Patent Application
20080020469
The present invention enables the operator of a combinational clinical analyzer to input samples in any order, without regard to the type or quantity of tests (i.e., assays) required of the samples, and not affect the operational performance of the device, i.e., the clinical analyzer will be able to process tests at its fully rated sample or volume throughput capacity without impacting the time to first result or the turnaround time associated with individual samples. For example, the present invention would allow an operator to input a set of patient samples, such as plasma, serum, whole blood, etc., that requires a very long time to complete for certain specific tests and then afterwards submit a series of samples requiring significantly less time to complete using other resources. The analyzer would evaluate the test requirements of the two sets as a whole and determine the schedule that minimizes the elapsed time to complete all analysis on all samples, generally referred to as the makespan, thereby maintaining the maximum possible sample throughput. A system of queues and priorities is used to make sure that the time to first result and associated turnaround time is not impacted for critical tests.
The benefits of the present invention are enhanced by the synergistic effects of two-dimensional random access to the input samples in that unlike some other clinical analyzers there is no requirement to process samples in sequential input order. Additionally, like systems, e.g., wet chemistry systems or dry chemistry systems, have the potential to share resources to maintain high operational efficiency. Such systems typically include a supply of consumables including thin-film slides, reaction vessels (cuvettes or streptavidin coated microwells), etc., a plurality of sensiometric or measuring devices including electrometers, reflectometers, luminescence, light transmissivity, photon detection, and the like for measuring specific aspects of the sample, incubator(s) for heating the samples, a supply of reagents, and a plurality of reagent delivery subsystems, all of which can be accessed and used at any time. Furthermore, the analyzer has an aliquot buffer which is a device capable of temporarily holding samples to enable re-testing or to allow sample trays, where the vast majority of required tests have been completed, to be removed from the metering zone thereby allowing a new tray of samples to be transferred into the metering area from the load/unloading area.
For a general understanding of the present invention, reference is made to the drawings. In the drawings, like reference numerals have been used to designate identical elements. In describing the present invention, the following term(s) have been used in the description.
The term “data” refers herein to physical signals that indicate or include information. When an item of data can indicate one of a number of possible alternatives, the item of data has one of a number of “values.” For example, a binary item of data, also referred to as a “bit,” has one of two values, interchangeably referred to as “1” and “0” or “ON” and “OFF” or “high” and “low.” An N-bit item of data has one of 2N values. A “multi-bit” item of data is an item of data that includes more than one bit.
The term “data” includes data existing in any physical form, and includes data that are transitory or are being stored or transmitted. For example, data could exist in the form of a one-dimensional or two-dimensional bar code that is subsequently scanned by a laser and optically converted to electronic or other transmitted signals, or as signals stored in electronic, magnetic, or other form.
To “obtain” or “produce” or “determine” an item of data is to perform any combination of operations that begins without the item of data and that results in the item of data. An item of data can be “obtained” or “produced” or “determined” by any operations that result in the item of data. An item of data can be “obtained from” or “produced from” or “determined from” other items of data by operations that obtain or produce the item of data using the other items of data.
An operation or event “transfers” an item of data from a first component to a second if the result of the operation or event is that an item of data in the second component is the same as an item of data that was in the first component prior to the operation or event. The first component “provides” the data, and the second component “receives” or “obtains” the data. Frequently various computer languages use the construct of a software subroutine or function call with parameters to achieve the data transfer.
An “operand” is an item of data on which an operation is performed.
An operation is performed “using an operand” when the operation is performed on the operand.
An “arithmetic operation” is an operation that obtains a numerical result that depends upon the value of an operand. Addition, subtraction, multiplication, and division are examples of arithmetic operations.
The term “sample horizon” refers to the number of input patient samples being considered by the scheduling algorithm. This number can vary from one to the maximum number of input patient samples possible.
A “feasible” schedule is a schedule that specifies a set of start times for each test such that no system constraints are violated and the schedule can be implemented operationally without error.
An “infeasible” schedule is a schedule that specifies a set of start times for each test, but at least one system constraint is violated and the schedule cannot be implemented operationally without error.
