1432

IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control ,

vol. 59, no. 7,

July

2012

An FPGA-Based Open Platform for Ultrasound Biomicroscopy

Weibao Qiu, Yanyan Yu, Fu Keung Tsang, and Lei Sun
AbstractUltrasound biomicroscopy (UBM) has been extensively applied to preclinical studies in small animal models. Individual animal study is unique and requires different utilization of the UBM system to accommodate different transducer characteristics, data acquisition strategies, signal processing, and image reconstruction methods. There is a demand for a flexible and open UBM platform to allow users to customize the system for various studies and have full access to experimental data. This paper presents the development of an open UBM platform (center frequency 20 to 80MHz) for various preclinical studies. The platform design was based on a fieldprogrammable gate array (FPGA) embedded in a printed circuit board to achieve B-mode imaging and directional pulsedwave Doppler. Instead of hardware circuitry, most functions of the platform, such as filtering, envelope detection, and scan conversion, were achieved by FPGA programs; thus, the system architecture could be easily modified for specific applications. In addition, a novel digital quadrature demodulation algorithm was implemented for fast and accurate Doppler profiling. Finally, test results showed that the platform could offer a minimum detectable signal of 25V, allowing a 51dB dynamic range at 47dB gain, and real-time imaging at more than 500frames/s. Phantom and in vivo imaging experiments were conducted and the results demonstrated good system performance.

I. Introduction igh-resolution, noninvasive visualization of living tissues is an indispensable technique for observing physiological activity on a miniature scale. Diseases associated with small human structures such as eyes, skin, and blood vessels can benefit from this technique. In addition, this technique can also facilitate preclinical research of small animal models of human diseases in mice, rats, zebrafish, etc. Preclinical study of small animal models can improve the understanding of physiological and functional mechanisms for pathologies and treatment strategies. Current high-resolution imaging modalities include micro-computed tomography (CT) [1], micro-magnetic resonance imaging (MRI) [2], micro-positron emission tomography (PET) [3], optical coherence tomography (OCT) [4], and ultrasound biomicroscopy (UBM) [5],[6]. UBM is able to delineate small structures with fine spatial resolution on the order of tens of micrometers [7]. It is
Manuscript received May 14, 2011; accepted January 4, 2012. Financial support from the Hong Kong Research Grant Council (RGC) General Research Fund (GRF; PolyU 5301/09E), and The Hong Kong Polytechnic University (A-PJ84) are gratefully acknowledged. The authors are with the Department of Health Technology and Informatics, The Hong Kong Polytechnic University, Hong Kong SAR, China (e-mail: htsunlei@inet.polyu.edu.hk). DOI http://dx.doi.org/10.1109/TUFFC.2012.2344 08853010/$25.00

clinically used for ophthalmology [8], dermatology [9], and intravascular diseases [10]. Preclinical small animal model research has also been significantly propelled by UBM, for example, in malignant tumor diagnosis [11], [12], cardiac diseases [13], [14], and embryonic developmental biology [15], [16]. Recently, UBM has been combined with optical methods such as OCT or fluorescence spectroscopy as a dual-modality imaging technique and has demonstrated great potential in biomedical studies [17][19]. Other advanced imaging techniques also take advantage of highresolution UBM and extended its biomedical applications, such as photoacoustic imaging [20], [21], contrast enhanced imaging [22], coded excitation imaging [23], 3-D imaging [24], elastography imaging [10], [25], and harmonic imaging [26]. Each study is unique in nature, and requires different utilization of the UBM system. UBM systems with fixed specifications, such as transducer characteristics, data acquisition strategy, signal processing method, and image display approach, do not satisfy broad preclinical study requirements. Researchers need a highly flexible device to best suit their specific investigations. In addition, access to the raw experimental data is also important to scientific discovery. Therefore, an open and flexible UBM system allowing users and researchers to customize the system for a specific biomedical study is necessary. Currently, several UBM systems have been reported with different implementations [5], [27][31], including a commercially available UBM system (Vevo 770, VisualSonics Inc., Toronto, Canada). These systems utilize single-element transducers attached to a mechanical motor to perform a 2-D scan. A typical setup for such a UBM system is shown in Fig. 1. A single-element transducer is translated by a motor probe to acquire a cross-sectional image. A pulse generator is employed to produce a highvoltage pulse to excite the transducer to generate ultrasonic waves. A transmit and receive (T/R) switch controls the transmission of ultrasonic waves and the reception of the echo signal backscattered from the tissue. The imaging platform typically includes an amplifier to boost the weak echo signal, a band-pass filter (BPF) to remove noise, an analog-to-digital converter (ADC) to digitize the amplified signal, and a computer to process the digital signal and display the ultrasound images. Such a system normally uses module-based components, a commercial ADC or oscilloscope for signal digitization, and a computer for data processing and image display. Although a modulebased strategy can achieve a fast implementation, such systems have fixed characteristics including circuitry and specifications, and require post-processing for visualization. In addition, RF data transmission and processing

2012 IEEE

qiu et al.: an FPGA-based open platform for ultrasound biomicroscopy

1433

Fig. 1. Typical setup for single-element transducer-based ultrasound biomicroscopy.

