One area of medical research that has been gaining a lot of attention is shear-wave elastography imaging. This technique has enormous potential for medical diagnosis due to its higher contrast properties compared to traditional B-mode or harmonic imaging. In shear-wave elastography imaging, after generating a push pulse, a shear wave is created in the region of interest in the tissue. This imaging technique is done today using one-dimensional linear arrays that produce a less satisfactory two-dimensional slice. Capturing the shear-wave with a high-channel-count (HCC) ultrasound matrix array, and then generating and processing volumetric data, may lead not only to research breakthroughs in this field, but eventually clinical adoption of new diagnosis methods for many diseases, including breast cancer.
Shear-wave elastography, as well as cutting-edge ultrasound research in many fields (see On the frontlines), requires an HCC ultrasound system that can interface to large matrix arrays. In the past, researchers seeking such a high channel-count system mostly built the system on their own. However, academic institutes found such an endeavor to be very costly and time-consuming, often running way over budget and taking years longer than planned to build. After all, universities need to focus on conducting research using the system; they are not set up to build the system itself. Further, support and continuous maintenance of a one-of-a-kind in-house-built system is very costly, and can result in lengthy downtime during which no research can be conducted with the system.
An HCC system needs to be, in essence, both a flexible ultrasound generator and a high-bandwidth data acquisition system (DAS). Any HCC system is comprised of the five major electronics subsystems seen in Fig.1 .
Fig. 1: The high-level block diagram for a high-channel-count system shows its five major subsystems: (counterclockwise from upper left) the ultrasound generator, the ultrasound receiver, the data transport & interconnect subsytem, the host, and the power supply unit.
The HCC system's ultrasound generator (TX) includes all the transmit circuitry and digital beam formation required to generate ultrasound pulses to the transducer. The ultrasound receiver (RX) provides all analog conditioning circuitry, the ADCs, and (optional) digital receive beam formation required to sample RF data from each channel and, optionally, generate beam data from it. The data transport and interconnect subsystem fundamentally includes the multiple high-speed data links, their control, protocol, cabling and connectors required to transport the receive data from the RX subsystem to the host subsystem.
The host computer (Host) is the master controller for the system hardware, including TX and RX, memory, HDD, and so forth. The Host contains very high-end graphic processing units (GPUs) to provide researchers with fast memory and I/O data paths, as well as access to the high-performance computation needed to run complex research algorithms. Lastly, there is the Power Supply Unit (PSU), which supplies all power rails to the system, including the ultrasound-specific high-voltage rails required by the ultrasound generator.
A realization of such an HCC system will require a hierarchical architecture. A baseline system of some number of channels (typically a power or two such as 64, 128, or 256 channels) can be designed. This baseline system will include a number of transmit and receive channels along with all their control and timing circuitry, and ultrasound-specific high-voltage power supplies. In this baseline system, the TX subsystem consists of a number of transmitter modules that include a fine delay resolution and multiple-level pulsing circuitry for each channel allowing users to generate virtually any arbitrary transmit sequence at various voltage outputs. Since most ultrasound transducers can use the same elements both for transit and receive, the baseline RX subsystem will consist of the same number of receiver modules implementing the same number of channels. This symmetrical approach simplifies the wiring between the electronics and the transducer connector. Receive channels will be amplified with common time/gain compensation (TGC) to account for signal attenuation with depth in the tissue. Next, at the channel level, sampling of the analog signal is accomplished by a dedicated ADC. Sampling rates should range up to 65 Msamples/s to comfortably oversample signals in the ultrasound bandwidth.
HCC systems inherently imply transport of huge amounts of data. For example, a 4,096-channel system, using 12-bit ADCs sampling at 65 MHz will produce approximately 3.2 Tbits of raw RF channel data per second. Many research applications indeed require all this data, and thus the system needs to be designed for such a “worst-case” scenario. But, transporting and storing so much data further upstream can be expensive, and pointless when all that data is not needed to accurately conduct the research at hand. Thus, the RX module should also implement several signal-processing functions including demodulation, filtering, and decimation. These functions can be used to reduce data rates. Further, for quick imaging or for probe placement verification, a beam-formed data path can be provided as well.
A number of such baseline systems are then properly interconnected to create larger systems. For example, four 1,024-channel systems can be used to build a 4,096-channel system.
The data transport and interconnect subsystem is designated as its own subsystem due to its complexity and the very high throughput demands placed on it, as previously discussed. Although often overlooked, the design and implementation of such a subsystem is non-trivial. The performance, robustness, flexibility, and viability of the system depend on it. The subsystem needs to support continuous data transfer without much limitation. In some extreme situations where this is not feasible with raw data, a capture-and-store option should be supported. For example, in a 128-channel RX module, a 12-cm scan with 6,400 12-bit samples at a 40-MHz sampling rate will produce 9.83 Mb per firing (12 bits/samples x 6,400 samples x 128 channels). Using a fast 9-Gbit DRAM, 1,000 firings can be stored in the front-end. Finally, a scalable protocol for properly handling and tagging the various sources types of data (channel, beam, and so on) needs to be implemented as part of any such subsystem (Fig. 2 ).
Fig. 2; The front end of an actual HCC ultrasound system shows some of the unique electronics needed to implement such a system.
The Host complex in a baseline unit of a very-large-channel-count system is responsible for communicating user commands to the front-end controller and timing circuitry, and receiving the very-high-bandwidth data coming from the RX subsystem across the data transport and interconnect subsystem. Very high-end GPUs in the Host subsystem are responsible for some or all of the researcher's demanding computation needs. In larger systems using two, four, or more baseline systems hierarchically, another CPU/GPU server is needed to aggregate these baseline systems.
The PSU has to realize the ultrasound-specific high-voltage dc/dc converter modules. Ultrasound power supplies are nontrivial. Extreme care must be taken to control noise in the design. A noisy PSU can contaminate the acquired data in many ways and may even compromise the research being conducted.
Finally, the HCC system must provide researchers full control over all of the system's configurability in real time. This interface preferably should abstract the hardware and low-level register access and allow parameters to be defined and set in high-level structures such as scans, frames, beams, firings, patterns, frequencies, filter coefficients, and the like. The acquired ultrasound data is available to the user via a shared memory interface.
The 1,024- to 4,096-channel cQuest Griffin ultrasound research system recently introduced by Cephasonics employs much of the above-mentioned critical architectural aspects. This HCC system also combines a MATLAB interface to promote a fast learning curve for system operation, and with proven, reliable technology to offer a much-needed 'buy' alternative to in-house implementations. Systems such as this will enable worldwide ultrasound researchers to focus entirely on conducting cutting-edge research that will hopefully lead to revolutionary new diagnostic clinical procedures.
Advertisement
Learn more about Cephasonics
