Real-time sample aspiration failure detection and control

Abstract

Methods for early real-time detection or prediction of aspiration faults in automated diagnostic analysis systems include artificial intelligence algorithms configured to use either cluster analysis or probabilistic graphical modeling based on aspiration pressure measurement signal waveforms. Aspiration faults can include under-aspiration and unwanted gel collection. These methods allow the aspiration process to be terminated in a timely manner to avoid or minimize possible adverse downstream consequences, such as erroneous sample test results and / or equipment downtime for inspection and cleaning. Apparatus for early real-time detection or prediction of aspiration faults is provided as well as other aspects.

Description

Technical Field

[0001] Cross - Reference to Related Applications This application claims the benefit of U.S. Provisional Patent Application No. 63 / 221,450, filed Jul. 13, 2021, entitled “REAL - TIME SAMPLE ASPIRATION FAULT DETECTION AND CONTROL”, the entire disclosure of which is incorporated herein by reference for all purposes.

[0002] This disclosure relates to the aspiration of liquids in an automated diagnostic analysis system.

Background Art

[0003] In medical testing, an automated diagnostic analysis system can be used to analyze a biological sample to identify analytes or other components in the sample. Biological samples can be, for example, urine, whole blood, serum, plasma, interstitial fluid, cerebrospinal fluid, etc. Such biological liquid samples are usually contained in sample containers (e.g., test tubes, vials, etc.) and can be transported to and from various imaging, processing, and analyzer stations within the automated diagnostic analysis system via container carriers and automated tracks.

[0004] An automated diagnostic analysis system typically includes one or more automated aspiration and dispensing devices, which are configured to aspirate (draw in) a liquid (e.g., a sample of biological liquid, or a liquid reagent, acid, or base to be mixed with the sample) from a liquid container and dispense the liquid into a reaction vessel (e.g., a cuvette, etc.). The aspiration and dispensing device typically includes a probe (e.g., a pipette), which is attached to a movable robotic arm or other automated mechanism that performs the aspiration and dispensing functions and transfers the sample or reagent to the reaction vessel.

[0005] During the aspiration process, a movable robotic arm that can be controlled by a system controller or a processor positions a probe above a liquid container and then lowers the probe into the container until the probe is partially immersed in a liquid (e.g., a biological liquid sample or a liquid reagent). Thereafter, a pump or other aspiration device is activated to aspirate (draw in) a portion of the liquid from the container into the probe. The probe is then withdrawn from the container and moved so that the liquid can be transferred to a reaction vessel for processing and / or analysis and dispensed into the reaction vessel.

[0006] During or after aspiration, the aspiration pressure signal is analyzed to determine whether any abnormalities have occurred, i.e., to check for clogging (e.g., collection of gel or other unwanted substances from the liquid container) or whether the amount of aspirated liquid is insufficient (hereinafter sometimes referred to as short - volume aspiration or malfunction).

SUMMARY OF THE INVENTION

PROBLEMS TO BE SOLVED BY THE INVENTION

[0007] Conventional aspiration detection systems may be able to detect some abnormal aspirations, but such conventional detections may not be sufficient to avoid some adverse results. Therefore, there is a need for improved methods and devices for detecting and / or predicting aspiration malfunctions so as to avoid or minimize such possible adverse results.

MEANS FOR SOLVING THE PROBLEM

[0008] In some embodiments, a method for detecting or predicting aspiration disorders in an automated diagnostic analysis system is provided. The method includes performing aspiration pressure measurement via a pressure sensor while a liquid is being aspirated in the automated diagnostic analysis system. The method also includes analyzing the aspiration pressure measurement signal waveform via a processor that executes an artificial intelligence (AI) algorithm. The AI algorithm is configured to perform cluster analysis of the aspiration pressure measurement signal waveform or probabilistic graphical modeling based on the aspiration pressure measurement signal waveform. The method further includes identifying and responding to an aspiration disorder via the processor in response to the analysis.

[0009] In some embodiments, an automated aspiration pipetting device is provided that includes a robotic arm, a probe coupled to the robotic arm, a pump coupled to the probe, a pressure sensor configured to perform aspiration pressure measurement while a liquid is being aspirated via the probe, and a processor configured to execute an artificial intelligence (AI) algorithm to detect or predict an aspiration disorder during the aspiration process and respond thereto. The AI algorithm is configured to analyze the aspiration pressure measurement signal waveform derived from the pressure sensor using cluster analysis or probabilistic graphical modeling.

[0010] In some embodiments, a non-transitory computer-readable storage medium includes an artificial intelligence (AI) algorithm configured to detect or predict an aspiration disorder based on analysis of an aspiration pressure measurement signal waveform. The analysis may be cluster analysis of the aspiration pressure measurement signal waveform or may use probabilistic graphical modeling based on the aspiration pressure measurement signal waveform.

[0011] Still other aspects, configurations, and advantages of the present disclosure may become readily apparent from the following detailed description and illustrative examples of numerous embodiments and implementation modes, including the best mode contemplated for carrying out the invention. The present disclosure can also enable other different embodiments, and some of the details thereof can be changed in various respects without departing from the scope of the present invention. For example, although the following description relates to an automated diagnostic analysis system, the aspiration disorder detection and / or prediction methods and apparatuses disclosed herein can be easily applied to other automated systems that benefit from early and accurate real-time detection and / or prediction of aspiration disorders. The present disclosure is intended to embrace all modifications, equivalents, and alternative forms within the scope of the claims.