needs fast data transfer and powerful calculation capabilities. For this reason, the frame rate is limited. Even though the graphics processing unit (GPU) was used in some UBM systems [31], [32], ultrasound data transfer is still challenging because of the limited speed of computer interfaces. High-frequency array transducer-based UBM systems [6], [26], [33][38] can achieve electronic dynamic focusing to increase the depth of field, including a commercially available UBM system (Vevo 2100, VisualSonics Inc.). In addition, electronic beam steering of linear array systems can avoid mechanical scanning, which allows for better safety and imaging speed. However, Madsen et al. [39] discovered that single-element transducer-based systems still have a better resolution in the focus zone than array system. In addition, the design and fabrication of highfrequency array transducers and electronic systems are highly complex, challenging, and costly. To a large extent, single-element transducer-based systems provide a viable and cost-effective solution. Field-programmable gate arrays (FPGAs) play an increasingly important role in modern electronics as programmable logic devices. They support accurate, fast signal processing. Traditional hardware circuitry and computer software algorithms can be replaced by FPGAs for field computation. Furthermore, FPGA algorithms provide reconfigurable and programmable features which have great potential for flexible implementation of UBM systems. This paper describes the development of a compact, highly flexible open UBM platform to satisfy various biomedical investigations. An FPGA serves as the core processor and replaces hardware circuitry with high processing speed and programmability. We demonstrate that both B-mode imaging and directional pulsed-wave (PW) Doppler can be realized on this platform. In addition, a novel high-frequency digital quadrature demodulation algorithm using the Hilbert transform was implemented to

achieve fast and accurate PW Doppler profiling. A coordinated rotation digital computer (Cordic) algorithm was employed for envelope detection together with Hilbert transform. Moreover, low-noise, high-speed analog electronics were used to achieve a high SNR and high sensitivity. A 64-bit PCI bus was incorporated in this platform as a high-speed data transmission interface for image data or raw RF data transfer to a computer through direct memory access operation. All of the electronics, such as amplifiers, filters, ADC, and FPGA, were incorporated in a single printed circuit board (PCB) for compactness and cost-effectiveness. Finally, phantom and in vivo measurements were conducted to demonstrate the system performance. II. Materials and Methods A. Hardware Implementation The block diagram of the open platform is shown in Fig. 2. To achieve a compact design, all of the electronic components are implemented in a single PCB. The echo signal received from the transducer is sent to the platform through a coaxial cable. After low-noise amplification and band-pass filtering, the echo signal is converted to a digital signal with a high-speed ADC. As the core processor of the platform, an FPGA is used to process the digitized signal and form real-time images. The algorithms implemented in the FPGA can easily adjust imaging settings through GUI software for different application requirements. The user can also choose different imaging strategies and reprogram the FPGA to satisfy various customized investigations. An external RAM with large capacity and high-speed interface is configured to the FPGA to support programmable applications such as a flexible digital scan converter (DSC). A 64-bit high-speed PCI interface is employed for real-time imaging and fast RF

1434

IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control ,

vol. 59, no. 7,

July

2012

Fig. 2. Schematic of the open ultrasound biomicroscopy (UBM) platform.

data transfer. A series of flexible communication ports are incorporated into the platform to interface with other external components such as motor controllers or pulse generators for synchronization, configuration, and communication. The following sections explain the detailed design and implementation of this open platform. 1) Front-End Analog Electronics: The quality of UBM images is significantly influenced by the front-end analog electronics. Poor electronic components selection and circuit design can increase the noise level and corrupt the ultrasound signals. This is especially serious for UBM, in which the echo signal is very weak. If the signal is distorted in the front-end, it can no longer be recovered by postprocessing. In this compact design, special attention was paid to reducing the noise level and to achieving high sensitivity. A small-size low-noise amplifier (SMA231, Tyco Electronics Co., Berwyn, PA) was used as the first stage amplifier, also called the pre-amplifier. The maximum noise figure was 1.7dB and the minimum gain is 26dB with gain flatness lower than 0.2dB in the frequency range from 10 to 250MHz. A low-pass filter (RLP83+, Mini-Circuits, Brooklyn, NY) with a cut-off frequency of 93MHz was selected for anti-aliasing. The insertion loss of the low-pass filter was less than 1dB. A low-noise, lowdistortion amplifier (THS4509, Texas Instruments Inc., Dallas, TX) was employed as the second-stage amplifier for additional 20dB amplification. The second-stage amplifier was also used to convert the single-ended signal to differential pair for the later ADC connection. 2) Analog-to-Digital Converter: According to the Nyquist sampling theorem, the sampling frequency of the ADC must be at least twice the maximum frequency of the ultrasound echo signal to avoid signal distortion. The quadrature sampling technique could minimize the sampling frequency to half of the Nyquist requirement by using two separate ADCs [36], but this complicates the design and compromises the ADC performance. However, it may be an option for cases in which the ADC sampling frequency is not sufficiently high. In this platform design, we utilize a single ADC (ADS5517, Texas Instruments Inc.) with a maximum sampling frequency of 200 MHz,

which is sufficiently high for the UBM platform. The typical effective number of bits of this ADC is 10.8bits. After digitization, the digital signal is transferred to the FPGA through a low-voltage differential signaling (LVDS) bus. An integrated voltage-controlled oscillator and a low-jitter clock generator (CDCM61002, Texas Instruments Inc.) was carefully selected to serve as accurate clock source for high-precision data acquisition. 3) Field-Programmable Gate Array: An FPGA is an integrated circuit which contains a hierarchy of reconfigurable interconnections designed to be configured by the user. Reconfigurable properties and high-speed implementation make this type of device very suitable for the design of flexible and programmable system. The FPGA programming may replace hardware circuits and achieve a similar functionality. Moreover, in such implementations, the function can easily be changed or modified by reprogramming the FPGA. In this platform design, a high-performance FPGA (Stratix II EP2S60F672I4, Altera Corp., San Jose, CA) with great signal integrity was employed to support highspeed signal processing, high programmability, and fast data transfer. The FPGA has 60 440 equivalent logic elements and 493 user I/O pins. It includes 36 digital signal processing (DSP) blocks (a total of two-hundred-eightyeight 9 9 multipliers), which can efficiently implement various signal processing algorithms. High-order filtering, Hilbert transform, envelope detection, digital scan conversion, and other algorithms can be implemented. The onchip LVDS termination reduces reflections and maintains signal integrity with a compact PCB size. A 64-MB RAM combined with two DDR2 SDRAMs (MT47H16M16, Micron Technology Inc., Boise, ID) is used for temporary buffering of data. The image data or raw RF data are transferred to a computer through a 66-MHz 64-bit PCI bus for storage or post-processing. In the current platform, B-mode imaging and directional PW Doppler have been realized in this platform. When other imaging methods should be used, such as coded excitation imaging, the FPGA can be reprogrammed without any hardware circuit changes.

