Cameras in Life Science Applications
Cameras provide vision to robotic and or motion control automation to handle DNA or lab samples for analysis, storage, packaging, or preparation. Bar code reading cameras in amino acid machines perform array analysis and testing. Clinical lab automation cameras help manage medications, biological specimens and handle samples and glass medical vials that have colored stoppers. Cameras are also used in medical devices, ophthalmology, biomedical research, and pathology.
Cameras check if biological samples are present, if needles contain liquid to make sure they are not broken. In these applications, cameras need to be small, compact, self-contained, and integrated with its own lighting and communication interfaces to save data to a database or send pass/fail acceptance notes to a PLC.
The rapid emergence of new sensor technology has a profound effect on product development. Choosing the correct sensor for life sciences imaging platforms can be challenging. The number of pixels, the true sensitivity of the sensor, the tradeoffs between competing technologies, the length of product life cycles, and long term costs are just a few of the areas that need to be considered.
When selecting a camera, users should consider at a high level the business challenge that the machine vision system will help them meet. Whether adding a machine vision camera to increase throughput or yield on a production line or to meet a federal regulation will help determine a specific camera.
The physical size and form of the camera for the specific use is the next most important consideration. Users should know if it needs to be compact, small, or integrated into an existing machine without modification.
The environment and scenarios where the camera will be used and whether the system relies on ambient light is the next important consideration. Operating environment is often overlooked but affects robustness and stability across a variety of conditions or locations, which can be a critical requirement of the camera. Camera interface, bandwidth requirement and the proximity of the processor to the camera location are also important factors.
Longevity of the platform is another critical consideration and is considered the most important factor for users in the life sciences by far. Customers who use devices in surgery, ophthalmology, and medical diagnostic sold into labs and hospitals around the world will trade off having the latest version of a technology for something that will be around long term. Guaranteeing stability of the product's platform throughout upgrades and the lifetime of the system is only one aspect of stability to consider.
For customers whose products require FDA approval, which can take years, long term stability from component suppliers is critical. Finding suppliers who think in terms of very long product lifecycles is a key consideration. Look for those that have financial strength and a track record of stable production, and who will freeze versions of their cameras for customers to ensure products don't become obsolete.
Finding suppliers who offer flexibility to adapt the camera to the customer's needs is also important. Often, cameras must fit into standardized medical housing formats with specific colors, materials, and labeling required by hospitals or other medical offices.
Next, each application has its own key considerations, as cameras can range from entry level for educational purposes to higher end cameras that count electrons in extremely low light. Resolution, pixel size and pitch may come into play in markets where cameras are used on microscopes, such as in research, as opposed to cameras used in industrial machine vision applications. Bright field, dark field, phase contrast, and fluorescence microscopy all have different needs, and the sensitivity depends on the objectives that are used and the application.
In color applications, it is important that cameras accurately reproduce the colors, have high dynamic range, and high quantum efficiency (QE). They also need sensors with low noise to have good sensitivity under low light. During longer exposures, sensors should have low dark current noise, the thermally related noise that accumulates on the sensor and cameras, and cameras should have good heat management to keep the device cool.
Depending on the type of defect the system will detect, a camera has to meet minimum resolution requirements to determine if the item belongs in the image, making a camera's resolution very important.
A camera's aperture and frames per second are also important, but both are highly application dependent. In the life sciences, if samples are fixed, such as stained slides, frame rate is not important (and instead, color reproduction becomes a critical element). Tracking moving samples, such as fish or blood flow, fast frame rates will be critical in a camera. But, the processor affects camera speed more than the sensor. There are techniques for achieving higher frame rates, such as running dual lines or using higher processing capability with a vision processor.
Field of view also makes a difference depending on the type of images that are taken. When taking tilted images of large samples, having a larger FOV allows users to take fewer pictures, completing the process faster than using a smaller FOV.
From a customer's perspective, the type of sensor is least important since they're interested in the overall solution, not usually the technology used to get there, but this varies depending on how camera savvy they are. In applications where low noise is critical, such as fluorescence, sensor type becomes important, but it is not in other applications, such as bright field imaging.
In scientific grade cameras where it is critical to have no distortions or aberrations in the field of view, using high grade glass can become important. Industrial grade cameras use lower grade glasses and many labs using them on microscopes will run into quality issues.
Sensitivity in the infrared is also key, as more engineering audiences are interested in robust IR solutions regardless of the ambient conditions.