[0012] The drawings described below are for illustrative purposes only and are not necessarily drawn to scale. Accordingly, the drawings and description should be regarded as illustrative in nature and not as restrictive. The drawings are not intended to limit the scope of the present invention in any way.

Brief Description of the Drawings

[0013]

Figure 1

Figure 2A

Figure 2B

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15A

Figure 15B

Figure 16

Figure 17

DETAILED DESCRIPTION OF THE INVENTION

[0014] Some conventional systems may be able to detect some abnormal aspirations, but such detections are not made early enough in the aspiration process to avoid possible adverse downstream results, such as inaccurate test results due to insufficient aspiration volume, and / or equipment downtime for inspection and cleaning of probes and other affected mechanisms and subsystems due to collection of gel or other undesirable substances.

[0015] Accordingly, the embodiments described herein provide a method and apparatus for accurately detecting or predicting aspiration failures in real time at an early stage of the aspiration process. Early real-time detection or prediction of aspiration failures advantageously avoids or minimizes any possible consequences downstream of a faulty aspiration, such as, for example, equipment downtime and / or incorrect analysis results, by timely terminating the aspiration process and / or by enabling the execution of suitable error state procedures. In some embodiments, an aspiration failure can be advantageously detected or predicted within the first 100 milliseconds after the start of the aspiration process. Detectable and / or predictable aspiration failures can include, for example, gel (or other undesirable substance) collection failures and / or insufficient aspiration volume failures. An insufficient aspiration volume failure occurs when the aspiration fails to draw in a sufficient amount of liquid, which can be caused, for example, by the liquid container not containing a sufficient amount of liquid and / or by a defect in the equipment (e.g., a defect in the aspiration pump or aspiration tube, improper positioning of the probe within the liquid container due to a defect in the robotic arm, blockage, etc.). A gel collection failure can also be caused by a defect in the equipment (e.g., improper positioning of the probe within the liquid container due to a defect in the robotic arm, the probe contacting or getting too close to the gel separator between sample components within the liquid container, or the bottom layer of gel or red blood cells within the liquid container, resulting in the gel being aspirated by the probe).

[0016] In some embodiments, early and accurate real-time detection or prediction of aspiration disorders can be implemented via a software or firmware learning-based AI (artificial intelligence) algorithm executed on a system controller, processor, or other similar computer device of an automated diagnostic analysis system or an automated aspiration dispensing device. In some embodiments, the AI algorithm can be configured to perform a cluster analysis of the aspiration pressure signal waveform using only two metrics. One metric obtains the rate of change of pressure of the aspiration pressure measurement signal waveform, and the other metric obtains the inflection characteristics of the aspiration pressure measurement signal waveform. Using these two metrics, a detection threshold (classification boundary) is established based on training data that includes pressure signal waveform samples representing both normal aspiration and abnormal aspiration (of different types). In other embodiments, three or more metrics can be used. The training data may be unlabeled (i.e., not labeled), and the class boundary for normal aspiration can be established based on K-means clustering using a four-cluster classification based on only two metrics. Alternatively, a supervised classification technique using a support vector machine can also be used.

[0017] In other embodiments, early and accurate real-time detection or prediction of aspiration disorders can be implemented via a software or firmware learning-based AI algorithm configured to perform probabilistic graphical modeling based on the aspiration pressure measurement signal waveform. In some of these embodiments, a Hidden Markov Model (HMM) can be used to predict aspiration disorders based on an examination of metrics derived from the aspiration pressure measurement signal waveform. A 3-state left-to-right HMM architecture can be used, and separate HMM models trained for "normal aspiration" and those trained for "abnormal aspiration" can be used. In some embodiments, N HMM models can be used, one trained for "normal aspiration" and each of the others trained for a specific type of "abnormal aspiration". Each model can be trained using a machine learning method and labeled (supervised or unsupervised) sample training data. In a 2-model embodiment, both HMMs can be run simultaneously in real-time against the measured aspiration pressure signal waveform. The sequence output probabilities over a continuous sequence of pressure signal values are calculated using both models. Classification as "normal" or "abnormal" aspiration can be performed by comparing the relative sequence likelihoods (P SEQUENCE NORMAL / P SEQUENCE ABNORMAL ), or alternatively, for "normal" and "abnormal" aspiration, by comparing the sequence likelihoods using the "normal" HMM model and the "abnormal" HMM model against their respective threshold sets.

[0018] In these embodiments where unlabeled sample training data with randomly mixed normal and abnormal aspiration pressure waveforms is available, an unsupervised classification method such as K-means clustering can be used to automatically classify the aspiration signal samples into a preferably selected number of groups. And the determination of which group can be considered normal aspiration can be made by examining one or more samples from each group and relying on prior knowledge of what a normal aspiration waveform should look like.

[0019] Advantageously, both the cluster analysis and the probabilistic graphical modeling embodiments for detecting or predicting aspiration disorders can be implemented online and in real time, starting from the beginning of the aspiration process, with low computational complexity (O(N)) and low memory requirements, and can thus be easily implemented in firmware or software. The training aspects of the cluster analysis and the probabilistic graphical modeling (for determining the failure threshold) can be executed offline. The trained transition probabilities and output probabilities of the probabilistic graphical modeling can be stored in the memory of a system controller, a processor, or other similar computer devices, and can then be used online to evaluate the failure state of each aspiration pressure measurement waveform in real time at the sampled points.