qiu et al.: an FPGA-based open platform for ultrasound biomicroscopy

1435

B. Algorithms As the core processor, the FPGA algorithms and implementation determine the performance of system. Fig. 3 shows the structure of the implemented algorithms for real-time B-mode imaging and directional PW Doppler. For B-mode imaging, BPF, digital time-gain compensation (TGC), and envelope detection are performed one after the other for data preconditioning. After removing noise with the BPF, the data are digitally amplified by configurable coefficients to compensate for ultrasound attenuation in the tissue. The acquired envelope data then undergo DSC and logarithmic compression. TGC, DSC, and logarithmic compression are all reconfigurable by the user through a GUI. For directional PW Doppler, the BPF is also employed, but in a different configuration optimized for flow measurement. A novel digital quadrature demodulation is used to extract the flow information from the echo signal. The data are then saved in the dual-port RAM for Doppler gate selection. After a train of data is accumulated, the spectrum extraction is started to estimate the flow velocity. Finally, image data or spectrogram data are transferred to the computer through the PCI bus for display and storage. The following sections give more details on the individual algorithms. 1) Reconfigurable Band-Pass Filter: Ultrasound echo signals usually contain noise, which interferes with the echo signal and disturbs imaging. Typically, analog filters are employed to eliminate undesirable noise [36]. However, analog filters are difficult to change and are normally fixed for a specific frequency range. Software-based digital BPF, on the other hand, is flexible, stable, small in size, and easy to change. Taking advantage of the FPGA, softwarebased reconfigurable BPF was implemented in this platform. With symmetric of finite impulse response (FIR) filter coefficients, the number of multipliers can be cut in half, which saves limited resources in the FPGA. To achieve fast processing, the FPGA employs internal high-speed

DSP blocks, which are faster than adaptive logic module (ALM)-based implementations. Although a higher-order FIR filter shows large SNR improvements, it results in longer ripples which can decrease the axial resolution. A 31-tap FIR filter was selected in the FPGA with a satisfactory trade-off between SNR and ripples. More than 40dB noise suppression was achieved with this filter. The FPGA simulation results showed that the maximum clock frequency of this FIR filter was higher than 287MHz. The coefficients of the FIR filter are flexible and reconfigurable by the user through a GUI to meet the requirements of different applications. The coefficients are also adjustable for transducers with different center frequency and bandwidth. 2) Envelope Detection: The amplitude of echo signals, namely the envelope, is utilized for B-mode ultrasound imaging. B-mode images represent the backscattering of ultrasound from the tissue. Because ultrasound echo signals are broadband, the Hilbert transform is used to extract the envelope. In the FPGA, the Hilbert transform was realized with an FIR filtering scheme [40]. Fig. 4 shows a simplified architecture of the designed envelope detection method. The in-phase (I) and quadrature (Q) signal were acquired by the Hilbert transform. The modulus of I/Q signals was calculated to remove the carrier frequency and to obtain the envelope. This function could be realized by several algorithms [41]. To achieve a high processing speed for real-time imaging, the Cordic algorithm was implemented. Cordic is an iterative process, using a series of adders and shifters [42]. It can support an efficient solution with high performance for modulus calculation by the FPGA. The FPGA simulation results showed that the maximum clock frequency of the designed envelope detection algorithm was higher than 280MHz. 3) Flexible Digital Scan Conversion: Image reconstruction depends on the transducer scanning mechanism. For example, when the transducer is translated in a sector or rotation scan, the ultrasound signal is acquired at differ-

Fig. 3. Schematic of algorithms implemented in the field-programmable gate array (FPGA) for real-time B-mode imaging and directional pulsedwave Doppler.

1436

IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control ,

vol. 59, no. 7,

July

2012

Fig. 4. Block diagram of envelope detector based on Hilbert transform and Cordic algorithm.

ent angles in polar coordinates. The ultrasound image, on the other hand, is displayed in Cartesian coordinates on the monitor. Therefore, scan conversion from polar to Cartesian coordinates is needed for accurate visualization. A high-speed DSC with linear interpolation was previously designed in the FPGA [43]. However, the implementation of DSC for large images requires enormous RAM memory, which is limited in the FPGA. Therefore, a novel effective method based on high-speed DDR2 SDRAM is proposed in this paper. Fig. 5 shows a functional block diagram of the flexible DSC using the linear interpolation. In this implementation, all angle and position data are stored in a DDR2 SDRAM at 267MHz. By means of this, the size of the image can easily be adjusted. Three dual-port RAMs in the FPGA are used in turn to buffer the high-speed data and to improve the computational efficiency of interpolation. The configurable imaging parameters, such as the transducer focus length, swing angle of the transducer, and number of scan-lines can easily be adjusted by computer software in this implementation. 4) Digital Quadrature Demodulation: Traditionally, analog demodulators are the choice for designing a PW Doppler scanner [5], [44]. This involves electronic components such as mixers, sample-and-hold devices, and low-pass

