Gi-Duck Kim

A single FPGA-based portable ultrasound imaging system for point-of-care applications

We present a cost-effective portable ultrasound system based on a single field-programmable gate array (FPGA) for point-of-care applications. In the portable ultrasound system developed, all the ultrasound signal and image processing modules, including an effective 32-channel receive beamformer with pseudo-dynamic focusing, are embedded in an FPGA chip. For overall system control, a mobile processor running Linux at 667 MHz is used. The scan-converted ultrasound image data from the FPGA are directly transferred to the system controller via external direct memory access without a video processing unit. The potable ultrasound system developed can provide real-time B-mode imaging with a maximum frame rate of 30, and it has a battery life of approximately 1.5 h. These results indicate that the single FPGA-based portable ultrasound system developed is able to meet the processing requirements in medical ultrasound imaging while providing improved flexibility for adapting to emerging POC applications.

A System-on-Chip Solution for Point-of-Care Ultrasound Imaging Systems: Architecture and ASIC Implementation

In this paper, we present a novel system-on-chip (SOC) solution for a portable ultrasound imaging system (PUS) for point-of-care applications. The PUS-SOC includes all of the signal processing modules (i.e., the transmit and dynamic receive beamformer modules, mid- and back-end processors, and color Doppler processors) as well as an efficient architecture for hardware-based imaging methods (e.g., dynamic delay calculation, multi-beamforming, and coded excitation and compression). The PUS-SOC was fabricated using a UMC 130-nm NAND process and has 16.8 GFLOPS of computing power with a total equivalent gate count of 12.1 million, which is comparable to a Pentium-4 CPU. The size and power consumption of the PUS-SOC are 27×27 mm2 and 1.2 W, respectively. Based on the PUS-SOC, a prototype hand-held US imaging system was implemented. Phantom experiments demonstrated that the PUS-SOC can provide appropriate image quality for point-of-care applications with a compact PDA size ( 200×120×45 mm3) and 3 hours of battery life.

2K-1 A Hand-Held Ultrasound Imaging System for Point-of-Care Applications

As demands for point-of-care (POC) services have increased in and out of hospitals, more and more attention has been drawn to developing portable medical diagnosis systems with imaging capability. We have recently developed a prototype of PDA-type hand-held device which is designed to have the functions and portability required for POC applications by combining an ultra small ultrasound scanner and a full custom PDA using Linux as its OS. The prototype now only supports a phased array but was designed to support various probe types. It is powered by a Lithium polymer battery to achieve the maximum scanning time of about one and half hours. We will present the images obtained with the system and application results of other features. The images will be compared to images obtained from a commercial product using 32 channels in both transmit and receive

A new smart probe system for a tablet PC-based point-of-care ultrasound imaging system: Feasibility study

There is a growing interest in a hand-held ultrasound (US) system for a point-of-care diagnosis. In this paper, we present the hand-held smart US probe system, which includes analog and digital front-ends, mid-processor, and interface. In the analog front-end, 16 channels could be implemented using two eight-channel high-voltage pulsers and analog-to-digital converters (ADC). The following digital front-end was implemented using a single low-cost field programmable gate array chip (FPGA). The output data from the digital front-end are transferred to the commercial tablet PC via the USB 3.0 interface. The tablet PC performs the back-end processing and image display on a graphical user interface (GUI). The smart probe system developed can provide the real-time B-mode images up to 22 Hz of frame rate with 7 W of estimated maximum power consumption. The weight and dimension of the developed system are 73 g and 150 × 40 × 10 mm3 in length, width, and height, respectively. To evaluate the performance of the proposed smart probe system, the phantom and in vivo experiments were conducted.

The new efficient multi-beamforming method based on multiple-access register block on a post-fractional filtering architecture

In medical ultrasound imaging, a multi-beamforming (MBF) method is used for supporting high frame rate imaging or functional imaging where multiple scanlines are reconstructed from a single excitation event. For efficient MBF, a time-sharing technique (i.e., MBF-TS) can be applied. However, the MBF-TS could degrade image quality due to the decreased beamforming frequency. In this paper, the multi-access register-based MBF (MBF-MAR) method running on the post-fractional filtering (PFF) architecture is presented. In PFF-MBF-MAR, instead of lowering beamforming frequency, a multi-access register at each channel is utilized for generating multiple scanlines simultaneously. To evaluate the performance of the proposed PFF-MBF-MAR method, the phantom experiment was conducted where 64- channel pre-beamformed radio-frequency (RF) data were captured from a tissue mimicking phantom by using a modified commercial ultrasound system (SONOLINE G40, Siemens Inc., USA) using a 3-MHz phased array probe. From the phantom experiment, the PFF-MBF-MAR method showed 4.7 dB and 0.6 improvements in the signal-to-noise ratio (SNR) and the contrast-to-noise ratio (CNR), respectively, compared to the PFF-MBF-TS method, while slightly increasing the hardware complexity (<5.2%). The similar results were achieved with the in vivo thyroid data. These results indicate that the proposed PFF-MBF-MAR method can be used for high frame rate imaging or functional imaging without sacrificing image quality while slightly increasing the hardware complexity.

