Systems and methods for detecting and classifying pre-analytical errors in clinical laboratory diagnostics

Abstract

In some embodiments, a method of creating a machine learning based automated authentication system for clinical laboratory diagnostics includes obtaining a biomarker training dataset including biomarker values for a plurality of biomarkers. The method also includes training the auto-encoder with the biomarker training data set as both an input and a target of the auto-encoder to create a trained auto-encoder. The method further includes deploying the trained auto-encoder in a hospital or laboratory. Numerous other embodiments are provided.

Description

(19) State Intellectual Property Office (12) Invention Patent Application (10) Application Publication Number (43) Application Publication Date (21) Application Number 202480054995.1 (22) Application Date 2024.07.30 (30) Priority Data 63 / 579060 2023.08.28 US (85) PCT International Application Entering National Phase Date 2026.02.26 (86) PCT International Application Application Data PCT / US2024 / 040126 2024.07.30 (87) PCT International Application Publication Data WO2025 / 049026 EN 2025.03.06 (71) Applicant: Siemens Medical Diagnostics, Inc., USA Address: New York, USA (72) Inventors: T. Dudikevich, J. Savina, D. Wood, P. Senas, C. Cooper (74) Patent Agency: China Patent Agency (Hong Kong) Limited, 72001 Patent Attorney: Cai Zhuo, Liu Chunyuan (51) Int.Cl. G01N 33 / 53 (2006.01) G01N 33 / 48 (2006.01) G01N 33 / 64 (2006.01) G16H 50 / 30 (2006.01) (54) Invention Title: System and Method for Detecting and Classifying Pre-analytical Errors in Clinical Laboratory Diagnostics (57) Abstract: In some embodiments, a method for creating a machine learning-based automated certification system for clinical laboratory diagnostics includes obtaining a biomarker training dataset comprising biomarker values ​​of multiple biomarkers. The method further includes training the autoencoder using a biomarker training dataset as both input and target to create a trained autoencoder. The method further includes deploying the trained autoencoder in a hospital or laboratory. Many other embodiments are provided. Claims 4 pages, Description 14 pages, Drawings 12 pages. CN 121844209 A 2026.04.10 CN 1 21 84 42 09 A 1. A method for creating a machine learning-based automated authentication system for clinical laboratory diagnostics, comprising: obtaining a biomarker training dataset, the biomarker training dataset including biomarker values ​​of a plurality of biomarkers; training the autoencoder using the biomarker training dataset as both input and target to create a trained autoencoder; and deploying the trained autoencoder in a hospital or laboratory. 2. The method of claim 1, wherein obtaining the biomarker training dataset comprises obtaining a list of biomarkers to be automatically authenticated from a hospital or laboratory, and creating the biomarker training dataset based on the biomarker list.3. The method of claim 2, wherein creating the biomarker training dataset comprises obtaining a biomarker dataset from a hospital or laboratory and creating the biomarker training dataset based on the biomarker dataset and a biomarker list. 4. The method of claim 3, wherein the biomarker training dataset comprises at least three biomarker values ​​for each biomarker in the biomarker list. 5. The method of claim 2, wherein the biomarker list comprises a list of biomarkers in a biomarker group. 6. The method of claim 5, wherein the biomarker group comprises biomarkers from at least one of a lipid group, a complete blood cell count (CBC) group, a basal metabolic group (BMP) group, a comprehensive metabolic group (CMP) group, a thyroid group, a urine group, and an immunoassay group. 7. The method of claim 5, wherein the biomarker group comprises routine diagnostic tests. 8. The method of claim 7, wherein the routine diagnostic tests comprise biomarkers from the lipid group, the complete blood cell count (CBC) group, and the comprehensive metabolic group (CMP). 9. The method of claim 1, further comprising deploying an error calculation module in a hospital or laboratory, the error calculation module being configured to determine a reconstruction error between a measured biomarker input to a trained autoencoder and a reconstructed biomarker output from the autoencoder in response to the measured biomarker. 10. The method of claim 1, further comprising employing the deployed autoencoder on biomarkers measured for patient samples. 11. The method of claim 10, wherein employing the deployed autoencoder comprises: obtaining a patient sample from a patient; performing at least one test on the patient sample to generate a measured biomarker from the patient sample; inputting the measured biomarker into the autoencoder to generate an output biomarker; and determining whether the measured biomarker is valid based on a comparison of the measured biomarker with the output biomarker. 