[0020] According to one or more embodiments, with reference to FIGS. 1 to 17, methods and apparatuses for early and accurate real-time detection or prediction of aspiration disorders will be described in more detail below.

[0021] FIG. 1 shows an automatic diagnostic analysis system 100 according to one or more embodiments. The automatic diagnostic analysis system 100 can be configured to automatically process and / or analyze a biological sample contained in a sample container 102. The sample container 102 can be received in one or more racks 104 provided in a loading area 106 in the system 100. A robotic container handler 108 can be provided in the loading area 106 to grip the sample container 102 from one of the racks 104 and load the sample container 102 into a container carrier 110 located on an automatic track 112. The sample container 102 can be transported through the entire system 100 via the automatic track 112 to, for example, a quality check station 114, an aspiration dispensing station 116, and / or one or more analyzer stations 118A-118C.

[0022] Quality check station 114 can pre-screen for contaminants or other undesirable characteristics of a biological sample to determine whether the sample is suitable for analysis. After passing the pre-screening, the biological fluid sample can be mixed with liquid reagents, acids, bases, or other solutions at aspiration dispensing station 116 to enable and / or facilitate analysis of the sample at one or more analyzer stations 118A - 118C. Analyzer stations 118A - 118C can analyze the sample for the presence, quantity, or functional activity of a target entity (analyte), such as DNA or RNA for example. Analytes commonly tested for can include enzymes, substrates, electrolytes, specific proteins, drugs of abuse, and therapeutic drugs. In system 100, more or fewer analyzer stations 118A - 118C can be used, and system 100 can include other stations (not shown), such as a centrifugation station and / or a decapping station.

[0023] The automated diagnostic analysis system 100 can also include a computer 120, or alternatively, may be configured to communicate remotely with an external computer 120. The computer 120 may be, for example, a system controller, and may have a microprocessor-based central processing unit (CPU) and / or other suitable computer processors. The computer 120 can include suitable memory, software, electronics, and / or device drivers for operating and / or controlling the various components of the system 100 (including the quality check station 114, the aspiration and dispensing station 116, and the analyzer stations 118A - 118C). For example, the computer 120 can control the movement of the carrier 110 between the loading area 106, around the track 112, between the quality check station 114, the aspiration and dispensing station 116, and the analyzer stations 118A - C, and between other stations and / or components of the system 100. One or more of the quality check station 114, the aspiration and dispensing station 116, and the analyzer stations 118A - C may be directly connected to the computer 120, or may communicate with the computer 120 via a network 122 such as a local area network (LAN), a wide area network (WAN), or other suitable communication networks including wired and wireless networks. The computer 120 may be housed as part of the system 100, or may be remote from the system 100.

[0024] In some embodiments, the computer 120 can be connected to a laboratory information system (LIS) database 124. The LIS database 124 can include, for example, patient information, tests to be performed on the biological sample, the date and time the biological sample was obtained, medical facility information, and / or tracking and routing information. Other information can also be included.

[0025] Computer 120 can be connected to a computer interface module (CIM) 126. CIM 126 and / or computer 120 can be connected to a display 128 that can include a graphical user interface. CIM 126, in conjunction with display 128, enables a user to access various control and status display screens and input data into computer 120. These control and status display screens can display and enable control of some or all aspects of quality check station 114, aspiration / dispensing station 116, and analyzer stations 118A - C for pre-screening, preparing, and analyzing a biological sample within sample container 102. CIM 126 can be used to facilitate interaction between the user and system 100. Using display 128, a menu including icons, scroll bars, boxes, and buttons can be displayed, through which a user (e.g., a system operator) can interface with system 100. The menu can include a number of functional elements programmed to display and / or operate the functional aspects of system 100.

[0026] Figures 2A and 2B each show a sample container 202A and 202B, and each figure represents the sample container 102 of FIG. 1. The sample containers 202A and 202B can include a blood collection tube, a test tube, a sample cup, a cuvette, or other containers such as transparent or translucent containers, including other containers that can hold a biological sample therein and allow the biological sample to be pre-screened, processed (e.g., aspirated), and analyzed. As shown in FIG. 2A, the sample container 202A can include a tube 230A and a cap 232A. The tube 230A can include thereon a label 234A that can indicate patient, sample, and / or test information in the form of a barcode, alphabetic characters, numeric characters, or a combination thereof. The tube 230A can contain therein a biological sample 236A that can include a serum or plasma portion 236SP, a sedimented blood portion 236SB, and a gel separator 216GA located therebetween. As shown in FIG. 2B, the sample container 202B, which can be structurally identical to the sample container 202A, can contain a homogeneous biological sample 236B therein, and the tube 230B has a gel bottom 236GB.

[0027] As will be described in more detail below, during the aspiration process, an improperly positioned probe contacting either the gel separator 236GA or the gel bottom 236GB (e.g., red blood cells) can result in an aspiration failure that can have adverse consequences such as incorrect test results and / or system downtime.

