Speed, accuracy and cost are what the doctor ordered.

A worker checks the surgical instruments that her colleagues have restored to like-new condition for hospitals. Vision technology plays a major role in the company’s business, both in creating a model of an ideal device and in checking restored instruments against it. Photo: Alliance Medical Corp.

As the medical industry struggles to contain soaring costs, it is actively seeking more and better products for less money. For this reason, hospitals are turning to suppliers like Alliance Medical Corp. (Phoenix) for innovative ways of deploying manufacturing technologies to do just that. This company resharpens and restores used surgical instruments to their original condition, and then sterilizes and packages them so hospitals can use them again, instead of throwing them away.

The inspection processes at Alliance, and other members of the Association of Medical Device Reprocessors (AMDR, Washington, D.C.), are crucial to proving to the hospitals and the Food and Drug Administration (FDA) that the instruments are as good as new. Vision technology plays a central role in establishing that proof. Engineers often use vision technology to measure and create a

computer-aided design (CAD) model of a predicate instrument, which is a new, unused specimen of the particular tool being restored. Once the factory resharpens or remanufactures a used instrument, inspectors measure and compare it to the model of the predicate instrument.

Alliance and its fellow members of the AMDR are representative of suppliers in the medical industry that are exploiting the power of modern vision technology. "The speed of video processing has increased commensurate with the speed of processors in computers," says Mark Glowacky, president, RAM Optical Instrumentation (Tempe, AZ). The processing speed gives vision technology the computing power necessary to collect and manipulate the large amount of data, not only in creating and using CAD models but also in handling the vast amount of measurements required for mass-produced devices.

Better processors also allow suppliers of vision technology to pack their products with much more sophisticated software than they could in the past. For example, "edge detection algorithms have gotten smarter," says Glowacky. "Using prior knowledge, these tools know what they are looking at ahead of time, automatically converting data points into features."

Typically, setup is a matter of choosing the type of feature from a drop-down menu and clicking on an image of the part to mark the feature. Consider a diameter. "The operator simply places the ‘circle' or ‘arc' tool over the feature, tells it whether it is scanning from a light area to a dark area, and clicks the mouse to enter the data points," says Simon Cosham, product manager, Starrett Precision Optical Ltd. (Skipton, England). "With one click of the mouse, the operator can enter as many as 300 data points."

The Galileo line of video measurement systems ranges from manual models to automatic machines using edge-finding software to measure distances between features. Photo: Starrett Precision Optical Ltd.

Better accuracy

Measurement accuracy has increased too. Advancements in zoom optics and high-resolution cameras during the past decade also have made vision technology well suited for the tiny devices and features that many manufacturers in the medical industry make. Encoders on movable stages, that actually record the measurements, also have become more accurate during the same period. "A tenth of a micron is almost common in encoding devices; whereas 10 years ago, their accuracy was only 1 micron," says Glowacky.

These advancements in processing power and measurement accuracy have allowed simple vision systems to evolve into optical measurement systems. In the past, the cameras on such devices simply magnified parts so the inspector could see small features. By adding translation stages with encoders and edge detection, simple magnification devices are transformed to accurate vision measurement systems.

"Because a high percentage of medical devices are quite small and require high magnification, this is where video comes into its own," says Cosham. "You are able to view the whole component at low magnification and then zoom to high magnification to see the area you want to measure. And systems typically use high-resolution color cameras, so we can now get an image close to what you would see with any microscope."

Probing patterns

Despite the emergence of high-end video measurement systems as a type of low-end vision system, conventional machine vision continues to find application in inspecting medical devices. Also benefiting from advances in technology, it not only measures parts but also shines in tasks that require feature recognition. It can check complex patterns on stents and other devices. Consequently, these vision systems can check both the reinforced weaving on a stent and the dimensions at the joint where the stent attaches to the tooling.

Faster processing speed and better software were responsible for the success that Invotec Engineering Inc. (Miamisburg, OH) had with a vision system that it designed and built for one of its medical customers. The advanced technology was needed to inspect 2-inch diameter tubes made of several pieces of rubber-like material that are woven together. The trick was that each piece has a different pattern, making them nearly impossible to inspect manually.

