorithms. In high dimensions, indexing structures become increasingly ineffective at pruning the search, making the self-join challenging to compute efficiently. We advance a GPU-accelerated self-join algorithm targeted towards high dimensional data. The massive parallelism afforded by the GPU and high aggregate memory bandwidth makes the architecture well-suited for data-intensive workloads. We leverage a grid-based GPU-tailored index to perform range queries, and propose the following optimizations: (i) a trade-off between candidate set filtering and index search overhead by exploiting properties of the index; (ii) reordering the data based on variance in each dimension to improve the filtering power of the index; and (iii) a pruning method for reducing the number of expensive distance calculations. Our algorithm generally outperforms a parallel CPU state-of-the-art approach.
... It makes use of a graphical processing unit (GPU) to perform both signal processing and display rendering, effectively leveraging the dual role that GPUs play as a many-core parallel processor and as a hardware-accelerated graphics renderer. Note that, from an ultrasound computing standpoint, GPU is known to be more energy-efficient than the central processing unit [25], and its programing is less labor-intensive than that for hardwareoriented OPs that use field-programmable gate array (FPGA) as the primary computing device [26]. These points of merit have motivated us to develop our live CESI scanning platform using a software-oriented approach. ...
Complex flow patterns are prevalent in the vasculature, but they are difficult to image noninvasively in real time. This paper presents the first real-time scanning platform for a high-frame-rate ultrasound technique called color-encoded speckle imaging (CESI) and its use in visualizing arterial flow dynamics
in vivo
. CESI works by simultaneously rendering flow speckles and color-coded flow velocity estimates on a time-resolved basis. Its live implementation was achieved by integrating a 192-channel programmable ultrasound front-end module, a 4.8-GB/s capacity data streaming link, and a series of computing kernels implemented on the graphical processing unit (GPU) for beamforming and Doppler processing. A slow-motion replay mode was also included to offer coherent visualization of CESI frames acquired at high frame rate [3000 frames per second (fps) in our experiments]. The live CESI scanning platform was found to be effective in facilitating real-time image guidance (at least 20 fps for live video display with 55-fps GPU processing throughout).
In vivo
pilot trials also showed that live CESI, when running in replay mode, can temporally resolve triphasic flow at the brachial bifurcation and can reveal flow dynamics in the brachial vein during a fist-clenching maneuver. Overall, live CESI has potential for use in routine investigations
in vivo
that seek to identify complex flow dynamics in real time and relate these dynamics to vascular physiology.
... The introduction of GPU processing makes it possible to perform all image reconstruction entirely in the software. Several research groups have reported successful ultrasound scanners with real-time imaging construction [17][18][19]. ...
... Other interesting areas of application are flow, contrast dynamics, and functional ultrasound imaging [31]. It is also important to mention that the implementation of this high frame rate imaging method has also been made possible thanks to the developments of GPU technologies, which provide the computational speed that is required to process the amount of data generated during plane wave imaging [32][33][34]. ...
Starting from key ultrasound imaging features such as spatial and temporal resolution, contrast, penetration depth, array aperture, and field-of-view (FOV) size, the reader will be guided through the pros and cons of the main ultrasound beam-forming techniques. The technicalities and the rationality behind the different driving schemes and reconstruction modalities will be reviewed, highlighting the requirements for their implementation and their suitability for specific applications. Techniques such as multi-line acquisition (MLA), multi-line transmission (MLT), plane and diverging wave imaging, and synthetic aperture will be discussed, as well as more recent beam-forming modalities.
... While it is possible to perform processing by leveraging the on-board central processing unit (CPU) [19], its processing capacity is fundamentally limited by the CPU's clock speed, and thus, the processing would need to be done on a retrospective basis. To overcome this issue, GPU has been leveraged as an enabling technology to facilitate high-throughput parallel processing of raw data samples [68]. The key benefit of using GPUs is that each of these computing devices contains thousands of processor cores (more than 3000 cores with latest technology), so it is well suited for high-throughput execution of single-instruction, multiple-thread computing algorithms [69], [70]. ...
Open platform (OP) ultrasound systems are aimed primarily at the research community. They have been at the forefront of the development of synthetic aperture, plane wave, shear wave elastography, and vector flow imaging. Such platforms are driven by a need for broad flexibility of parameters that are normally preset or fixed within clinical scanners. OP ultrasound scanners are defined to have three key features including customization of the transmit waveform, access to the prebeamformed receive data, and the ability to implement real-time imaging. In this paper, a formative discussion is given on the development of OPs from both the research community and the commercial sector. Both software- and hardware-based architectures are considered, and their specifications are compared in terms of resources and programmability. Software-based platforms capable of real-time beamforming generally make use of scalable graphics processing unit architectures, whereas a common feature of hardware-based platforms is the use of field-programmable gate array and digital signal processor devices to provide additional on-board processing capacity. OPs with extended number of channels (>256) are also discussed in relation to their role in supporting 3-D imaging technique development. With the increasing maturity of OP ultrasound scanners, the pace of advancement in ultrasound imaging algorithms is poised to be accelerated.
... While it is possible to perform processing by leveraging the on-board central processing unit (CPU) [19], its processing capacity is fundamentally limited by the CPU's clock speed and thus the processing would need to be done on a retrospective basis. To overcome this issue, GPU has been leveraged as an enabling technology to facilitate high-throughput parallel processing of raw data samples [68]. The key benefit of using GPUs is that each of these computing devices contains thousands of processor cores (more than 3000 cores with latest technology), so it is well suited for high-throughput execution of singleinstruction, multiple-thread computing algorithms [69], [70]. ...
Open platform (OP) ultrasound systems are aimed primarily at the research community. They have been at the forefront of the development of synthetic aperture, plane wave, shear wave elastography and vector flow imaging. Such platforms are driven by a need for broad flexibility of parameters that are normally pre-set or fixed within clinical scanners. OP ultrasound scanners are defined to have three key features including customization of the transmit waveform, access to the pre-beamformed receive data and the ability to implement realtime imaging. In this paper, a formative discussion is given on the development of OPs from both the research community and the commercial sector. Both software and hardware based architectures are considered, and their specifications are compared in terms of resources and programmability. Software based platforms capable of real-time beamforming generally make use of scalable graphics processing unit (GPU) architectures, whereas a common feature of hardware based platforms is the use of fieldprogrammable gate array (FPGA) and digital signal processor (DSP) devices to provide additional on-board processing capacity. OPs with extended number of channels (>256) are also discussed in relation to their role in supporting 3-D imaging technique development. With the increasing maturity of OP ultrasound scanners, the pace of advancement in ultrasound imaging algorithms is poised to be accelerated.