[0028] FIG. 3 shows an aspiration pipetting device 316 according to one or more embodiments. The aspiration pipetting device 316 can be or represent a part of the aspiration pipetting station 116 of the automated diagnostic analysis system 100. Optionally, the aspiration pipetting device 316 can be or be adjacent to a part of one or more of the analyzer stations 118A - 118C. Note that the methods and apparatus described herein for detecting or predicting aspiration failures can be used with other embodiments of aspiration pipetting devices.

[0029] The aspiration dispensing device 316 can aspirate a biological sample (e.g., sample 236A and / or 236B), a reagent, etc. and dispense it into the reaction vessel, enabling or facilitating the analysis of the biological sample at one or more analyzer stations 118A - C. The aspiration dispensing device 316 can include a robot 338 configured to move a probe assembly 340 within the aspiration dispensing station. The probe assembly 340 can include a probe 340P configured to aspirate a reagent 342 from, for example, a reagent packet 344 as shown. The probe assembly 340 can also be configured to aspirate a biological sample 336 from a sample container 302 located in the aspiration dispensing device 316 via, for example, an automatic track 112 (after its cap is removed as shown). A portion of the reagent 342, other reagents, and the sample 336 can be dispensed into a reaction vessel such as a cuvette 346 by the probe 340P. In some embodiments, the cuvette 346 can be configured to hold only a few microliters of liquid. Another portion of the biological sample 336 can be dispensed by the probe 340P into other cuvettes (not shown) together with other reagents or liquids.

[0030] The operation of some or all of the components of the aspiration dispensing device 316 can be controlled by a computer 320. The computer 320 can include a processor 320A and a memory 320B. The memory 320B can store a program 320C executable by the processor 320A. The program 320C can include algorithms for positioning the probe assembly 340 and for controlling and / or monitoring the aspiration and dispensing of liquid by the probe assembly 340. The program 320C can also include an artificial intelligence (AI) algorithm 320AI configured to detect or predict aspiration failures, as will be further described later. In some embodiments, the computer 320 can be a separate computing / control device connected to the computer 120 (system controller). In other embodiments, the configuration and functions of the computer 320 can be implemented in and executed by the computer 120. Also, in some embodiments, the positioning of the probe assembly and / or the aspiration / dispensing function of the probe assembly can be implemented in a separate computing / control device.

[0031] Robot 338 can include one or more robot arms 342, a first motor 344, and a second motor 346 configured to move probe assembly 340, for example, within aspiration dispensing station 116 of system 100. Robot arm 342 can be coupled to probe assembly 340 and first motor 344. First motor 344 can be controlled by computer 320 to move robot arm 342, and thus probe assembly 340, to a position above a liquid container. Second motor 346 can be coupled to robot arm 342 and probe assembly 340. Second motor 346 can also be controlled by computer 320 to move probe 340P vertically in and out of a liquid container and aspirate liquid therefrom or dispense liquid thereto. In some embodiments, robot 338 can also include one or more sensors 348, such as, for example, current, vibration, and / or position sensors, coupled to computer 320 to provide feedback and / or facilitate the operation of robot 338.

[0032] Aspiration dispensing device 316 can also include a pump 350 mechanically coupled to conduit 352 and controlled by computer 320. Pump 350 can generate a vacuum or negative pressure (e.g., aspiration pressure) in conduit 352 to aspirate liquid and can generate a positive pressure (e.g., dispensing pressure) in conduit 352 to dispense liquid.

[0033] The aspiration dispensing device 316 can further include a pressure sensor 354 configured to measure the aspiration and dispensing pressures within conduit 352 and generate pressure data accordingly. The pressure data can be received by computer 320 and used to control pump 350. The aspiration pressure measurement signal waveform (versus time) can be derived by computer 320 from the received pressure data and input into AI algorithm 320AI to detect or predict aspiration failures in probe assembly 340 during the aspiration process. The AI algorithm 320AI can also be trained using the aspiration pressure measurement signal waveform derived from the received pressure data from pressure sensor 354 to detect or predict aspiration failures. The pressure sensor 354 can be placed at any suitable location within the fluid path to sense pressure.

[0034] FIG. 4 shows a method 400 for detecting or predicting an aspiration failure in an automated diagnostic analysis system according to one or more embodiments. At process block 402, method 400 can begin by performing an aspiration pressure measurement via a pressure sensor while liquid is being aspirated in the automated diagnostic analysis system. For example, the aspiration pressure measurement can be performed by pressure sensor 354 of the aspiration dispensing device 316 (FIG. 3), which can be part of the aspiration dispensing station 116 of the automated diagnostic analysis system 100 (FIG. 1). Optionally, an aspiration dispensing device similar or identical to aspiration dispensing device 316 can be incorporated or disposed as part of one or more of analyzer stations 118A-118C.

[0035] In process block 404, method 400 can include analyzing the suction pressure measurement signal waveform via a processor configured to execute an artificial intelligence (AI) algorithm that performs either (A) cluster analysis of the suction pressure measurement signal waveform or (B) probabilistic graphical modeling based on the suction pressure measurement signal waveform. In some embodiments, the cluster analysis can include using only two metrics based on the suction pressure measurement signal waveform. In some embodiments, the probabilistic graphical modeling can implement two probabilistic graphical models that are executed simultaneously on the suction pressure measurement signal waveform. In other embodiments, the cluster analysis may include three or more metrics based on the suction pressure measurement signal waveform. In still other embodiments, the probabilistic graphical modeling may implement three or more probabilistic graphical models (one related to normal suction and each of the others related to a different type of abnormal suction) that are executed simultaneously on the suction pressure measurement signal waveform.

