Assessing the quality of a fluid process using machine learning

Abstract

A system for assessing a quality of a fluid process performed on a fluid including one of a fluid aspiration and a fluid dispense is provided. The system includes pump; a probe; a connection, wherein the pump and the probe are attached to the connection and a fluid path is formed between the pump and the probe at least in part by the connection; a pressure sensor in fluid communication with the fluid path and configured to sense pressures of the pump during the fluid process, wherein a pressure measurement data comprises the sensed pressures of the pump during the fluid process; a processor; and a memory. The memory comprising instructions that are executed by the processor to cause the processor to receive pressure measurement data during the fluid process, and determine, using a trained machine learning algorithm, the quality of the fluid process.

Description

ASSESSING THE QUALITY OF A FLUID PROCESS USING MACHINE LEARNINGCROSS REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 517,864, entitled "ASSESSING THE QUALITY OF A FLUID PROCESS USING MACHINE LEARNING"’ filed August 4, 2023. the disclosure of which is hereby incorporated by reference in its entirety for all purposes.TECHNICAL FIELD

[0002] The present disclosure relates generally to assessing the quality of a fluid processes and, in particular, to assessing the quality of a fluid aspiration or a fluid dispense using machine learning.BACKGROUND

[0003] In laboratory diagnostic instruments, a human sample is mixed with chemical reagents and the chemical reaction is studied / measured to predict the analyte or the health condition of the patient. The reagents are stored in small plastic reagent packs whereas the sample is stored in test tubes. A probe connected to a pump is used to aspirate or “draw’’ the required volume of reagent from the reagent pack or the sample from the test tube. Since the volume of the reagent or sample drawn is critical for a successful diagnostic test, a pressure sensor is used to prepare a pressure vs. time curve. The shape of the pressure vs. time curve is used to predict if the full volume was aspirated or whether only a part of the volume was aspirated with this condition being called a “short” aspiration. This monitoring of the pressure is used to reduce the risk of reporting an incorrect patient result due to incorrect reaction volumes.

[0004] Conventional methods identify key points or regions on the pressure curve. Those points may then be directly compared to a threshold value (such as a minimum pressure) or used in a calculation to determine a value that can then also be compared to a threshold value(s). For example, a best-fit slope may be calculated across a region of the curve, or a comparison of a difference between points (or averages of points) at two or more regions of the pressure curve. Often each curve is evaluated against multiple criteria, each of which is optimized to find a specific feature or failure mode.

[0005] Some of the aspiration quality assessment techniques include a method detecting aspiration quality’ is to compare the aspiration pressure signal with that of a predetermined value of air or other liquid or gas, such as disclosed in U.S. Patent No. 7,867,769, which is hereby incorporated by reference herein in its entirety. Another is to analyze the pressure signal to detect abnormalities, such as disclosed in U.S. Patent Nos. 6,370,942 and 7,634,378, which are hereby incorporated by reference herein in their entireties. FIG. 1 illustrates how a pressure trace may be analyzed to assess the quality of an aspiration, according to a conventional method. FIG. 2 is pressure trace annotated with examples of using data from the pressure trace to calculate aspiration quality values and using them to assess aspiration quality, according to a conventional method. FIG. 3 is a pair of calculated parameters (metrics) from a group of pressure traces annotated to illustrate how parameters may be analyzed to assess the quality of the fluid process, according to a conventional method. Specifically, FIG. 3 is an example of two criteria, a “clog” metric and a “viscosity” metric used in combination to determine if a given aspiration is acceptable or not.

[0006] Typically, the thresholds (i.e., limits) used for comparison to evaluate a pressure curve to yield a pass / fail result are determined empirically. In other words, a large number (tens to thousands) of pressure curves are generated at each condition, and a statistical distribution of the calculated values is established. These distributions can be used to determine a pass / fail limit that has an appropriate level of consumer / producer risk. This determination must be performed for each set of conditions (e.g., flow rate, volume, sample type).

[0007] It is desired to provide a faster and more accurate quality assessment of fluid processes than conventional methods.

[0008] The present disclosure is directed to overcoming these and other problems of the prior art.SUMMARY

[0009] Embodiments of the present invention address and overcome one or more of the above shortcomings and drawbacks, by providing systems, methods, and computer program products for assessing the quality of a fluid aspiration or a fluid dispense using machine learning. Additional features and advantages of the invention will be made apparent from thefollowing detailed description of illustrative embodiments that proceeds with reference to the accompanying drawings.

[0010] In an exemplary embodiment, a system for assessing a quality of a fluid process performed on a fluid including one of a fluid aspiration and a fluid dispense is provided. The system can include pump; a probe; a connection, wherein the pump and the probe are attached to the connection and a fluid path is formed between the pump and the probe at least in part by the connection; a pressure sensor in fluid communication with the fluid path and configured to sense pressures of the pump during the fluid process, wherein a pressure measurement data comprises the sensed pressures of the pump during the fluid process; a processor; and a memory. The memory comprising instructions that are executed by the processor to cause the processor to receive, from the pressure sensor, pressure measurement data during the fluid process, wherein the pressure measurement data comprises pressure data and time data, and determine, by the processor using a trained machine learning algorithm, the quality of the fluid process based on the received pressure measurement data.

[0011] In some embodiments, the system further comprises an analyzer configured to analyze a mixture comprising the fluid when the quality of the fluid process is satisfactory. In some embodiments, the instructions further cause the processor to cause the pump to operate such that a subsequent fluid process is performed in response to determining that the quality of the fluid process is abnormal. In some embodiments, the trained machine learning algorithm comprises a sensitivity threshold, wherein the instructions further cause the processor to adjust the sensitivity threshold prior to determining the quality of the fluid process. In some embodiments, the trained machine learning algorithm comprises a machine learning algorithm trained on a training set, and the training set comprises a plurality of pressure traces of normal fluid processes and a plurality of pressure traces of abnormal fluid processes at a ratio of at least 5: 1.

[0012] In some embodiments, the trained machine learning algorithm comprises a machine learning algorithm trained on a training set, and the training set comprises a plurality of pressure traces conforming to a common domain such that the trained machine learning algorithm can be used to determine fluid process qualities of fluid processes performed on a plurality of volumes of fluids that also conform to the common domain. In some embodiments, each of the plurality7of pressure traces comprises a plurality of data points, wherein the common domain comprises a number of data points. In some embodiments, thecommon domain comprises one of a maximum time value and a minimum pressure value. In some embodiments, the training set comprises a plurality of pressure traces of aspirations and a plurality of pressure traces of dispenses.

[0013] In some embodiments, trained machine learning algorithm comprises a machine learning algorithm trained on training set, and the training set comprises at least one modified pressure trace, wherein the at least one modified pressure trace comprises an unmodified pressure trace having at least one portion excluded. In some embodiments, the at least one portion excluded comprises one of a portion associated with pump deceleration and a portion associated with pump acceleration. In some embodiments, the trained machine learning algorithm comprises a machine learning algorithm trained on training set, and the training set comprises at least one modified pressure trace, wherein the at least one modified pressure trace comprises a subset of an unmodified pressure trace associated with an information-rich portion of the unmodified pressure trace.

[0014] In another exemplary embodiment, a method of quality assessment of a fluid process is provided. The method includes performing, by a fluid processing system, the fluid process on a fluid, wherein the fluid process comprises one of a fluid aspiration and a fluid dispense; receiving, by a processor, pressure measurement data during the fluid process, wherein the pressure measurement data comprises pressure data and time data; and determining, by the processor using a trained machine learning algorithm, the quality assessment of the fluid process based on the received pressure measurement data.

[0015] In some embodiments, the method further includes analyzing a mixture comprising the fluid in response to determining that the quality assessment of the fluid process is satisfactory. In some embodiments, the method further includes aspirating, by the fluid processing system, an additional fluid process in response to determining the quality assessment of the fluid process is abnormal. In some embodiments, the trained machine learning algorithm comprises a sensitivity threshold, and the method further includes adjusting the sensitivity threshold prior to determining the quality assessment of the fluid process. In some embodiments, the method further includes modifying the pressure measurement data to conform to a common domain, wherein the trained machine learning algorithm can be used to determine fluid process qualities of the fluid process performed on a plurality of volumes of fluids that conform to the common domain.

[0016] In some embodiments, the trained machine learning algorithm comprises a machine learning algorithm trained on a training set, and the training set comprises a plurality of pressure traces of normal fluid processes and a plurality of pressure traces of abnormal fluid processes at a ratio of at least 5: 1. In some embodiments, the trained machine learning algorithm comprises a machine learning algorithm trained on training set, the training set comprises at least one modified pressure trace, and the at least one modified pressure trace comprises one of an unmodified pressure trace having at least one portion excluded and a subset of an unmodified pressure trace associated w ith an information-rich portion of the unmodified pressure trace.

[0017] In yet another exemplary embodiment, a computer program product embodied in a computer readable storage medium is provided. The computer readable storage medium comprises software that when executed by a processor cause the processor to receive, from a pressure sensor, pressure measurement data during a fluid process performed on a fluid, wherein the pressure measurement data comprises pressure data and time data; and determine, by the processor using a trained machine learning algorithm, a quality assessment of the fluid process based on the received pressure measurement data.