An “objective function” is a weighted linear combination of costs or resources, e.g., dollars or time, respectively, that when evaluated produces a number indicative of the relative “goodness” of a particular solution to a problem. Principally, the value of an objective function is either minimized, e.g., minimum time, or maximized, e.g., maximum revenue. If the solution to the problem produces either a desired minimum or maximum then the solution is said to be optimal.
An “optimal” schedule is a feasible schedule that specifies a set of start times for each test such that an objective function is maximized or minimized. For example, a specific objective function is one that computes the elapsed time from the start of the first test to the end of the last test. This specific objective function usually is minimized. Usually optimal schedules are generated using mathematical programming techniques such as linear programming, zero-one programming, dynamic programming, or variations thereof. These specific techniques are defined in Harvey M. Wagner, Principles of Operations Research, 2nd Edition, Prentice-Hall, 1975, which is hereby incorporated by reference.
A “sub-optimal” schedule is a feasible schedule that specifies a set of start times for each test in an attempt to optimize a specific objective function but does not guarantee achieving the minimum or maximum of the specific objective function. Usually sub-optimal schedules are the result of applying a heuristic optimizing method.
A “heuristic optimizing method” is an algorithm that satisfactorily works in many situations.
To access samples at “random” means that samples may be accessed in any order within the metering zone.
The ability to access samples “sequentially” means that samples may be accessed only in a predefined order within the metering zone. Often “sequential” access means that samples may be accessed only in time input order.
By “combinational” it is meant that the analyzer includes at least two chemistry systems that can encompass any combination of “dry” and/or “wet” chemistry systems. In brief and in a typical “dry” chemistry system, a patient sample and/or other fluids are aspirated from a fluid supply and deposited onto a dry slide or receiving element such as those described in U.S. Pat. No. 3,992,158 to Przybylowicz et al. The dry slide or receiving element is incubated and the amount or presence of at least one analyte in the sample metered onto the element is determined, such as through use of an electrometer, reflectometer, or other suitable testing device. In one variation of a “wet” chemistry system, e.g., the Wet Chemistry A tests referenced herein, a patient sample is aspirated from a fluid supply and deposited into a chemically coated e.g., streptavidin microwell or similar receiving element. A reagent is aspirated and added to the sample in the microwell, the resulting mixture is incubated, and subsequently washed. Additional sequences of reagent addition, incubation, and washing are performed until finally a signal reagent is added followed by incubation and measurement of the test result based upon chemiluminescence. In another variation of a “wet” chemistry system, e.g., the Wet Chemistry B tests referenced herein, a patient sample is aspirated from a fluid supply and deposited into an uncoated cuvette or similar receiving element. A reagent is aspirated and added to the sample in the cuvette and the resulting mixture is incubated, Additional sequences of reagent addition and incubations may be performed until a measurement of the test result is made based upon color change or other changes which affect optical transmission through the cuvette.
The “volume throughput capacity” of a clinical analyzer is defined as the total number of tasks that the clinical analyzer can complete in a stated unit of time. For example, a combinational clinical analyzer might have a “volume throughput capacity” of 100 tests per hour under certain specific conditions.
A clinical analyzer is said to operate at its “fully rated” volume throughput capacity when the incoming mix of samples and associated tests are in the appropriate proportions such that the absolute maximum number of tests are completed in the shortest possible time. This is a situation rarely achieved during operation of the analyzer.
The normal “operational performance” of a clinical analyzer is the average number of tests that can be complete per unit time when the incoming mix of samples and tests are in proportions reflective of ordinary hospital operations. This number is always lower than the “fully rated” volume throughput capacity of the analyzer.
The “time to first result” is defined as the elapsed time between when the clinical analyzer's power is first switched on and when the first unit of completed work (i.e., result of analysis of the first sample) is available. During this time the analyzer must warm up, perform all calibrations, and make sure that all aspects of the machine are ready for operation. Only then can the analyzer accept samples and start generating test results.
