Life Science Technology
In traditional machine vision applications, a camera requires a frame grabber to interface with it controlling computer. When a camera doesn't have onboard processing power and results are needed from data in as fast a time as possible, frame grabbers give system designers the ability to process the data where customer wants. They give more control or the ability to view or analyze results in real time.
Frame grabbers evolved from being analog to PCMCIA Type I to PCMCIA Type II to card bus types that can go into desktop and laptop computers as the push towards tiny form factors has continued. Now, there are thousands of analog, HD, and HD-SDI frame grabbers, and each board is meant for a different use.
Instantaneous results are encouraged in life science applications like genome sequencing using X-Rays, where frame grabbers are the only solutions without interruptions, and scanning 32-64 images of an eye in an ophthalmology doctor's office.
The first factor that makes an imaging board suitable for an application in any industry is the type of output from the camera it will be used with. Camera outputs can be analog or one of two types of digital, and different types of boards are used with each type.
Analog output comes from cameras that have a single RCA or SVIDEO cable, using NTSC or PAL, and are still used in some ultrasound and inspection cameras. However, the resolution is limited with analog signals providing interlaced video, so most cameras are in the digital domain to serve higher resolution images and video.
The two types of digital cameras provide much higher density outputs, Camera Link and High Definition, Serial Digital Interface (HD-SDI), which is an interface with standards for pre-defined resolutions formed by the Society of Motion Picture and Television Engineers. HD-SDI cameras provide standards 720p or 1080p video. High definition cameras are more commonly used than analog cameras in high resolution life science applications such as on surgical cameras or inspections, whether with standard visible cameras, IR technology, or specialized hyper-spectra cameras.
How long the board will be available is critical to customers from life science applications who have to deal with lengthy FDA approval processes that can take years. Manufacturers that design boards to last at least a decade or longer chose components and software that will be around for a long time. Linux operating software is the customer's choice over Windows for machine vision in the life sciences.
In applications that produce extremely large volumes of data from large sensors or 12-20 megapixel cameras at 180fps, like in genome sequencing, features like NVIDIA's GPUDirect help process up to 2 GB or more per second. Beneficial with CoaXPress and some Camera Link hardware, this option processes data directly in a GPU, leaving the CPU running at a minimal load, increasing throughput and reducing latency.
The next consideration is whether the system will go into a laptop or desktop PC-based system or some other platform. One oncology application providing radiation for cancer treatment uses desktop pc-based systems with high resolution cameras and PCI frame grabbers, like the Matrox Meteor-II MC/4.
PCI Express (PCIe) boards are also used in many laptop or other space-constrained applications, such as in robotic surgery where more than one camera will be used on each robot and precise triggering and timing is needed to provide real-time video while the surgeon works. For customers who want small and portable systems, laptops are still the platform of choice, but that is already changing as cameras are going into many other small systems and embedded PCs that have a few PCIe slots available for boards. There are more USB 3.0 frame grabbers available now, but they are seldom used in the life sciences arena where precise triggering or real-time processing and analytics are needed at the same time.
Some applications require frame grabbers, and many don't. Mid- to higher-end applications typically use them, and it's mainly the resolution, frame rates, and high bit color information that determines whether they're needed or not. SEM applications do use frame grabbers. GigE cameras plugged into laptops are an example of a low end application for frame grabbers.
Applications that take static or still images with an area camera don't pose any unique requirements for imaging boards, looking at fluorescence or genetic mutations in slides, or automation in pharmacy applications checking tablet sizes. If PLCs are used, it is usually a PCI-based imaging board in a desktop or embedded system. With GigE cameras used in package testing that have an Ethernet interface, the board also goes into the computer. And when smart cameras are used, very common in lab automation applications, the pc-base is already in the camera so no frame grabbers are needed.
It also depends if users require compression or not. Hardware compression is typically faster than software compression. Anyone that has to record over long periods of times, whether it is minutes or hours, depending on how much disk space is available, may need boards that perform compression tasks.
The imaging boards used in life science applications are similar to those used in other industries. Frame grabbers are used when there are large amounts of data because they can get high resolution and fast transfer rates. Every life science application offers its own challenges, and the boards need to be a reliable part of the stringent quality chain of components that go into devices that will have quality and other standards in order to obtain certification.
Legacy products tend to stick with PC platforms. New products are moving towards using SDI boards but may not need a frame grabber if it only needs to display an image rather than perform analytics, like a surgeon who watches a monitor not the patient.
It is more likely that life sciences applications need portability compared to other industries. Laptops help keep several operating rooms going at the same time in a hospital or desktop systems that are on carts. Users don't need an ultrasound machine in every room in every doctor's office and often they are wheeled around as they are needed to save time and money. Portable systems are easier to update and swap when better systems come along.