[0018] This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter. Additional features and advantages of the disclosed technology’ will be made apparent from the following detailed description of illustrative embodiments that proceeds with reference to the accompanying drawings.BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The foregoing and other aspects of the present invention are best understood from the following detailed description when read in connection with the accompanying drawings. For the purpose of illustrating the invention, there are shown in the drawings embodiments that are presently preferred, it being understood, however, that the invention is not limited to the specific instrumentalities disclosed. Included in the drawings are the following Figures:

[0020] FIG. 1 illustrates how a pressure trace may be analyzed to assess the qualify of an aspiration, according to a conventional method;

[0021] FIG. 2 is pressure trace annotated with examples of using data from the pressure trace to calculate aspiration quality values and using them to assess aspiration quality’, according to a conventional method;

[0022] FIG. 3 is select pressure trace metrics annotated to illustrate how a pressure trace may be analyzed to assess the quality of the fluid process, according to a conventional method;

[0023] FIG. 4 is an aspiration system with pressure monitoring, according to an embodiment of the disclosure;

[0024] FIG. 5 is a graph of mismatch probability, according to an embodiment of the disclosure;

[0025] FIG. 6 is a graph of representative pressure traces for various aspiration volumes and at different aspiration speeds, according to an embodiment of the disclosure;

[0026] FIG. 7 illustrates numerous aspirations of air and liquid, according to an embodiment of the disclosure;

[0027] FIG. 8 is a flow chart a method of assessing the quality7of a fluid process using machine learning, according to an embodiment of the disclosure;

[0028] FIGS. 9A-9C are graphs showing an unmodified pressure trace and a pressure trace modified using linear interpolation, according to an embodiment of the disclosure;

[0029] FIG. 10 is a table of linear interpolation modifying pressure vs. time data, according to an embodiment of the disclosure;

[0030] FIG. 11 is a graph of the unmodified pressure trace and modified pressure trace from FIG. 10, according to an embodiment of the disclosure;

[0031] FIGS. 12A and 12B are a diagram of the CNN used in the Use Case described herein, according to an embodiment of the disclosure;

[0032] FIG. 13 is a confusion matrix of the output of the CNN on the test dataset, according to an embodiment of the disclosure;

[0033] FIG. 14A and 14B is a comparison between an embodiment of the CNN algorithm and a conventional solution from the CI 1900 I A Reagent Volume Check, according to an embodiment of the disclosure;

[0034] FIGS. 15A and 15B is an example of classified traces, according to an embodiment of the disclosure; and

[0035] FIG. 16 illustrates an exemplary computing environment within which embodiments of the invention may be implemented.DETAILED DESCRIPTION

[0036] Independent of the grammatical term usage, individuals with male, female or other gender identities are included with this term.

[0037] It is noted that a “pressure trace’' is typically defined as a graph of pressure as a function of time. However, for ease of drafting, as used herein, a “pressure trace” is to be understood as either a graph of pressure as a function of time or the pressure and time data that could be plotted to create a graph. Thus, for example, when the present disclosure describes that a machine learning algorithm is trained on pressure traces, it is to be understood that the machine learning algorithm is trained on pressure trace graphs, the pressure and time data that could be plotted to create a graph, or both. Similarly, when the present disclosure describes that a machine learning algorithm analyzes a pressure trace, it is to be understood that the machine learning algorithm analyzes a pressure trace graph, the pressure and time data that could be plotted to create a graph, or both. For references to a pressure trace’s “shape” with respect to pressure and time data that could be plotted to create a graph (but isn’t), its shape is the shape that would results if the data were plotted.

[0038] In some embodiments, the systems, methods, and computer program products described herein can be used, for example, in an in vitro diagnostics (“IVD”) environment, using IVD equipment, or as part of IVD methods. For example, in some embodiments, the systems, methods, and computer program products described herein can be used with diagnostic or chemistry analyzers like automated clinical diagnostic analyzers and automated clinical chemistry analyzers. These analyzers can process hundreds of thousands of human sample diagnostic tests per year. These tests can be prescribed to a patient by a Primary CarePhysician (e.g., general health checkup), a specialist (e.g.. heart, cancer), and in a hospital setting (e.g., prior to treatment or surgery).

[0039] Determining the quality of a metered sample or reagent transfer is an important function of a clinical diagnostic instrument, for example, in an IVD environment. Typical failure modes that are desirable to detect include probe clogs and insufficient sample or reagent in the liquid vessel, among others. A common method for determining the quality of fluid process, e.g., an aspiration or dispense, is monitoring the pressure in the aspiration system (such as illustrated in FIG. 4) during a fluid process, analyzing that pressure, and assessing of the quality of the fluid process. FIG. 4 is a fluid process system with pressure monitoring, according to an embodiment of the disclosure. In some embodiments, the fluid process system 400 can include a pump 401, tubing 402, a pressure transducer 403, a probe 404, and a liquid vessel 405. The fluid process system 400 can be used for one or more fluid processes, e.g., fluid aspiration or fluid dispense.

[0040] The present disclosure describes systems, methods, and computer program products for assessing the quality of a fluid process using machine learning. In some embodiments, a trained machine learning algorithm is used to assess the uality of a fluid process like a fluid aspiration or a fluid dispense, performed using as aspiration system by analyzing the fluid process’ pressure trace.

[0041] It is noted that a “pressure trace’' is ty pically defined as a graph of pressure as a function of time. However, for ease of drafting, as used herein, a “pressure trace” is to be understood as either a graph of pressure as a function of time or the pressure and time data that could be plotted to create a graph. Thus, for example, when the present disclosure describes that a machine learning algorithm is trained on pressure traces, it is to be understood that the machine learning algorithm is trained on pressure trace graphs, the pressure and time data that could be plotted to create a graph, or both. Similarly, when the present disclosure describes that a machine learning algorithm analyzes a pressure trace, it is to be understood that the machine learning algorithm analyzes a pressure trace graph, the pressure and time data that could be plotted to create a graph, or both.

[0042] Using a machine learning algorithm to make this assessment can have several advantages over conventional methods: It can be faster (e.g., more than one assessment per one second) and more accurate. It can provide objective results that do not rely on a technicalexpert’s analysis, which can be subjective. It can provide objective results in contrast to “feature engineering” in which a technical expert is required to determine discriminative features in the data for the algorithm to process. Feature engineering can be subjective, unavoidably limited in scope, and inherit the biases of the technical expert.

[0043] Another advantage is that the machine learning algorithm can continue to learn and improve. For example, in some embodiments, the machine learning algorithm can use unsupervised training and / or reinforcement learning to learn and improve as it is used. Alternatively or additionally, in some embodiments, the machine learning algorithm can be retrained with additional training examples for which ground truth is known. For example, as more and more pressure traces become available, these pressure traces can be added to the training set used to train the machine learning algorithm. Thus, the machine learning algorithm can evolve with additional training and continues to improve over time. In some embodiments, this training can be controlled by one or more subject matter experts so that the system evolution can improve prediction accuracy.

[0044] In addition, using a machine learning algorithm to assess the quality of a fluid process can provide a robust solution that can be minimally affected, or not affected at all, by variations in pressure traces caused by, for example, differences in the fluid process equipment. Because some conventional solutions use static values of slopes and plateaus on the pressure traces they are susceptible to variations in the pressure traces even for the same aspiration volume. In other words, the variations in the pressure traces can negatively affect the accuracy of conventional solutions. The variations in pressure traces can occur due to, for example, variations in the viscosity of various reagents, variations in the pressure sensors, variations in the fluidics hardware (e.g., probes, pumps, tubing, fittings), variations in the customer environment (e.g., laboratory temperature, altitude of the hospital laboratory etc.) and manufacturing variability of all the components and subsystems. Since machine learning algorithms use training data that has been manually processed, such variations can be injected as noise or edge cases during training. This can improve robustness.

[0045] Another advantage of using a machine learning algorithm is that it provides the opportunity’ for users to adjust the sensitivity’ of the analysis, which is not practical, and in some cases not possible, using existing methods. A machine learning algorithm, e.g., Convolutional Neural Network (“CNN”), can determine the category’ for the output of the input pressure trace by determining a probability' of that pressure trace falling into eachoutput category. This probability is in essence how strongly the algorithm calculates the categorization. For example, a pressure trace that very closely fits the "‘normal” profile might have a probability of 99.99%, whereas a pressure trace with some deviations might have a probability of 60%. In some embodiments, whichever probability out of all the categories is greatest results in the determined category. However, using a machine learning algorithm allows for the implementation of an additional feature: algorithm sensitivity. By adjusting the threshold probability for a “normal” (or. alternatively, “abnormal”) classification, the detection of abnormal samples can be made more or less sensitive. A default “factory” setting may be determined, for example, 80%. In this case, a sample can only be determined to be “normal” if the algorithm probability is greater than 80%. However, since each testing laboratory has their own requirements and risk tolerance, as well as typical sample quality, they may wish for the algorithm to have higher or lower sensitivity. Thus, the user may choose to adjust the sensitivity (in some embodiments, within pre-defined boundaries). For higher sensitivity, the level may be raised to, for example, 90%. Likewise, for lower sensitivity, the level may be reduced to, for example, 70%. This is a capability that is difficult to implement in conventional solutions because of the number of interacting thresholds involved but can be an advantageous feature versus a one-size-fits-all approach.

