Michael I. Fuller

A portable, low-cost, highly integrated, 3D medical ultrasound system

A design is presented, with prototype component experimental results, of a portable, low-cost 3D ultrasound system that utilizes innovative integrated circuit (IC) topologies and beamforming algorithms to condense the beamforming process into a small custom IC coupled with a generic, commercially available DSP chip. Furthermore, the system includes a low-cost, fully sampled 2D transducer array placed directly adjacent to the receive circuitry allowing precise impedance matching and improved SNR.

The sonic window: second generation prototype of low-cost, fully-integrated, pocket-sized medical ultrasound device

This paper will focus on the development of the second generation Sonic Window prototype, which will occupy a package the size of a deck of cards and be capable of forming C-mode ultrasound images in real-time. An array of 3,600 full receive channels—consisting of protection circuitry, a preamplifier, tuneable bandpass filter, sampler bank, and 8-bit ADC—is formed by flip-chip attaching twelve identical 300-channel custom ICs onto a double-layer flex circuit substrate. The receive channel pitch is 200 µm x 275 µm and the entire receive circuitry array has a form factor of only 1.9 cm x 1.8 cm. The flex circuit provides I/O fanout to a beamforming PCB containing bias circuitry, programmable logic, a DSP, and LCD for image display. The receive circuitry array connects through a z-axis electrically conducting interface to a 60x60 element, 300-micron- pitch, fully-sampled transducer array inexpensively fabricated on a double-layer PCB substrate. Experimental test results from a test chip containing receive circuitry components are presented, along with a description of the 3,600-channel receive circuitry array development.

Real time imaging with the Sonic Window: A pocket-sized, C-scan, medical ultrasound device

The Sonic Window is a pocket-sized, C-scan ultrasound device designed to expand ultrasound into clinical settings and applications that have yet to benefit from its utility. The low component cost, compact form factor with integrated 2D transducer array, and an intuitive C-scan image format displayed directly over the anatomy of interest are what distinguish this device from conventional ultrasound systems. A prototype of the Sonic Window was designed and constructed consisting of a fully sampled 60 × 60 transducer array, custom integrated circuits (ICs) containing 3,600 front-end receive channels, a high-voltage transmit circuit, an off-the-shelf digital signal processor (DSP), and a liquid crystal display. The 2D transducer array was fabricated on one side of a 2-layer PCB and the front-end ICs were attached onto the other side using a flip-chip process. The front-end ICs acquire the data for one C-scan slice per transmit event and send digital data to the DSP, which uses the Direct-Sampled In-phase Quadrature (DSIQ) method for efficient beamforming. The prototype fits in a 6-cm × 15-cm × 3.5-cm enclosure containing a 1500 mAh rechargeable battery and weighs 6 oz. (170 g). It can form real-time C-scan images at 43 FPS with a 2-hour scan time on battery power. Real-time volume capture enables B-mode image formats as well as 3D imaging. Lateral resolution is approximately 0.75 mm and cystic contrast is 43 dB. Phantom target and in vivo images are presented to demonstrate the imaging capability of the device.

Novel transmit protection scheme for ultrasound systems

The problem of protecting or isolating extremely sensitive receive circuitry from high-voltage transmit circuitry is commonly addressed through the use of diode bridges, transformers, or high-voltage switches, which prove to be prohibitively expensive, bulky, and power consuming for use in portable, low-cost, battery-powered systems. These approaches also compound the interconnect difficulties associated with two-dimensional (2-D) transducer arrays. In this paper we present a novel transmit protection scheme that allows compact MOSFET shunting devices to be brought on-chip within each receive channel implemented in a standard CMOS integrated circuit process. During transmit, the high voltage transmit pulse is driven onto the common connection of the transducer array and the on-chip MOSFET devices shunt the current to ground. During receive, these devices are turned off, the common connection of the transducer array is shunted to ground, and the received echo can be detected as usual. The transmit protection scheme was experimentally shown to shunt a 16 mA peak current resulting from the equivalent of a 100-V, 25-ns-risetime transmit pulse through a 4 pF transducer element. The scheme was also incorporated into a prototype 1024-channel, low-cost, ultrasound system successfully used to form pulse echo images.

Portable, low-cost medical ultrasound device prototype

The first generation prototype of a portable, low-cost medical ultrasound device is described along with experimental results. The prototype system consists of a fully sampled 2D transducer array, sixteen custom receive circuitry chips multiplexed into two bandpass filter channels, and an onboard programmable logic device. A PC fitted with a commercially available data acquisition card is used for data collection and analysis. Beamforming is performed using the direct-sampled in-phase/quadrature method. Pulse-echo images obtained with the prototype are presented and future work is discussed.

