The work station robot, already hard at work in industrial andpharmaceutical laboratories, has begun to appear in clinicallaboratories. This herald of a new phase in automated medicaltechnology provides a host of advantages in routine analysis. During thecoming decade, it will profoundly change the way we perform tests. Over the past 20 years, laboratory medicine has moved towardprogressively greater levels of automation. Today, practically everyclinical chemistry technologist uses at least one instrument requiringonly placement of specimens in a sample tray and a push of the startbutton. From that point, the instrument calibrates itself withappropriate standards, analyzes the specimens, and prints a report thatcan often be appended directly onto the patient’s chart. Many instrument also provide accuracy and precision estimates andreports that indicate whether the result is within normal range. Inhighly computerized hospitals, analyzers interface with hospitalinformation systems to generate and forward reports electronically.
Somesystems can combine reports from various work stations. Specimen collection and preparation, in fact, are the only steps inthe entire process from requisition to billing that are still unassistedby automation. It isn’t yet feasible to automate the collection ofpatient specimens, but a new breed of robot work stations can largelytake over preparation of specimens for analysis, thus extendinglaboratory automation one crucial step further. In this article, we will examine the many benefits of robotics inthe clinical laboratory, the basic technical elements involved, and theactual process of putting robots to work in a variety ofapplications–including our own experience with a mechanical helper atthe bench. The idea of robot work stations isn’t new. Back in 1971, Dr.Raymond Gambino predicted, “As wage rates continue to rise andskilled help becomes scarcer, the robot’s utility will become moreattractive.” Robotic technology is built into most of thehigh-throughput instruments we use today, and even the earliestautomated analyzers included basic robotic components.
Laboratory robotics evolved from the manually operated manipulationdevices developed to handle hazardous–and particularly,radioactive–materials. These early robots were just pivoted gripperhands enclosed in an isolated area and operated by a technologist fromwithout. Ultimately, the development of electronic and sensor controlsallowed walkaway operation, and the stand-alone robot was born. The heart of a prototype robot work station is the arm/handassembly, an electromechanical device mounted in the center of a workbench. The arm, a horizontal bar attached to a vertical column, isplaced on a turntable attached to a base unit. The robot’scomputer controller rotates the turntable through 360 degrees and movesthe arm assembly up and down the column and in and out from the center. A mechanical hand attachment performs the actual tasks.
This handhas three types of motion, or degrees of freedom, to position itselfanywhere within a defined reach of the base unit. Visualize a cylinder;within it coordinate system, the hand can rotate, move up or down, andmove in and out. In the most versatile arrangement, the hand itself can rotate aboutthe arm in a wristlike motion.
Interchangeable hands can be attached toperform a variety of functions. For example, a unit with two clamp-likefinger devices can open or close upon electronic command to grasp vials,test tubes, and other items. Another hand can be fitted with a syringewhose plunger is operated by the computer controller, adding a fingermotion or fifth type of movement to the robot’s repertoire. Formaximum flexibility, the arm should be able to change hands undercomputer control during the course of operation. Other lab equipment is permanently fastened to the work surfacesurrounding the arm unit–standard accessories including test tube racksand reagent bottles, along with other familiar devices modified forremote control operation (electronic balances, vortex mixing stations,automatic liquid dispensers, heating baths, centrifuges, and sampleshakers).
Finally, the work station incorporates some devices unique torobotics, such as stands in which to park idle robot hands, racks forpipet tips and other robot supplies, stations to uncap bottles orspecimen containers, and mechanical or optical sensors enabling therobot to verify its operations. This verification process is critical because the robot cannot”see” what it is doing. It depends on knowing the preciselocation of its supporting equipment and specimens.
To that end, whenit picks up a sample tube from a rack it can be programmed to touch thetube against a switch sensor, producing a signal that tells the computera tube was indeed lifted from that location. Programming the robot through a computer control unit calls foringenuity and patience. First, it must be taught the location of allsupporting apparatus and its movement coordinates. The softwarecontrolling this process must provide a convenient way to identify thesepoints within the robot’s cylinder-shaped range of motion. In thesystem we use, the programmer steps the arm and hand to the desiredposition by push button. Each position is assigned a name that isstored in a dictionary in the computer’s memory.
Next, the programmer defines a series of unit activities that canbe performed anywhere in the work space. A typical routine–say, for apouring action–would consist of a wrist rotation with the robot handholding an open test tube. Routines and locations can then be combinedto form procedures.
In an “empty to waste” procedure, forexample, the hand would move over a waste beaker and execute the pourroutine. Using this programming method, we can develop unit operations toperform the functions of weighing, pipetting, centrifuging, decanting,mixing, heating, performing column chromatography, extracting,filtering, and other ordinary laboratory tasks. In the final step, theprogrammer combines locations, routines, and procedures to create a”method” for the complete process of specimen preparation. The process control software also permits other typical computerfunctions like the ability to loop, or go back and perform the same taskrepeatedly; to branch into different program segments depending oncertain variables; to index to different test tube rack positions; toperform calculations; to control peripheral devices; and to read andwrite to a disk and communicate with other computers. Our familiarity with manual methods of specimen preparation maymake this process seem like more trouble than it’s worth, butremember that initial programming is a one-time chore. Once the robothas been taught its tasks, it offers a number of key advantages overmanual performance–especially to financially constrained laboratoriesof the future: * Accuracy.