[0046] To illustrate this, consider FIG. 5. FIG. 5 is a graph of mismatch probability, according to an embodiment of the disclosure. FIG. 5 shows the probabilities determined by a machine learning algorithm, using 50% as the sensitive threshold. Light great is data predicted correctly (matches the “truth”), dark grey are the mismatches. The Y-axis is the algorithm calculated probability of the trace being a member of the class, and the x-axis are the two classes (False = OK. True = Abnormal). This graph illustrates how adjusting the probability threshold up or down will catch more mismatches in one class or the other, but also then create more of the opposite type of mismatch (producer / consumer risk).

[0047] A user may wish to adjust the machine learning algorithm's sensitivity for any number of reasons: the user's risk tolerance in general, the machine learning algorithm has more false negatives compared to false positives (or vice versa), the volume of fluid, and the sample’s quality control, for example. A user may wish to adjust the machine learning algorithm based on her risk tolerance. For example, in some situations, the consequence that the quality assessment is incorrect is small. In such situations, a user can lower the sensitivitythreshold, e.g., to 40%. In other situations, the consequence that the quality assessment is incorrect is great. In such situations, a user can raise the sensitivity threshold, e.g.. to 80%.

[0048] Additionally, a user may wish to adjust the machine learning algorithm because the machine learning algorithm has more false negatives compared to false positives (or vice versa). Based on the specific conditions in a particular hospital laboratory, it is possible that the machine learning algorithm may have more false negatives (i.e., full aspirations detected as shorts) compared to false positives (i.e., shorts detected as full aspirations) or vice versa. Use of machine learning approach allows the possibility' of letting the customer or field service engineer choose this balance (i.e., the inconvenience of repeating a good aspiration versus risking a short reaction volume), in some embodiments within a predetermined sensitivity band. The user can adjust the sensitivity to optimize performance.

[0049] In addition, a user may wish to adjust the machine learning algorithm’s sensitivity based on the volume of fluid. For example, the user can require greater sensitivity for larger volumes and vice versa. In other embodiments, the opposite may be true: the user can require lower sensitivity for larger volumes and vice versa.

[0050] Even more, a user may wish to adjust the machine learning algorithm based on the sample’s quality control. For example, in some laboratories, all samples are analyzed prior to a fluid process and are only subjected to the fluid process if they are of good quality.Because the quality of samples is good, then there can less “noise” across samples. When there is less noise across samples, a user may want increased scrutiny of the aspirations to be more discriminating. Alternatively, since the samples are prescreened, a user may decide that the probability of a bad sample is low, and so then require less sensitivity / scrutiny on the aspiration of the sample, to reduce the false error rate. In these laboratories, a user may require a greater sensitivity. In other laboratories, not all samples are analyzed prior to a fluid process. In these laboratories, a user may require a lower sensitivity.

[0051] Yet another advantage of using a machine algorithm for this purpose is that the machine learning algorithm's training can be controlled using the concept of down sampling to improve the odds of detection of short aspirations over full aspirations by rebalancing the training set to have a larger proportion of abnormal aspirations compared to their occurrence in use. This can improve accuracy while minimizing false positives (i.e., shorts detected asfull aspiration) and false negatives (i.e., full aspirations detected as short). This is not an option in conventional solutions.

[0052] In some embodiments, the machine learning algorithm can be a deep learning algorithm used to analyze the pressure trace and predict if the reagent aspiration was normal / full or abnormal / short. Deep learning is a class of machine learning algorithms based on artificial neural networks. The algorithm can be trained by providing a large set of normal aspiration and abnormal aspiration traces that have been processed and labelled. In some embodiments, processing and labelling is performed manually. Once trained, the algorithm can analyze an unknown trace and predict whether the reagent was successfully aspirated, or if it was abnormal.

[0053] In some embodiments, the systems and methods disclosed herein can use a Convolutional Neural Network (“CNN’') supervised machine learning algorithm to separate “normal” aspiration pressure traces from “abnormal” traces that can occur, e.g., due to a clog or other obstruction, or from running short of liquid to aspirate. How ever, as one of ordinary skill in the art will appreciate, other machine learning algorithms that can accomplish the same predictions can be used.

[0054] A neural network is a type of machine learning algorithm that uses weighted connections between artificial “neurons”, or nodes, in multiple layers to identify and classify patterns in the input data. A neural network consists of an input layer of nodes that each accept an element of the incoming data, an output layer that classifies that incoming data, and one or more “hidden” layers that process the input data into the output classification. A CNN uses a series of filters that are convoluted with the input data. These filters help to identify' key features in the data, creating a feature map associated w ith each filter w ithin the layer. Use case examples w ith 2-dimensional data include hand-written character recognition, face recognition, and object recognition. In some embodiments of the present disclosure, the input data can be a 1 -D time series, so one-dimensional convolutional filters can be used. Use of the CNN can eliminate the need for human-determination of key features in the pressure trace, and for specific algorithms to identify those features. The algorithm itself, during training and tuning, can determine the best features to use. This approach can also eliminate the need for manually determining appropriate threshold limits for each of those features in each of the use cases.

[0055] In some embodiments, the machine learning algorithm can be iteratively trained on one or more training sets until its accuracy exceeds a predefined threshold. The training set can include pressure traces of normal fluid aspirations as well as pressure traces from various forms of abnormal fluid processes, such as air aspirations, partial aspirations, and clogged aspirations, for example. The pressure traces in the training set can be labeled, e.g., categorized with a “truth” state. The label can indicate the assessment of the fluid process represented by the pressure trace. For example, the fluid process may have been normal or abnormal, and its pressure trace would be labeled accordingly. To provide another example, a pressure trace can be labeled as “Full Aspiration” or “Abnormal Aspiration.” In some embodiments, pressure traces from abnormal fluid processes may be, additionally or alternatively, labeled with the type of abnormality , e.g., clog, air aspiration, or insufficient, which can be helpful for an end user to identify the source of the problem, and also may be used to direct the instrument software to take appropriate recovery action. In some embodiments, abnormalities that are difficult to classify can be classified simply as “abnormal.” The output layer can consist of a node for each category / label.

[0056] As described above, in some embodiments, one machine learning algorithm can classify a pressure trace as normal or abnormal and can further classify the abnormal pressure traces according to their abnormality. In other embodiments, a first machine learning algorithm can classify a pressure trace as normal or abnormal and a second machine learning algorithm can subclassify the abnormal pressure traces according to their abnormality. In such embodiments, the first machine learning algorithm can be trained to classify' pressure traces as “full” or “abnormal,” and the second machine learning can be trained to subclassify abnormal pressure traces, which is less important to get right. While using two machine learning algorithms in this way can improve accuracy in the subclassification, it can also increase computational overhead.

[0057] In yet other embodiments, one machine learning algorithm can classify a pressure trace as normal or abnormal and then conventional techniques can be used to classify the abnormal pressure traces according to their abnormality' to provide diagnostic feedback. In such embodiments, the machine learning algorithm may only be trained to classify pressure traces as “full” or “abnormal,” which can maximize training per class because there are only two classes (i.e., full and abnormal). This approach can have the advantage of both maintaining the training available per class to maximize the accuracy of detecting abnormaltraces (because splitting the current abnormal class into multiple pieces can significantly reduce the number of training examples per class, thus potentially reducing overall accuracy), while still providing some indicator of the type of failure, as the conventional algorithms often do. The additional computational complexity can be minimal.

[0058] In some embodiments, the training set can include only pressure traces of one type of fluid process. For example, in some embodiments, the training set can include only pressure traces from fluid aspirations. This may be the case if a user only intends to use the trained machine learning algorithm to assess the quality of the type of fluid process on which the machine learning algorithm was trained. However, in some embodiments, the training set can include pressure traces from multiple fluid processes, e.g., fluid aspirations and fluid dispenses. Including multiple fluid processes can improve the accuracy of the trained machine learning algorithm in assessing the quality of either fluid process.

[0059] As described above in the Background section, it can be important in assessing the quality of a fluid process to identity abnormal fluid processes. However, training a machine learning algorithm on a training set comprised of randomly selected pressure traces may not be effective in training the machine learning algorithm to accurately identify abnormal fluid processes. This is due to the fact that, in practice, there are far more normal fluid processes than abnormal fluid processes, and, while the pressure traces of normal fluid processes (for a given volume at a given flow rate) all look substantially the same, pressure traces of abnormal fluid processes can look substantially different from each other because there are several different types of abnormal fluid processes (e.g., clog, air aspiration, partial aspiration) that each manifest differently on the pressure trace. As a result, a machine learning algorithm training on randomly selected pressure traces is trained mostly on the pressure traces of normal fluid processes and trained minimally on the pressure traces of each type of abnormal fluid process.

[0060] To address this concern, in some embodiments, a training set with a higher percentage of pressure traces from abnormal fluid processes than would be present in a set of randomly selected pressure traces can be used. For example, in some embodiments, the ratio of pressure traces from abnormal fluid processes to normal fluid processes can be, for example, five to one. Using a training set with more pressure traces from abnormal fluid processes can increase the chances the machine learning algorithm will accurately identify a pressure trace from an abnormal fluid process.

[0061] Additionally or alternatively, to address this concern, in some embodiments, “edge cases” can be added in the training set. For example, one or more edge cases can be added to a training set that is otherwise formed from randomly selecting pressure traces. An edge case is a pressure trace that can be difficult to assess. It can be a “close call.” In some embodiments, when the machine learning algorithm incorrectly assesses the quality' of a fluid process, the corresponding pressure trace is added to the training set as an edge case. For example, if the machine learning algorithm determines that the fluid process was normal, but that was incorrect, the pressure trace could be labeled as “abnormal” and added to the training set.