... These open-architecture systems have enabled researchers to readily implement different variants of unfocused pulsing sequences that are essential for realizing HiFRUS [26]. Second, high-throughput computing hardware such as graphical processing units have greatly matured [27]. These parallel processing devices have served well to achieve real-time execution of HiFRUS-related computation tasks, such as pixel-by-pixel beamforming [28][29][30], Doppler processing [31][32][33], and post hoc filtering [34]. ...
Advancements in diagnostic ultrasound have allowed for a rapid expansion of the quantity and quality of non-invasive information that clinical researchers can acquire from cardiovascular physiology. The recent emergence of high frame rate ultrasound (HiFRUS) is the next step in the quantification of complex blood flow behavior, offering angle-independent, high temporal resolution data in normal physiology and clinical cases. While there are various HiFRUS methods that have been tested and validated in simulations and in complex flow phantoms, there is a need to expand the field into more rigorous in vivo testing for clinical relevance. In this tutorial, we briefly outline the major advances in HiFRUS, and discuss practical considerations of participant preparation, experimental design, and human measurement, while also providing an example of how these frameworks can be immediately applied to in vivo research questions. The considerations put forward in this paper aim to set a realistic framework for research labs which use HiFRUS to commence the collection of human data for basic science, as well as for preliminary clinical research questions.
... For each transmit firing event, raw pre-beamformed data were acquired from all array channels through a channel data acquisition system (Cheung et al. 2012). As explained in our earlier work (Yiu et al. 2014), the acquired data sets were processed offline via a high-throughput beamforming and flow estimation platform that was based on graphics processing unit (GPU) technology (So et al. 2011; Yiu et al. ...
Voiding dysfunction that results from bladder outlet (BO) obstruction is known to alter significantly the dynamics of urine passage through the urinary tract. To non-invasively image this phenomenon on a time-resolved basis, we pursued the first application of a recently developed flow visualization technique called vector projectile imaging (VPI) that can track the spatiotemporal dynamics of flow vector fields at a frame rate of 10,000 fps (based on plane wave excitation and least-squares Doppler vector estimation principles). For this investigation, we designed a new anthropomorphic urethral tract phantom to reconstruct urinary flow dynamics under controlled conditions (300 mm H2O inlet pressure and atmospheric outlet pressure). Both a normal model and a diseased model with BO obstruction were developed for experimentation. VPI cine loops were derived from these urinary flow phantoms. Results show that VPI is capable of depicting differences in the flow dynamics of normal and diseased urinary tracts. In the case with BO obstruction, VPI depicted the presence of BO flow jet and vortices in the prostatic urethra. The corresponding spatial-maximum flow velocity magnitude was estimated to be 2.43 m/s, and it is significantly faster than that for the normal model (1.52 m/s) and is in line with values derived from computational fluid dynamics simulations. Overall, this investigation demonstrates the feasibility of using vector flow visualization techniques to non-invasively examine internal flow characteristics related to voiding dysfunction in the urethral tract.
... Deep regions in the human body are typically investigated by transmitting long pulses that inherently involve poor resolution. Even though the possible advantages of coded transmission are evident [7][8][9], no details on the implementation of pulse compression methods in real-time imaging systems using array probes have been presented in the literature for long time. Even more surprising is the situation in medical Doppler ultrasound, where the possible SNR improvement offered by pulse compression could be crucial to allow the visualization of blood flow in deeply located vessels [10]. ...
Initially proposed for radar applications, pulse compression is a technique that enables the transmission of long coded pulses and matched filtering in reception in order to improve the range without sacrificing the resolution. Even if the effectiveness of pulse compression was demonstrated also for medical ultrasound systems, no details on the implementation of pulse compression methods in real-time imaging or Doppler systems using array probes have been presented for long time. In this work, the pulse compression technique was implemented in a low-cost research scanner and it is shown that it can be suitably used for both imaging and spectral Doppler investigations in real-time. An in vivo example is also presented, and it highlights that high SNR gain (up to 13 dB) can be obtained.
... Two factors have dissuaded researchers from employing 2-D estimators in elastography: (i) the high computation burden, and (ii) the lack of an established framework for predicting how factors influence performance. To improve computational efficiency, we and others have developed graphics processing unit (GPU)-accelerated 2-D displacement estimators (Idzenga et al. 2014;So et al. 2011;Verma and Doyley 2014). Concerning the second issue, researchers have developed frameworks for evaluating 1-D estimators (Varghese and Ophir 1998;Walker and Trahey 1995), but these are not appropriate for evaluating 2-D estimators. ...
We derived the Cramér Rao lower bound for 2-D estimators employed in quasi-static elastography. To illustrate the theory, we modeled the 2-D point spread function as a sinc-modulated sine pulse in the axial direction and as a sinc function in the lateral direction. We compared theoretical predictions of the variance incurred in displacements and strains when quasi-static elastography was performed under varying conditions (different scanning methods, different configuration of conventional linear array imaging and different-size kernels) with those measured from simulated or experimentally acquired data. We performed studies to illustrate the application of the derived expressions when performing vascular elastography with plane wave and compounded plane wave imaging. Standard deviations in lateral displacements were an order higher than those in axial. Additionally, the derived expressions predicted that peak performance should occur when 2% strain is applied, the same order of magnitude as observed in simulations (1%) and experiments (1%–2%). We assessed how different configurations of conventional linear array imaging (number of active reception and transmission elements) influenced the quality of axial and lateral strain elastograms. The theoretical expressions predicted that 2-D echo tracking should be performed with wide kernels, but the length of the kernels should be selected using knowledge of the magnitude of the applied strain: specifically, longer kernels for small strains (<5%) and shorter kernels for larger strains. Although the general trends of theoretical predictions and experimental observations were similar, biases incurred during beamforming and subsample displacement estimation produced noticeable differences.
... Two factors have dissuaded researchers from employing 2-D estimators in elastography: (i) the high computation burden, and (ii) the lack of an established framework for predicting how factors influence performance. To improve computational efficiency, we and others have developed graphics processing unit (GPU)-accelerated 2-D displacement estimators (Idzenga et al. 2014;So et al. 2011;Verma and Doyley 2014). Concerning the second issue, researchers have developed frameworks for evaluating 1-D estimators (Varghese and Ophir 1998;Walker and Trahey 1995), but these are not appropriate for evaluating 2-D estimators. ...
In this study, we present a theoretical framework for characterizing the performance of two-dimensional displacement and strain estimators. Specifically, we derived the Cramer-Rao lower bound for axial and lateral displacements estimated from radio frequency echo data. The derived analytical expressions include the effects of signal decorrelation, electronic noise, point spread function (PSF), and signal processing parameters (window size and overlap between the successive windows). We modeled the 2-D PSF of pulse-echo imaging system as a sinc-modulated spatial sine pulse in the axial direction and as a sinc function in the lateral direction. For validation, we compared the variance in displacements and strains, incurred when quasi-static elastography was performed using conventional linear array (CLA), plane wave (PW) and compounded plane wave (CPW) imaging techniques. We also extended the theory to assess the performance of vascular elastograms. The modified analytical expressions predicted that CLA and CPW should provide the worst and best elastographic performance, respectively, which was confirmed both in simulations and experimental studies. Additionally, our framework predicted that the peak performance should occur when 2 \% strain is applied, the same order of magnitude as observed in simulations (1 \%) and experiments (1 \% -- 2 \%).