filters. These components not only introduce extra noise, but are also limited to a specific transducer with a given frequency range. A completely digital implementation of PW Doppler has many advantages, such as low noise, stability, flexibility, and small size. In this platform, the directional PW Doppler uses the same front-end electronics and signal transmission path used for B-mode imaging. Previously, a digital mixer was used to acquire Doppler waveform [45]. However, the involvement of mixer made the implementation complicated and was subject to a precise selection of the demodulation frequency. Here, we propose a simple digital quadrature demodulation scheme which is directly applied to the RF data, and which eliminates the mixer by employing the Hilbert transform for precise I/Q acquisition. The method of Hilbert transform implementation is similar to that described previously in envelope detection section. After acquiring the I/Q signals, two separate RAMs are employed to buffer the I/Q data for Doppler gate selection. One data point is chosen corresponding to the depth at which the Doppler signal is interrogated. Afterward, a train of such data points from consecutive A-lines is collected and transferred to the spectrum extraction module for flow velocity estimation. 5) Spectrum Extraction: Spectral analysis is conducted to estimate the flow velocity and the blood flow pattern

Fig. 5. Functional block diagram of flexible digital scan converter with linear interpolation.

qiu et al.: an FPGA-based open platform for ultrasound biomicroscopy

1437

Fig. 6. Block diagram of spectrum extraction for directional pulsed-wave Doppler.

after quadrature demodulation. Fourier transform is employed as an essential spectral analysis tool. The FPGA is capable of high-speed handling of the Fourier transform. Fig. 6 shows the block diagram for the spectrum extraction in the FPGA. A programmable wall filter is inserted to remove the low-frequency signal caused by non-moving tissue or motion of the surrounding tissue. I/Q demodulated data are transferred into the first-in, first-out (FIFO) buffers after the wall filtering. Two separate FIFO buffers work in turns for each I/Q channel to process the data in a sliding rectangular window. Complex fast Fourier transform (CFFT) is employed to calculate the frequency spectrum of the windowed data. The Cordic algorithm discussed previously is used to calculate the modulus of I/Q data. Afterward, the results are transferred to the computer through the PCI interface for display and storage. The number of data points used for spectral analysis is adjustable by the user through a GUI. Typically, the more the data points are used, the better the velocity resolution is. However, the temporal resolution gets worse with increasing number of points and the processing speed slows down. An appropriate number of data points gives a good compromise among frequency resolution, temporal resolution, and processing speed. 256-point CFFT was employed in this design. The FPGA simulation results showed that the maximum clock frequency was higher than 240MHz. III. Evaluation and Results Fig. 7 shows a photograph of the open platform prototype. A 12-layer PCB with a size of 16.8 10.6cm includes low-noise analog electronics, ADC, DDR2 SDRAM, high-speed FPGA, and PCI bus. A. Performance of Front-End Electronics The performance of the front-end electronics, such as the gain flatness, was tested by a set of RF instruments, including a signal generator (HP8656B, Hewlett-Packard Co., Palo Alto, CA) and a spectrum analyzer (HP8591E, Hewlett-Packard Co.). The spectrum analyzer was used to receive and analyze the amplified signal in different frequencies sent by the signal generator and processed by the front-end electronics. Another test was conducted to

measure the minimum detectable signal and the dynamic range. A five-cycle sinusoidal signal produced by a function generator (AFG 3251, Tektronix Inc., Beaverton, OR) was attenuated by a series of attenuators (Mini-Circuits) and fed into the platform. The signal level at which the weak signal passing through the platform could be distinguished from the background noise was used to determine the minimum detectable signal of the system. Fig. 8 shows the system gain as a function of frequency. The front-end electronics shows a good flatness of gain with fluctuations smaller than 1.2dB in the range from 10 to 90MHz. Test results indicated that the platform has a minimum detectable signal of 25V. Given the input range of the ADC (2Vpp), the system has a 51dB dynamic range with a gain of 47dB and a 50MHz center frequency. No digital filtering was used in this test. B. Performance of Digital Band-Pass Filter The performance of the digital BPF was tested as follows. Twenty cycles of a 50-V sinusoidal signal were generated with a function generator and sent to the platform. The unprocessed raw RF data and the data processed by the BPF, shown in Fig. 9, were transferred to the PC. It can be seen that the SNR after the BPF is significantly

Fig. 7. Photograph of the open ultrasound biomicroscopy platform on a 12-layer printed circuit board.

1438

IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control ,

vol. 59, no. 7,

July

2012

Fig. 8. Gain flatness of the front-end electronics.

improved. In addition, dc offsets are eliminated by the filter. Quantitative analysis showed approximately 4.8dB improvement in SNR by the 31-tap FIR filter, which improves the dynamic range of the system to approximately 56dB. C. Test of Imaging Speed To evaluate the image processing and data transfer speed of the platform, a burst of 50-MHz sinusoidal signal with high pulse repetition frequency was generated by a function generator and sent to the platform. A frame counter in the FPGA calculated the speed of the platform in terms of the number of processed images per second. The test result showed that the platform could process more than 500 images (256 256 pixels) per second. D. Phantom Imaging A customized multi-functional pulse generator, shown in Fig. 10(a), was employed to generate high-voltage pulses for evaluation of this platform. A high-speed MOSFET is used to generate several high-frequency bipolar or spike pulses for B-mode imaging and PW Doppler. A high-speed sector probe and a digital servo controller [29], shown in Fig. 10(b), are incorporated for the evaluation of the platform. The setup is capable of supporting up to

Fig. 10. (a) Photos of the multi-functional pulse generator and (b) highspeed sector probe with digital servo controller.

Fig. 9. (a) Unprocessed raw data and (b) processed data with a 31-tap band-pass filter implemented in the field-programmable gate array.

