Using NI FlexRIO for Next-Generation Real-Time Ultrasound Array Imaging for Non-Destructive Testing

  Print

"NI FlexRIO with customisable external hardware modules, combined with simplified programming in LabVIEW, removed constraints such as high channel count, data rate, and cost."

- David Lines, Diagnostic Sonar Ltd.

The Challenge:
Creating a flexible and scalable architecture for acquiring ultrasonic array data and processing it in real time.

The Solution:
Using modular, reconfigurable field-programmable gate array (FPGA)-based NI FlexRIO hardware to acquire multiple data streams in parallel.

Author(s):
David Lines - Diagnostic Sonar Ltd.
James Wharrie - Diagnostic Sonar Ltd.

Background

Flexibility is just one of the many benefits of acquiring ultrasound data directly into a PC and then performing application-specific processing in software. However, the large number of channels in ultrasound array imaging systems introduces data throughput and frontend connectivity challenges. These challenges can be addressed by combining channels and reducing data rate through hardware-based processing local to the digitisers. Until now, this required custom circuitry, which greatly restricted the flexibility of localised processing and defeated one of the main goals.

Conventional Imaging

Steering and focusing is achieved by applying differential delays in the excitations to each ultrasonic sourcing element and in the received signals from each element. The delays compensate for the varying path lengths between the elements and the focal point to produce constructive interference at this point (Figure 1).

Volumetric scanning to produce rendered 3D images is done by mechanically scanning a standard array. Therefore, we have a growing interest in 2D arrays for electronic volumetric scanning, as well as improved 2D focusing. This significantly increases the number of channels even if sparse array techniques are used, but any new architecture should be able to handle this expansion.

Alternative Pulse-Echo “Holographic” Imaging

For many years, Diagnostic Sonar Ltd (DSL) has used a full raw data (FRD) technique for data acquisition and processing to investigate optimised focusing and imaging. Rather than exciting a large group of elements to produce a transmit beam with a well-defined focus, a much smaller group—typically just one element—is activated to produce a widely divergent sound field. Signals received from all elements are collected as usual, but rather than being delayed and combined to form the receive beam, they are stored for later processing.

By selecting the appropriate sample from each of the transmit-receive combinations for every pixel, every point is optimally focused when it is transmitted and when it is received, resulting in an improved image. These novel imaging approaches have remained in the laboratory due to the high cost of the customised hardware, as well as the long processing time—several seconds or more—which precludes their use for area scanning.

Data Rate Considerations

Each ultrasound channel needs to be sampled with a resolution of 12 bits or more at a rate of 50 MHz or higher, resulting in a 3.2 GB/s data rate for 32 channels. The physics of the ultrasound propagation in the target medium places restrictions on the maximum beam repetition rate, so local buffering reduces the overall data rate from this value. However, due to the scalability requirements, a number of these parallel 32-channel modules may be needed. Localised processing is even more critical to avoid bus bandwidth constraints.

The Solution

FPGAs, which are ICs containing a massive resource of low-level logic blocks that can each be configured by software, are linked up by programmable interconnects to achieve customised hardware functions. The speed and inherent parallelism of FPGAs means these devices are well-suited for multichannel ultrasound processing and are now widely used to provide the frontend processing in medical imagers. The ability to reprogram these devices under software control provides the desired flexibility, but the need to use a hardware description language (HDL) to code the FPGA has limited the use to these customised imagers.

The NI FlexRIO product family combines a PXI-based FPGA module with application-specific frontend adapters. The FPGA can be programmed using graphical programming with the NI LabVIEW FPGA Module to provide the combination of power and flexibility to address these issues.

The FRD acquisition system uses two PXI slots—one for the NI FlexRIO module and digitiser and the other for a high-voltage power supply and transmitter, receivers and multiplexers. Currently, the data is transferred to the host PC for image reconstruction, and even on a PXI system, the data is running at around 20 to 30 frames per second—a big improvement over the several seconds previous solutions took to reconstruct just a single image. A further 4X increase in speed is anticipated when this reconstruction code is moved into the FPGA, which greatly reduces the amount of data that needs to be transferred to the host. The development time, from defining the architecture to demonstrating real-time images, was just three months (Figures 2 and 3).

Conventional beamforming requires multiple pulsers. We implemented a compatible 32-channel tri-level pulser in a separate adapter module for use with an additional NI FlexRIO board. Therefore, this 3-slot PXI subsystem is a general purpose, 32-channel module that can be used in parallel with additional modules for ultimate performance or, alternatively, used with expandable multiplexer modules. Conventional receive beamforming also can be achieved within the FPGA, again minimising data transfer and the processing requirements on the host PC.

This modular approach helps us confidently assess the number of channels needed for a specific detection performance using FRD mode, even on something as small as a base system with multiplexers. The desired acquisition speed then determines the level of parallelism needed and informs us if we should use PXI or PXI Express.

Summary

PC-based acquisition hardware, integrated with custom software developed in LabVIEW, provides an ideal environment for developing new imaging systems. The same hardware may also be suitable for deployment in applications with relatively small production quantities. Various constraints such as high channel count, data rates, and cost, have prevented this approach until now. The recent introduction of PXI-based FPGA devices with customisable external hardware modules, combined with simplified programming in LabVIEW, removed these constraints. This resulted in a suite of ultrasonic hardware and software modules that can address a wide range of real-time imaging applications from conventional beamformers to FRD acquisition and display.

Author Information:
David Lines
Diagnostic Sonar Ltd.
Baird Road, Kirkton Campus
Livingston EH54 7BX
United Kingdom
dave.lines@diagnosticsonar.com

Bookmark and Share


Explore the NI Developer Community

Discover and collaborate on the latest example code and tutorials with a worldwide community of engineers and scientists.

‌Check‌ out‌ the‌ NI‌ Community


Who is National Instruments?

National Instruments provides a graphical system design platform for test, control, and embedded design applications that is transforming the way engineers and scientists design, prototype, and deploy systems.

‌Learn‌ more‌ about‌ NI