... The acquired high-frame-rate data were post-processed with MATLAB to generate B-mode image sequences and flow information that could be overlaid on the B-mode image. The raw data were first beamformed into IQ data using custom delay-and-sum GPU beamforming tools [7], [8], each set of a plane-wave transmissions was summed together (i.e., compounded) and then saved to the host computer. This reduced the total number of image frames to M xN and the effective PRF to f PRF ef f = f PRF a /a. ...
Clinical ophthalmic ultrasound is currently performed with mechanically scanned, single-element probes, but these are unable to provide useful information about blood flow with Doppler techniques. Linear arrays are well-suited for the detection of blood flow, but commercial systems generally exceed FDA ophthalmic safety limits. A high-speed plane-wave ultrasound approach with an 18-MHz linear array was utilized to characterize blood flow in the orbit and choroid. Acoustic intensity was measured and the plane-wave mode was within FDA limits. Data were acquired for up to 2 sec and up to 20,000 frames/s with sets of steered plane-wave transmissions that spanned 2*θ degrees where 0 degrees was normal to the array. Lateral resolution was characterized using compounding from 1 to 50 transmissions and -6-dB lateral beamwidths ranged from 320 to 180 μm, respectively. Compounded high-frame-rate data were post-processed using a singular value decomposition spatiotemporal filter and then flow was estimated at each pixel using standard Doppler processing methods. A 1-cm diameter rotating scattering phantom and a 2-mm diameter tube with a flow of blood-mimicking fluid were utilized to validate the flow-estimation algorithms. In vivo data were obtained from the posterior pole of the human eye which revealed regions of flow in the choroid and major orbital vessels supplying the eye.
... Nevertheless, with the increases in the computational power of PCs and the rapid development in parallel computing technique, it is full of possibility of completing the high-quality algorithms in real-time. By taking advantage of the large number of parallel executing cores in modern GPU [54], many researchers have used GPU as accelerators across a range of application domains [55], including the 3D US. Dai et al. [49] processed the PTL interpolation with compounding on the GPU in parallel and achieved a realtime reconstruction of up to 90 frames/s. ...
Real-time three-dimensional (3D) ultrasound (US) has attracted much more attention in medical researches because it provides interactive feedback to help clinicians acquire high-quality images as well as timely spatial information of the scanned area and hence is necessary in intraoperative ultrasound examinations. Plenty of publications have been declared to complete the real-time or near real-time visualization of 3D ultrasound using volumetric probes or the routinely used two-dimensional (2D) probes. So far, a review on how to design an interactive system with appropriate processing algorithms remains missing, resulting in the lack of systematic understanding of the relevant technology. In this article, previous and the latest work on designing a real-time or near real-time 3D ultrasound imaging system are reviewed. Specifically, the data acquisition techniques, reconstruction algorithms, volume rendering methods, and clinical applications are presented. Moreover, the advantages and disadvantages of state-of-the-art approaches are discussed in detail.
... The acquired high-frame-rate data were post-processed with MATLAB to generate B-mode image sequences and flow information that could be overlaid on the B-mode image. The raw data were first beamformed into IQ data using custom delay-and-sum GPU beamforming tools [7], [8], each set of a plane-wave transmissions was summed together (i.e., compounded) and then saved to the host computer. This reduced the total number of image frames to M xN and the effective PRF to f PRF ef f = f PRF a /a. ...
... Podemos optar entre muchas formas de composición de imagen en función de cómo se realice la insonificación, la adquisición, o cómo se combinen las señales recogidas (Szabo, 2004). El problema clásico de conformación de haces en recepción y tiempo real para un array de N elementos e I puntos en la imagen tiene un nivel de complejidad O(N × I), y a día de hoy se han desarrollado distintos tipos de soluciones tanto a nivel de hardware (Siritan and et al., 2013;Wall and Lockwood, 2005;Camacho and et al., 2007), como de software (So et al., 2011). ...
Resumen: Este trabajo analiza la implementación software en un sistema de imagen ultrasónica del Total Focusing Method para la compensación dinámica en tiempo real de los tiempos de vuelo para emisión y recepción de todos los puntos de la imagen. Para ello, haciendo uso de técnicas GPGPU, se analizan dos diferentes alternativas de implementación, mostrando como una planificación adecuada de acceso a los datos permite mejorar los tiempos de ejecución del algoritmo. Abstract: This paper studies the software implementation in an ultrasonic imaging system of Total Focusing Method. In order to accomplish real-time requirements parallel programming techniques have been used. Then, using GPGPU techniques, two different implementation alternatives are analysed, showing how proper planning of access to data improves the performance of the algorithm. Palabras clave: Imagen Ultrasónica, Procesamiento de Señal, Computación Paralela, GPU, CUDA, Keywords: Ultrasonic imaging GPU Signal Processing Parallel computing CUDA
Ultrafast ultrasound imaging can achieve high frame rate by emitting planewave (PW). However, the image quality is drastically degraded in comparison with traditional scanline focused imaging. Using adaptive beamforming techniques can improve image quality at cost of real-time performance. In this work, an adaptive beamforming based on minimum variance (ABF-MV) with deep neural network (DNN) is proposed to improve the image performance and to speed up the beamforming process of ultrafast ultrasound imaging. In particular, a DNN, with a combination architecture of fully-connected network (FCN) and convolutional autoencoder (CAE), is trained with channel radio-frequency (RF) data as input while minimum variance (MV) beamformed data as ground truth. Conventional delay-and-sum (DAS) beamformer and MV beamformer are utilized for comparison to evaluate the performance of the proposed method with simulations, phantom experiments, and in-vivo experiments. The results show that the proposed method can achieve superior resolution and contrast performance, compared with DAS. Moreover, it is remarkable that both in theoretical analysis and implementation, our proposed method has comparable image quality, lower computational complexity, and faster frame rate, compared with MV. In conclusion, the proposed method has the potential to be deployed in ultrafast ultrasound imaging systems in terms of imaging performance and processing time.