[0036] Analyzing the suction pressure measurement signal waveform to detect or predict a suction malfunction is based on distinguishable differences between the characteristics indicated by the pressure measurement signal waveform of normal suction and the characteristics indicated by the pressure measurement signal waveform of abnormal suction (representing a malfunction state).

[0037] For example, FIG. 5 shows a suction pressure signal versus time graph 500 of approximately 30 or more measured pressure signal waveforms of normal suction (substantially overlapping each other) according to one or more embodiments. Normal suction may have been performed, for example, in a suction dispensing device such as suction dispensing device 316 (of FIG. 3).

[0038] FIG. 6 shows a suction pressure signal versus time graph 600 of approximately 30 or more pressure signal waveforms of abnormal suction according to one or more embodiments. Abnormal suction may have occurred, for example, in a suction dispensing device such as suction dispensing device 316 (of FIG. 3).

[0039] The distinguishable characteristic differences between the signal waveform of graph 500 and the signal waveform of graph 600 can be detected in real time during the suction process by a trained AI algorithm, such as AI algorithm 320AI. The AI algorithm 320AI executable by the processor 320A can be implemented in any suitable form of artificial intelligence programming, including but not limited to neural networks, such as convolutional neural networks (CNNs), deep learning networks, recurrent networks, or other types of machine learning algorithms or models. Therefore, it should be noted that the AI algorithm 320AI is not, for example, a simple look-up table. Rather, the AI algorithm 320AI can be trained to detect or predict one or more types of suction obstacles and can improve (make more accurate decisions or predictions) without being explicitly programmed.

[0040] In some embodiments, the detection of the suction obstacle state by the AI algorithm 320AI can be based on a cluster analysis that uses only two metrics derived from the suction pressure measurement signal waveform during the suction process. These two metrics are used as predictors of suction obstacles. The detection threshold (having a classification boundary or range) is based on the two metrics, the training data, and in some embodiments, only K-means clustering. Other suitable clustering algorithms are also possible. The training data represents both normal suction and different types of abnormal suction. The time-varying classification range is established based on statistical measures.

[0041] The first of the two metrics is as follows:

Number

Number

[0042] The second of the two metrics is as follows:

Number

[0043] Metric 1 obtains the rate of change of pressure over time, and metric 2 obtains the inflection characteristics of the suction pressure waveform. Metrics 1 and 2 have been found to be reliable predictors of impending or early suction failure.

[0044] FIG. 8 shows a method 800 for training an AI algorithm 320AI for cluster - based analysis of a suction pressure signal waveform according to one or more embodiments. The training of the AI algorithm 320AI can be implemented offline. The determination of the optimal number of clusters and the tuning of parameters such as P(threshold), moving average filter parameters, ε, etc. can be done through post - training verification. The determination of a suitable sample rate for real - time sampling of pressure measurements can be based on the sensitivity to the time to failure detection. For the robustness of suction failure prediction, in some embodiments, multiple failure states can be counted over a set of time steps before identifying suction as abnormal.

[0045] Method 800 can begin with input data block 802 where a training set of aspiration pressure signal waveforms is provided. In process block 804, for each sampled aspiration pressure measurement, Metric 1 and Metric 2 are calculated.

[0046] In process block 806, for each of Metric 1 and 2, global statistics are calculated over a detection time window. Figures 5 and 6 show an example of the detection time window (see the outlined "sample aspiration phase"). The global statistics include upper and lower threshold values based on the global minimum and maximum values of the metric over the detection time window. For example, FIG. 9 shows a graph 900 of normal and abnormal aspiration pressure signal waveforms versus time, where upper threshold values 958 and 959 based on the global minimum and maximum values of Metric 1 and lower threshold values 960 and 961 based on the global minimum and maximum values of Metric 1 are calculated according to one or more embodiments (note that lower threshold value 960 happens to be approximately the same as upper threshold value 959). FIG. 10 shows a graph 1000 of normal and abnormal aspiration pressure signal waveforms versus time, where upper threshold values 1058 and 1059 based on the global minimum and maximum values of Metric 2 and lower threshold values 1060 and 1061 based on the global minimum and maximum values of Metric 2 are calculated according to one or more embodiments.

[0047] In block 808, to identify the clusters corresponding to normal aspiration, clustering is performed on the sample aspirations based on the global statistics calculated in process block 806. FIG. 11 shows a 4 - cluster classification 1100 based on Metric 1 according to one or more embodiments, and FIG. 12 shows a 4 - cluster classification 1200 based on Metric 2 according to one or more embodiments. As shown in FIGS. 11 and 12, clustering based on the global minimum and maximum values of Metric 1 and Metric 2 effectively separates normal aspiration samples, such as those shown in Cluster 1, from abnormal aspiration samples, such as those shown in Clusters 2, 3, and 4.

[0048] In block 810, method 800 can include calculating statistics (e.g., mean and standard deviation) for each of metrics 1 and 2 for the normal suction cluster (cluster 1) at each instant. These statistics can be used to classify samples as normal or abnormal. Normal suction samples exhibit the lowest variation in mean and standard deviation among all samples.