The process of defining the “test requirements” associated with a sample or series of samples is composed of determining which tests on specific individual samples are to be performed, what priority is to assigned to each test, and what resources, including processing time, are required to complete specific tests. For example, a specific sample may require three Wet Chemistry A tests, four Wet Chemistry B tests, and two Dry Chemistry C tests. In particular, the tests could be:
Adding a compilation of the individual test's priority and other factors, such as time required to complete a specific test and the number of disposable metering tips available over the sample horizon, required for each of the above tests would complete the test requirements. Some of these requirements are system constraints that are inherent in the clinical analyzer system.
The “turnaround time” for a test or series of tests associated with a sample processed by a clinical analyzer is defined as the elapsed time from time from when a sample is input to the clinical analyzer to the time that the results are available for the test or series of tests during normal operation of the analyzer, i.e., a non-startup situation.
The “makespan” of a set of tests on a group of samples is defined as the elapsed time between input of the group of samples into the clinical analyzer and when the value for the last test on the last sample is reported. If the set of test finish times (starting at time 0) on a group of samples is denoted by a vector f={f1, f2, f3, . . . , fn} where n is defined as the total number of tests required, then the makespan is defined to be the maximum element of f, or max(f). Alternatively, the makespan can be calculated by:
If some time intervals comprising the makespan are more important than others, then it is possible to weight the time intervals in the above equation to obtain a more favorable schedule. The weights can be imposed as follows:
where the weights, ai, are constrained as follows:
A “wet” chemistry system for purposes of the description that follows includes a reaction vessel which receives predetermined volumetric quantities of sample, reagent, and other fluids which are appropriately metered into the reaction vessel in order to perform an assay(s). The assay is incubated as the fluids are added to the assay(s) and specific analysis is performed, such as through luminescence detectors, light transmissivity detectors, photon detection, and the like using suitable measuring apparatus.
The analyzer that is described herein is a combinational clinical analyzer having a plurality of “dry” chemistry systems and a plurality of “wet” chemistry systems. It will be understood from the discussion that follows, however, that several variations and modifications are possible which embody the essential concepts of the present invention. For example, the analyzer can include a pair of “dry” chemistry systems and a pair of “wet” chemistry systems. Furthermore, like systems, e.g., “wet” chemistry systems or “dry” chemistry systems, can be integrated such that one system can use the resources of another system should it prove to be an operational advantage.
Referring now to
The present invention provides a method to define an optimal schedule of operation for the combinational clinical analyzer based upon the number and type of tests required to be performed on each sample, the sample horizon, and the automated inner working of the analyzer. In particular, the individual sample test requirements are known and may be transferred automatically to the memory of the scheduling computer through the console computer interface via the hospital's patient information network interface or manually by operator intervention. In addition, the sample horizon is a number that can vary between 1 and k, where k=62 in this example, which is the maximum number of input patient samples to be considered by the scheduler; in certain cases the value of the sample horizon may be pre-determined or in other cases determined in real time by operational circumstances or parameters. Furthermore, the amount of time required to perform each test is fixed by the inner working of the analyzer; in the simplest case, it takes a seconds to perform the Wet Chemistry test A, β seconds to perform the Wet Chemistry test B, and γ seconds to perform the Dry Chemistry test C. Normally, if a is smaller than either β or γ, then (β/α) is an integer and (γ/α) is an integer. For the preferred embodiment, α=4.75 seconds, β=38.0 seconds, and γ=19.0 seconds. The integer time ratios required for each test are then (β/α)=8 time units and (γ/α)=4 time units; this provides a natural system of timing in terms of time units (each unit being 4.75 seconds) which will be used below. Each chemistry can conduct a significant number of different assays, and in this example the Wet Chemistry test A can perform about 45 assays, the Wet Chemistry test B can perform about 25 assays, and the Dry Chemistry test C can perform about 60 assays. Resource sharing is possible between the various chemistry systems, e.g., one of the metering arms is shared between the Wet Chemistry A incubator, the Wet Chemistry B incubator, and the reagent supply.