"If you placed them together in the wrong order, like one on top of the other, instead of the other way around, it would not work correctly," explains John Hanna, president of Invotec, a manufacturer of custom assembly and test equipment. "When you looked at it with your eye, it was extremely difficult to tell how they were woven together. You'd almost have to pull them apart."

With appropriate lighting, the vision system was able to distinguish each pattern from the other and determine whether they were assembled correctly. Just as important, it was able to carry out its task in 4 seconds, which allowed for on-the-fly inspection as the pieces came from the machine.

To simplify the inspection of devices made of flexible materials, Cognex Corp. (Natick, MA) has developed software called PatFlex that can recognize the device after it has been deformed. Because stents, catheters and other flexible devices deform easily, they will not look exactly the same when an inspector presents them to the camera. A conventional vision system would need either a fixture designed to hold each part the same way or extensive programming to compensate for deviation.

A vision system using PatFlex, however, can identify an unfixtured part in a deformed state and project it back to the original state electronically to check the features. Setup is a matter of training the software with the known good part in an undeformed state. "You have the best of both worlds," says Steve Cruickshank, product manager of Cognex's PC-based vision systems. "You can use minimal fixturing and not need excessive programming to find the object that you're looking for."

This vision system verifies the presence of the large number of small metallic inserts in a plastic medical component. The nester and actuator, shown here, position the components under the camera as well as the camera itself. Photo: Invotec Engineering Inc.

Distributed vision

Cheaper and smaller hardware is another way that conventional vision systems have benefited from advances in electronics. A number of vendors have developed devices called vision sensors, cameras that contain their own processors for local processing. These compact, self-contained units run independently of the PCs. Consequently, adding a camera is no problem when an application is too complex or time-

consuming for the computing capacity of one vision unit.

Systems integrator ABCO Automation Inc. (Brown Summit, NC) is exploiting this advantage to help a manufacturer of ostomy bags automate its inspection processes to improve quality control and cut costs. One of ABCO's machines uses 12 Cognex In-Sight vision sensors in three stations to check both the flanges for attaching the bags to the patients' skin and the blister packages containing them. All of this is done at a rate of 120 packages per minute. Although having three stations triples the cost for cameras and programming, the low cost of the sensors makes the expense reasonable, especially when the resulting system is much more reliable than it would be otherwise.

The 4-inch square flanges come to the machine from a process that applies a smooth plastic film to a gummy elastomer that sticks to the skin and cuts a hole in the center of each assembly. Because inspecting each flange inside its package is too risky, the machine checks it and the package before it seals it in the package. "If the system were to look at the flange with the paper on the package, the print could cause false rejects or false accepts by tricking it into thinking either a flange was there when it wasn't, or it was contaminated when it was clean," says Perry Cornelius, a consulting engineer in ABCO's Advanced Systems Group.

Rather than attempting to design one station with an elaborate lighting scheme, Cornelius decided on using three simpler, more reliable methods. The first station checks the paper backing, illuminating the area so the camera can check the print for legibility before the paper is applied to the blister. The second station shines a strong back light through the slightly translucent flange as it sits in the blister so the camera can verify its presence, check its size, look for contaminants and ensure a hole was cut in the center. "Here, we're using simple ‘blob' analysis for detecting contaminants, which appear as areas of dark pixels," says Cornelius. "Sometimes grease drops off a machine, for example." The third station comes after the package is sealed so it can check the integrity of the seal and ensure that the product is not caught in it.

To set up the vision sensor to inspect one of its catheters, an engineer used the spreadsheet interface that comes with the sensor. The process involved selecting vision tools and parameters from drop-down menus. The software automatically fills the spreadsheet with the measurement values. Photo: Cognex

Need nice needles?