Gigabit Ethernet (GigE) is by far the most common and preferred interface for machine vision cameras because there is a networking infrastructure already in place and many people are familiar with IP addressing and registering devices on a network. In applications that have more than a single camera or large distances between instruments, like in lab automation, GigE is the most in demand for new design wins. It offers advantages with long cable distances and the ability to handle multiple distributed cameras.
Some higher end camera applications may offer dedicated interfaces, such as high frame rate cooled cameras for microscopy that require frame grabber cards in the host computer to capture images. But since 2014, USB 3.0 is emerging as the technology to use for higher frame rate cameras and applications that need fast image transfer rates.
With its shorter cable length restrictions there is a trade-off requiring the camera to be closer if speed is desired. Typically if an application uses a single camera that needs to connect to a host PC and is a short distance away, USB 3.0 can drive faster data rates, which could be higher resolutions or higher frame rates. But when more than one camera is needed or devices are spread around a larger area, USB 3.0 isn't practical and GigE offers an advantage.
There is still a strong presence in Firewire as a digital interface, which is an older technology but still important mainly in medical and life science markets because of longevity. In terms of long term supply, many customers are not willing to change.
As a technology, Firewire cameras dropped off after USB 2.0 was introduced. Now, USB interfaces offer plug and play with almost every computer. Many cameras in the life science market use the USB 2.0 interface when it is appropriate for the amount of data being passed. Users needing an interface that delivers higher bandwidth while still offering plug and play and being readily available will use USB 3.0 with its higher throughput.
One decision to be made with camera interfaces is whether to use proprietary interface or a standardized interface. If one uses a proprietary interface, you usually are locked in to the camera and software provided by the camera manufacturer. In contrast, if you use a standardized interface, you have the flexibility to use any camera or software that meets the standard. This flexibility gives users a wider range of solutions and even the ability to change all or parts of the system in the future with very little investment. Additionally, standardized equipment is often less costly as they are sold to larger markets.
AIA and other leading global vision associations have put together an excellent reference to these standards called “Global Machine Vision Interface Standards – Understanding today’s digital camera interface options”. This can be downloaded by clicking here.
The following are the leading, modern digital interface standards. Clicking on each will take you to more information on the standard.
Camera Link, Camera Link HS and CoaXpress require a frame grabber while GigE Vision and USB3 Vision operate through a PC’s native bus adapter.
Using color or monochrome cameras depends on a user's specific application. In microscopy, 80% of the time users look at color samples with bright field imaging and don't worry about sensitivity. In these cases they can use color cameras depending on budget constraints.
In segments like lab automation, bar code reading, and flow control processes, color is not as important as the ability to provide the precise positioning and handling or contrast information that monochrome cameras are capable of. As is the norm in other industries, 95% of the cameras used in life sciences are grey scale and only about 5% are color.
All camera designs start with monochrome sensors. Color cameras use Bayer or other filters over the sensors that produce red, green, and blue pixels and only capture 1/3 of the light hitting that pixel, reducing the amount of overall light reaching the sensor. If the same camera is available in both mono and color versions, generally the unfiltered mono has a higher quantum efficiency (QE) with a higher ability to convert photons to electrons.
Most fluorescence applications are low light and need more sensitive monochromatic cameras. Monochromatic cameras are also favored for fluorescence applications because they have more accurate resolution. They provide sharp edges, contours, and details compared to color cameras which estimate and interpolate neighboring pixels to produce a final color pixel. The processing causes a lower resolution than what is seen in the sensor of a monochromatic camera.
Color cameras are important in medical diagnostic or bright fluorescence applications and critical when inspecting vial caps. Ophthalmology inspections examine the topology of the eye with color cameras. But, many applications that use color cameras actually need monochrome because they provide more sensitivity.
Digital Cameras Prevalent but Analog Still Important
Most cameras in the life sciences are now digital since they provide the opportunity to produce higher resolution images at higher frame rates and use digital CMOS sensors.
Low power analog cameras are still important in specialized life science markets that have size restrictions and don't have stringent resolution requirements. One such example is cameras used for endoscopy – very small cameras or sensors embedded at the head of an endoscope with a tight bundle of optics in a lipstick-sized tube that will go into a human body.
CCD sensors have been the most popular option until recently, but the industry is transitioning as CMOS sensors are starting to become more attractive. Before, CMOS sensors were not often used in life science markets because they were seen as less expensive and lower quality. They went into entry-level and budget conscious applications such as educational cameras. CCDs were reserved for higher end imaging in diagnostic or research applications performing critical image analyses where precise high quality images, true color reproduction, and lower noise was more important.