At analyzer startup, the metering zone of the analyzer does not contain any input sample trays or input samples. As it would be inefficient to have the analyzer wait for the metering zone to fill to its capacity of 40 samples, the scheduling algorithm is initiated as soon as the first samples appear in the metering zone. The first step is to define the number and type of tests that will be performed on these samples. This definition could be manifested in a computer file stored either in the scheduling computer's memory or stored on an auxiliary device such as a hard disk drive. For the samples in question, the combinational clinical analyzer, either with or without the assistance of the operator, would determine the appropriate test requirements. Preferably, information on all samples would be available to the analyzer as records in the hospital's patient information system which the analyzer can selectively reference.
The Appendix contains an example of what a computer file of input patient sample information for a specific input tray might look like in computer memory. The file consists of a series of twenty records, two for each patient sample. In the first record, the patient is identified by a unique operator-assigned or hospital-assigned sequence number, first initial and last name plus another unique eight-digit number (e.g., social security number). The total number of tests for Wet Chemistry A, Wet Chemistry B, and Dry Chemistry C are represented by the next three numbers, and specific, identified tests of each type are subsequently identified by a series of test numbers, shown as two digits here. The format is such that up to seven Wet Chemistry A test numbers, four Wet Chemistry B test numbers, and six Dry Chemistry C test numbers can be contained in one record. These test numbers have a one-to-one correspondence with computer-based information specifying specific tests to be performed on the sample. In the second record, the initial patient identification information is repeated and is followed by a series of seventeen numbers representing the amount of time required to process each test specified in the previous record, respectively. The amount of time represented by the number is comprised of the individual times required for all operational steps required to complete the test including movement of the metering arm, reagent addition, incubation of the sample, and analysis readout. This test time requirement is used by the scheduling procedure to determine the optimal order in which the tests should be conducted.
During normal operation, the clinical analyzer routinely completes all tests associated with a particular set of samples in a particular input tray. To aid in the removal of trays of samples from the metering zone to the load/unloading zone, the aliquot buffer is used as an intermediate storage area for samples that are awaiting scheduling but have not yet been metered. A portion of one or more samples in a tray in the metering zone, which may have had nearly all samples and tests completed, is transferred to the aliquot buffer (the aliquot buffer also may receive samples based upon operator requests for retests and other non-routine operations). As this transfer is concluded, the completed sample tray is removed from the metering zone to the load/unloading zone, and a new tray of input patient samples is moved into the metering zone. At this point in time, up to 62 input samples, 40 in the input trays in the metering zone and 22 samples in the aliquot buffer, may be waiting to have tests performed; however, some of the input samples may have already had all of their tests completed and may be waiting for the tests on other samples in their trays to be finished. Also, depending upon the value of the sample horizon N, there may be some samples waiting to be included in the current horizon. As one or more samples may finish their testing regime at a specific point in time, up to N new samples may be added into the new samples being scheduled at any time. The usual case will be to add M<N new samples into the group being scheduled. Operationally, the analyzer software will have to update the test requirements of the samples already in progress and append the test requirements for the samples being added. The resulting new scheduling problem is then submitted to the scheduling algorithm for processing.
Another aid for the scheduling algorithm is the establishment of virtual queues to which the input patient samples are assigned automatically based upon information in the hospital's patient information system or manually by the operator. Typically there are three prioritized queues: the Wet Chemistry A queue, the Wet Chemistry B and Dry Chemistry C queue, and a Wet Chemistry B queue in this example formed from samples in the aliquot buffer.
To further aid the scheduling algorithm in finding feasible schedules, within each queue, the operator or information contained in the hospital's patient information system may optionally indicate a priority for each input patient sample or test. In particular, the following priorities are possible: STAT, which has the highest priority, softSTAT, which has the second highest priority, REFLEX, which has the third highest priority, and ROUTINE, which has the lowest priority. A predefined numerical weight can be added to the input patient sample record for use by the scheduling algorithm. For example, STAT might have a weight of 4, softSTAT might have a weight of 3, REFLEX might have a weight of 2, and ROUTINE might have a weight of 1. Within a queue and within a priority, the samples that have been in the system the longest have the highest priority.