The number of and types of drivers available are an important consideration for users who wish to use frame grabbers in their systems. Software that is properly written can communicate with many different types of boards. Using common drivers instead of sdk's and as a communication layer between the frame grabber and computer like DirectShow capture support make it easy to reuse the technology on different types of systems, such as a laptop and desktop. What's most important for developers is if the board has many drivers and a very large up-to-date sdk library that can grow and be responsive to customers' current needs and future updates. Also, sdks should have functions that are easy to write, and a very developed sdk library can help that too.
As the industry sees smaller and faster cameras and as camera sensor resolution has advanced, a need for card with higher frame rates that can handle higher resolution has developed. The trend has been towards using CoaXPress frame grabbers, whose high speeds have allowed new applications to open that traditionally didn't have a machine vision solution before.
CoaXPress technology was introduced in 2009 and standards were introduced in 2011. CoaXPress frame grabbers can handle data that is too fast or too large for GigE or USB 3.0 and allow long cable lengths up to 60 meters away. Computers could be located outside a clean operating room or away from hazardous chemicals in biomedical applications making it safe for humans to control large numbers of cameras for 3D systems.
They operate like one-way spigots, offering video out plus bidirectional control without using a separate cable. It is similar to SDI or GigE in that it has power over the cable, and up/downlink and video out, and is used for high speed applications 100 frames per second and above, which is where medical devices are heading.
Most life science applications that use imaging boards don't require a tremendous amount of memory and most don't need extensive processing. They create measurements from images, of how much plaque is in an artery in the chest or measuring the size of a visual representation of a cancer cell.
In low quantities boards range at the low end from around $400 for frame grabbers that go into laptops with standard analog single inputs, to $2500 for the highest end boards that go into desktop systems with CoaXPress and multiple HD channels. Boards in this industry cost the same as for other industries, since there is really nothing unique about the boards serving life science applications. If the board can do the job, pricing is secondary.
For board manufacturers, the main challenge life science applications pose is the question of compatibility. Users have to know the camera's technology because the frame grabber's technology has to match. Which version of Camera Link and how many cables are coming out of the camera are important. When a customer that needs a 6GB 1080p camera and 3GB SDI approaches and asks if there is a frame grabber that can handle the data rate they need, the challenge is finding out if a board is available, and whether camera files or drivers need to be written.
There is a whole sub-industry of converters, but not everyone wants to add them to their system because they add cost and complexity. Usually converters update old technologies to new, but typically if a new system is being built today, it uses the newest technology.
Manufacturers don't want their boards to be the slowest item when compared to the computer and the camera. Their challenge across all industries is how to move the data fast enough. Typically boards have always been faster than cameras.
Emerging markets for imaging boards 20 years ago were bottle cap checking on assembly lines and license plate reading. Document processing and mail sorting is still large, but today, life sciences applications are the fastest growing and stronger markets. Within this market, anything related to research is taking off, as more sensors are coming out and users want faster and more detail and boards are able to meet those standards.
Robotic surgery and diagnostics is another strong area, with 3D biopsies and remote cameras that are located on one continent and controlled by doctors on another as a brand new market that is taking off. During kidney surgery, two cameras guide probes so the surgeon can see what's happening in 3D. The growth is being driven by the fact that there are not enough doctors to serve where they are needed, both in remote areas and foreign countries and in the United States where there are huge demands as baby boomers age.
The most important factor to consider when considering an imaging board for a life science application is whether it offers the capabilities that are needed. A common problem board manufacturers encounter is that the designers specifying the boards are not machine vision specialists and don't understand its limitations. Large life sciences OEMs may have a vision engineering type department, but at smaller companies, the design task is typically done by electrical engineers who may understand circuits but don't know this different field of engineering.
In order to reach the high speeds needed in many life science applications, a board's latency is another factor to consider. After an image is taken, if a single frame is taken and then processed into memory, it may take a few frames for the computer to render the image. At 30fps, the image could be about 40 frames behind. 4-5 frames very quickly can turn into 10-15 frames if anything is slow during buffering, which is half of a second. This type of latency could look very bad in applications that demand real-time feedback to make decisions.
In the medical space, boards with good latency are needed in applications like body scans, watching fluid flow, testing that chemo drugs are going to the right parts of the body or knowing where to make an incision during surgery as a camera is moved around. In biological applications, buffering and dropping frames may mean missing part of a reaction.
Some board manufacturers offer different strategies to avoid problems like these, such as double triggering modes that take two frames very close in time.
Another factor to consider is vendor support, because over time, the platforms will change. Computers will get smaller, be able to do more, and be faster, and cameras will be smaller, faster, and less expensive too. So will the boards. These days, average employees work 2.5 years at one job before moving to a new position. Most companies in life sciences are not able to hire the expertise they need, and they are in constant transition. They rely on vendors more and more for support, and may or may not get it. This happens across all industries, but it is a unique consideration for life science industry which is more time (and longevity) critical than other industries. It is important to know a vendor's history and whether they will be around tomorrow, and whether support will be available if something goes wrong during development or the lifetime of the product.