是一篇由NI公司市场开拓经理John Hottenroth撰写到署名**，介绍了利用FPGA快速构建高性能医疗成像系统到方法和技巧，对医疗电子开发者来说有实用价值。
Time to market is of critical importance for medical instruments. A difference of even a few months in the release of a product can significantly impact the ROI of the project, both from missed revenue and because of market dynamics. Yet medical imaging instrument developers are required to integrate the latest technology to build systems with excellent analog performance, complex processing and visualization, and high data throughput resulting from higher speed ADCs and increasing channel counts.NI FlexRIO is an example of a product that combines a user-programmable FPGA with customizable I/O.While time-to-market pressures combined with integration of new technologies can create design challenges, reconfigurable field-programmable gate array (FPGA) technologies coupled with flexible integration platforms are helping to develop prototype imaging systems more quickly, while still delivering new innovations to the market. Specifically, developers are combining modular FPGA hardware, higher-level design tools, and industry standard platforms to create highly flexible, scalable, and customizable imaging systems.
Off-the-Shelf FPGA Hardware Provides Flexibility to Try New Ideas
FPGAs can enable a lot of design flexibility to try out new ideas and reduce risk earlier in the development of a system. Traditionally, demonstrating hardware-based processing required a custom application-specific integrated circuit (ASIC), but ASIC development is expensive and functionality is fixed. Since FPGAs are reconfigurable through software, a designer can save a lot of development time, demonstrating hardware-based processing while being able to reprogram the FPGA to accommodate modifications that are unknown during initial specification. While FPGA board designs can be complex, modular off-the-shelf FPGA boards can help by providing hardware to build around, with infrastructure components for I/O connectivity, bus interfacing, and DRAM communication. Developing these components in-house can consume a lot of time in a project, and can distract developers from new innovations where they add the most value.
Santec, a Japanese company, demonstrated these benefits on a swept source optical coherence tomography (SS-OCT) imaging system. Using National Instruments’ (NI) FlexRIO, which combines customizable I/O with a user-programmable FPGA module, they were able to quickly evaluate a new hardware architecture that gave them significantly faster imaging rates over their existing system. Santec leveraged validated, off-the-shelf FPGA hardware, along with firmware for the PCI Express bus interface, which allowed them to concentrate on the algorithms and other parts of the system where they could add the most value. For I/O, Santec used a custom adapter module design that combined a high-speed ADC for image acquisition with the DAC circuitry for the laser scanner control. Prototyping using modular FPGA hardware allowed them to quickly get a working system and determine required hardware changes, since the I/O was decoupled from the FPGA back end.
Santec was able to acquire real data from their prototype system and compare with images from their existing system to prove the algorithms before moving code to the FPGA. Once the hardware and firmware were proven, they were able to quickly move to a more deployable PCI Express board, reusing many parts of the design and further reducing project risk. Overall, Santec achieved an image display rate of 40 frames per second with the new FPGAbased system, a 4X performance improvement over their previous system. Leveraging FPGA-based processing, they were also able to reduce the cost and size of their computer, enabling them to address new markets that need a smaller, lower-cost system for portable applications.
Higher Level FPGA Design Tools Enable Faster Development
Another challenge of using FPGAs for prototyping is that programming a system using a traditional hardware description language like VHDL can be very time-consuming, lengthening project timelines; however, recent advancements in development tools have made FPGA programming more efficient by allowing higher level graphical tools to be used for overall system design, which can also leverage existing VHDL IP (Xilinx CORE Generator™, in-house developed, 3rd party, etc) where appropriate. When used properly, these tools can enable very fast development of a prototype system so that algorithms and hardware performance can be evaluated and refined.
A UK-based company, Diagnostic Sonar, was able to demonstrate this concept on a novel phased array ultrasound imaging system. Using NI FlexRIO FPGA hardware with NI LabVIEW FPGA, a graphical design language for FPGA programming, they were able to go from specification of the architecture to a working prototype system showing real-time imaging in less than three months. To save time, they focused their hardware design efforts on a 32-channel, high-voltage pulser module for FlexRIO that pairs with an off-the-shelf receiver module. While their prototype system had 32 channels, their architecture can easily scale to 64, 128, or other multiples of 32 channels for both transmit and receive, as well as accommodate a variety of ultrasonic arrays.
Algorithms were initially developed using LabVIEW on the host — including beamforming, filtering, and rectification — along with a graphical user interface (GUI) to visualize data. After demonstrating the prototype system, they were able to use LabVIEW FPGA to move algorithms to the FPGA to accelerate the processing performance. In the end, Diagnostic Sonar was able to leverage modular FPGA hardware and graphical software to create a high-performance, multi-channel ultrasound acquisition and processing system that was scalable and customizable for a range of applications.
Industry-Standard Platforms Allow Efficient System Integration
Santec used NI FlexRIO to prototype a new architecture and FPGA-based image processing for their low-cost, portable OCT imaging system.A system that is flexible and scalable to meet future requirements can be highly valuable for a development team, but these features are often some of the first to go when a development schedule slips. One way to tackle this problem is to build modular systems using an industry-standard platform with a wide variety of I/O. This way, a developer can leverage a combination of custom and off-the-shelf hardware early in the design process to quickly show a working prototype. Ideally, this platform should be flexible enough to combine off-the-shelf CPU and FPGA-based processing, scale to add higher channel counts or even other imaging modalities, and include a variety of I/O for high performance and lower speed signal acquisition, generation, and control.
Researchers at Kitasato University in Japan recently showed the power of using a flexible and scalable platform when they demonstrated the world’s first real-time 3D Optical Coherence Tomography (OCT) imaging system. While their specific OCT research area may not apply to other imaging system designers, the more important takeaway is that they chose a platform that could integrate the multiple technologies required for their high-performance design. Specifically, they chose the PXI platform, which provided high throughput data transfers over PCI Express, accurate timing and synchronization of multiple modules, a wide variety of I/O, and the ability to create “peer-to-peer data streams” that connect multiple FPGA modules over direct memory access (DMA) without ever needing to involve the host.
Since the research team’s goal is to move toward real-time optical biopsy, their target was real-time display of 3D OCT images at a rate of 12 volumes per second. Their system contained a total of 22 FPGA modules, which combined data from 320 channels (each acquiring data at 10 MS/s), as well as performing noise subtraction, windowing and FFT processing. To achieve their 3D imaging capabilities, the two highest performance processing FPGAs in the system computed over 700,000 512-point FFTs every second.
Using LabVIEW to integrate and control the different parts of the system, they transferred data from the FPGA subsystem to a quad-core PC with an NVIDIA FX3800 Graphics Processing Unit (GPU) to perform real-time 3D rendering and display. They also needed to log data for extended time periods since they wanted to be able to conduct group screening tests for cancer. While their architecture doesn’t limit the image acquisition time, the research team enabled logging of up to 100 minutes on their prototype system, which required a little more than 3 TB of hard drive space.
Bringing Everything Together
Researchers at Kitasato University in Japan combined FPGA and GPU processing to create the world’s first system capable of real-time, 3D OCT imaging. Snapshots of finger skin (top and side view) from the system are shown.The need for imaging system designers to deliver innovative new products under tight timelines is unlikely to change, but FPGA technology combined with the right integration platform can help make this process more efficient. Diagnostic imaging modalities continue to advance with new algorithms, higher performance processing, and better hardware. By combining modular, off-the-shelf FPGA hardware with high-level design tools, system developers can create flexible and scalable systems to meet the needs of these next-generation imaging systems.
 |
楼主
标题置顶
标题高亮