In one embodiment, the present invention provides a methodology utilizing a first heuristic optimizing procedure to provide a test-by-test starting time schedule to process a set of input samples in a combinational clinical analyzer. The resulting starting time schedule will minimize or nearly minimize the makespan required to complete the tests. After defining the test requirements for the samples in the metering zone and assigning the tests to priority queues, the first scheduling heuristic is based upon the execution of scheduling logic for each of the three queues, i.e., the Wet Chemistry A queue, the Wet Chemistry B queue, and the Dry Chemistry C queue, at predefined time intervals. Overall, the scheduling activity is controlled by the top-level scheduler logic as shown in
Wet Chemistry A Queue: On each four (4) time unit increment, the first sample is selected from the queue, and the algorithm attempts to schedule the first test at 4 time units from the current time. If a conflict is detected, i.e., perhaps another test has been scheduled already, the algorithm attempts to schedule the test at subsequent four (4) time unit intervals into the future, until the test is successfully scheduled. This process continues until all tests in the sample are scheduled. The algorithm always attempts to schedule each test at the next 4 time unit interval from the current time. After the sample is completely scheduled, the algorithm checks for two conditions: (a) there is no test from the sample scheduled in the next four (4) time unit interval or (b) if the scheduled sample requires more then a configurable number of disposable metering tips to process (an inherent system constraint). If either condition is true, all the tests from the sample are removed from the schedule, and the next sample from the queue is attempted to be scheduled. This process continues until a sample is found with a test which can be scheduled in the next 4 time unit interval and does not require more than a configurable number of disposable metering tips to process. If the algorithm reaches the end of the prioritized queue and no sample meets both conditions, then no Wet Chemistry A sample is scheduled in the next 4 time unit interval. A flowchart of this logic is shown in
Wet Chemistry B and Dry Chemistry C Queue: At each time unit (4.75 second interval), the first sample from the queue is selected and all tests from that sample are scheduled. The algorithm, as indicated by the logic in
Wet Chemistry B Queue: On each eight time unit interval the algorithm attempts to schedule all tests within all samples in the aliquot buffer, if it contains samples, using the same logic as applied to scheduling samples in the Wet Chemistry B and Dry Chemistry C queue. A flowchart of this equivalent logic is shown in
It is important to have as many samples as possible in the metering zone and to keep samples moving through the analyzer as quickly as possible and to keep the samples in the metering zone for as long as possible to enable re-testing if required. It also is desired to minimize the number of gaps between trays and to position as many samples as possible in the metering zone. The first goal should be to maintain maximal workflow, and a secondary goal is to enable re-testing. To maintain maximal workflow, new sample trays should be indexed into the metering zone whenever possible. To enable re-testing, samples are not indexed out of the sample handler if no new work is in the load/unload area. A flow chart of sample handler rotation logic is shown in
By definition, the first heuristic scheduling procedure produces an improved schedule that may provide an improved makespan, preferably the minimum makespan.
For example, Appendix A contains a representative file of patient and sample information for ten input samples that might be created and stored as an internal computer file. This files consists of twenty records, two for each patient sample. The first record contains the patient's name, an eight-digit identification number, e.g., social security number, the number of Wet Chemistry A tests, the number of Wet Chemistry B tests, and the number of Dry Chemistry C tests, a series of seven numbers indicating specific test numbers for Wet Chemistry Tests, a series of four numbers indicating specific test numbers for Wet Chemistry B tests, and six numbers indicating specific test numbers for Dry Chemistry C tests, respectively. The specific test numbers are one-to-one linked to internal computer records uniquely identifying the test to be performed. The second record repeats the patient's name and the eight-digit identification number plus has a series of seventeen numbers specifying the time requirement (in time units) for each of the previous records' Wet Chemistry A, Wet Chemistry B, and Dry Chemistry C tests. Considering only the first five patient samples, this information defines a series of 44 tests. If the same priority is assigned to all tests, the above first heuristic algorithm produces the following feasible schedule:
where the makespan for the above schedule is 544 time units based upon test 19 being the last test to finish.