High-frame-rate ultrasound imaging uses unfocused transmissions to insonify an entire imaging view for each transmit event, thereby enabling frame rates over 1,000 fps. At these high frame rates, it is naturally challenging to realize real-time transfer of channel-domain raw data from the transducer to the system back-end. Our work seeks to halve the total data transfer rate by uniformly decimating the receive channel count by 50% and, in turn, doubling the array pitch. We show that, despite the reduced channel count and the inevitable use of a sparse array aperture, the resulting beamformed image quality can be maintained by designing a custom convolutional encoder-decoder neural network to infer the radiofrequency (RF) data of the nullified channels. This deep learning framework was trained with in vivo human carotid data (5 MHz plane wave imaging; 128 channels; 31 steering angles over a 30° span; 62,799 frames in total). After training, the network was tested on an in vitro point target scenario that was dissimilar to the training data, in addition to in vivo carotid validation datasets. In the point target phantom image beamformed from inferred channel data, spatial aliasing artifacts attributed to array pitch doubling were found to be reduced by up to 10 dB. For carotid imaging, our proposed approach yielded a lumen-to-tissue contrast that was on average within 3 dB compared to the full-aperture image, whereas without channel data inferencing, the carotid lumen was obscured. When implemented on an RTX-2080 GPU, the inference time to apply the trained network was 4 ms which favors real-time imaging. Overall, our technique shows that, with the help of deep learning, channel data transfer rates can be effectively halved with limited impact on the resulting image quality.
The aim of this paper is to present a new ultrasound platform for medical imaging research, in which all the image formation, capture and emission parameters are user-defined. The requirement to be fully configurable was necessary to enable the evaluation of new signal processing and beamforming algorithms developed by researchers. Since it is a new equipment, different subsystems (synchronism, human–machine interface, platform configuration, image formation, hardware and mechanics) were designed and developed. The degree of flexibility required resulted in a large amount of configuration parameters to be described by the user. The JSON file format provided a well-structured and clear way to describe the different parameters. The hardware design is multi-board. The various functions of the platform are performed by the Synchronism, Multiplex, Tx and Rx boards, mounted in a rack that communicates with a PC. The transmitted beam and the receiver beamforming are generated and implemented in programmable hardware, using FPGAs. Some design techniques reduced the cost and development time, like the use of FPGAs in commercial modules. The RF signal is processed in a GPU using four pipelines that run sequentially to produce the images. First results using different configurations has proven the efficiency of the solution.KeywordsUltrasound real-time platformFlexible configurationMulti-board systemGPU-based medical imaging
Fast and efficient imaging techniques are important for real-time ultrasound imaging. The delay and sum (DAS) beamformer is the most widely-used strategy in focused ultrasound imaging (FUI) modality. However, calculating the time delays and coherently summing the amplitude response in DAS is computationally expensive and generally require a high-performance processor to realize real-time processing. In this study, an efficient spectrum beamformer, namely full-matrix capture (FMC)-stolt, is proposed in FUI system with a linear phased array. The imaging performance of FMC-stolt was validated with the point-scatter simulation and in vitro point and cyst phantoms, and then compared with that of five beamformers, that is, Multiline acquisition (MLA), retrospective transmit beamforming (RTB) in the FUI modality, as well as DAS, Garcia’s frequency-wavenumber (f-k), Lu’s f-k in the coherent plane wave compounding imaging (CPWCI) modality, under specific conditions. We show that the imaging performance of FMC-stolt is better than MLA-DAS in non-transmit-focal regions, and comparable with RTB-DAS at all imaging depths. FMC-stolt also shows better discontinuity alleviation than MLA and RTB. In addition, FMC-stolt has similar imaging characteristics (e.g., off-axis resolution, computational cost) as the f-k beamformers. The computational complexity and actual computational time indicate that FMC-stolt is comparable to Garcia’s f-k, Lu’s f-k, and faster than RTB and CPWCI-DAS if the transmitting numbers are close for FUI and CPWCI. The study demonstrates that the proposed FMC-stolt could achieve good reconstruction speed while preserving high-quality images and thus provide a choice for software beamforming for conventional B-mode ultrasound imaging, especially for hand-held devices with limited performance processors.
Two main metrics are usually employed to assess the quality of medical ultrasound (US) images, namely the contrast and the spatial resolution. A number of imaging algorithms have been proposed to improve one of those metrics, often at the expense of the other one. This paper presents the application of a correlation-based ultrasound imaging method, called Excitelet, to medical US imaging applications and the inclusion of a new Phase Coherence (PC) metric within its formalism. The main idea behind this algorithm, originally developed and validated for Non-Destructive Testing (NDT) applications, is to correlate a reference signal database with the measured signals acquired from a transducer array. In this paper, it is shown that improved lateral resolutions and a reduction of imaging artifacts are obtained over the Synthetic Aperture Focusing Technique (SAFT) when using Excitelet in conjuction with a PC filter. This novel method shows potential for the imaging of specular reflectors, such as invasive surgical tools. Numerical and experimental results presented in this paper demonstrate the benefit, in terms of contrast and resolution, of using the Excitelet method combined with PC for the imaging of strong reflectors.
Increasing attention has been attracted to the research of ultrasound computed tomography (USCT). This article reports the design considerations and implementation details of a novel USCT research system named UltraLucid, which aims to provide a user-friendly platform for researchers to develop new algorithms and conduct clinical trials. The modular design strategy is adopted to make the system highly scalable. A prototype has been assembled in our laboratory, which is equipped with a 2048-element ring transducer, 1024 transmit (TX) channels, 1024 receive (RX) channels, two servers, and a control unit. The prototype can acquire raw data from 1024 channels simultaneously using a modular data acquisition and a transfer system, consisting of 16 excitation and data acquisition (EDAQ) boards. Each EDAQ board has 64 independent TX and RX channels and 4-Gb Ethernet interfaces for raw data transmission. The raw data can be transferred to two servers at a theoretical rate of 64 Gb/s. Both servers are equipped with a 10.9-TB solid-state drive (SSD) array that can store raw data for offline processing. Alternatively, after processing by onboard field-programmable gate arrays (FPGAs), the raw data can be processed online using multicore central processing units (CPUs) and graphics processing units (GPUs) in each server. Through control software running on the host computer, the researchers can configure parameters for transmission, reception, and data acquisition. Novel TX–RX scheme and coded imaging can be implemented. The modular hardware structure and the software-based processing strategy make the system highly scalable and flexible. The system performance is evaluated with phantoms and
in vivo
experiments.