A new nonlinear zone-based beamforming method for point-of-care ultrasound: Algorithms and implementation

It is important to reduce the hardware complexity of a dynamic receive beamformer for point-of-care ultrasound imaging systems. In this paper, a new nonlinear zone-based beamforming (ZBF-NLI) method, which can regulate delay errors throughout the imaging depth while reducing the hardware complexity, is proposed. For quantitative evaluation, normalized pixel intensity error (NPE) and mean square error (MSE) were measured from the Field-II simulation and phantom experiments with the allowance error of 0.25 for 16-f0 delay focusing resolution in ZBF-NLI method. Also, the identical number of zones was utilized for both ZBF-LI and ZBF-NLI methods. As a results, the dynamic delays calculated by the ZBF-NLI method shows 0 of MSE, while the ZBF-LI method indicated 0.52. Also, from the 8-bit US images of Field-II simulation and phantom experiment, the 6.16 and 8.39 of normalized pixel intensity error of the ZBF-LI method could be eliminated, respectively, while utilizing the identical number of zones. The ZBF-NLI method was implemented on the low-cost FPGA chip (Spartan 6, Xilinx Inc., CA, USA) of the custom-built portable US imaging system, which indicated that the proposed ZBF-NLI method only requires 9.14% of memory compared to that of the ZBF-LI method (i.e., 16.84 KB vs. 184.32 KB).

Context modeling based lossless compression of radio-frequency data for software-based ultrasound beamforming

Abstract A new lossless compression method using context modeling for ultrasound radio-frequency (RF) data is presented. In the proposed compression method, the combination of context modeling and entropy coding is used for effectively lowering the data transfer rates for modern software-based medical ultrasound imaging systems. From the phantom and in vivo data experiments, the proposed lossless compression method provides the average compression ratio of 0.45 compared to the Burg and JPEG-LS methods (0.52 and 0.55, respectively). This result indicates that the proposed compression method is capable of transferring 64-channel 40-MHz ultrasound RF data with a 16-lane PCI-Express 2.0 bus for software beamforming in real time.

A Point-of-care diagnosis system for emergency ultrasound: Prototype system implementation

In Point-of-care (POC) applications, the need of a diagnosis imaging tool (e.g., portable ultrasound) combined with a physiological monitoring device (e.g., echocardiogram) is rapidly growing since it is essential for effectively evaluating patients condition. In this paper, a new POC diagnosis system is presented, which can simultaneously provide ultrasound imaging and vital biosignal information for immediate diagnosis on the urgent signs of internal injuries, e.g., hemoperitoneum, hemopericardium, hemothorax, and pneumothorax.

A Method for Elimination of Common Grating Lobes for Designing Optimum Periodic Sparse Arrays

We present analytic methods for designing optimum periodic sparse array schemes consisting of periodic transmit and receive sparse arrays with different periodicities. The purpose of this work is to minimize the number of active array elements while suppressing grating lobe levels below a certain desired level. For this purpose, the fundamental conditions for eliminating common grating lobes, ones that occur at the same position on transmit and receive, are derived for a general model of periodic sparse arrays. The proposed design methods are applied to the design of some optimum periodic sparse arrays, of which the results are demonstrated.

A new directional demodulation method for vector Doppler imaging

An ultrasound vector Doppler imaging is useful for detecting flow components normal to the ultrasound beam direction. However, the conventional vector Doppler imaging method suffers from the bias of the time interval between samples caused by the mismatch between transmit and receive directions during demodulation. In this paper, a new directional demodulation method, in which demodulation is performed with a modified sample interval depending on the receive beam steered angle to reduce the bias occurred in a conventional ultrasound vector Doppler imaging is presented. To evaluate the performance of the proposed directional demodulation method, the pre-beamformed radio-frequency (RF) data from in-vitro experiments were obtained using a commercial ultrasound system and a Doppler string phantom. The true flow velocity of the phantom was 0.3 m/s. The center frequency of 5 MHz and the pulse repetition frequency of 4 kHz were used for the experiments. Also, a 32-element sub-aperture on a 128-element 7.2-MHz linear array probe were used for emission and reception while changing the flow direction from -45 degrees to 0 degree by a step of 5 degrees. The proposed directional demodulation method successfully visualizes all flow directions. In addition, it lowers a bias on flow estimation compared to the conventional method (i.e., 0.0255±0.0516 m/s vs. 0.0248±0.0469 m/s of error of velocity, 2.4862±3.8911 degrees vs. 2.4857±3.5115 degrees of error of direction, respectively). These results indicate that the proposed directional demodulation method can enhances the accuracy in flow estimation for vector Doppler imaging.