In another embodiment, the present invention provides a method to utilize a second heuristic optimizing procedure to produce a test-by-test starting time schedule to process a set of input samples in a combinational clinical analyzer. The resulting starting time schedule will minimize or nearly minimize the makespan required to complete the tests. This second scheduling heuristic is based upon looking at all tests for all samples that are within a predefined or operator-defined sample horizon. The first step is to define the test requirements of the samples in the sample horizon, which consists of accessing input sample records in the computer memory and organizing the data into a form suitable for transfer to the scheduling algorithm. For mathematically oriented software like the second heuristic method or mathematical programming algorithms, like linear programming, zero-one programming, or dynamic programming, a linear algebra organization in the form of vectors or matrices is usually preferred.
The information file shown in Appendix A is read into the scheduling computer and the information manipulated into an operand consisting of a matrix consisting of 44 rows (the number of tests) by 3 columns, where column 1 is the test number (from record one), column 2 is the time requirement (in time units from record two), and column 3 is an indicator variable where 1 denotes a Wet Chemistry A test, 2 denotes a Wet Chemistry B test, and 3 denotes a Dry Chemistry C test. If the sample horizon is set at 5 samples, the operand formed by the software in Appendix B for the first five patients' sample information (ten records) in Appendix A is as follows:
where the matrix operand consists of three vectors in the software as follows:
OPERAND=[testv reqv type]
After the construction of the vector comprising the operand, the Appendix B software calls the function schedule to transfer the operand to the second heuristic scheduling procedure and generate the associated schedule based upon assigning the earliest start times to the tests requiring the longest time. While not guaranteeing a minimum makespan, in practice, this method produces excellent schedules. Furthermore, when higher priority tests are scheduled before lower priority tests, time to first result is often minimized. The schedule is generated and returned by the scheduling software in Appendix B for type 1 (Wet Chemistry A) test and type 2 (Wet Chemistry B) tests is as follows:
This schedule completes all type 1 and type 2 tests with a makespan of 472 time units or 2,242 seconds which is optimal, i.e., the minimum makespan possible for this set of tests. Compared to the makespan of the schedule for the same samples using the first heuristic method above, this is an improvement of (544−472)=72 time units or about 13%. Application of the same scheduling algorithm to the type 3 (Dry Chemistry C) tests, which is not subject to resource sharing, and hence runs unobstructed, produces a schedule as follows:
where the makespan for the type 3 tests is only 108 time units, which is much less that the 472 time units required for the type 1 and type 2 tests.
In another embodiment, the present invention provides a method to utilize a mathematical programming procedure to produce a test-by-test starting time schedule for processing a set of input samples in a combinational clinical analyzer. Using the same methodology of defining the test requirements, transferring the test requirements to the scheduling algorithm, and subsequent generation of the schedule, this embodiment results in an optimal schedule. The resulting starting time schedule (which may not be unique, i.e., there may exist multiple schedules achieving the same, minimal makespan) will minimize the makespan required to complete the tests. The mathematical programming procedure employed is based upon a technique called implicit enumeration that is appropriate for the solution of problems formulated in a zero-one programming configuration. However, because many different formulations of the combinational clinical analyzer scheduling problem are possible, other mathematical programming methods such as linear programming, dynamic programming, branch and bound procedures among others, can be utilized to obtain optimal schedules.
The substitution of an implicit enumeration algorithm to determine a schedule for the first five patient samples of Appendix A produces the same schedule as the second heuristic scheduling procedure above.
The scheduling method according to the present invention can be implemented by a computer program, having computer readable program code, interfacing with the computer controller of the analyzer as is known in the art. The scheduling method can also be incorporated in an article of manufacture that includes a computer usable medium having computer readable program code configured to conduct the method of the present invention. The computer usable medium can include such known mediums such as an optical disk, or a hard drive.
It will be apparent to those skilled in the art that various modifications and variations can be made to the methods and processes of this invention. Thus, it is intended that the present invention cover such modifications and variations, provided they come within the scope of the appended claims and their equivalents.
The disclosure of all publications cited above is expressly incorporated herein by reference in their entireties to the same extent as if each were incorporated by reference individually.
This application claims priority from U.S. Provisional Application 60/832,191, filed on Jul. 20, 2006, which is incorporated by reference in its entirety.
| Number | Date | Country | |
|---|---|---|---|
| 60832191 | Jul 2006 | US |