200 frames of B-mode images per second. The entire system was built according to the architecture shown in Fig. 1. GUI software was designed to receive and display the ultrasound images for real-time visualization. An ultrasound image of a wire phantom in water is shown in Fig. 11 with a dynamic range of 50dB. The wire phantom consists of five 20-m-diameter tungsten wires (California Fine Wire Co., Grover Beach, CA) with about 1.6 and 1.0mm distance intervals along the vertical and horizontal directions, respectively. In this experiment, a 35-MHz polyvinylidene fluoride (PVDF) single-element transducer (focus length 13.6 mm, relative bandwidth 88%, f-number 2.7) was employed. No noise is visible in the image. The appearance of the wires in the image is similar to the result from a previous module-based system[46].

qiu et al.: an FPGA-based open platform for ultrasound biomicroscopy

1439

Fig. 11. Wire phantom image using the open platform. A 35-MHz polyvinylidene fluoride single-element transducer (focus length 13.6mm, relative bandwidth 88%, f-number 2.7) was employed.

A tissue-mimicking phantom was fabricated to further test the image quality of the open platform. The phantom fabrication procedure followed Madsens method [39]. It consists of a mixture of deionized water, high-grade agarose, preservative, propylene glycol, filtered bovine milk, and glass beads to generate tissue mimicking attenuation and backscattering. Anechoic spheres were fabricated separately and dispersed in the phantom to test the system resolution [39]. The size of the anechoic spheres was in the range of 180 to 280m, controlled by a dedicated sieve (Fisherbrand sieves, Fisher Scientific, Pittsburgh, PA). Fig. 12 shows ultrasound images of the tissue phantom acquired by the platform prototype and a commercial UBM scanner (Vevo 770, 55MHz, VisualSonics Inc.). The black circular dots in the images are the anechoic spheres. The measured diameter of the spheres in the images is approximately 180m, which is in good agreement with the actual sphere size. The transducer for this experiment was a 50-MHz LiNbO3 transducer with a focal length of 9.3mm and a relative bandwidth of 47% [47]. The overall image quality of the two systems in terms of noise level, penetration depth, spatial resolution, and dynamic range is similar. E. Pulsed-Wave Doppler For quantitative verification of directional PW Doppler measurement, a movable stage, controlled by a high-precision servo motor (TSL-2504, Shenzhen Techsoft Motion Technology Co. Ltd., Shenzhen, China), was employed. The speed could be accurately controlled and the position precision was as high as 5m. The movable stage was controlled by a servo motor driver and traveled back and forth with a repetitive trapezoidal profile, shown in Fig. 13(a). The acceleration rate was 30mm/s2, slew speed was 15mm/s and travel distance was 20mm. A reflector, which was similar to the aforementioned tissue-mimicking phantom, was put into a water bath container attached

to the movable stage. The 50-MHz LiNbO3 transducer mentioned previously was fixed above the reflector in the water bath with an angle of approximately 72 between the ultrasound beam and the stage moving direction. A seven-cycle 40Vpp sinusoidal pulse was generated by a multifunctional pulse generator with a pulse repetition frequency of 3.9kHz. Fig. 13(b) shows the obtained spectrogram, demonstrating the velocity profile of the backand-forth moving stage measured by PW Doppler. After the Doppler angle correction, the measured Doppler waveform agreed well with the actual trapezoidal velocity pattern. Fig. 14 shows an in vivo spectrogram acquired from a vein at the back of a human hand. A similar test was also performed in [48]. A needle transducer (42.5MHz) [44] was placed close to the vein, coupled by ultrasound gel, at an angle of approximately 72. A seven-cycle 40-Vpp sinusoidal pulse was generated with a pulse repetition frequency of 1.95kHz. The blood actually flowing away from the transducer confirmed the negative velocity profile.

Fig. 12. Images of tissue phantom with anechoic spheres. (a) 48dB dynamic range. The trigger pulse is a single-cycle 150Vpp bipolar pulse. A 50-MHz LiNbO3 transducer with a focus length of 9.3mm and a relative bandwidth of 47% was employed. (b) Image of the same phantom using a 55-MHz Vevo 770 transducer.