[0062] As mentioned above, the machine learning algorithm can be trained using one or more training sets. However, it can be expensive to train a machine learning algorithm because the training takes a large amount of computational resources and data, which can be expensive. To address this concern, in some embodiments, the machine learning algorithm can be trained using a subset of a pressure trace rather than the entire pressure trace. The training set can include no pressure trace subsets, some pressure trace subsets, or be comprised only of pressure trace subsets.

[0063] Certain portions of a pressure trace can be more helpful in assessing the quality of the fluid process than other portions. For example, while the portion of the pressure trace corresponding to pump acceleration and pump deceleration can contain helpful information, it can be difficult to detect abnormalities in these portions since nearly every' trace appears identical as the pump acceleration dominates the pressure change (same for the deceleration phase). Therefore, in some embodiments, these portions, e.g., the portions of the pressure trace that are less helpful, are removed, and the machine learning algorithm is trained using the portions of the pressure trace that remain after the less helpful portions are removed. In other embodiments, the opposite is done - portions of the pressure trace that are more helpful are extracted, and the machine learning algorithm is trained using the extracted portions. Portions of the pressure trace that can be more helpful include, for example, referring to FIG. 1, the comer just prior to deceleration 102, the overall flat portion when the pump is running at constant speed 101, and the very' end of the pressure trace 103. By training the machine learning algorithm on only certain portions of the pressure trace, the machine learning algorithm is focused on the portions of the pressure trace that are usually more helpful in assessing the fluid process’ quality'.

[0064] In some embodiments, the amount of data and computational resources needed to train a machine learning algorithm is reduced by reducing the number of machine learning algorithms by providing one machine learning algorithm that can be used to assess the quality of a fluid process performed on a fluid, regardless of flow rate or fluid volume. For example, in some embodiments, the machine learning algorithm is trained on a training set of pressure traces (irrespective of the volume or flow rate of the fluid process) that conform to a common scheme. The trained machine learning algorithm can be used to assess the quality of a fluid process of any volume or flow rate, provided that the pressure trace conforms to the same common scheme.

[0065] FIG. 5 is a graph of representative pressure traces for various aspiration volumes and at different aspiration speeds, according to an embodiment of the disclosure. As illustrated in FIG. 5, the shapes of pressure traces for aspirations will be different depending on the fluid and the volume. As one of ordinary skill in the art would expect, as flow rate increases, pressure decreases, whether aspirating liquid or air. The difference between an air or liquid aspiration increases proportionately as well. Note that aspirations of varying volumes but using the same flow rate have the same pressure drop.

[0066] FIG. 6 illustrates numerous aspirations of air and liquid, according to an embodiment of the disclosure. In FIG. 6, each pressure trace is labeled with an alphanumeric identifier having three parts, each separated by a hyphen. The first part is a client-specific identifier. The second part refers to an internal fluid ty pe. The third part represents the volume aspirated. As illustrated in FIG. 6, during two of those liquid aspirations, insufficient liquid was available, resulting in a "‘short” aspiration. Photometric analysis of the transferred material revealed that these two aspirations were 40% and 16% short, respectively. While the 40% short aspiration is clearly separated from the norm at the end of the steady-state aspiration (-140ms), the 16% short aspiration is somewhat more difficult to distinguish.

[0067] Note also that there are many other factors that can affect the shape of a pressure trace. For example, pressure traces have different lengths (i.e., along the x-axis) and heights (i.e., along the y-axis) for a variety of reasons including, for example, the volume of fluid subject to the fluid process. Because the length and / or height of the pressure trace can differ for each volume, a machine learning algorithm is needed for each volume. In other words, to assess the fluid process for three different volumes, three different machine learning algorithms may be needed, one for each volume. Each machine learning algorithm can betrained using a training set comprised of pressure traces from fluid processes of one fluid volume, and then the trained machine learning algorithm can be used to assess the quality of a fluid process performed on the same volume of fluid.

[0068] For example, consider FIG. 7. FIG. 7 illustrates numerous aspirations of air and liquid. During two of those liquid aspirations, insufficient liquid was available, resulting in a ■‘short” aspiration. Photometric analysis of the transferred material revealed that these two aspirations were 40% and 16% short, respectively. While the 40% short aspiration is clearly separated from the norm at the end of the steady-state aspiration (-140ms), the 16% short aspiration is somewhat more difficult to distinguish. How ever, as one of ordinary skill in the art would appreciate, a machine learning algorithm trained using pressure traces from air aspirations may not be accurate in assessing the quality of a liquid aspiration because the shapes of pressure traces are substantially different.

[0069] As mentioned previously, there are many factors that can affect the shape of a pressure trace. For example, the pressure drop during a fluid process, e.g., an aspiration, and probe and tubing diameters and lengths. The pressure drop during a fluid process can be dictated by the following factors: (1) liquid or gas properties (e.g., density, dynamic viscosity', etc.) (2) flow rate, and (3) probe and tubing diameters and lengths. In one embodiment, this can be demonstrated with an analysis using the Bernoulli equation for the steady state portion of the aspiration as follows:P = Pressure p = Density / g = Gravitational constant a = Energy correction v = Velocity’ h = Height factorwhere:P is the effect of fluid density1 1 \ is the effect of tubing and pipettorP 42l 42) geometry pg(h2~ ^i) is the effect of height of liquid columnp 2 V is the effect of flow resistance2Q2J 42"

[0070] Further analysis can provide more details about the effects of probe diameters, fluid viscosities. In the equation below, which shows main influencers on the pressure drop, it can be seen that viscosity directly influences the pressure drop during aspiration, and that probe diameter has a very significant effect.

[0071] As described above, many factors can affect the shape of a pressure trace, and because the shape of a pressure trace can differ for each volume subjected to a fluid process, a machine learning algorithm can be needed for each volume. However, the slopes, plateaus, and other features of a pressure trace, regardless of the volume of fluid aspirated, will substantially align when the pressure trace is compressed (or expanded) to a fixed length in the time domain or height in the pressure domain. Because of this, a machine learning algorithm that is trained on pressure traces that share a fixed length in the time domain or height in the pressure domain can be used to assess the quality of a fluid process performed on any volume, as long as that volume's pressure trace also shares the fixed length in the time domain or height in the pressure domain. In other words, instead of having to train multiple machine learning algorithms on multiple training sets for use in assessing fluid processes for multiple volumes, only one machine learning algorithm is trained on one trained set but can be used in assessing the fluid process of multiple volumes. Therefore, in some embodiments, the machine learning algorithm is trained on a training set of pressure traces (irrespective of the volume or flow rate of the fluid process) that conform to a common scheme, i.e., a fixed length in the time domain or height in the pressure domain. The trained machine learning algorithm can be used to assess the quality of a fluid process of any volume or flow rate, provided that the pressure trace also conforms to the fixed length in the time domain or height in the pressure domain.

[0072] One of ordinary skill in the art might be dissuaded from modifying a pressure trace to conform to a common scheme prior to analysis for fear that the modification would negatively affect the results of the subsequent analysis. However, when a high enough sampling frequency is used during the aspiration / dispense, the pressure trace’s shape ismaintained after modification, and the results are therefore not negatively affected by the compression.

[0073] A pressure trace of a fixed length in the time domain or a fixed height in the pressure domain could also be achieved by adjusting the sampling rate for each fluid process so that a fixed number of datapoints is output. However, this approach is not practical because the total duration of the fluid process is not always known ahead of time. Modifying the pressure trace after the fluid process effectively provides a variable sampling rate without having to adjust the sampling rate for each fluid process.

[0074] In addition to reducing the amount of computation resources and data needed, training a machine learning algorithm on pressure traces that conform to a common scheme has another advantage: it permits accurate analysis of pressure traces for which there is insufficient data to generate accurate machine learning algorithms. For less common volumes (e.g., 38uL), there may not be sufficient data to train a machine learning algorithm to accurately assess fluid processes of that volume. But, with this solution, the trained machine learning algorithm works for all volumes (provided that they are modified to the common scheme), including volumes that for which insufficient data exists to train a machine learning algorithm. Said another way. using a machine learning algorithm on modified pressure trace and makes assessing of certain volumes using a machine learning algorithm possible that were previously impossible.

[0075] FIG. 8 is a flow chart of a method of assessing the quality of a fluid process using machine learning, according to an embodiment of the present disclosure. At step 801, the method 800 can include collecting a plurality of labeled pressure traces. In some embodiments, this step can include collecting pressure traces and labeling them. At step 802, the method 800 can include modifying each of the plurality of labeled pressure traces to conform to a common scheme. The common scheme can be. for example, in the time or pressure domain. At step 803, the method 800 can include training a machine learning algorithm on the modified pressure traces. In some embodiments, the method can optionally include determining the accuracy of the trained machine learning algorithm, comparing the accuracy to a predetermined threshold, and iteratively retraining the machine learning algorithm until the machine learning algorithm’s accuracy exceeds the predetermined threshold. At step 804, the method can include performing a fluid process, e.g., an aspiration or a dispense, of a fluid. The fluid can be anything that creates a pressure profile, forexample, a biological sample, a liquid reagent, or air. At step 805, the method 800 can include measuring pressure as a function of time while the fluid process is performed. In some embodiments, pressure can be measured by a pressure sensor connected to tubing between a pump and a probe. Pressure can be measured in time stamps or duration.