Test results are only as accurate as the preparationof specimens and standards. Once the rest of the test process isautomated, specimen preparation often becomes the chief source of error. It’s part of human nature for people to make mistakes. A stepmay be skipped, a reagent may be delivered incorrectly, a specimen maybe accidentally contaminated by an unwashed apparatus, or a balance maybe misread.
Day and night shift personnel may vary slightly in habitsor practices, such as reading the miniscus of a pipet differently,leading to small but systematic differences in test results. By performing tasks identically every time without proceduralmistakes, robots can improve method accuracy. Once a procedure has beensuccessfully adapted to the robot and programming glitches have beeneliminated, the robot only makes mistakes when a component is worn orfails. And these errors are almost always obvious immediately, usuallybringing the procedure to a halt. Subtle and easily overlooked roboterrors are rare.
* Precision. A robot processes each specimen with much lowervariance than a person does. A technologist may have to answer thetelephone or be distracted by a co-worker at a critical moment and leavesome samples in an incubator bath a few minutes too long.
Robots, whichalways follow the exact same program, eliminate many of these randomdeviations. * Reliability. If the results of a complex procedure are needed atany time of the day or night, as with certain toxicology procedures, arobot is always on duty. Technologists who are most skilled andexperienced in a particular procedure will not be at the bench round theclock.
* Documentation. Specimen preparation may take different pathsdepending on the situation. If the initial result is too high, thesample may have to be diluted and run again. Or the supplied volume maybe too small for standard treatment. The robot computer, working inconcert with the instrument’s computer, can be programmed to handlethese variations and document the exact method of preparation for eachspecimen. A technologist needn’t be present to make the decisionand record the results.
* Cost. The typical robot work station currently costs about thesame as 1-1/2 to 2 years of an experienced technologist’s salaryand benefits. The system we use cost approximately $47,000, whichincludes a base price of $25,000 and an additional $22,000 in neededaccessories, such as shakers, centrifuges, vortexes, and cappingstations. Depending on the procedures the robot will perform,accessories can run from $20,000 to $40,000 over the basic arm/hand unitcost. Obviously, the robot represents a significant investment, but onewith the potential to pay for itself in a reasonably short period. Itenhances productivity by allowing more tests to be run at existing stafflevels and frees personnel for more challenging activities requiringhuman hudgment.
Particular work stations may be especially appropriate for robotspecimen preparation. High-volume work stations are good candidates,even if the specimen treatment process is simple, since such repetitivetasks are boring and tiring. Unpopular tasks, such as fecal analysis,are also likely targets. So are complex procedures with a high marginfor error or dangerous procedures involving radioactive or other toxicreagents.
To evaluate the feasibility of using routine robotics in laboratorymedicine, our biochemistry department at The Cleveland Clinic Foundationrecently installed a Zymark Robotic System, manufactured by ZymarkCorp., Hopkinton, Mass. The robot’s first task was a complex one:sample preparation for therapeutic drug monitoring.
We trained it toperform all steps in the liquid-solid extraction of tricyclicantidepressant drugs from plasma prior to their analysis byhigh-performance liquid chromatography (HPLC). Here is how weincorporated the robot into routine use for the analysis of tricyclicantidepressants. Before teaching the robot any task, you must define every operationinvolved, down to the smallest step. Figure I presents a dual flowdiagram for our manual and robotic tricyclic procedures. Each movement involved in extracting the specimen must beprogrammed as an absolute or relative position where the robot willperform a specific step. The most efficient way is to program thesystem “from the top down”–that is, by delineating the majorsteps necessary to reach the goal, and then writing subroutines to reachthe end point of each major step. The robot performs the entire procedure by combining andintegrating these subroutines.
Figure II shows the string of roboticcommands necessary to perform a tricyclic antidepressant assay from thetime of specimen delivery to the robot until the specimen is ready forinjection on the HPLC. This program may look quite simple, but keep inmind that each step involves numerous subroutines composed of manyindividual steps. The Zymark robot system and its auxiliary equipment for drugextraction are shown as a schematic diagram in Figure III. The robotcomputer controller, a separate component that functions as the roboticbrain, is programmed in a language called Easy-Lab that uses English tocommunicate with the programmer. First, we program each step in a routine. Using a manualcontroller, the operator moves the robot arm, with appropriate handattached, with appropriate hand attached, to a desired position andassigns each location a name that is stored, along with the hand’sexact coordinates, in the computer dictionary.
From this point on, thehand will travel only to that position when instructed to do so. Eachlocation can have only one name; if the same location is used in adifferent subroutine, it cannot be renamed. This initial process is somewhat time-consuming.
It usually takesthree to five minutes to define one position. We estimate that it tookabout 72 hours to program the entire tricyclic assay. A completeprintout of the program requires 1,436 lines, one line for each commandthe robot is given to complete the process. Once the initial subroutinepositions have been named, we don’t have to reprogram theirlocations coordinates for use in other procedures.