1440

IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control ,

vol. 59, no. 7,

July

2012

Although PCI was chosen for the design, a USB interface scheme is a possible alternative. The imaging platform, together with a newly designed pulse generator and a sector motor probe [29], can significantly reduce the UBM size and make the equipment compact and portable. Furthermore, such a device can be upgraded to connect a monitor and hard disk directly for real-time imaging and image data saving without a CPU. The platform design may also be expanded to support array transducer-based UBM systems. An imaging system supporting an annular array transducer with 5 to 8 elements [49][53] is under development. It can also incorporate a catheter transducer for intravascular or endoscopic imaging. References
[1] C. T. Badea, M. Drangova, D. W. Holdsworth, and G. A. Johnson, In vivo small animal imaging using micro-CT and digital subtraction angiography, Phys. Med. Biol., vol. 53, no. 19, pp. R319R350, 2008. [2] B. Driehuys, J. Nouls, A. Badea, E. Bucholz, K. Ghaghada, A. Petiet, and L. W. Hedlund, Small animal imaging with magnetic resonance microscopy, ILAR J., vol. 49, no. 1, pp. 3553, 2008. [3] U. Schnckel, S. Hermann, L. Stegger, M. Law, M. Kuhlmann, O. Schober, K. Schfers, and M. Schfers, Small-animal PET: A promising, non-invasive tool in pre-clinical research, Eur. J. Pharm. Biopharm., vol. 74, no. 1, pp. 5054, 2010. [4] J. G. Fujimoto, Optical coherence tomography for ultrahigh resolution in vivo imaging, Nat. Biotechnol., vol. 21, no. 11, pp. 1361 1367, 2003. [5] F. S. Foster, M. Y. Zhang, Y. Q. Zhou, G. Liu, J. Mehi, E. Cherin, K. A. Harasiewicz, B. G. Starkoski, L. Zan, D. A. Knapik, and S. L. Adamson, A new ultrasound instrument for in vivo microimaging of mice, Ultrasound Med. Biol., vol. 28, no. 9, pp. 11651172, 2002. [6] F. S. Foster, J. Mehi, M. Lukacs, D. Hirson, C. White, C. Chaggares, and A. Needles, A new 1550MHz array-based micro-ultrasound scanner for preclinical imaging, Ultrasound Med. Biol., vol. 35, no. 10, pp. 17001708, 2009. [7] F. S. Foster, C. J. Pavlin, K. A. Harasiewicz, D. A. Christopher, and D. H. Turnbull, Advances in ultrasound biomicroscopy, Ultrasound Med. Biol., vol. 26, no. 1, pp. 127, 2000. [8] R. H. Silverman, J. A. Ketterling, and D. J. Coleman, High-frequency ultrasonic imaging of the anterior segment using an annular array transducer, Ophthalmology, vol. 114, no. 4, pp. 816822, 2007. [9] M. Vogt and H. Ermert, In vivo ultrasound biomicroscopy of skin: Spectral system characteristics and inverse filtering optimization, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 54, no. 8, pp. 15511559, 2007. [10] C. L. de Korte, A. F. W. van der Steen, E. I. Cespedes, G. Pasterkamp, S. G. Carlier, F. Mastik, A. H. Schoneveld, P. W. Serruys, and N. Bom, Characterization of plaque components and vulnerability with intravascular ultrasound elastography, Phys. Med. Biol., vol. 45, no. 6, pp. 14651475, 2000. [11] A. M. Cheung, A. S. Brown, L. A. Hastie, V. Cucevic, M. Roy, J. C. Lacefield, A. Fenster, and F. S. Foster, Three-dimensional ultrasound biomicroscopy for xenograft growth analysis, Ultrasound Med. Biol., vol. 31, no. 6, pp. 865870, 2005. [12] W. Goessling, T. E. North, and L. I. Zon, Ultrasound biomicroscopy permits in vivo characterization of zebrafish liver tumors, Nat. Methods, vol. 4, no. 7, pp. 551553, 2007. [13] Y. Q. Zhou, F. S. Foster, B. J. Nieman, L. Davidson, X. J. Chen, and R. M. Henkelman, Comprehensive transthoracic cardiac imaging in mice using ultrasound biomicroscopy with anatomical confirmation by magnetic resonance imaging, Physiol. Genomics, vol. 18, no. 2, pp. 232244, 2004. [14] J. Du, C. Zhang, J. Liu, C. Sidky, and X. P. Huang, A point mutation (R192H) in the C-terminus of human cardiac troponin I causes diastolic dysfunction in transgenic mice, Arch. Biochem. Biophys., vol. 456, no. 2, pp. 143150, 2006.

Fig. 13. Pulsed-wave Doppler image of a trapezoidally moving reflector. A 50-MHz LiNbO3 transducer with a focus length of 9.3mm and a relative bandwidth of 47% was employed.

The cyclic pattern of the spectrogram correlated well to the heartbeat of the subject. The heart rate of 80beats/ min measured from the spectrogram agreed with the heart rate of 78beats/min of the subject. IV. Conclusions A programmable open platform for real-time UBM based on a high-speed FPGA was developed and evaluated. It was implemented in a compact and cost-effective PCB scheme. This open platform demonstrated high image quality with a good spatial resolution for preclinical small animal imaging. The flexible and programmable design makes the open platform suitable for various biomedical studies.

Fig. 14. Pulsed-wave Doppler waveform of a vein in the back of a human hand. A 42.5-MHz needle transducer was employed.