[0076] In some embodiments, raw pressure vs. time data can be filtered to reduce or eliminate noise. In some embodiments, an outlier electrical noise prefilter can be applied to raw data points (i.e., datapoints with a delta of more than 2,500 can be replaced with the moving average of the preceding three datapoints). This can be limited to 10% of raw datapoints. In some embodiments, the Butterworth low pass filter can also be applied to raw data in each of the three motion phases with different coefficients (i.e., for cutoff frequency) for each phase. For example, for the acceleration delay phase, the filtering can be undamped with a cutoff of 0.05; for the slew phase, the filtering can be critically damped with a cutoff of 0.04; and for the deceleration delay phase the filtering can be underdamped with a cutoff of 0.075. Some of these methods are explained in U.S. Patent No. 9,909.945, which is hereby incorporated herein in its entirety. In some embodiments, the pressure as a function of time data, filtered or raw, can be plotted.

[0077] At step 806, the method can include modifying the pressure trace such that it conforms to a common scheme. In some embodiments, the common scheme is defined by a maximum time value. For example, the maximum time value can be no values over 100 ms in duration. In some embodiments, the common scheme is defined by a period of time, 3:00 pm to 3:05 pm. for example. In other embodiments, the common scheme is defined by a minimum pressure value. For example, the minimum pressure value can be no values less than -400 counts.

[0078] In some embodiments, the common scheme is defined by a number of data points. For example, a common scheme may include 100 data points. FIGS. 9A-9C illustrate an example of this embodiment. FIGS. 9A-9C are graphs showing an unmodified pressure trace and a pressure trace modified using linear interpolation, according to an embodiment of the disclosure. FIG. 9A is a graph of unmodified pressure vs. time data for 38uL, 50uL, and lOOuL pressure traces, according to an embodiment of the disclosure. FIG. 9B is a graph with the pressure trace in FIG. 9A modified to conform to common time scheme that includes 100 data points from 1 to 100. FIG. 9C is a graph with both the unmodified and modified pressure traces in FIGS. 9A and 9B, respectively.

[0079] Pressure traces can be conformed to a common time scheme by any method known in the art. In some embodiments, the pressure traces can be compressed or expanded to the common time scheme. For example, this can be visualized as stretching, or compressing, as applicable, the pressure trace along the x-axis (or the y-axis, or both the x- axis and the y-axis) and stopping when the desired pressure trace length is achieved.

[0080] In some embodiments, datapoints can be removed from or added to the pressure trace. For example, if the common time scheme 100 datapoints and the unmodified pressure trace has 200 datapoints, modifying the pressure trace can include removing every other datapoint to reduce the pressure trace to 100 datapoints. For another example, if the common time scheme is 100 datapoints and the pressure trace has 50 datapoints. modifying the pressure trace data can include adding a datapoint between each raw datapoint to increase the pressure vs. time data to 100 datapoints. The additional datapoints can be determined, for example, using interpolation, as one of ordinary skill in the art will appreciate.

[0081] In some embodiments, interpolation can be used to generate new data for the pressure trace based on the unmodified pressure trace. For filtered data, interpolation can be performed before or after filtering. Interpolation can help maintain the shape of the pressure trace. Interpolation can be linear or non-linear. FIG. 10 is a table of linear interpolation modifying pressure vs. time data, according to an embodiment of the disclosure. In FIG. 10, column “38-xl” includes the raw time data and column “38-yl” includes the raw7pressure data. Column “38-xnew” includes the modified time data. In the embodiment illustrated in FIG. 10, the new time data are integers beginning with zero and ending with seventeen. Column “38-ynew” includes the modified pressure data. The modified pressure data is calculated using linear interpolation. Columns “slope-m,” “yintercept-c,” and “ycomp” include values helpful to calculate the modified pressure data, as one of ordinary skill in the art will appreciate. FIG. 11 is a graph of the unmodified pressure trace and modified pressure trace from FIG. 10, according to an embodiment of the disclosure.

[0082] In some embodiments, generating new7data for the pressure trace can be done using the interpolate.interpld(x, y) function in the Python SciPy library. However, the same can be achieved using any other software program. To generate new data for the pressure trace, for example. Python SciPy’s interpolate.interpld(x, y), the following steps can be performed: The pressure trace values y are mapped to the unmodified time interval x. Next,the x is scaled to the required length (e.g., 100). Y values are interpolated to the integers on this new compressed x-scale, yielding a uniform number of points for every pressure trace.

[0083] Returning to FIG. 8, in some embodiments, the method 800 can optionally include shifting or offsetting the pressure trace so its starting value is zero. This can be done, for example, by subtracting the difference between the pressure trace’s starting value and zero for every datapoint in the pressure trace, either before or after it is conformed to a common domain. This can have the effect of removing pressure sensor offset differences, thus removing one source of input “noise” that the algorithm would otherwise need to resolve.

[0084] At step 807, the method 800 can include assessing the quality of the fluid process by analyzing the modified pressure trace using the trained machine learning algorithm. If the quality7of the fluid process is good, in some embodiments, the fluid can be added to a mixture, and the mixture can be analyzed. If the quality of the fluid process is bad, in some embodiments, an additional fluid process is performed instead of adding the fluid to a mixture to be analyzed.

[0085] In another embodiment, a method of assessing the quality of a fluid process using machine learning can include one or more of the following: collecting pressure traces from multiple instruments, manually cleaning / filtering / processing the traces, labeling the traces as full aspiration or as short, splitting the dataset in to training and testing sets, fitting machine learning model on the training dataset, performing hyperparameter tuning on the machine learning model, making predictions on unseen or test dataset, studying misclassifications using a confusion matrix, inspecting the misclassified traces and correcting labels if necessary7, if label is correct, injecting the edge case in the training set, and iteratively repeating the above steps until the required accuracy is achieved.

[0086] In some embodiments, the method can include using an auxiliary input, like the fluid volume or flow rate, to the machine learning algorithm. In some embodiments, this can allow for more accurate analysis. This can be helpful in situations in which the user is limited in time, computational resources, or both.

[0087] USE CASE

[0088] The following example use case describes an example of a machine learning algorithm used to assess the quality of aspirations. This section is intended solely for explanatory purposes and not in limitation.

[0089] In one example, a machine learning algorithm uses 100 input data points from the trace (i.e., conforms the input pressure trace to a scheme having 100 data points), with volume passed as an auxiliary input. The machine learning is sequential with 2 hidden layers using ID convolution. Each layer is followed by batch normalization (which stabilizes and increases the speed of neural network training by normalizing the mean and variances of the input data to each layer), relu activation (a piecewise function that will output the input directly if it is positive, otherwise, it will output zero) and dropout (which helps to prevent overfitting of the trained model). Layers may also be followed by an average or max pooling operation that reduces the size of the input to the following layer. Global Average pooling (which calculates the average output of each filtered feature map in the final layer to prepare the output for the final categorization) is performed at the end before the zero or more fully connected (dense) layers, which in conjunction with the final softmax output layer, performs the classification of the input pressure trace. The model (diagramed in FIGS. 12A and 12B) is created in Py thon with the TensorFlow Keras, Hy peropt, and ML flow libraries. The architecture of the neural network includes parameters such as the number of hidden layers (for example 2 or 3). the number of filters in the each of the layers of the CNN (for example, between 64 to 256), kernel size of the convolution filter in each hidden layer (for example, between 4 and 12), the number of fully connected layers, the learning rate of the stochastic gradient decent algorithm, and the optimizer algorithm for the stochastic gradient descent (for example. "Adadelta" or "‘Adam”).

[0090] Hyperparameter tuning is the process of selecting specific values of the above defined parameters for the neural network. This is done by perfonning an automated factorial-based evaluation of the various settings.

[0091] FIG. 13 is a confusion matrix of the output of the CNN on the test dataset, according to an embodiment of the disclosure. Upper left and lower right boxes indicate the number of pressure traces that the algorithm correctly assessed, whereas the upper right and lower left illustrate those which that algorithm incorrectly assessed. The upper left box indicates 731,847 aspirations that were normal, and that the algorithm also categorized as normal. Likewise, the lower right box shows 5,465 aspirations that were abnormal, and thealgorithm also correctly categorized as abnormal. There were 11 aspirations that were normal but the algorithm mistakenly flagged as abnormal (upper right box), and 25 aspirations that were abnormal but the algorithm flagged as normal (lower left box). Note the overall very low percentage of missed cases. Overall accuracy of predictions can be calculated as 100 x (Normal aspirations prediction as normal + Abnormal predictions predicted as abnormal) / All traces 100 x (731,847 + 5,465) / 737,348 = 99.995 %.

[0092] FIGS. 14A and 14B are a comparison between an embodiment of the CNN algorithm and a conventional solution from the CI1900 IA Reagent Volume Check, according to an embodiment of the disclosure. This data is presented in the form of a set of confusion matrices, comparing the CNN (ML) algorithm to the conventional (instrument) results. The actual or ‘'truth” result is presented on the y-axis, and the algorithm result is presented on the x-axis.

[0093] It is apparent that the CNN (ML) algorithm correctly classified far more short aspirations than the conventional algorithm, 25 vs 1,195. This type of misclassification can allow processing reactions with insufficient reagent, that can have negative results on instrument performance. Note that these charts present an apples-to-apples comparison, in that the conventional algorithm has been limited to using the same data as the CNN. In practice, the conventional algorithm also uses the dispense data, which improves its accuracy. Dispense data may be used in the CNN algorithm to improve accuracy in a future enhancement. Even without the dispense data, the CNN algorithm is capable of detecting smaller missing volumes than possible with the conventional algorithms.