[0049] In block 812, method 800 can include calculating normal suction statistics for each of metrics 1 and 2 for cluster 1 at each instant to determine a classification range. The following statistical procedure can be used to establish the range corresponding to the class (cluster 1) of normal suction for metrics 1 and 2 at each time sample: Metric UpperLimit (t) = Metric 75percentile (t) + α Metric IQR Metric LowerLimit (t) = Metric 25percentile (t) - α Metric IQR (t) where Metric IQR (t) = Metric 75percentile (t) - Metric 25percentile (t) (Note: t = time) Here, α = 1.5 or 3 (the optimal value of α can be adjusted by validation after initial training), the sampling time Ts = 1 / fs, where fs is the sampling rate of the pressure measurement during the suction process. A suitable sampling time can be, for example, 1 millisecond. Other suitable sampling rates can be used.

[0050] Figure 13 shows a baseline classification zone or range 1300 of metric 1, cluster 1, based on calculations performed in process block 812 with α = 3, according to one or more embodiments. The baseline classification range 1300 includes an upper bound curve 1362, an average curve 1363, and a lower bound curve 1364.

[0051] Figure 14 shows a baseline classification zone or range 1400 of metric 2, cluster 1, based on calculations performed in process block 812 with α = 3, according to one or more embodiments. The baseline classification range 1400 includes an upper bound curve 1462, an average curve 1463, and a lower bound curve 1464.

[0052] Note that the baseline classification ranges 1300 and 1400 are determined and stored in a non-parametric form, as shown. In a first alternative embodiment, the baseline classification range can be parameterized by a global representation using polynomials, B-splines, autoregressive moving average (ARMA) models, or other suitable basis functions. In a second alternative embodiment, the suction phase can be subdivided into sub-phases (e.g., four), and the baseline classification ranges over each sub-phase can be individually parameterized via a local representation using polynomials, B-splines, ARMA models, or other suitable basis functions. The parametric forms of the first and second alternative embodiments can reduce the memory required compared to the non-parametric form, but there may be additional computational costs.

[0053] Once established, the baseline classification ranges 1300 and 1400 can be used in cluster analysis of the suction pressure measurement signal waveform to detect / predict suction disturbances, as will be described later.

[0054] Returning to FIG. 4, method 400 can continue in process block 406 by identifying and responding to a suction obstruction via a processor in response to the clustering analysis performed in process block 404. The suction obstruction can be identified by determining for each sampled instant of the suction pressure measurement signal waveform, t = t n ∈ detection time window For each of metrics 1 and 2, metric i (t = t n ) ≤ metric i,UpperLimit and metric i (t = t n ) ≥ metric i,LowerLimit whichever is the case.

[0055] If the above conditions are not met within the detection window, the suction can be identified as abnormal, the suction process can be terminated, and / or system procedures for an error state can be followed.

[0056] In other embodiments, instead of performing cluster analysis in process block 404 as described above, the analysis performed in process block 404 can alternatively include probabilistic graphical modeling based on the aspiration pressure measurement signal waveform, in which case two probabilistic graphical models are executed simultaneously for the aspiration pressure measurement signal waveform. To model the dynamics of sample aspiration, a Hidden Markov Model (HMM) can be used. The training data sets for normal and abnormal aspiration samples are first identified either in a supervised manner (i.e., expert-based labeling of the data) or in a supervised manner using a machine learning method such as K-means clustering described above. Since the state transitions during sample aspiration are sequential such that a state transition can occur from the current state value to an adjacent higher state value, a left-to-right HMM architecture can be used. The HMM model is trained using the Expectation-Maximization (EM) algorithm for separate HMM models, i.e., one for normal aspiration and one for abnormal aspiration. The "expectation" step of the EM algorithm is applied in the form of the Baum-Welch (forward-backward) algorithm. Once training is complete, the AI algorithm 320AI is configured to simultaneously execute the normal aspiration HMM and the abnormal aspiration HMM in real time for the measured aspiration pressure signal waveform. The sequence output probabilities over consecutive sequences of pressure signal values are calculated using both HMMs. The classification as normal or abnormal aspiration is based on the relative sequence likelihood: P SEQUENCE,NORMAL / P SEQUENCE,ABNORMAL It may also be based on a comparison of [[ID=]], or alternatively, for normal and abnormal aspirations, it can also be by comparing sequence likelihoods using normal HMMs and abnormal HMMs for their respective threshold sets. Training can be performed in an offline mode using training data sets for normal and abnormal aspirations. Once trained, the HMM model can be implemented online in firmware or software DML (Definitive Media Library) for real-time aspiration disorder detection / prediction.

[0057] Figures 15A and 15B show two-cluster classifications 1500A, 1500B for each metric according to one or more embodiments, with the first cluster of each metric shown in Figure 15A and the second cluster of each metric shown in Figure 15B. Training of the normal HMM and abnormal HMM includes the following: (1) In some embodiments, detection time windows 1566A and 1566B, which may each be 200 milliseconds (other detection time windows may be used), and calculating metric 1 and metric 2 for each of cluster 1 (normal aspiration) and cluster 2 (abnormal aspiration) within a predetermined detection time window; (2) calculating the average vector of the metrics for normal aspiration samples (cluster 1) and subtracting the calculated average vector from the average vector of each aspiration sample in the training set of the samples; and (3) creating a discrete set of quantized (integer-valued) ranges of the output by defining the number of outputs / outputs (i.e., evenly dividing the minimum-maximum range of the metric residuals from step (2)).