Synthetic aperture (SA) beamforming is a principal technology of modern medical ultrasound imaging. In that the use of focused transmission provides superior signal-to-noise ratio (SNR) and is promising for cardiovascular diagnosis at the maximum imaging depth of about 160 mm. But there is a pitfall in increasing the frame rate to more than 80 frames per second (frames/s) without image degradation by the haze artifact produced when the transmit foci (SA virtual sources) placed within the imaging field. We hypothesize that the source of this artifact is a grating lobe caused by coarse (decimated) multiple transmission and manifesting in the low brightness region in the accelerated-frame-rate images. We propose an intertransmission coherence factor (ITCF) method suppressing haze artifacts caused by coarse-pitch multiple transmission. The method is expected to suppress the image blurring because the SA grating lobe signal is less coherent than the main lobe signals. We evaluated an ITCF algorithm for suppressing the grating artifact when the transmission pitch is up to four times larger than the normal pitch (equivalent to 160 frames/s). In
in-vitro
and
in-vivo
experiments, we demonstrated the strong relation of haze artifact with the grating lobe due to the coarse-pitch transmission. Then, we confirmed that the ITCF method suppresses the haze artifact of a human heart by 15 dB while preserving the spatial resolution. The ITCF method combined with focused transmission SA beamforming is a valid method for getting cardiovascular ultrasound B-mode images without making a compromise in the trade-off relationship between the frame rate and the SNR.
Two delay-and-sum beamformers for 3-D synthetic aperture imaging with row-column addressed arrays are presented. Both beamformers are software implementations for graphics processing unit (GPU) execution with dynamic apodizations and 3rd order polynomial subsample interpolation. The first beamformer was written in the MATLAB programming language and the second was written in C/C++ with the compute unified device architecture (CUDA) extensions by NVIDIA. Performance was measured as volume rate and sample throughput on three different GPUs: a 1050 Ti, a 1080 Ti, and a TITAN V. The beamformers were evaluated across 112 combinations of output geometry, depth range, transducer array size, number of virtual sources, floating point precision, and Nyquist rate or inphase/ quadrature beamforming using analytic signals. Real-time imaging defined as more than 30 volumes per second was attained by the CUDA beamformer on the three GPUs for 13, 27, and 43 setups, respectively. The MATLAB beamformer did not attain real-time imaging for any setup. The median, single precision sample throughput of the CUDA beamformer was 4.9, 20.8, and 33.5 gigasamples per second on the three GPUs, respectively. The CUDA beamformer’s throughput was an order of magnitude higher than that of the MATLAB beamformer.
Digital ultrasound probes integrate the analog front end in the housing of the probe handle and provide a digital interface instead of requiring an expensive coaxial cable harness to connect. Current digital probes target the portable market and perform the bulk of the processing (beamforming) on the probe, which enables the probe to be connected to commodity devices, such as tablets or smartphones, running an ultrasound app to display the image and control the probe. Thermal constraints limit the number of front-end channels as well as the complexity of the processing. This prevents current digital probes to support advanced modalities, such as vector flow or elastography, requiring high-frame-rate imaging. In this paper, we present Light Probe, a digital ultrasound probe, which integrates a 64-channel 100-V
<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">pp</sub>
TX/RX front end, including analog-to-digital conversion (up to 32.5 MS/s at 12 bit), and is equipped with an optical high-speed link (26.4 Gb/s) providing sustainable raw samples access to all the channels, which allows the processing to be performed on the connected device without thermal power constraints. By connecting the probe to a graphics processing unit-equipped PC, we demonstrate the flexibility of software-defined B-mode imaging using conventional and ultrafast methods. We achieve plane-wave and synthetic aperture imaging with frame rates from 30 up to 500 Hz consuming between 5.6 and 10.7 W. By using a combination of power and thermal management techniques, we demonstrate that the probe can remain within operating temperature limits even without active cooling, while never having to turn the probe off for cooling, hence providing a consistent quality of service for the operator.
An ultrasonic transducer technology to generate wideband impulses using a one-dimensional chain of spheres was previously presented. The Hertzian contact between the spheres causes the nonlinearity of the system to increase, which transforms high amplitude narrowband sinusoidal input into a train of wideband impulses. Generation of short duration ultrasonic pulses is desirable both in diagnostic and therapeutic ultrasound. Nevertheless, the biggest challenge in terms of adaptation to biomedical ultrasound is the coupling of the ultrasonic energy into biological tissue. An analytical model was created to address the coupling issue. Effect of the matching layer was modelled as a flexible thin plate clamped from the edges. Model was verified against hydrophone measurements. Different coupling materials, such as glass, aluminium, acrylic, silicon rubber, and vitreous carbon, was analysed with this model. Results showed that soft matching layers such as acrylic and rubber inhibit the generation of higher order harmonics. Between the hard matching materials, vitreous carbon achieved the best results due to its acoustic impedance.
Purpose:
High-intensity focused ultrasound (HIFU) is becoming an effective and noninvasive treatment modality for cancer and solid tumors. In order to avoid the cancer relapse and guarantee the success of ablation, there should be no gaps left among all HIFU-generated lesions. However, there are few imaging approaches available for detecting the HIFU lesion gaps in real time during ablation.
Methods:
Transient axial shear strain elastograms (ASSEs) were proposed and evaluated both numerically and experimentally to detect the lesion gaps immediately after the cessation of therapeutic HIFU exposure. Acoustic intensity and subsequent acoustic radiation force were first calculated by solving the nonlinear Khokhlov-Zabolotskaya-Kuznetzov (KZK) equation. Motion of being- and already-treated lesions during and after HIFU exposure was simulated using the transient dynamic analysis module of finite element method (FEM). The corresponding B-mode sonography of tissue-mimicking phantom with two HIFU lesions inside was simulated by FIELD II, and then axial strain elastograms (ASEs) under static compression and transient ASSEs were reconstructed. An ultrasound imaging probe was integrated with the HIFU transducer and used to obtain radio-frequency (RF) echo signals at high frame rate using plane wave imaging (PWI). The resulting strains were mapped using the correlation-based method and block search strategy.
Results:
Acoustic radiation force from the therapeutic HIFU burst is sufficiently strong to produce significant displacement. As a result, large and highly localized axial shear strain appears in the gap zone between two HIFU-generated lesions and then disappears after sufficient HIFU ablation (no gap between them). Such capability of detecting the lesion gap is validated at the varied acoustic radiation force density, gap width, and the size of the lesion. In contrast, conventional ASEs using the static compression cannot distinguish whether a gap exists between lesions. Static ASEs and transient ASSEs reconstructed using both high-speed photography and sonography in the gel phantom show the same conclusion as that in the simulation. Ex vivo tissue experiments further confirmed that the presence of large axial shear strain in the gap zone. The ratios of axial shear strain in the porcine kidney and liver samples had statistical differences for two HIFU-generated lesions without and with a gap (p < 0.05).
Conclusions:
Large axial shear strain induced by the acoustic radiation force from therapeutic HIFU burst only appears between two HIFU-generated lesions with a gap between them. Transient ASSEs reconstructed immediately after the cession of HIFU exposure can easily, reliably, and sensitively detect the gap between produced lesions, which would provide real-time feedback to enhance the success of HIFU ablation. This article is protected by copyright. All rights reserved.