The benefits of modern vision technology, both video and vision systems, are not limited to reverse engineering projects and small-lot inspection. The technology also has been successful in the mass production of hypodermic needles, where manufacturers can make a million needles each month. The ability to inspect large numbers of tiny needle points rapidly is crucial because the manufacturers work to a no-defect standard to prevent their products from ripping and tearing tissue, causing excessive pain or other problems. Because of the no-defect standard, sample sizes are quite large. If the sample size is 10%, the inspection system on a line making 1 million needles must check 100,000 points each month.

The form and size of hypodermic needles are crucial to their function. Manufacturers grind the tips of stainless steel tubes to an angle to create edges that pierce and then cut the skin. Not only must each point be sharp to pierce the skin, but the rest of edges around the tubes also must be sharp, smooth and clean to separate the skin without tearing it. Consequently, the inspection process must measure size and check the condition of the surfaces.

Vision technology can make all of the relevant measurements quickly. Normally, manufacturers load between 100 and 200 needles in a fixture and feed the whole lot to the measurement machine. The inspection system can then check all of them quickly. Glowacky at RAM reports that one hypodermic needle point per second is the goal of one customer application.

Another attribute of vision technology that needle manufacturers, as well as other medical suppliers, are learning to exploit is its ability to automate the storage of inspection data for traceability. For this, the camera can capture serial numbers etched on a part. This ability fits high-volume production, such as needle manufacturing, particularly well because of large amount of data being collected and stored. "It grabs the serial number of the part, its image and the measurement data," says Glowacky. "Then it stores all that on an optical disk" for retrieval, if a problem were to surface.

Catheters produced for critical and cardiac care are being inspected with a machine vision sensor. To ensure that the pieces fit together correctly at assembly, the vision system measures the distance between the end of a sheath body and a flex-straw-like piece of corrugated plastic extrusion that is insert molded to the body. Too big a gap could cause the corrugated extrusion to kink instead of bending at the corrugation during use. Photo: Cognex

Team spirit

Despite the technological advancements, today's vision systems can't do it all. Many times, vendors will combine video and conventional vision with other inspection technologies to create multisensor inspection equipment. Mixing and matching technologies allows manufacturers to assign the best one to each inspection task, a strategy that is especially effective in high-volume applications. The goal is to keep cycle times as short as possible, yet ensure that measurements remain within tolerances.

"Video is ideal for measuring edges and things that you can see," says Bill Verwys, manager, applications engineering, Optical Gaging Products Inc. (Rochester, NY). "Touch probes are very good at measuring the sides of prismatic features, such as cylinder walls, and lasers are very good for measuring free-form contours." Multisensor inspection offers all of these benefits on one machine.

A common multisensor application is inspecting injection-molded plastic components such as syringe cylinders, blood vials, needle covers and valves. Video usually can measure the outer dimensions and most exterior features, such as threads, effectively. When it comes time to measuring the inside accurately, touch probes are the best means of reaching in the cylinder and taking the measurements video cameras can't see.

Inspecting hip replacements and other prosthetics is another common multisensor application. This application typically requires combining laser and video to measure the intricate shapes on these devices. Lasers work best on their free-form surfaces. Touching all the points along the surfaces with a touch probe would be too time consuming, and there are no distinct edges to offer the contrast necessary for vision to work.

Video is useful during setup. "We set them [prosthetics] up in single-

station fixtures on a rotary axis because they are so intricate that you want to look at them from multiple views," says Verwys. "So you need to do what we call, ‘carrying the coordinate system,' which is rotating the coordinate system along with the part. Because the datum point is on the opposite side of the intricate shape, you also need to set it up from the bottom." He establishes the coordinate system usually by video and sometimes by touch probe. The laser scans the cross-sections, collecting a cloud of data points that the software assembles and overlays the corresponding sections on a CAD model of the part.

Whether used alone or with other sensors, vision technology has evolved to the point where it can be a cost-

effective tool for the medical industry. As a computer-dependent technology, image processing only continues to become cheaper as computing power falls in price. Consequently, modern vision devices can not only meet the stringent quality control requirements set by medical suppliers, but also can automate the record keeping mandated by the FDA without breaking the back of a health care system. And that's what the doctor ordered. MMP