12. The method of claim 11, wherein the measured biomarker comprises at least one of a lipid group, a complete blood cell count (CBC) group, a basal metabolic group (BMP), a comprehensive metabolic group (CMP), a thyroid group, a urine group, and an immunoassay group. 13. The method of claim 11, wherein the comparison includes determining a reconstruction error between the measured biomarker and the output biomarker. 14. The method of claim 13, wherein determining the reconstruction error includes determining at least one mean squared error value. Claims 1 / 4 page 2 CN 121844209 A 15. The method of claim 11, further comprising determining the cause of the difference between the measured biomarker and the output biomarker.16. The method of claim 15, wherein determining the cause of the difference between the measured biomarker and the output biomarker includes determining a classification of the representation of the measured biomarker. 17. The method of claim 16, wherein the representation of the measured biomarker includes a latent spatial representation of the measured biomarker generated by an autoencoder. 18. The method of claim 16, wherein determining the classification of the representation of the measured biomarker includes: clustering the representation of the measured biomarker into one of a plurality of clusters using a clustering algorithm; and classifying the difference between the measured biomarker and the output biomarker based on said plurality of clusters. 19. The method of claim 18, wherein said plurality of clusters includes a clinical condition cluster and a pollution cluster. 20. A method for creating a machine learning-based automated authentication system for clinical laboratory diagnostics, comprising: obtaining a first biomarker list for automated authentication; obtaining a first training dataset for the first biomarker list, the first training dataset including biomarker values ​​for each biomarker in the first biomarker list; training a first autoencoder using the first training dataset as both input and target to create a first trained autoencoder; obtaining a second biomarker list for automated authentication; obtaining a second training dataset for the second biomarker list, the second training dataset including biomarker values ​​for each biomarker in the second biomarker list; training a second autoencoder using the second training dataset as both input and target to create a second trained autoencoder; and deploying the first trained autoencoder and the second trained autoencoder. 21. The method of claim 20, wherein: obtaining the first biomarker list and the second biomarker list comprises obtaining the first biomarker list and the second biomarker list from a hospital or laboratory; obtaining the first training dataset and the second training dataset comprises obtaining at least one biomarker dataset from a hospital or laboratory, and creating the first training dataset and the second training dataset based on the at least one biomarker dataset; and deploying the first trained autoencoder and the second trained autoencoder comprises deploying the first trained autoencoder and the second trained autoencoder in a hospital or laboratory. 22. The method of claim 21, wherein the first biomarker list comprises a list of biomarkers in a first biomarker group, and the second biomarker list comprises a list of biomarkers in a second biomarker group.23. The method of claim 22, wherein the first and second biomarker groups comprise different sets of biomarkers from at least one of the lipid group, the complete blood cell count (CBC) group, the basal metabolic group (BMP), the comprehensive metabolic group (CMP), the thyroid group, the urine group, and the immunoassay group. 24. A machine learning-based automated authentication system for clinical laboratory diagnostics, comprising: a processor; a memory coupled to the processor, the memory including an autoencoder trained on the biomarker training dataset using both input and target data as an autoencoder; and computer program instructions stored in the memory, which, when executed by the processor, cause the processor to: receive one or more biomarkers measured for a patient; input the measured biomarkers to the autoencoder to generate output biomarkers; and determine whether the measured biomarkers are valid based on a comparison of the measured biomarkers input to the autoencoder with the output biomarkers of the autoencoder. 25. The automated authentication system of claim 24, wherein the biomarker training set comprises a biomarker dataset from a hospital or laboratory. 