At The heart of a vehicle’s operation is the engine. The same metaphoric relation applies between a research innovation and an experimental method. It is safe to say that innovations in biomedical ultrasonics cannot be materialized without foundational advances in experimental methods, tools, and protocols. Indeed, because biomedical ultrasonics naturally spans a diverse research space—thanks to its versatile applicability in both diagnostics and therapeutics—a broad range of investigative methods are correspondingly required to serve a multitude of experimental needs.
Flow phantoms with anatomically realistic geometry and high acoustic compatibility are valuable investigative tools in vascular ultrasound studies. Here, we present a new framework to fabricate ultrasound-compatible flow phantoms to replicate human vasculature that is tortuous, non-planar and branching in nature. This framework is based upon the integration of rapid prototyping and investment casting principles. A pedagogical walkthrough of our engineering protocol is presented in this paper using a patient-specific cerebral aneurysm model as an exemplar demonstration. The procedure for constructing the flow circuit component of the phantoms is also presented, including the design of a programmable flow pump system, the fabrication of blood mimicking fluid, and flow rate calibration. Using polyvinyl alcohol (PVA) cryogel as the tissue mimicking material, phantoms developed with the presented protocol exhibited physiologically relevant acoustic properties (attenuation coefficient: 0.229±0.032 dB/(cm∙MHz); acoustic speed: 1535±2.4 m/s), and their pulsatile flow dynamics closely resembled the flow profile input. As a first application of our developed phantoms, the flow pattern of the patient-specific aneurysm model was visualized by performing high-frame-rate color-encoded speckle imaging (CESI) over multiple time-synchronized scan planes. Persistent recirculation was observed, and the vortex center was found to shift in position over a cardiac cycle, indicating the 3-D nature of flow recirculation inside an aneurysm. These findings suggest that phantoms produced from our reported protocol can serve well as acoustically-compatible test-beds for vascular ultrasound studies, including 3-D flow imaging.
The graphics processing unit (GPU) has become an integral part of today's mainstream computing systems. Over the past six years, there has been a marked increase in the performance and capabilities of GPUs. The modern GPU is not only a powerful graphics engine but also a highly parallel programmable processor featuring peak arithmetic and memory bandwidth that substantially outpaces its CPU counterpart. The GPU's rapid increase in both programmability and capability has spawned a research community that has successfully mapped a broad range of computationally demanding, complex problems to the GPU. This effort in general-purpose computing on the GPU, also known as GPU computing, has positioned the GPU as a compelling alternative to traditional microprocessors in high-performance computer systems of the future. We describe the background, hardware, and programming model for GPU computing, summarize the state of the art in tools and techniques, and present four GPU computing successes in game physics and computational biophysics that deliver order-of-magnitude performance gains over optimized CPU applications.
Synthetic transmit aperture (STA) imaging gives the possibility to acquire an image with only few emissions and is appealing for 3D ultrasound imaging. Even though the number of emissions is low, the change in position of the scatterers prohibits the coherent summations of ultrasound echoes and leads to distortions in the image. In order to develop motion compensation and/or velocity estimation algorithms a thorough and intuitive understanding of the nature of motion artifacts is needed. This paper proposes a simple 2D broad band model for STA images, based on the acquisition procedure and the beamformation algorithm. In STA imaging a single element transmits a cylindrical wave. All elements are used in receive, and by applying different delays a low resolution image (LRI) is beamformed. A Fourier relation exists between the aperture function and all points in the beamformed LRI. This relation is used to develop an approximation of the point spread function (PSF) of a LRI. It is shown that the PSF of LRIs obtained by transmitting with different elements can be viewed as rotated versions of each other. Summing several LRIs gives a high resolution image. The model approximates the PSF of a high resolution image as a sum of rotated PSFs of a single LRI. The approximation is validated with a Field II simulation. The model predicts and explains the motion artifacts, and gives an intuitive feeling of what would happen for different velocities.
Compact ultrasound technology has facilitated growth in point-of-care uses in many specialties. This review includes videos demonstrating the use of ultrasonography to guide central venous access, detect pneumothorax, detect evidence of hemorrhage after trauma, and screen for abdominal aortic aneurysm.
Recently, significant improvement in image resolution has been demonstrated by applying adaptive beamforming to medical ultrasound imaging. In this paper, we have used the minimum-variance beamformer to show how the low sidelobe levels and narrow beamwidth of adaptive methods can be used, not only to increase resolution, but also to enhance imaging in several ways. By using a minimum-variance beamformer instead of delay-and-sum on reception, reduced aperture, higher frame rates, or increased depth of penetration can be achieved without sacrificing image quality. We demonstrate comparable resolution on images of wire targets and a cyst phantom obtained with a 96-element, 18.5-mm transducer using delay-and-sum, and a 48-element, 9.25-mm transducer using minimum variance. To increase frame rate, fewer and wider transmit beams in combination with several parallel receive beams may be used. We show comparable resolution to delay-and-sum using minimum variance, 1/4th of the number of transmit beams and 4 parallel receive beams, potentially increasing the frame rate by 4. Finally, we show that by lowering the frequency of the transmitted beam and beamforming the received data with the minimum variance beamformer, increased depth of penetration is achieved without sacrificing lateral resolution.
The emergence of ultrafast frame rates in ultrasonic imaging has been recently made possible by the development of new imaging modalities such as transient elastography. Data acquisition rates reaching more than thousands of images per second enable the real-time visualization of shear mechanical waves propagating in biological tissues, which convey information about local viscoelastic properties of tissues. The first proposed approach for reaching such ultrafast frame rates consists of transmitting plane waves into the medium. However, because the beamforming process is then restricted to the receive mode, the echographic images obtained in the ultrafast mode suffer from a low quality in terms of resolution and contrast and affect the robustness of the transient elastography mode. It is here proposed to improve the beamforming process by using a coherent recombination of compounded plane-wave transmissions to recover high-quality echographic images without degrading the high frame rate capabilities. A theoretical model is derived for the comparison between the proposed method and the conventional B-mode imaging in terms of contrast, signal-to-noise ratio, and resolution. Our model predicts that a significantly smaller number of insonifications, 10 times lower, is sufficient to reach an image quality comparable to conventional B-mode. Theoretical predictions are confirmed by in vitro experiments performed in tissue-mimicking phantoms. Such results raise the appeal of coherent compounds for use with standard imaging modes such as B-mode or color flow. Moreover, in the context of transient elastography, ultrafast frame rates can be preserved while increasing the image quality compared with flat insonifications. Improvements on the transient elastography mode are presented and discussed.