[0058] In some embodiments, detection or prediction of aspiration disorders can each be performed by an HMM having a left-to-right architecture 1600 as shown in Figure 16, having six states k = 1 to 6, constrained state transitions, twelve output states, and a 15-time step sequence.

[0059] In an alternative embodiment, instead of running the HMM over the entire detection time window of the suction process, the detection time window can be subdivided into sub-phases (e.g., four), and for each sub-phase, separate HMM models for normal and abnormal suction can be trained and then run against the suction pressure measured to detect / predict a suction disorder. By subdividing the detection time window into sub-phases, the complexity and / or size of each of the HMMs can be reduced, and thus the overall computational cost can be reduced.

[0060] Returning to FIG. 4, method 400 can continue in process block 406 by identifying and responding to a suction disorder via a processor in response to the probabilistic graphical modeling performed in process block 404. The suction disorder can be identified by running two HMMs simultaneously, one trained for normal suction and the other trained for abnormal suction. In some embodiments, as follows, to identify normal suction, at each sampled instant of the suction pressure measurement signal waveform, first and second threshold values of sequence likelihood can be applied: P SEQUENCE,NORMAL > first threshold value; and P SEQUENCE,ABNORMAL < second threshold value; where, in some embodiments, the first threshold value may be 0.90 and the second threshold value may be 0.50 (other suitable values may be used for the first and second threshold values).

[0061] The sequence likelihood is calculated based on the composite probability of an observation sequence (length = 15) conditioned on the corresponding known state sequence.

[0062] Advantageously, both the cluster analysis and the probabilistic graphical modeling as described herein detected and / or predicted aspiration failures within the first 100 milliseconds from the start of the aspiration process (see, e.g., FIG. 17 showing a histogram 1700 of aspiration failures detected among a dataset of aspiration samples according to one or more embodiments). This early real-time aspiration failure detection or prediction advantageously enables timely termination of the aspiration process and / or execution of suitable error state procedures so as to avoid or minimize any possible consequences downstream of a faulty aspiration, such as, for example, equipment downtime and / or incorrect analysis results.

[0063] Furthermore, it has been found that both the cluster analysis and the probabilistic graphical modeling advantageously perform an efficient and accurate binary classification of normal or abnormal sample aspiration in real time. Both embodiments have low computational complexity (O(N)) and memory requirements during online execution and can thus be easily implemented in firmware or software. Although the computational cost may be high during the training phase of the AI algorithm 320AI (of FIG. 3) for both embodiments, the training phase can be executed offline.

[0064] Note that the methods and apparatuses described herein are not limited to any particular type of aspiration failure. For example, in addition to insufficient aspiration volume and aspiration failure due to collection of gel or unwanted substances, provided that there are a sufficient number of sample aspiration pressure waveform samples (i.e., training data) for the particular type of aspiration failure to be detected / predicted, other failures caused by, for example, inaccurate titration by an aspiration pump, software-related errors during titration, deterioration of the flow state of the fluidic manifold upstream of the probe, and / or electrical noise and / or environmental effects that affect titration can also be detected / predicted. And by applying the methods and apparatuses described herein, distinct metric profiles associated with each type of aspiration failure can be identified. Thus, the type and number of metrics used to detect / predict aspiration failure can be based on the particular aspiration profile of the particular aspiration failure to be detected / predicted. New metrics can be derived for any new aspiration profile. Similarly, the number of cluster classifications selected for analysis is determined by the number of predicted categories of the state and / or type of failure, as well as the availability of training data and the ability of the clustering analysis to classify the different aspiration failure states present in the training data within an acceptable accuracy range. Thus, for example, if a higher level of false negatives is acceptable, the 4-cluster classification described above can be reduced to a 3-cluster classification, for example. Thus, the methods and apparatuses described herein are not limited to any particular type or number of metrics and / or clusters for detecting / predicting aspiration failure.

[0065] While the present disclosure is susceptible to various modifications and alternative forms, specific method and apparatus embodiments have been shown by way of example in the drawings and are described in detail herein. However, it should be understood that the specific methods and apparatuses disclosed herein are not intended to limit the present disclosure or the claims below.

Claims

1. A method for detecting or predicting aspiration disorders in an automated diagnostic analysis system, comprising: In an automated diagnostic analysis system, performing aspiration pressure measurement via a pressure sensor while liquid is being aspirated; Analyzing the aspiration pressure measurement signal waveform via a processor configured to execute an artificial intelligence (AI) algorithm configured to perform: Cluster analysis of the aspiration pressure measurement signal waveform, or Probabilistic graphical modeling based on the aspiration pressure measurement signal waveform; In response to said analyzing, identifying an aspiration disorder via the processor and responding thereto; wherein the AI algorithm is configured to perform cluster analysis of the aspiration pressure measurement signal waveform or probabilistic graphical modeling based on the aspiration pressure measurement signal waveform; wherein the cluster analysis comprises using only two metrics based on the aspiration pressure measurement signal waveform; wherein a first metric of the only two metrics is to obtain a rate of change of pressure of the aspiration pressure measurement signal waveform including a moving average of the aspiration pressure gradient, and a second metric of the only two metrics is to obtain an inflection characteristic of the aspiration pressure measurement signal waveform; wherein the probabilistic graphical modeling comprises two probabilistic graphical models executed simultaneously on the aspiration pressure measurement signal waveform; wherein a first probabilistic graphical model of the two probabilistic graphical models is trained using normal aspiration data, and a second probabilistic graphical model of the two probabilistic graphical models is trained using abnormal aspiration data.

