|
By Charles Hawker
Charles Hawker
06/16/08
Charles D. Hawker, Ph.D., MBA, FACB
Scientific Director of Automation and Special Projects
ARUP Laboratories
Associate Professor of Pathology (Adjunct)
University of Utah School of Medicine , Salt Lake City
Technologies that have been used in industrial settings for the past two decades are now starting to make their way into clinical laboratory automation. Known generally as “ machine vision,” this is a group of technologies that could potentially replace human inspection of specimens. In the typical clinical laboratory, there are many pre-analytic or post-analytic activities that simply involve someone picking up a specimen and looking at it. They may be checking to see if it is the correct specimen for the ordered test or if it is at the correct temperature. They may be verifying that it is from the correct patient or for the correct test by comparing a handwritten label from a doctor’s office to an overlaying computer-printed label. They may be inspecting specimens for clots, fibrin, hemolysis, icterus, or lipemia. These are all human inspection steps, many of which could potentially be automated.
Practically every product purchased today was assembled and packaged with greater efficiency and quality through the use of machine vision or similar automated technologies. In breweries, wineries, and other bottling plants, machines check fill levels and reject under-filled cans or bottles on rapidly moving conveyors. The Anheuser-Busch breweries, for example, can detect as little as a 0.05 ounce under-fill in a 12-ounce can of beer on a conveyor moving in excess of 1,000 cans per minute using a gamma ray detection system. (By comparison, today’s fastest laboratory automation conveyor systems generally range from several hundred specimens per hour to perhaps 2,000 specimens per hour.) Other bottling plants may check fill levels using weight or X-rays instead of gamma ray systems. In pharmaceutical plants, sophisticated optical systems inspect bottles on conveyor lines to be sure labels are on straight and plastic seals on caps are uniform and tight. Camera systems with optical character verification (OCV) capability read stock numbers, lot numbers, and expiration dates, comparing them to what is expected. Other systems assure there are no red pills in bottles of white pills or vice versa. Some automated plants inspect packages of disposable diapers to make sure they are correctly sealed, while others inspect jars of marmalade to make sure there are no glass shards in them. Some plants use machine vision to make sure there are no ¼ inch bolts or nuts mixed into a package of ½ inch or 3/8 inch products. The list is endless. These inspections are done at speeds that far exceed what humans can do and with accuracies of virtually 100 percent, which is simply not possible with human inspections .
In clinical laboratory automation, manufacturers are beginning to employ machine vision systems to improve quality and replace human inspections. Beckman Coulter’s AutoMate and Olympus America’s OLA2500 both have optical systems that can determine the diameter of the specimen tube, the height of packed red cells, and the height of the serum meniscus through the side of the tube, even when the tube is covered by specimen labels, and from those data, calculate the volume of serum or plasma above the cells. This information is then used by the system to determine if there is enough serum for the aliquots that have been requested of that specimen. Motoman and PVT have incorporated optical systems into their automation that can inspect tubes for cap color, height, and diameter, matching this information against an expectation communicated to the automated system by the LIS. The PVT RSA Pro Workstation can also inspect through the open side of a tube (if it is not completely covered with a label) and determine the presence and degree of hemolysis, lipemia, or icterus, in addition to determining the height of packed red cells and calculating the volume of serum or plasma above the cells. Information from these inspections in all of these lab automation systems can be used to route tubes, verify correctness of the specimen for an ordered test, prompt a report comment, or cause the tube to be set aside for manual intervention. The latter step is more valuable than it sounds. If 95 percent of tubes pass an automated inspection, then the technologist only has to inspect 5 percent of the tubes instead of 100 percent to determine an appropriate next step.
ARUP Laboratories is owned by the University of Utah and has developed a working relationship with faculty in the university’s College of Engineering to provide projects for engineering graduate students that can lead to their degrees, but also provide ARUP with valuable new technologies. One of these collaborations has led to the development of an optical system that can “see” through the side of a tube with up to six thicknesses of labels and differentiate between liquid (serum, urine, or water) and air. This system will be used robotically to ascertain that the tube has a certain minimum volume of specimen for testing or does not exceed a certain maximum volume, both of which could cause problems in other robotic steps. A variation of this concept is now being developed that will determine the height and diameter of the tube, locate the meniscus of the specimen that is present, and calculate the total specimen volume in the tube. This has value for both pre-analytic and post-analytic activities.
Another ARUP project involves the development of a high-speed camera system that can robotically inspect the labels on a specimen tube and use sophisticated optical character verification (OCV) software to compare the patient names on different labels (e.g., a hospital label versus an overlaying reference lab label) to determine that they are the same and that the specimen was not inadvertently mislabeled. To accomplish this inspection, obviously, the patient names on both labels must be visible to the camera. This system is still in development, and, although it may not reach fruition, we are cautiously optimistic that it will be successful.
To this observer, there is an interesting parallel with the development of machine vision systems for the clinical laboratory and the development of laboratory automation systems in general. Automobile assembly plants and other manufacturers began to use automation and robotics as long as 40 years ago, whereas it was only 25 years ago that Dr. Sasaki and his team at the Kochi Medical School began to apply industrial automation concepts in the clinical laboratory. [ Felder RA. Masahide Sasaki (Obituary). Clin Chem 2006;52:791-2.] Moreover, transferring these early automation ideas from Dr. Sasaki’s laboratory to large numbers of clinical laboratories through the efforts of vendors has mostly occurred since 1995, substantially lagging behind the industrial use of automation and robotics. Somewhat parallel to that, machine vision systems for inspection of products and processes have been in place in industrial settings for at least two decades. Of course, they are much faster and more sophisticated today than in those early years. Nevertheless, it seems interesting that, once again, the clinical laboratory industry is roughly 20 years behind in applying these specific industrial technologies in the laboratory setting that could make highly valuable contributions to quality and efficiency, and help ameliorate the impending shortage of medical technologists.
|