We have applied the minimum variance (MV) adaptive beamformer to medical ultrasound imaging and shown significant improvement in image quality compared to delay-and-sum (DAS). We demonstrate reduced mainlobe width and suppression of sidelobes on both simulated and experimental RF data of closely spaced wire targets, which gives potential contrast and resolution enhancement in medical images. The method is applied to experimental RF data from a heart phantom, in which we show increased resolution and improved definition of the ventricular walls. A potential weakness of adaptive beamformers is sensitivity to errors in the assumed wavefield parameters. We look at two ways to increase robustness of the proposed method; spatial smoothing and diagonal loading. We show that both are controlled by a single parameter that can move the performance from that of a MV beamformer to that of a DAS beamformer. We evaluate the sensitivity to velocity errors and show that reliable amplitude estimates are achieved while the mainlobe width and sidelobe levels are still significantly lower than for the conventional beamformer.
Preface. Acknowledgements. 1. Introduction. 2. Energy Efficient Design. 3. Microprocessor System Architecture. 4. Circuit Design Methodology. 5. Energy Driven Design Flow. 6. Microprocessor and Memory IC's. 7. DC-DC Voltage Conversion. 8. DC-DC Converter IC for DVS. 9. DVS System Design and Results. 10. Software and Operating System Support. 11. Conclusions. Index.
The introduction of real time 3D echocardiography allows for deformation tracking in three dimensions, without the limitations of 2D methods. However, the processing needs of 3D methods are much higher. The aim of the study was to optimize the performance of 3D block matching by using a Single Instruction Multiple Data (SIMD) model, a technique employed to achieve data level parallelism. Two implementations of SIMD have been tested. The first is based on Streaming SIMD Extensions (SSE), the second uses CUDA, a SIMD architecture proposed by NVIDIA, which is available on several graphics cards. The proposed methods have been validated on synthetic and patient data. With the use of SIMD architecture, the overall processing time is significantly reduced, thus making 3D speckle tracking feasible in a clinical setting. Apart from the implemented sum of square differences (SAD), other matching criteria (e.g. sum of squared differences, normalized cross correlation) can be implemented efficiently (especially on the GPU) thus further improving the accuracy of the block matching process.
Ultrasound scanner design has changed dramatically with variations in clinical capability and with the technologies available for achieving that capability. There have been transitions from manual scanning to mechanical and electronic scanning and from bistable black and white images to full gray scale images with electronic scan converters and the migration to digital hardware and to software-based systems. This article reviews most of these developments and tries to anticipate the clinical role of ultrasound scanning in the context of the current trends of development.
Ultrasound (US) has become widely used in clinical medicine for the diagnosis of a variety of disease processes. The unique ability of US to provide accurate information through an efficacious, painless, portable, and nonionizing method has expanded its role and application in diverse medical settings. Given the current economic environment and the related interest in creating the greatest value for health care expenditures, US has been evaluated to compare its clinical accuracy/efficacy and cost-effectiveness versus other imaging modalities. The following literature review reports the results of research studies aimed at comparing the accuracy/efficacy and cost of US versus alternative imaging modalities, including magnetic resonance imaging, computed tomography, contrast angiography, and single-photon emission computed tomography.
One of the key additions to clinical ultrasound (US) systems during the last decade was the incorporation of three-dimensional (3-D) imaging as a native mode. Compared to previous-generation 3-D US imaging systems, today's systems offer easier volume acquisition and deliver superior image quality with various visualization options. This has come as a result of many technological advances and innovations in transducer design, electronics, computer architecture, and algorithms. While freehand 3-D US techniques continue to be used, mechanically scanned and/or two-dimensional (2-D) matrix-array transducers are increasingly adopted, enabling higher volume rates and easier acquisition. More powerful computing engines with instruction-level and data-level parallelism and high-speed memory access support new and improved 3-D visualization capabilities. Many clinical US systems today have a 3-D option that offers interactive acquisition and display. In this paper, we cover the innovations of the last decade that have enabled the current 3-D US systems from acquisition to visualization, with emphasis on transducers, algorithms, and computation.
High frame-rate ultrasound RF data acquisition has been proved to be critical for novel cardiovascular imaging techniques, such as high-precision myocardial elastography, pulse wave imaging (PWI), and electromechanical wave imaging (EWI). To overcome the frame-rate limitations on standard clinical ultrasound systems, we developed an automated method for multi-sector ultrasound imaging through retrospective electrocardiogram (ECG) gating on a clinically used open architecture system. The method achieved both high spatial (64 beam density) and high temporal resolution (frame rate of 481 Hz) at an imaging depth up to 11 cm and a 100% field of view in a single breath-hold duration. Full-view imaging of the left ventricle and the abdominal aorta of healthy human subjects was performed using the proposed technique in vivo. ECG and ultrasound RF signals were simultaneously acquired on a personal computer (PC). Composite, full-view frames both in RF- and B-mode were reconstructed through retrospective combination of seven small (20%) juxtaposed sectors using an ECG-gating technique. The axial displacement of the left ventricle, in both long-axis and short-axis views, and that of the abdominal aorta, in a long-axis view, were estimated using a RF-based speckle tracking technique. The electromechanical wave and the pulse wave propagation were imaged in a cineloop using the proposed imaging technique. Abnormal patterns of such wave propagation can serve as indicators of early cardiovascular disease. This clinical system could thus expand the range of applications in cardiovascular elasticity imaging for quantitative, noninvasive diagnosis of myocardial ischemia or infarction, arrhythmia, abdominal aortic aneurysms, and early-stage atherosclerosis.
Although they show potential to improve ultrasound image quality, plane wave (PW) compounding and synthetic aperture (SA) imaging are computationally demanding and are known to be challenging to implement in real-time. In this work, we have developed a novel beamformer architecture with the real-time parallel processing capacity needed to enable fast realization of PW compounding and SA imaging. The beamformer hardware comprises an array of graphics processing units (GPUs) that are hosted within the same computer workstation. Their parallel computational resources are controlled by a pixel-based software processor that includes the operations of analytic signal conversion, delay-and-sum beamforming, and recursive compounding as required to generate images from the channel-domain data samples acquired using PW compounding and SA imaging principles. When using two GTX-480 GPUs for beamforming and one GTX-470 GPU for recursive compounding, the beamformer can compute compounded 512 × 255 pixel PW and SA images at throughputs of over 4700 fps and 3000 fps, respectively, for imaging depths of 5 cm and 15 cm (32 receive channels, 40 MHz sampling rate). Its processing capacity can be further increased if additional GPUs or more advanced models of GPU are used.
GPU computing is at a tipping point, becoming more widely used in demanding consumer applications and high-performance computing. This article describes the rapid evolution of GPU architectures-from graphics processors to massively parallel many-core multiprocessors, recent developments in GPU computing architectures, and how the enthusiastic adoption of CPU+GPU coprocessing is accelerating parallel applications.