[0094] In this run, the CNN misclassified 11 successful aspirations as failures, whereas the conventional algorithm misclassified 505. This category7does not impact analytic performance but can require the need to repeat a test. It is expected that with further refinement, these classifications will be further reduced.

[0095] FIGS. 15 A and 15B are an example of classified traces, according to an embodiment of the disclosure. The grey traces are abnormal that were detected as abnormal, the black traces are normal that were detected as normal, and the light yellow trace was misclassified. Referring specifically to FIG. 15 A, in this case, one trace (the light grey curve) was categorized as normal, but in reality7it was slightly short of a full aspiration. This represents an edge case that can be difficult to differentiate. Referring specifically to FIG.15B, many of the grey traces in FIG. 15B are quite close to appearing normal yet are still detected correctly.

[0096] The following table example results in terms of percent of total aspirations for an expanded dataset. Again, note that the CNN algorithm is capable of detecting significantly more short aspirations than conventional algorithms using the same aspiration-only data set (some conventional algorithms, such as the one shown, uses additional pressure data from the fluid dispense to supplement the evaluation, helping to improve accuracy, as seen in the right two columns). The difference illustrated here is over 48x improvement.

[0097] It should be noted that these results are from a CNN with which training sets and even architecture have yet to be fully optimized, and as such it is expected that the results can be improved further. For example, a smaller, but more information rich, portion of the trace may be used in the CNN either alone or supplemental. Different prefiltering methods can be employed, and enhanced training set employed. Also, a few of the CNN false positive results shown in the table below are due to errors in the labels for the actual aspiration result, and that dataset is still under the process of being updated.

[0098] The table below shows that the machine learning algorithm is 46X times better than conventional solutions at producing false positives, and 48X better at producing falsenegatives, (i. e. , looked at another way, it produces 48X fewer false negatives than conventional algorithm in apples-to-apples comparison, and 19x fewer when conventional algorithm also uses dispense data.)

[0099] It is clear from the above results that the CNN algorithm is superior in assessing aspiration quality to conventional methods, with a 19x improvement in detected short aspirations (reduced false negatives) over a conventional algorithm, and an 48x improvement when comparing with equal pressure data sets (aspiration only). Further improvement is expected as the method is refined.

[0100] In addition to the current implementation (using a 1 dimensional CNN with multiple hidden layers and a pressure trace that is compressed along the time axis and with aspiration volume as an auxiliary input), the following alternate implementations are also being investigated:

[0101] The example described above may be modified to include one or more of the following: a neural network architecture, including, for example, recurring neural network, long short term memory (LSTM), and transformer architectures; pressure traces that conform to both a common time domain and a common pressure domain; use of an auxiliary input, e.g., flow rate or aspiration speed; and excluding zones of the pressure trace with limited information to maximize accuracy on the most informative zones (for example, excluding zones with pump acceleration).

[0102] The presently disclosed subject matter is primarily described in the context of pressure traces the shape of which change depending on the volume of fluid aspirated. However, as one of ordinary skill in the art will appreciate, a pressure trace's shape can alsochange depending on a variety of other variables, including, for example, aspiration duration, aspiration speed, and geometry of the probe and tubing. It is to be understood that the systems, methods, and computer program products described herein are functional for all aspiration quality assessment of all pressure traces, regardless of the variable that drives their change in shape.

[0103] FIG. 16 illustrates an exemplary computing environment within which embodiments of the invention may be implemented. For example, this computing environment 1600 may be configured to execute a method of placing an item having irregular dimensions. The computing environment 1600 may include computer system 1610, which is one example of a computing system upon which embodiments of the invention may be implemented. Computers and computing environments, such as computer system 1610 and computing environment 1600, are known to those of skill in the art and thus are described briefly here.

[0104] As shown in FIG. 16, the computer system 1610 may include a communication mechanism such as a bus 1605 or other communication mechanism for communicating information within the computer system 1610. The computer system 1610 further includes one or more processors 1620 coupled with the bus 1605 for processing the information. The processors 1620 may include one or more central processing units (CPUs), graphical processing units (GPUs), or any other processor known in the art.

[0105] The computer system 1610 also includes a system memory 1630 coupled to the bus 1605 for storing information and instructions to be executed by processors 1620. The system memory 1630 may include computer readable storage media in the form of volatile and / or nonvolatile memory. such as read only memory7(ROM) 1631 and / or random access memory (RAM) 1632. The system memory RAM 1632 may include other dynamic storage device(s) (e.g.. dynamic RAM, static RAM. and synchronous DRAM). The system memory ROM 1 31 may include other static storage device(s) (e.g., programmable ROM, erasable PROM, and electrically erasable PROM). In addition, the system memory 1630 may be used for storing temporary7variables or other intermediate information during the execution of instructions by the processors 1620. A basic input / output system (BIOS) 1633 containing the basic routines that help to transfer information between elements within computer system 1610, such as during start-up, may be stored in ROM 1631. RAM 1632 may contain data and / or program modules that are immediately accessible to and / or presently being operatedon by the processors 1620. System memory 1630 may additionally include, for example, operating system 1634, application programs 1635, other program modules 1636 and program data 1637.

[0106] The computer system 1610 also includes a disk controller 1640 coupled to the bus 1605 to control one or more storage devices for storing information and instructions, such as a hard disk 1641 and a removable media drive 1642 (e.g., floppy disk drive, compact disc drive, tape drive, and / or solid state drive). The storage devices may be added to the computer system 1610 using an appropriate device interface (e.g., a small computer system interface (SCSI), integrated device electronics (IDE), Universal Serial Bus (USB), or FireWire).

[0107] The computer system 1610 may also include a display7controller 1665 coupled to the bus 1605 to control a display 1666, such as a cathode ray tube (CRT) or liquid crystal display (LCD), for displaying information to a computer user. The computer system 1610 includes an input interface 1660 and one or more input devices, such as a keyboard 1662 and a pointing device 1661, for interacting with a computer user and providing information to the processor 1620. The pointing device 1661, for example, may be a mouse, a trackball, or a pointing stick for communicating direction information and command selections to the processor 1620 and for controlling cursor movement on the display 1666. The display 1666 may provide a touch screen interface which allows input to supplement or replace the communication of direction information and command selections by the pointing device 1661.

[0108] The computer system 1610 may perform a portion or all of the processing steps of embodiments of the invention in response to the processors 1620 executing one or more sequences of one or more instructions contained in a memory7, such as the system memory 1630. Such instructions may be read into the system memory 1630 from another computer readable medium, such as a hard disk 1641 or a removable media drive 1642. The hard disk 1641 may contain one or more datastores and data files used by embodiments of the present invention. Datastore contents and data files may be encry pted to improve security7. The processors 1620 may also be employed in a multi-processing arrangement to execute the one or more sequences of instructions contained in system memory 1630. In alternative embodiments, hard-wired circuitry may be used in place of or in combination with software instructions. Thus, embodiments are not limited to any specific combination of hardware circuitry7and software.

[0109] As stated above, the computer system 1610 may include at least one computer readable medium or memory for holding instructions programmed according to embodiments of the invention and for containing data structures, tables, records, or other data described herein. The term “computer readable medium” as used herein refers to any medium that participates in providing instructions to the processor 1620 for execution. A computer readable medium may take many forms including, but not limited to, non-volatile media, volatile media, and transmission media. Non-limiting examples of non-volatile media include optical disks, solid state drives, magnetic disks, and magneto-optical disks, such as hard disk 1641 or removable media drive 1642. Non-limiting examples of volatile media include dynamic memory', such as system memory' 1630. Non-limiting examples of transmission media include coaxial cables, copper wire, and fiber optics, including the wires that make up the bus 1605. Transmission media may also take the form of acoustic or light yvaves, such as those generated during radio yvave and infrared data communications.

[0110] The computing environment 1600 may further include the computer system 1610 operating in a networked environment using logical connections to one or more remote computers, such as remote computer 1680. Remote computer 1680 may be a personal computer (laptop or desktop), a mobile device, a server, a router, a network PC, a peer device or other common network node, and typically includes many or all of the elements described above relative to computer system 1610. When used in a networking environment, computer system 1610 may include modem 1672 for establishing communications over a network 1671, such as the Internet. Modem 1672 may be connected to bus 1605 via user netyvork interface 1670, or via another appropriate mechanism.[OHl] Network 1671 may be any netyvork or system generally known in the art, including the Internet, an intranet, a local area netyvork (LAN), a yvide area netyvork (WAN), a metropolitan area network (MAN), a direct connection or series of connections, a cellular telephone network, or any other netyvork or medium capable of facilitating communication between computer system 1610 and other computers (e.g., remote computer 1680). The netyvork 1671 may be wired, yvireless or a combination thereof. Wired connections may be implemented using Ethernet, Universal Serial Bus (USB), RJ-11 or any other wired connection generally known in the art. Wireless connections may be implemented using WiFi, WiMAX, and Bluetooth, infrared, cellular networks, satellite or any other wireless connection methodology generally knoyvn in the art. Additionally, several networks maywork alone or in communication with each other to facilitate communication in the network 1671.