Application of X-Y separable 2-D array beamforming for increased frame rate and energy efficiency in handheld devices

Two-dimensional arrays present significant beamforming computational challenges because of their high channel count and data rate. These challenges are even more stringent when incorporating a 2-D transducer array into a battery-powered hand-held device, placing significant demands on power efficiency. Previous work in sonar and ultrasound indicates that 2-D array beamforming can be decomposed into two separable line-array beamforming operations. This has been used in conjunction with frequency-domain phase-based focusing to achieve fast volume imaging. In this paper, we analyze the imaging and computational performance of approximate near-field separable beamforming for high-quality delay-and-sum (DAS) beamforming and for a low-cost, phase-rotation-only beamforming method known as direct-sampled in-phase quadrature (DSIQ) beamforming. We show that when high-quality time-delay interpolation is used, separable DAS focusing introduces no noticeable imaging degradation under practical conditions. Similar results for DSIQ focusing are observed. In addition, a slight modification to the DSIQ focusing method greatly increases imaging contrast, making it comparable to that of DAS, despite having a wider main lobe and higher side lobes resulting from the limitations of phase-only time-delay interpolation. Compared with non-separable 2-D imaging, up to a 20-fold increase in frame rate is possible with the separable method. When implemented on a smart-phone-oriented processor to focus data from a 60 × 60 channel array using a 40 × 40 aperture, the frame rate per C-mode volume slice increases from 16 to 255 Hz for DAS, and from 11 to 193 Hz for DSIQ. Energy usage per frame is similarly reduced from 75 to 4.8 mJ/frame for DAS, and from 107 to 6.3 mJ/frame for DSIQ. We also show that the separable method outperforms 2-D FFT-based focusing by a factor of 1.64 at these data sizes. This data indicates that with the optimal design choices, separable 2-D beamforming can significantly improve frame rate and battery life for hand-held devices with 2-D arrays.

A prototype low-cost handheld ultrasound imaging system

We describe a very low cost handheld ultrasound system that we are currently developing for routine applications such as image guided needle insertion. We provide a system overview and focus discussion on our beamforming strategy, direct sampled I/Q (DSIQ) beamforming. DSIQ beamforming is a low cost approach that relies on phase rotation of in-phase/quadrature (I/Q) data to implement focusing. The I/Q data are generated by directly sampling the received radio frequency (RF) signal, rather than through conventional baseband demodulation. We describe our efficient hardware implementation of the beamformer, which results in significant reductions in beamformer size and cost. We also present the results of experiments and simulations that compare the DSIQ beamformer to more conventional approaches, namely time delay beamforming and traditional complex demodulated I/Q beamforming. Results that show the effect of an error in the direct sampling process, as well as dependence on signal bandwidth and system f number (f#) are presented. These results indicate that the image quality and robustness of the DSIQ beamformer are adequate for routine applications.

Separable 2D direct-sampled in-phase quadrature beamforming for greatly increased computational efficiency

Two dimensional arrays present a significant engineering challenge due to high channel count and data rates. In addition, for 2D arrays in handheld devices, power consumption is also very important in order to achieve multi-hour battery life. Previous work indicates that a 2D beamforming operation can be decomposed into two successive 1-D focusing operations for computationally efficient volume imaging. In this paper, we analyze the application of this method to a power-efficient 2D C-mode beamforming method using phase rotation focusing, known as direct-sampled in-phase quadrature (DSIQ). We show that the separable approach applied to DSIQ does not discernably degrade imaging performance in most practical conditions. Using this method increases C-mode frame rates in a handheld device 17-fold from 11.4 Hz to 193 Hz. The energy cost of beamforming a frame of data falls from 107.2 mJ to 6.2 mJ. When this power efficiency is allied with the low-duty cycle strategy used by the DSIQ front-end ASICs, a very power efficient system is achieved, capable of beamforming a frame of data from 3600 channels, including front-end power, using <; 10mJ of energy. This level of energy efficiency is essential to enabling multi-hour battery life in handheld ultrasound systems with 2D transducer arrays.

40 MHz 0.25 um CMOS embedded 1T bit-line decoupled DRAM FIFO for mixed-signal applications

An embedded 40 MHz FIFO buffer for use in mixed-signal information processing applications is presented. The buffer design uses a 1T DRAM topology for its unit memory cell component, a sense amplifier, and two circular shift registers for implementing refresh and read-write pointers. The sense amplifier uses bit-line decoupling to improve readout performance. Our particular application requires the storage of 800 samples of a received ultrasound signal that pass through 48 channels consisting of a preamplifier, a sample-and-hold, and an 8-bit ADC. Data is written into memory in parallel in a sequential, burst-mode fashion and read sequentially at leisure, with interspersed refresh of the memory cells. Layout and design issues concerning implementing memory in a standard 0.25 um process are discussed and simulation results are presented.