SAFT techniques are based on the sequential activation, in emission and reception, of the array elements and the post-processing of all the received signals to compose the image. Thus, the image generation can be divided into two stages: (1) the excitation and acquisition stage, where the signals received by each element or group of elements are stored; and (2) the beamforming stage, where the signals are combined together to obtain the image pixels. The use of Graphics Processing Units (GPUs), which are programmable devices with a high level of parallelism, can accelerate the computations of the beamforming process, that usually includes different functions such as dynamic focusing, band-pass filtering, spatial filtering or envelope detection. This work shows that using GPU technology can accelerate, in more than one order of magnitude with respect to CPU implementations, the beamforming and post-processing algorithms in SAFT imaging.
Synthetic aperture (SA) schemes provide effective means to improve
SNR and resolution of ultrasound images. Since SA schemes are subjective
to motion artifacts due to their long data acquisition time, however,
there have been efforts to compensate for the motion artifacts. In this
paper, we investigate the effects of a commonly used motion compensation
method when it is applied to a SA scheme, called bidirectional
pixel-based focusing (BiPBF). From the results, we verified that axial
motion is a dominant factor in causing the motion artifacts and the
axial motion can be compensated successfully at the sacrifice of
sidelobe levels
This review is about the development of three-dimensional (3D) ultrasonic medical imaging, how it works, and where its future lies. It assumes knowledge of two-dimensional (2D) ultrasound, which is covered elsewhere in this issue. The three main ways in which 3D ultrasound may be acquired are described: the mechanically swept 3D probe, the 2D transducer array that can acquire intrinsically 3D data, and the freehand 3D ultrasound. This provides an appreciation of the constraints implicit in each of these approaches together with their strengths and weaknesses. Then some of the techniques that are used for processing the 3D data and the way this can lead to information of clinical value are discussed. A table is provided to show the range of clinical applications reported in the literature. Finally, the discussion relating to the technology and its clinical applications to explain why 3D ultrasound has been relatively slow to be adopted in routine clinics is drawn together and the issues that will govern its development in the future explored.
Color Doppler ultrasound is a routinely used diagnostic tool for assessing blood flow information in real time. The required signal processing is computationally intensive, involving autocorrelation, linear filtering, median filtering, and thresholding. Because of the large amount of data and high computational requirement, color Doppler signal processing has been mainly implemented on custom-designed hardware, with software-based implementation--particularly on a general-purpose CPU--not being successful. In this paper, we describe the use of a graphics processing unit for implementing signal-processing algorithms for color Doppler ultrasound that achieves a frame rate of 160 fps for frames comprising 500 scan lines x 128 range samples, with each scan line being obtained from an ensemble size of 8 with an 8-tap FIR clutter filter.
Medical images are created by detecting radiation probes transmitted through or emitted or scattered by the body. The radiation, modulated through interactions with tissues, yields patterns that provide anatomic and/or physiologic information. X-rays, gamma rays, radiofrequency signals, and ultrasound waves are the standard probes, but others like visible and infrared light, microwaves, terahertz rays, and intrinsic and applied electric and magnetic fields are being explored. Some of the younger technologies, such as molecular imaging, may enhance existing imaging modalities; however, they also, in combination with nanotechnology, biotechnology, bioinformatics, and new forms of computational hardware and software, may well lead to novel approaches to clinical imaging. This review provides a brief overview of the current state of image-based diagnostic medicine and offers comments on the directions in which some of its subfields may be heading.
Ultrasound imaging is now in very widespread clinical use. The most important underpinning technologies include transducers, beam forming, pulse compression, tissue harmonic imaging, contrast agents, techniques for measuring blood flow and tissue motion, and three-dimensional imaging. Specialized and emerging technologies include tissue characterization and image segmentation, microscanning and intravascular scanning, elasticity imaging, reflex transmission imaging, computed tomography, Doppler tomography, photoacoustics and thermoacoustics. Phantoms and quality assurance are necessary to maintain imaging performance. Contemporary ultrasonic imaging procedures seem to be safe but studies of bioeffects are continuing. It is concluded that advances in ultrasonic imaging have primarily been pushed by the application of physics and innovations in engineering, rather than being pulled by the identification of specific clinical objectives in need of scientific solutions. Moreover, the opportunities for innovation to continue into the future are both challenging and exciting.
The ways in which ultrasound is used in medical diagnosis are reviewed, with particular emphasis on the ultrasound source (probe) and implications for acoustic exposure. A brief discussion of the choice of optimum frequency for various target depths is followed by a description of the general features of diagnostic ultrasound probes, including endo-probes. The different modes of diagnostic scanning are then discussed in turn: A-mode, M-mode, B-mode, three-dimensional (3D) and 4D scanning, continuous wave (CW) Doppler, pulse-wave spectral Doppler and Doppler imaging. Under the general heading of B-mode imaging, there are individual descriptions of the principles of chirps and binary codes, B-flow, tissue harmonic imaging and ultrasound contrast agent-specific techniques. Techniques for improving image quality within the constraints of real-time operation are discussed, including write zoom, parallel beam forming, spatial compounding and multiple zone transmission focusing, along with methods for reducing slice thickness. At the end of each section there is a summarising comment on the basic features of the acoustic output and its consequences for patient safety.
The paper describes the use of synthetic aperture (SA) imaging in medical ultrasound. SA imaging is a radical break with today's commercial systems, where the image is acquired sequentially one image line at a time. This puts a strict limit on the frame rate and the possibility of acquiring a sufficient amount of data for high precision flow estimation. These constrictions can be lifted by employing SA imaging. Here data is acquired simultaneously from all directions over a number of emissions, and the full image can be reconstructed from this data. The paper demonstrates the many benefits of SA imaging. Due to the complete data set, it is possible to have both dynamic transmit and receive focusing to improve contrast and resolution. It is also possible to improve penetration depth by employing codes during ultrasound transmission. Data sets for vector flow imaging can be acquired using short imaging sequences, whereby both the correct velocity magnitude and angle can be estimated. A number of examples of both phantom and in vivo SA images will be presented measured by the experimental ultrasound scanner RASMUS to demonstrate the many benefits of SA imaging.
The CUDA programming model provides a straightforward means of describing inherently parallel computations, and NVIDIA's Tesla GPU architecture delivers high computational throughput on massively parallel problems. This article surveys experiences gained in applying CUDA to a diverse set of problems and the parallel speedups over sequential codes running on traditional CPU architectures attained by executing key computations on the GPU.
Three-Dimensional Ultrasound: From Acquisition to Visualization and from Algorithms to Systems
- K Karadayi
- R Manuguli
- Y Kim
K. Karadayi, R. Manuguli, and Y. Kim,
''Three-Dimensional Ultrasound: From Acquisition to Visualization and from Algorithms to Systems,'' IEEE Rev. Biomedical
Eng., vol. 2, 2009, pp. 23-39.