26. The automated authentication system of claim 25, wherein the autoencoder is deployed at the hospital or laboratory providing the biomarker dataset. 27. The automated authentication system of claim 25, wherein the autoencoder is deployed at a different hospital or laboratory than the hospital or laboratory providing the biomarker dataset. 28. The automated authentication system of claim 24, wherein the biomarker training set comprises biomarkers from a biomarker group. 29. The automated authentication system of claim 28, wherein the biomarker group comprises biomarkers from at least one of a lipid group, a complete blood cell count (CBC) group, a basal metabolic group (BMP) group, a comprehensive metabolic group (CMP) group, a thyroid group, a urine group, and an immunoassay group. 30. The automated authentication system of claim 24, wherein the biomarker training set comprises biomarkers from routine diagnostic tests. 31. The automated authentication system of claim 30, wherein the routine diagnostic test comprises biomarkers from the lipid group, the complete blood cell count (CBC) group, and the comprehensive metabolic group (CMP). 32. The automated authentication system of claim 24, further comprising computer program instructions stored in memory, which, when executed by a processor, cause the processor to determine a reconstruction error between the measured biomarker and the output biomarker.33. The automated authentication system of claim 32, further comprising computer program instructions stored in memory, which, when executed by a processor, cause the processor to determine at least one mean squared error based on the measured biomarker and the output biomarker of the autoencoder. 34. The automated authentication system of claim 24, further comprising computer program instructions stored in memory, which, when executed by a processor, cause the processor to predict the cause of the difference between the measured biomarker and the output biomarker of the autoencoder. 35. The automated authentication system of claim 34, further comprising computer program instructions stored in memory, which, when executed by a processor, cause the processor to determine the classification of the representation of the measured biomarker. 36. The automated authentication system of claim 35, wherein the representation of the measured biomarker includes a latent spatial representation of the measured biomarker, wherein the latent spatial representation is generated by an autoencoder. 37. The automated authentication system of claim 36, further comprising computer program instructions stored in memory, which, when executed by a processor, cause the processor to employ a clustering algorithm to cluster the representations of the measured biomarkers into one of a plurality of clusters; and, based on the plurality of clusters, classify the differences between the measured biomarkers and output biomarkers. 38. The automated authentication system of claim 37, wherein the plurality of clusters includes a clinical condition cluster and a contamination cluster. 39. The automated authentication system of claim 24, further comprising computer program instructions stored in memory, which, when executed by a processor, cause the processor to recommend manual processing to obtain a test for the one or more biomarkers measured for the patient, based on the differences between the measured biomarkers and the output biomarkers of the autoencoder. 40. A method for automatically certifying clinical laboratory diagnostics, comprising: receiving one or more biomarkers measured for a patient; inputting the measured biomarkers into an autoencoder to generate an output biomarker, wherein the autoencoder is trained on a biomarker training dataset using both the biomarker training dataset as input and target to the autoencoder; and determining whether the measured biomarker is valid based on a comparison of the measured biomarkers input to the autoencoder with the output biomarker of the autoencoder. 41. The method of claim 40, wherein the biomarker training dataset comprises a biomarker dataset from a hospital or laboratory, and wherein the autoencoder is deployed at the hospital or laboratory providing the biomarker dataset.42. The method of claim 40, wherein the biomarker training set comprises a biomarker dataset from a first hospital or laboratory, and wherein the autoencoder is deployed at a second hospital or laboratory different from the first hospital or laboratory providing the biomarker dataset. 43. The method of claim 40, wherein the autoencoder is a variational autoencoder. Claims 4 / 4 Page 5 CN 121844209 A System and method for detecting and classifying pre-analytical errors in clinical laboratory diagnostics