[0112] The embodiments of the present disclosure may be implemented with any combination of hardware and software. In addition, the embodiments of the present disclosure may be included in an article of manufacture (e.g., one or more computer program products) having, for example, computer-readable, non-transitory media. The media has embodied therein, for instance, computer readable program code for providing and facilitating the mechanisms of the embodiments of the present disclosure. The article of manufacture can be included as part of a computer system or sold separately.

[0113] While various aspects and embodiments have been disclosed herein, other aspects and embodiments will be apparent to those skilled in the art. The various aspects and embodiments disclosed herein are for purposes of illustration and are not intended to be limiting, with the true scope and spirit being indicated by the following claims.

[0114] An executable application, as used herein, comprises code or machine readable instructions for conditioning the processor to implement predetermined functions, such as those of an operating system, a context data acquisition system or other information processing system, for example, in response to user command or input. An executable procedure is a segment of code or machine readable instruction, sub-routine, or other distinct section of code or portion of an executable application for performing one or more particular processes. These processes may include receiving input data and / or parameters, performing operations on received input data and / or performing functions in response to received input parameters, and providing resulting output data and / or parameters.

[0115] A graphical user interface (GUI), as used herein, comprises one or more display images, generated by a display processor and enabling user interaction with a processor or other device and associated data acquisition and processing functions. The GUI also includes an executable procedure or executable application. The executable procedure or executable application conditions the display processor to generate signals representing the GUI display images. These signals are supplied to a display device which displays the image for viewing by the user. The processor, under control of an executable procedure or executable application, manipulates the GUI display images in response to signals received from theinput devices. In this way. the user may interact with the display image using the input devices, enabling user interaction with the processor or other device.

[0116] The functions and process steps herein may be performed automatically or wholly or partially in response to user command. An activity (including a step) performed automatically is performed in response to one or more executable instructions or device operation without user direct initiation of the activity.

[0117] While various illustrative embodiments incorporating the principles of the present teachings have been disclosed, the present teachings are not limited to the disclosed embodiments. Instead, this application is intended to cover any variations, uses, or adaptations of the present teachings and use its general principles. Further, this application is intended to cover such departures from the present disclosure that are within known or customary practice in the art to which these teachings pertain. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting.

[0118] In the above detailed description, reference is made to the accompanying drawings, which form a part hereof. In the drawings, similar symbols typically identify similar components, unless context dictates otherwise. The illustrative embodiments described in the present disclosure are not meant to be limiting. Other embodiments may be used, and other changes may be made, without departing from the spirit or scope of the subject matter presented herein. It will be readily understood that various features of the present disclosure, as generally described herein, and illustrated in the Figures, can be arranged, substituted, combined, separated, and designed in a wide variety of different configurations, all of which are explicitly contemplated herein.

[0119] Aspects of the present technical solutions are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the technical solutions. It w ill be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer readable program instructions.

[0120] These computer readable program instructions can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processingapparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. These computer readable program instructions can also be stored in a computer readable storage medium that can direct a computer, a programmable data processing apparatus, and / or other devices to function in a particular manner, such that the computer readable storage medium having instructions stored therein comprises an article of manufacture including instructions which implement aspects of the function / act specified in the flowchart and / or block diagram block or blocks.

[0121] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process, such that the instructions which execute on the computer, other programmable apparatus, or other device implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0122] The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present technical solutions. In this regard, each block in the flowchart or block diagrams can represent a module, segment, or portion of instructions, which includes one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks shown in succession can, in fact, be executed substantially concurrently, or the blocks can sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flowchart illustration, and combinations of blocks in the block diagrams and / or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts or cany7out combinations of special purpose hardware and computer instructions.

[0123] A second action can be said to be “in response to” a first action independent of whether the second action results directly or indirectly from the first action. The second action can occur at a substantially later time than the first action and still be in response to thefirst action. Similarly, the second action can be said to be in response to the first action even if intervening actions take place between the first action and the second action, and even if one or more of the intervening actions directly cause the second action to be performed. For example, a second action can be in response to a first action if the first action sets a flag and a third action later initiates the second action whenever the flag is set.

[0124] With respect to the use of substantially any plural and / or singular terms herein, those having skill in the art can translate from the plural to the singular and / or from the singular to the plural as is appropriate to the context and / or application. The various singular / plural permutations may be expressly set forth herein for sake of clarity.

[0125] It will be understood by those within the art that, in general, terms used herein are generally intended as “open” terms (for example, the term “including” should be interpreted as “including but not limited to,” the term “having” should be interpreted as “having at least.” the term “includes” should be interpreted as “includes but is not limited to,” et cetera). While various compositions, methods, and devices are described in terms of “comprising” various components or steps (interpreted as meaning “including, but not limited to”), the compositions, methods, and devices can also “consist essentially of’ or “consist of’ the various components and steps, and such terminology should be interpreted as defining essentially closed-member groups.

[0126] As used in this document, the singular forms “a,” “an,” and “the” include plural references unless the context clearly dictates otherwise. Unless defined otherwise, all technical and scientific terms used herein have the same meanings as commonly understood by one of ordinary skill in the art. Nothing in this disclosure is to be construed as an admission that the embodiments described in this disclosure are not entitled to antedate such disclosure by virtue of prior invention.

[0127] In addition, even if a specific number is explicitly recited, those skilled in the art will recognize that such recitation should be interpreted to mean at least the recited number (for example, the bare recitation of “two recitations.” without other modifiers, means at least two recitations, or two or more recitations). Furthermore, in those instances where a convention analogous to “at least one of A, B, and C, et cetera” is used, in general such a construction is intended in the sense one having skill in the art would understand the convention (for example, “a sy stem having at least one of A, B, and C” would include but notbe limited to systems that have A alone, B alone, C alone. A and B together, A and C together, B and C together, and / or A, B, and C together, et cetera). In those instances where a convention analogous to '‘at least one of A, B, or C, et cetera’’ is used, in general such a construction is intended in the sense one having skill in the art would understand the convention (for example, “a system having at least one of A, B, or C” would include but not be limited to systems that have A alone, B alone, C alone, A and B together, A and C together, B and C together, and / or A, B. and C together, et cetera). It will be further understood by those within the art that virtually any disjunctive word and / or phrase presenting two or more alternative terms, whether in the description, sample embodiments, or drawings, should be understood to contemplate the possibilities of including one of the terms, either of the terms, or both terms. For example, the phrase “A or B” will be understood to include the possibilities of '‘A” or “B” or ‘’A and B.”

[0128] In addition, where features of the disclosure are described in terms of Markush groups, those skilled in the art will recognize that the disclosure is also thereby described in terms of any individual member or subgroup of members of the Markush group.

[0129] As will be understood by one skilled in the art, for any and all purposes, such as in terms of providing a written description, all ranges disclosed herein also encompass any and all possible subranges and combinations of subranges thereof. Any listed range can be easily recognized as sufficiently describing and enabling the same range being broken down into at least equal halves, thirds, quarters, fifths, tenths, et cetera. As a non-limiting example, each range discussed herein can be readily broken down into a lower third, middle third and upper third, et cetera. As will also be understood by one skilled in the art all language such as ‘"up to,” “at least,” and the like include the number recited and refer to ranges that can be subsequently broken down into subranges as discussed above. Finally, as will be understood by one skilled in the art, a range includes each individual member. Thus, for example, a group having 1-3 components refers to groups having 1, 2, or 3 components. Similarly, a group having 1-5 components refers to groups having 1, 2, 3, 4, or 5 components, and so forth.NON-LIMITING ILLUSTRATIVE EMBODIMENTS

[0130] The following is a list of non-limiting illustrative embodiments disclosed herein:

[0131] Illustrative embodiment 1. A system for assessing a quality of a fluid process performed on a fluid including one of a fluid aspiration and a fluid dispense, the system comprising: a pump; a probe; a connection, wherein the pump and the probe are attached to the connection and a fluid path is formed between the pump and the probe at least in part by the connection; a pressure sensor in fluid communication with the fluid path and configured to sense pressures of the pump during the fluid process, wherein a pressure measurement data comprises the sensed pressures of the pump during the fluid process; a processor; and a memory comprising instructions that are executed by the processor to cause the processor to: receive, from the pressure sensor, pressure measurement data during the fluid process, wherein the pressure measurement data comprises pressure data and time data, and determine, by the processor using a trained machine learning algorithm, the quality of the fluid process based on the received pressure measurement data.

[0132] Illustrative embodiment 2. The system according to the preceding embodiment, wherein the system further comprises: an analyzer configured to analyze a mixture comprising the fluid when the quality of the fluid process is satisfactory.

[0133] Illustrative embodiment 3. The system according to one of the preceding embodiments, wherein the instructions further cause the processor to: cause the pump to operate such that a subsequent fluid process is performed in response to determining that the quality of the fluid process is abnormal.

[0134] Illustrative embodiment 4. The system according to one of the preceding embodiments, wherein the trained machine learning algorithm comprises a sensitivity threshold, wherein the instructions further cause the processor to: adjust the sensitivity threshold prior to determining the quality of the fluid process.

[0135] Illustrative embodiment 5. The system according to one of the preceding embodiments, wherein the trained machine learning algorithm comprises a machine learning algorithm trained on a training set, and wherein the training set comprises a plurality7of pressure traces of normal fluid processes and a plurality of pressure traces of abnormal fluid processes at a ratio of at least 5: 1.