For example, theposition programs for adding internal standards to plasma or loading anHPLC injection vial are the same for all assays. After the robot has learned the various subroutines, they must beedited so that the total program is integrated to provide a continuousflow of movement from one subroutine to the next. Programming is really quite straightforward, as long as yo rememberthat the robot moves in three dimensions and always travels between twopoints by the shortest possible route. When integrating subroutinessome distance apart, the robot may choose an unanticipated path–andcrash into any obstruction along the way. These mishaps occur quickly.By the time you realize a crash is imminent, it will already havehappened. Naturally, these crashes can damage robot hands. Avoid them byestablishing safe positions for the robot to reach during a subroutine.
It may take the arm/hand assembly a few seconds longer to go from pointA to point B via a safe route, but it will arrive at Point B withoutcrashing. The hardest part of teaching a robot, in fact, is recognizing thefact that it is a slavishly cooperative helper. It does exactly whatit’s told, and wrong moves usually result from programming errors.Once you discover the flaw in its instructions, the problem cangenerally be corrected. When the robot has been successfully taught to carry one specimenthrough the entire procedure without error, it can process any number ofspecimens identically. At this point, the program is ready forverification, to make sure it produces the same results as manualpreparation. We verified our tricyclic antidepressant assay in the usual manner,by comparing results obtained manually for standards, controls, andpatient specimens with the same specimens processed by the robot.
Thedata, excerpted in Figures IV and V, have clearly demonstrated thatreproducibility of the robot assay compares well with that of the manualmethod. We are now in the process of adapting ELISA-type immunoassaysand the determination of antiepileptic drugs, various drug metabolites,and porphyrins to robotic analysis. Our experience has confirmed our belief that robotics are indeedthe wave of the future, with a wide variety of potential applications.It’s a common misconception that only high-volume laboratories canuse robotics efficiently.
In reality, mid-size or even some smallhospital labs can use robots effectively which careful planning. Versatility is one key benefit to consider. Similar assays can beeasily performed at the same work station.
A robot performingliquid-solid extractions of one drug can conduct the same procedure onmany different drugs using the same basic ssytem and operatingprinciples. Switching assays, once the robot is programmed, usuallytakes no more time than is needed to change reagent bottles, rack in thenew set of specimens, and tell the robot to perform the new assay. Thisusually takes less than 15 minutes. The main limit on a robot’s repertoire is the space availableat the bench. The arm moves within a full circle and has access to anypoint within its diameter. The number of pieces of equipment andsupport racks this area will hold determines how many procedures can beperformed. Our robot performs both liquid-liquid and liquid-solid extractionsin its work area. The platform shaker and centrifuge are designatedonly for liquid-liquid extractions and are not used during theliquid-solid extraction phase.
Conversely, the column extractingstation is unused during the liquid-liquid extractions. The robot can function overnight without supervision, so it’spossible to start a day’s workload the night before by having thenight technologist rack the specimens, place them by the robot, tell itthe number of specimens to be processed, and push the start button.Robotic procedures aren’t necessarily faster than manual methods,as Figure IV also shows; their value lies in the employee time saved bywalkaway operation. So far, we haven’t focused on the one most important part ofrobot operations: people. Robots will work for anyone who can performthe setups, but they each need a special caretaker–one personresponsible for their care, preventive maintenance, and troubleshooting.In our lab, two technologists have been trained to program andtroubleshoot the robot.
Since we put our robot into practice last May, it has been downonly twice, both times within the first 60 days of implementation.Preventive maintenance consists of oiling the arm once a month. Wedon’t think that’s a bad track record for a workhorse that’s willing to run day and night. Unfortunately, the very idea of robots often triggers a response ofmistrust and anxiety from medical technologists, who picture their jobsbeing eliminated. We have found, however, that attitudes quickly changeonce the staff understands both the advantages and limitations ofrobotics. It doesn’t take long to recognize that the robotactually provides technologists with a new level of freedom. Rather than eliminating jobs, robots provide a new resource thatallows the expansion of laboratory services with no significant increasein personnel costs. Released from repetitive, routine procedures,technologists can pursue tasks that demand more of their skills andintellectual ability, such as improving communication between thelaboratory and clinical services, developing new analytical techniques,and pursuing clinically relevant research projects.
Robotics are justanother avenue toward better patient care. In years to come, we will see robots used extensively in clinicalpharmacology labs as well as routine clinical chemistry sections. Howwill widespread implementation of robotics change the clinical chemistrylaboratory of the future? It will provide a cost-effective way todeliver many services without a significant rise in labor and operatingcosts. More important, it will free more professional time and talentfor tasks that require human intervention. After working with our robot for nine months, we share the growingconviction that robots will be a major part of laboratory medicine inthe next few years, and not just in large institutions. Theirversatility, cost-effectiveness potential, and ease of operationguarantee an important role.
Getting to know them now, and implementingthem as soon as it’s feasible to do so, will improve our efficiencyand the quality of our patient care.