qiu et al.: an FPGA-based open platform for ultrasound biomicroscopy [15] Y. Q. Zhou, F. S. Foster, D. W. Qu, M. Zhang, K. A. Harasiewicz, and S. L. Adamson, Applications for multifrequency ultrasound biomicroscopy in mice from implantation to adulthood, Physiol. Genomics, vol. 10, no. 2, pp. 113126, 2002. [16] S. Kulandavelu, D. Qu, N. Sunn, J. Mu, M. Y. Rennie, K. J. Whiteley, J. R. Walls, N. A. Bock, J. C. Sun, A. Covelli, J. G. Sled, and S. L. Adamson, Embryonic and neonatal phenotyping of genetically engineered mice, ILAR J., vol. 47, no. 2, pp. 103117, 2006. [17] T. J. Rmer, J. F. Brennan III, G. J. Puppels, A. H. Zwinderman, S. G. van Duinen, A. van der Laarse, A. F. van der Steen, N. A. Bom, and A. V. Bruschke, Intravascular ultrasound combined with Raman spectroscopy to localize and quantify cholesterol and calcium salts in atherosclerotic coronary arteries, Arterioscler. Thromb. Vasc. Biol., vol. 20, no. 2, pp. 478483, 2000. [18] Y. Sun, J. Park, D. N. Stephens, J. A. Jo, L. Sun, J. M. Cannata, R. M. Saroufeem, K. K. Shung, and L. Marcu, Development of a dual-modal tissue diagnostic system combining time-resolved fluorescence spectroscopy and ultrasonic backscatter microscopy, Rev. Sci. Instrum., vol. 80, no. 6, art. no. 065104, 2009. [19] H. C. Yang, J. Yin, C. Hu, J. Cannata, Q. Zhou, J. Zhang, Z. Chen, and K. K. Shung, A dual-modality probe utilizing intravascular ultrasound and optical coherence tomography for intravascular imaging applications, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 57, no. 12, pp. 28392843, 2010. [20] L. V. Wang, Multiscale photoacoustic microscopy and computed tomography, Nat. Photonics, vol. 3, no. 9, pp. 503509, 2009. [21] S. Sethuraman, J. H. Amirian, S. H. Litovsky, R. W. Smalling, and S. Y. Emelianov, Spectroscopic intravascular photoacoustic imaging to differentiate atherosclerotic plaques, Opt. Express, vol. 16, no. 5, pp. 33623367, 2008. [22] D. E. Goertz, E. Cherin, A. Needles, R. Karshafian, A. S. Brown, P. N. Burns, and F. S. Foster, High frequency nonlinear B-scan imaging of microbubble contrast agents, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 52, no. 1, pp. 6579, 2005. [23] J. Mamou, O. Aristizbal, R. H. Silverman, J. A. Ketterling, and D. H. Turnbull, High-frequency chirp ultrasound imaging with an annular array for ophthalmologic and small-animal imaging, Ultrasound Med. Biol., vol. 35, no. 7, pp. 11981208, 2009. [24] L. A. Wirtzfeld, G. Wu, M. Bygrave, Y. Yamasaki, H. Sakai, M. Moussa, J. I. Izawa, D. B. Downey, N. M. Greenberg, A. Fenster, J. W. Xuan, and J. C. Lacefield, A new three-dimensional ultrasound microimaging technology for preclinical studies using a transgenic prostate cancer mouse model, Cancer Res., vol. 65, no. 14, pp. 63376345, 2005. [25] R. L. Maurice, M. Daronat, J. Ohayon, E. Stoyanova, F. S. Foster, and G. Cloutier, Non-invasive high-frequency vascular ultrasound elastography, Phys. Med. Biol., vol. 50, no. 7, pp. 16111628, 2005. [26] A. Needles, M. Arditi, N. G. Rognin, J. Mehi, T. Coulthard, C. Bilan-Tracey, E. Gaud, P. Frinking, D. Hirson, and F. S. Foster, Nonlinear contrast imaging with an array-based micro-ultrasound system, Ultrasound Med. Biol., vol. 36, no. 12, pp. 20972106, 2010. [27] E. Cherin, R. Williams, A. Needles, G. Liu, C. White, A. S. Brown, Y. Q. Zhou, and F. S. Foster, Ultrahigh frame rate retrospective ultrasound microimaging and blood flow visualization in mice in vivo, Ultrasound Med. Biol., vol. 32, no. 5, pp. 683691, 2006. [28] J. Liu, G. Jeng, T. Wu, and P. C. Li, ECG triggering and gating for ultrasound small animal imaging, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 53, no. 9, pp. 15901596, 2006. [29] L. Sun, X. Xu, W. D. Richard, C. Feng, J. A. Johnson, and K. K. Shung, A high-frame rate duplex ultrasound biomicroscopy for small animal imaging in vivo, IEEE Trans. Biomed. Eng., vol. 55, no. 8, pp. 20392049, 2008. [30] L. Sun, C. Lien, X. Xu, and K. K. Shung, In vivo cardiac imaging of adult zebrafish using high frequency ultrasound (4575 MHz), Ultrasound Med. Biol., vol. 34, no. 1, pp. 3139, 2008. [31] M. R. Bosisio, J. M. Hasquenoph, L. Sandrin, P. Laugier, S. L. Bridal, and S. Yon, Real-time chirp-coded imaging with a programmable ultrasound biomicroscope, IEEE Trans. Biomed. Eng., vol. 57, no. 3, pp. 654664, 2010. [32] M. Lewandowski and A. Nowicki, High frequency coded imaging system with RF software signal processing, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 55, no. 8, pp. 18781882, 2008. [33] C. Hu, X. Xu, J. M. Cannata, J. T. Yen, and K. K. Shung, Development of a real-time, high-frequency ultrasound digital beamformer for high-frequency linear array transducers, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 53, no. 2, pp. 317323, 2006.