[0136] Illustrative embodiment 6. The system according to one of the preceding embodiments, wherein the trained machine learning algorithm comprises a machine learning algorithm trained on a training set. and wherein the training set comprises a plurality ofpressure traces conforming to a common domain such that the trained machine learning algorithm can be used to determine fluid process qualities of fluid processes performed on a plurality of volumes of fluids that also conform to the common domain.

[0137] Illustrative embodiment 7. The system according to one of the preceding embodiments, wherein each of the plurality of pressure traces comprises a plurality of data points, wherein the common domain comprises a number of data points.

[0138] Illustrative embodiment 8. The system according to one of the preceding embodiments, wherein the common domain comprises one of a maximum time value and a minimum pressure value.

[0139] Illustrative embodiment 9. The system according to one of the preceding embodiments, wherein the training set comprises a plurality of pressure traces of aspirations and a plurality of pressure traces of dispenses.

[0140] Illustrative embodiment 10. The system according to one of the preceding embodiments, wherein the trained machine learning algorithm comprises a machine learning algorithm trained on training set, and wherein the training set comprises at least one modified pressure trace, wherein the at least one modified pressure trace comprises an unmodified pressure trace having at least one portion excluded.

[0141] Illustrative embodiment 11. The system according to one of the preceding embodiments, wherein the at least one portion excluded comprises one of a portion associated with pump deceleration and a portion associated with pump acceleration.

[0142] Illustrative embodiment 12. The system according to one of the preceding embodiments, wherein the trained machine learning algorithm comprises a machine learning algorithm trained on training set, and wherein the training set comprises at least one modified pressure trace, wherein the at least one modified pressure trace comprises a subset of an unmodified pressure trace associated with an information-rich portion of the unmodified pressure trace.

[0143] Illustrative embodiment 13. A method of quality' assessment of a fluid process, the method comprising: performing, by a fluid processing system, the fluid process on a fluid, wherein the fluid process comprises one of a fluid aspiration and a fluid dispense; receiving.by a processor, pressure measurement data during the fluid process, wherein the pressure measurement data comprises pressure data and time data; and determining, by the processor using a trained machine learning algorithm, the quality assessment of the fluid process based on the received pressure measurement data.

[0144] Illustrative embodiment 14. The method according to one of the preceding embodiments, further comprising: analyzing a mixture comprising the fluid in response to determining that the quality assessment of the fluid process is satisfactory.

[0145] Illustrative embodiment 15. The method according to one of the preceding embodiments, further comprising: aspirating, by the fluid processing system, an additional fluid process in response to determining the quality assessment of the fluid process is abnormal.

[0146] Illustrative embodiment 16. The method according to one of the preceding embodiments, wherein the trained machine learning algorithm comprises a sensitivity threshold, wherein the method further comprises: adjusting the sensitivity threshold prior to determining the quality assessment of the fluid process.

[0147] Illustrative embodiment 17. The method according to one of the preceding embodiments, further comprising: modify ing the pressure measurement data to conform to a common domain, wherein the trained machine learning algorithm can be used to determine fluid process qualities of the fluid process performed on a plurality of volumes of fluids that conform to the common domain.

[0148] Illustrative embodiment 18. The method according to one of the preceding embodiments, wherein the trained machine learning algorithm comprises a machine learning algorithm trained on a training set, and wherein the training set comprises a plurality of pressure traces of normal fluid processes and a plurality of pressure traces of abnormal fluid processes at a ratio of at least 5: 1.

[0149] Illustrative embodiment 19. The method according to one of the preceding embodiments, wherein the trained machine learning algorithm comprises a machine learning algorithm trained on training set, wherein the training set comprises at least one modified pressure trace, and wherein the at least one modified pressure trace comprises one of anunmodified pressure trace having at least one portion excluded and a subset of an unmodified pressure trace associated with an information-rich portion of the unmodified pressure trace.

[0150] Illustrative embodiment 20. A computer program product embodied in a computer readable storage medium comprising software that when executed by a processor cause the processor to: receive, from a pressure sensor, pressure measurement data during a fluid process performed on a fluid, wherein the pressure measurement data comprises pressure data and time data; and determine, by the processor using a trained machine learning algorithm, a quality assessment of the fluid process based on the received pressure measurement data.

[0151] Various of the above-disclosed and other features and functions, or alternatives thereof, may be combined into many other different systems or applications. Various presently unforeseen or unanticipated alternatives, modifications, variations or improvements therein may be subsequently made by those skilled in the art, each of which is also intended to be encompassed by the disclosed embodiments.

Claims

CLAIMSWe claim:1 . A system for assessing a quality of a fluid process performed on a fluid including one of a fluid aspiration and a fluid dispense, the system comprising: a pump; a probe; a connection, wherein the pump and the probe are attached to the connection and a fluid path is formed between the pump and the probe at least in part by the connection; a pressure sensor in fluid communication with the fluid path and configured to sense pressures of the pump during the fluid process, wherein a pressure measurement data comprises the sensed pressures of the pump during the fluid process; a processor; and a memory comprising instructions that are executed by the processor to cause the processor to: receive, from the pressure sensor, pressure measurement data during the fluid process, wherein the pressure measurement data comprises pressure data and time data, and determine, by the processor using a trained machine learning algorithm, the quality of the fluid process based on the received pressure measurement data.

2. The system of claim 1, wherein the system further comprises: an analyzer configured to analyze a mixture comprising the fluid when the quality7of the fluid process is satisfactory.

3. The system of claim 1, wherein the instructions further cause the processor to: cause the pump to operate such that a subsequent fluid process is performed in response to determining that the quality of the fluid process is abnormal.

4. The system of claim 1 , wherein the trained machine learning algorithm comprises a sensitivity threshold, wherein the instructions further cause the processor to: adjust the sensitivity7threshold prior to determining the quality7of the fluid process.

5. The system of claim 1,wherein the trained machine learning algorithm comprises a machine learning algorithm trained on a training set. and wherein the training set comprises a plurality of pressure traces of normal fluid processes and a plurality of pressure traces of abnormal fluid processes at a ratio of at least 5: 1.

6. The system of claim 1 , wherein the trained machine learning algorithm comprises a machine learning algorithm trained on a training set, and wherein the training set comprises a plurality of pressure traces conforming to a common domain such that the trained machine learning algorithm can be used to determine fluid process qualities of fluid processes performed on a plurality of volumes of fluids that also conform to the common domain.

7. The system of claim 6, wherein each of the plurality of pressure traces comprises a plurality of data points, wherein the common domain comprises a number of data points.

8. The system of claim 6, wherein the common domain comprises one of a maximum time value and a minimum pressure value.

9. The system of claim 6, wherein the training set comprises a plurality' of pressure traces of aspirations and a plurality of pressure traces of dispenses.

10. The system of claim 1, wherein the trained machine learning algorithm comprises a machine learning algorithm trained on training set, and wherein the training set comprises at least one modified pressure trace, wherein the at least one modified pressure trace comprises an unmodified pressure trace having at least one portion excluded.

11. The system of claim 10, wherein the at least one portion excluded comprises one of a portion associated with pump deceleration and a portion associated with pump acceleration.

12. The system of claim 1 ,wherein the trained machine learning algorithm comprises a machine learning algorithm trained on training set, and wherein the training set comprises at least one modified pressure trace, wherein the at least one modified pressure trace comprises a subset of an unmodified pressure trace associated with an information-rich portion of the unmodified pressure trace.

13. A method of quality assessment of a fluid process, the method comprising: performing, by a fluid processing system, the fluid process on a fluid, wherein the fluid process comprises one of a fluid aspiration and a fluid dispense; receiving, by a processor, pressure measurement data during the fluid process, wherein the pressure measurement data comprises pressure data and time data; and determining, by the processor using a trained machine learning algorithm, the quality assessment of the fluid process based on the received pressure measurement data.

14. The method of claim 13, further comprising: analyzing a mixture comprising the fluid in response to determining that the quality assessment of the fluid process is satisfactory.

15. The method of claim 13, further comprising: aspirating, by the fluid processing system, an additional fluid process in response to determining the qualify assessment of the fluid process is abnormal.

16. The method of claim 13, wherein the trained machine learning algorithm comprises a sensitivity threshold, wherein the method further comprises: adjusting the sensitivity threshold prior to determining the qualify assessment of the fluid process.

17. The method of claim 13, further comprising: modifying the pressure measurement data to conform to a common domain, wherein the trained machine learning algorithm can be used to determine fluid process qualities of the fluid process performed on a plurality of volumes of fluids that conform to the common domain.

18. The method of claim 13,wherein the trained machine learning algorithm comprises a machine learning algorithm trained on a training set. and wherein the training set comprises a plurality of pressure traces of normal fluid processes and a plurality of pressure traces of abnormal fluid processes at a ratio of at least 5: 1.

19. The method of claim 13, wherein the trained machine learning algorithm comprises a machine learning algorithm trained on training set, wherein the training set comprises at least one modified pressure trace, and wherein the at least one modified pressure trace comprises one of an unmodified pressure trace having at least one portion excluded and a subset of an unmodified pressure trace associated with an information-rich portion of the unmodified pressure trace.

20. A computer program product embodied in a computer readable storage medium comprising software that when executed by a processor cause the processor to: receive, from a pressure sensor, pressure measurement data during a fluid process performed on a fluid, wherein the pressure measurement data comprises pressure data and time data; and determine, by the processor using a trained machine learning algorithm, a quality assessment of the fluid process based on the received pressure measurement data.

Images