2. The method of claim 1, wherein the cluster analysis is based on unsupervised training data, or the probabilistic graphical modeling is based on supervised or unsupervised training data.

3. The method of claim 1, wherein the cluster analysis is based on labeled training data using a supervised learning algorithm to establish a threshold having a classification range.

4. The method of claim 1, wherein the cluster analysis comprises K-means clustering using a four-cluster classification based on only two metrics.

5. The method of claim 1, wherein the aspiration disorder is determined based on a comparison between an output from the first probabilistic graphical model and an output from the second probabilistic graphical model.

6.

7.

8. The probabilistic graphical modeling is the method according to claim 1, including a left-to-right hidden Markov model (HMM) architecture.

7. The left-to-right HMM architecture is: Six states; Twelve output states; A 15-time step sequence, The method according to claim 6, including.

8. Specifically responding further includes identifying and responding to a suction failure via a processor within 100 milliseconds after the liquid starts to be suctioned, the method according to claim 1.

9. The suction failure is the collection of gel or undesirable substances or the suction with insufficient suction volume, the method according to claim 1.

10. An automatic suction pipetting device, comprising: A robotic arm; A probe connected to the robotic arm; A pump connected to the probe; A pressure sensor configured to perform a suction pressure measurement when liquid is being suctioned through the probe; A processor configured to execute an artificial intelligence (AI) algorithm to detect or predict a suction failure during the suction process and respond thereto, the AI algorithm being configured to analyze a suction pressure measurement signal waveform derived from the pressure sensor using cluster analysis or probabilistic graphical modeling; Including, The cluster analysis includes using only two metrics based on the suction pressure measurement signal waveform, The first metric of the only two metrics is to obtain the rate of change of pressure of the suction pressure measurement signal waveform including the moving average of the suction pressure gradient, and the second metric of the only two metrics is to obtain the inflection characteristics of the suction pressure measurement signal waveform, The probabilistic graphical modeling includes two probabilistic graphical models executed simultaneously for the suction pressure measurement signal waveform, The first probabilistic graphical model of the two probabilistic graphical models is trained using normal suction data; and The second probabilistic graphical model of the two probabilistic graphical models is trained using abnormal suction data, the automatic suction pipetting device.

11. The cluster analysis includes K-means clustering using 4-cluster classification based on only two metrics, the automatic suction pipetting device according to claim 10.

12. The cluster analysis uses a plurality of cluster classifications, one of the plurality of cluster classifications represents normal aspiration data, and each of the other of the plurality of cluster classifications represents different types of abnormal aspiration data. The automatic aspiration dispensing device according to claim 10.

13. The probabilistic graphical modeling includes a plurality of probabilistic graphical models that are simultaneously executed on the aspiration pressure measurement signal waveform. One of the plurality of probabilistic graphical models is trained using normal aspiration data, and each of the other of the plurality of probabilistic graphical models is trained using different types of abnormal aspiration data. The automatic aspiration dispensing device according to claim 10.

14. The probabilistic graphical modeling includes a left-to-right hidden Markov model (HMM) architecture. The automatic aspiration dispensing device according to claim 10.

15. The left-to-right HMM architecture includes: Six states; Twelve output states; A 15 time step sequence, The automatic aspiration dispensing device according to claim 14.

16. The processor is configured to identify and respond to an aspiration failure within 100 milliseconds after liquid begins to be aspirated during the aspiration process by executing an AI algorithm. The automatic aspiration dispensing device according to claim 10.

17. An automatic diagnostic analysis system comprising: The automatic aspiration dispensing device according to claim 10; One or more analyzer stations for analyzing a biological sample; An automatic track for transporting sample containers and reaction vessels to and from the automatic aspiration dispensing device and one or more analyzer stations, The automatic diagnostic analysis system including the above.

18. A non-transitory computer-readable storage medium including an artificial intelligence (AI) algorithm configured to detect or predict an aspiration failure based on an analysis of an aspiration pressure measurement signal waveform using cluster analysis of the aspiration pressure measurement signal waveform or using probabilistic graphical modeling based on the aspiration pressure measurement signal waveform, The cluster analysis includes using only two metrics based on the aspiration pressure measurement signal waveform. Of the only two metrics, the first metric is to obtain the rate of change of pressure of the suction pressure measurement signal waveform including the moving average of the suction pressure gradient, and the second metric of the only two metrics is to obtain the inflection characteristics of the suction pressure measurement signal waveform. The probabilistic graphical modeling includes two probabilistic graphical models that are simultaneously executed on the suction pressure measurement signal waveform. The first probabilistic graphical model of the two probabilistic graphical models is trained using normal suction data; and The second probabilistic graphical model of the two probabilistic graphical models is trained using abnormal suction data, the non-transitory computer-readable storage medium.

Images