1441

[34] X. Xu, L. Sun, J. M. Cannata, J. T. Yen, and K. K. Shung, Highfrequency ultrasound Doppler system for biomedical applications with a 30-MHz linear array, Ultrasound Med. Biol., vol. 34, no. 4, pp. 638646, 2008. [35] L. Zhang, X. Xu, C. Hu, L. Sun, J. T. Yen, J. M. Cannata, and K. K. Shung, A high-frequency, high frame rate duplex ultrasound linear array imaging system for small animal imaging, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 57, no. 7, pp. 15481557, 2010. [36] J. A. Brown and G. R. Lockwood, A digital beamformer for highfrequency annular arrays, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 52, no. 8, pp. 12621269, 2005. [37] C. Hu, K. A. Snook, P. Cao, and K. K. Shung, High-frequency ultrasound annular array imaging, Part II: Digital beamformer design and imaging, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 53, no. 2, pp. 309316, 2006. [38] J. A. Ketterling, S. Ramachandran, and O. Aristizabal, Operational verification of a 40-MHz annular array transducer, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 53, no. 3, pp. 623630, 2006. [39] E. L. Madsen, G. R. Frank, M. M. McCormick, M. E. Deaner, and T. A. Stiles, Anechoic sphere phantoms for estimating 3-D resolution of very-high-frequency ultrasound scanners, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 57, no. 10, pp. 22842292, 2010. [40] P. Levesque and M. Sawan, Real-time hand-held ultrasound medical-imaging device based on a new digital quadrature demodulation processor, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 56, no. 8, pp. 16541665, 2009. [41] P. Montuschi and M. Mezzalama, Survey of square rooting algorithms, Proc. Inst. Elect. Eng. E, vol. 137, no. 1, pp. 3140, 1990. [42] J. E. Volder, The CORDIC trigonometric computing technique, IRE Trans. Electron. Comput., vol. EC-8, pp. 330334, 1959. [43] J. Chang, J. T. Yen, and K. K. Shung, High speed digital scan converter for high-frequency ultrasound sector scanners, Ultrasonics, vol. 48, no. 5, pp. 444452, 2008. [44] Q. Zhou, X. Xu, E. J. Gottlieb, L. Sun, J. M. Cannata, H. Ameri, M. S. Humayun, P. Han, and K. K. Shung, PMN-PT single crystal, high-frequency ultrasonic needle transducers for pulsed-wave Doppler application, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 54, no. 3, pp. 668675, 2007. [45] C. Hu, Q. Zhou, and K. K. Shung, Design and implementation of high frequency ultrasound pulsed-wave Doppler using FPGA, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 55, no. 9, pp. 21092111, 2008. [46] L. Sun, W. D. Richard, J. M. Cannata, C. C. Feng, J. A. Johnson, J. T. Yen, and K. K. Shung, A high-frame rate high-frequency ultrasonic system for cardiac imaging in mice, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 54, no. 8, pp. 16481655, 2007. [47] J. M. Cannata, T. A. Ritter, W. H. Chen, R. H. Silverman, and K. K. Shung, Design of efficient, broadband single-element (2080 MHz) ultrasonic transducers for medical imaging applications, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 50, no. 11, pp. 15481557, 2003. [48] M. Vogt, Direct sampling and baseband conversion in Doppler systems for high-frequency ultrasound blood flow measurements, Electron. Lett., vol. 41, no. 14, pp. 789790, 2005. [49] J. A. Brown, C. E. M. Demore, and G. R. Lockwood, Design and fabrication of annular arrays for high frequency ultrasound, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 51, no. 8, pp. 1010 1017, 2004. [50] J. A. Ketterling, O. Aristizabal, D. H. Turnbull, and F. L. Lizzi, Design and fabrication of a 40-MHz annular array transducer, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 52, no. 4, pp. 672681, 2005. [51] K. A. Snook, C. Hu, T. R. Shrout, and K. K. Shung, High-frequency ultrasound annular array imaging, Part I: Array design and fabrication, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 53, no. 2, pp. 300308, 2006. [52] E. J. Gottlieb, J. M. Cannata, C. Hu, and K. K. Shung, Development of a high-frequency (>50 MHz) copolymer annular-array, ultrasound transducer, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 53, no. 5, pp. 10371045, 2006. [53] H. R. Chabok, J. M. Cannata, H. H. Kim, J. A. Williams, J. Park, and K. K. Shung, A high-frequency annular-array transducer using an interdigital bonded 1-3 composite, IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 58, no. 1, pp. 206214, 2011.

1442

IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control ,

vol. 59, no. 7,

July

2012

Weibao Qiu was born in Jilin Province, China, in 1982. He received his B.S. degree from the Hefei University of Technology, Hefei, China, in 2004 and the M.S. degree from the State Key Laboratory of Precision Measurement Technology and Instruments, Tianjin University, Tianjin, China, in 2007. He worked at the RFID Research Center, ZTE Corporation, China, as a hardware/FPGA research engineer for two years after receiving his masters degree. He is currently a Ph.D. candidate in biomedical engineering in the Department of Health Technology and Informatics, The Hong Kong Polytechnic University, Hong Kong, China. His research interests include digital signal processing, FPGA algorithms, high-speed circuit design, high-frequency ultrasound imaging systems, digital beamformers, coded excitation imaging, IVUS imaging, and Doppler imaging.

Fu Keung Tsang was born in Dongguan, China, in 1987. He received the B.S. degree in biomedical engineering in 2010 from the Hong Kong Polytechnic University, Kowloon, Hong Kong. He is currently a research assistant in the Department of Health Technology and Informatics, The Hong Kong Polytechnic University, Hong Kong, China. His research interests include ultrasound coded excitation imaging and small animal imaging.

Yanyan Yu was born in Jilin Province, China, in 1983. She received her B.S. and M.S. degrees from the Hefei University of Technology, Hefei, China, in 2006 and 2009, respectively. She worked at the Hefei Meiya Optoelectronic Technology Inc., China, as an image processing software engineer for nearly one year after receiving her masters degree. She is currently working as a research assistant in the Department of Health Technology and Informatics, The Hong Kong Polytechnic University, Hong Kong, China. Her research interests include digital image processing, ultrasonic imaging, and the theory and applications of acoustic radiation force.

Lei Sun was born in Shenyang, China. He received his B.S. degree in electrical engineering from the University of Science and Technology of China, Hefei, China, in 1996; the M.S. degree in electrical engineering from the Chinese Academy of Sciences, Beijing, China, in 2000; and the Ph.D. degree in bioengineering from The Pennsylvania State University, University Park, PA, in 2004. Currently, he is an Assistant Professor in the Department of Health Technology and Informatics, Hong Kong Polytechnic University, Hong Kong, China. From 2004 to 2008, he was a postdoctoral fellow and a Research Associate at the National Institutes of Health (NIH) Resource Center for Medical Ultrasonic Transducer Technology, University of Southern California, Los Angeles, CA. His research interests include ultrasonic and Doppler imaging, high-frequency ultrasound, magnetic resonance guided high-intensity focused ultrasound (HIFU), small animal imaging, cardiac imaging, ultrasonic tissue characterization, biomedical instrumentation, and biomedical signal and image processing.