But, the CMOS sensor industry has changed over the last few years, and now offers better quality sensors in terms of lower noise, better sensitivity for long exposures in low light, and better color fidelity. The industry is rapidly switching, as camera manufacturers produce more CMOS cameras that provide extremely good images for high end imaging applications.
Some camera manufacturers focus on low end markets and use lower cost CMOS sensors and others primarily use CCD sensors. The decision of which to choose is application specific. CCD sensors still offer lower dark currents than CMOS making CCDs the prevalent option in some applications, such as wide field fluorescence microscopy.
There are low end, medium range, and high end cameras that span all price points used in life sciences. The application defines the type of camera used, and the quality of the build and components create the price differences. For example, higher grades of glass are used in cameras where better image quality is desired. Manufacturers find that life sciences and medical market users are willing to pay more for cameras that have to meet more stringent requirements. Users in life sciences are less sensitive to price because they are more sensitive to longevity, image quality, and adaptability of the product.
Traditional machine vision cameras that are used in high-volume life science markets can cost less than $500 when customers and applications can trade-off image quality for lower prices. Typically cameras in the industrial domain have large pixels and capture quickly. VGA cameras that provide basic information to ensure quality throughout to an automated fill process, checking if vials are properly sealed or filled, can sell for few hundred dollars.
The lowest end respectable cameras for life science educational purposes are around $500. Most cameras are in the $3,000-$10,000 range, and specialty cameras that serve niche applications, such as large resolution full-frame cameras, cost more. Electron multiplier CCD cameras with deep cooling to meet lower noise and higher QE requirements can cost more than $20,000.
Guaranteeing that a camera product will be available for the long term is an important challenge, and may be the industry's most difficult. To ensure longevity, camera manufacturers select components that go into camera hardware carefully, and work with vendors to ensure components will be sold for that long. Larger manufacturers with strong financial power can influence sensor suppliers to secure the products for the long term. During development, when customers use a specific sensor that has reached its end of life, integrators work with them to buy enough inventory to carry their product forward ten years or more.
Manufacturers also optimize camera designs for long term supply using more stable and long term type technologies to ensure they don't become dependent on a new or emerging technology that may not survive for the long run. They build and test their own software to ensure a product won't need updates after it has been validated by a customer. If a product does need to change, manufacturers carefully control product releases and provide early information to customers and consult with them in order to have plenty of time to react to their needs.
Ingress protection or IP rating is another challenge for camera manufacturers for cameras that need to be enclosed or will be used in pressure washed or chemical detergent environments. Cameras exist that meet IP67 standards, the more common standard required by users in the life sciences. But cameras required to meet the more stringent IP69k standard, most commonly requested by life science users working with biological samples who perform wash downs with chemical detergents between running samples or periodic maintenance on a room or system to ensure it is hygienic, are harder to find and pose a challenge for manufacturers. Manufacturers end up having to encase the camera in an enclosure since most cameras themselves are not built to meet the IP69k requirements.
No "Killer App" for Machine Vision Cameras in Life Sciences
Many segments within the life sciences markets tend to be cyclical, consumer demand driven, and related to economic cycles. However, the medical market is more continuous, and has shown steady demand with typically long-term customers, making it an attractive market. Growing segments within this area are medical diagnostics, ophthalmology, surgical equipment, dental scanners, microscopy, pathology, lab automation and a few other smaller research oriented areas like DNA and gene sequencing. But they appear to all be growing equally, with not one application standing out as a clear leader. Other areas showing good growth are educational markets that use inexpensive cameras and open micromanager software suitable for low budgets.
Cameras are built differently for different applications. But, across the board, If there was one crucial factor for machine vision camera users to consider above all else, consistency and reliability emerge above others. Users don't want cameras whose results change over time and prefer reproducible, consistent performance, and behavior.
Sensor selection, camera metrics and the system as a whole, including stable lighting, white balancing the camera and calibrating the monitor to make sure the colors are not off – can dramatically impact whether results are reproducible and reliable without any post processing. Machine vision cameras in life sciences capture and produce an accurate, reliable reproduction of what is seen, as opposed to consumer cameras that rely on post-processing.
Bar-code reading cameras provide images which are data about how and why something failed. In these cases, data management is the top factor, and more important to life science industries than the image itself. When users want to perform an inspection and read a bar code during the same operation, the ability to use the camera for both purposes becomes a crucial factor. Rather than installing two cameras, both bar code reading algorithms and machine vision algorithms should be present on a single camera to save time and provide traceability during the inspection.