- » View All
Medical Imaging Pushes Cameras, Processing to the Limits
by Winn Hardin - AIA Posted 05/05/2014Across the board — from individual imaging components including cameras, lights, and processing units to image processing software and network requirements — medical and scientific machine vision applications are pushing the limits when it comes to performance, accuracy, and security.
We’re living longer, and as a result, we’re consuming more medical services and products. And where there is increasing demand in a resource-constrained world, there is opportunity for automation and machine vision.
“The biomedical and scientific imaging industries are a hot area of interest right now for machine vision companies,” says Alex Shikany, AIA Director of Market Analysis. “Vision-related market opportunities in this space exist primarily with drug discovery, robotic surgeries, inspection systems, and detection technologies.”
In fact, AIA has been examining the life-sciences market in more detail as interest has risen and will be publishing a market opportunity study on this area in the second quarter of this year.
Artemis Vision (Denver, Colorado) is a machine vision integrator with extensive experience in medical markets, including manufacturing and packaging, as well as R&D and point-of-care systems. The two application areas are very different, but both are pushing the boundaries of machine vision system design, explains Tom Brennan, President of Artemis Vision.
On the manufacturing and packaging side, government regulations demand that companies take extreme measures to verify and document production. For example, Artemis has clients in both prescription drug fulfillment and distribution, where every order is different, and pharmaceutical manufacturing, where each line produces one medical product or pharmaceutical.
“Both applications have very strict change control,” Brennan says. “Any change to the system must have detailed documentation and be signed off by the IT department and plant manager, as well as quality control. Every change is hand-validated on thousands of units on a single machine before a solution or change can be rolled out to all production equipment. Every system has the strictest security and login requirements, so no matter what change is made, the system knows who did it and logs every keystroke. All inspection photos are saved for 90 days or more, and in many cases, there is an offline machine vision system developed by in-house engineers that verifies the work done by a machine vision integrator. The general belief is that if two systems developed independently reach the same result, then chances are the result is correct.”
To help companies manage all this data, Artemis Vision recently released Vision Wrangler, which takes data form all the inspection equipment and places it on a secure web server that can be accessed from any location. Brennan expects demand for Vision Wrangler will continue to increase as pharmaceutical and medical-device manufacturers prepare for the effects of the recently enacted U.S. Federal Drug Administration’s (FDA) Drug Supply Chain Security Act (DSCSA).
Managing Sensors, Bandwidth for Medical Applications
While Artemis generally uses smart camera solutions for medical and pharmaceutical production and packaging applications because they are compact, self-contained solutions that are easier to secure, point-of-care and similar medical microscopy systems are testing the limits of components suppliers and system designers.
Blood- and tissue-screening microscopy systems used in clinical labs and now operating rooms across the country often use some form of polarization subtraction imaging to separate cancerous cells from healthy tissue, for example, or make other chemical measurements in fluid (in vitro) or in living samples (in vivo). Polarization subtraction signals are very faint, requiring sensors and cameras with high sensitivity. High frame rates allow technicians to see dynamic events, which provide much more data than “snapshots,” while larger sensors can view larger tissue samples or multiple wells in a test tray. All of these requirements boost the data coming out of the sensor and put pressure on the processing elements.
XIMEA GmbH (Munster, Germany), a manufacturer of scientific-grade cameras, is preparing to release two new scientific lines of cameras: the xiB, CMOS cameras with 12 MP and 20 MP resolutions, and the xiD line of scientific CCD cameras with resolutions up to 12 MP and USB3 interface. Both camera lines use special architectures to improve camera speed and transfer rates while maintaining high sensitivities. The xiB is one of the first machine vision cameras to use a PCIe x4 Gen2 external interface with a total bandwidth of 20 Gb/s.
“Instead of CoaXpress or Camera Link, which both require a frame grabber, we put the frame grabber in an FPGA on the camera to convert data into PCI Express,” says Max Larin, Founder and CEO of XIMEA. “At the other end is a PCIe adapter card for as low as $30 that connects the PC to the camera, covering distances of up to 300 meters when fiber optic cable is utilized. The PCI Express board can cost 30 times less than a standard Camera Link or CoaXpress board to deliver more bandwidth over longer distances. The high bandwidth is important to our medical customers, and the long reach is important to our security and defense customers.”
The xiD line combines scientific-grade CCD cameras with a USB3 interface along with four data taps, allowing users to achieve 2.8 Mpix at 60 fps. “Faster cameras mean shorter exposures, so signal-to-noise ratios at the camera are critical,” Larin notes. “Going fast without having a sensor with the highest sensitivity and dynamic range doesn’t do you any good. You have to have both, and now you can.”
Handling High Data
Frame grabber and camera manufacturer Alacron (Nashua, New Hampshire) also is to paying attention to PCIe extenders, among other consumer-based high-bandwidth protocols that are less expensive and easier to wire than proprietary camera link and other industrial vision protocols. Just as important as getting the data to the processing unit is how to handle the data when it arrives.
For that reason, Alacron CEO Dr. Joe Sgro is excited about AMD’s new lines of APUs that combine a CPU and graphics processing unit (GPU) on a single die. “Since the majority of our customers are in the medical, military, and manufacturing fields, the adoption of higher-resolution cameras means you need a 10-times speed increase over what the standard PC or vision processors can do today,” Sgro says. “GPUs are well-suited to the task, delivering hundreds of gigaflops per second compared to about 30 gigaflops for a CPU alone.”
“Application-specific DSPs require a company to use a specific hardware-software tool set,” adds Alacron’s head of engineering, Paul Stanton. “Plus, the I/O tends to be limited and not as scalable as using a GPU. While you still need to have some sophisticated graphics programming knowledge to make the most of a GPU, things are getting better as Matlab and Open CL add GPU functionality. We’re still in the infancy of making graphics SIMD processors available to machine vision customers. But we’re definitely heading that way.”
Optics and lighting suppliers also are responding to medical applications. SCHOTT North America Inc. (San Jose, California) has seen a big increase recently in demand for LED fiber-delivered lights for medical microscopy applications. “Halogen light sources with fiber light guides have been the traditional choice for microscopy,” says Jason Baechler, Lighting and Imaging North America Business Manager. “By going to LEDs, users can select their wavelength for fluorescence and reduce the amount of light and heat coming from wavelengths they don’t want without the drawbacks of a laser illuminator. On the medical packaging side, we’re seeing high demand for IR LED lights because IR penetrates the thin plastic of blister packs and medical packaging better than visible light.”
SCHOTT also has developed hybrid macro lenses for medical and scientific manufacturing applications that reduce distortion across larger areas for the 1-in. sensors while maintaining a price point that falls between low-cost CCTV lenses and more-expensive telecentric lenses.
Across the board, machine vision suppliers and system designers are responding to demands from the medical and scientific communities for higher-performance tools at lower price points. Combined with higher-throughput automation, these developments will help the developed world to continue to improve the quality of life for an aging population without breaking the bank.
There are currently no comments for this article.
Leave a Comment:
All fields are required, but only your name and comment will be visible (email addresses are kept confidential). Comments are moderated and will not appear immediately. Please no link dropping, no keywords or domains as names; do not spam, and please do not advertise.