[0001] Cross-reference to related applications This application claims the benefit of U.S. Provisional Application No. 63 / 579,060, filed August 28, 2023, under 35 USC § 119(e). The entire contents of one or more of the patent applications cited above are hereby expressly incorporated herein by reference. Technical Field

[0002] This application relates to medical testing, and more particularly, to systems and methods for detecting and classifying pre-analytical errors. Background Art

[0003] Clinical laboratories play a vital role in the diagnosis and treatment of diseases by analyzing patient samples and providing accurate and reliable test results. However, the process of analyzing these samples is not without its challenges. One such challenge is detecting deviations from the expected patient reference range, which can be caused by a variety of factors, including pre-analytical errors.

[0004] Pre-analytical errors refer to any errors or inconsistencies that occur even before the sample is analyzed. These errors can include problems with sample collection, handling, and storage, as well as inadequate labeling or documentation. Pre-analytical errors can have a significant impact on the accuracy and reliability of test results, making it necessary for clinical laboratories to detect and address them before they cause any harm.

[0005] To address this challenge, clinical laboratories have developed a method called automated certification of laboratory results. Automated certification involves using rule-based software algorithms to automatically review and certify the accuracy of laboratory test results before they are reported to patients and healthcare providers. This helps ensure that only accurate and reliable results are reported, thereby reducing the risk of providing incorrect or misleading information.

[0006] Despite the benefits of automated certification, several challenges remain associated with this approach. One such challenge is the cost and scarcity of qualified laboratory personnel who need to manually review any results that do not meet automated accreditation criteria. These personnel are highly trained and experienced professionals, and the shortage makes it difficult for clinical laboratories to maintain a sufficient workforce to meet their patients' needs.

[0007] Furthermore, the manual review process can be both time-consuming and slow, which can further delay the availability of results to patients and healthcare providers.This delay can have serious consequences, especially in emergency situations where timely diagnosis and treatment are crucial.

[0008] Therefore, there is a need for improved methods and apparatus for detecting and resolving pre-analytical errors in clinical laboratories, which are more efficient, accurate, and cost-effective than current methods. Summary of the Invention

[0009] In some embodiments, a method for creating a machine learning-based automated authentication system for clinical laboratory diagnostics includes: obtaining a biomarker training dataset, the biomarker training dataset comprising biomarker values ​​of a plurality of biomarkers; training an autoencoder using both the biomarker training dataset as input and target to create a trained autoencoder; and deploying the trained autoencoder in a hospital or laboratory.

[0010] In some embodiments, a method for creating a machine learning-based automated authentication system for clinical laboratory diagnostics includes: obtaining a first biomarker list for automated authentication; obtaining a first training dataset for the first biomarker list, the first training dataset including biomarker values ​​for each biomarker in the first biomarker list; training a first autoencoder using the first training dataset as both input and target to create a first trained autoencoder; obtaining a second biomarker list for automated authentication; obtaining a second training dataset for the second biomarker list, the second training dataset including biomarker values ​​for each biomarker in the second biomarker list; training a second autoencoder using the second training dataset as both input and target to create a second trained autoencoder; and deploying the first trained autoencoder and the second trained autoencoder.

[0011] In some embodiments, a machine learning-based automated authentication system for clinical laboratory diagnostics includes: a processor; a memory coupled to the processor, the memory including an autoencoder trained on the biomarker training dataset by using a biomarker training dataset as both input and target of the autoencoder; and computer program instructions stored in the memory, which, when executed by the processor, cause the processor to: receive one or more biomarkers measured for a patient; input the measured biomarkers into the autoencoder to generate an output biomarker; and determine whether the measured biomarkers are valid based on a comparison of the measured biomarkers input into the autoencoder with the output biomarkers of the autoencoder.

[0012] In some embodiments, a method for automatically certifying clinical laboratory diagnostics includes: receiving one or more biomarkers measured for a patient; inputting the measured biomarkers into an autoencoder to generate an output biomarker, wherein the autoencoder is trained on the biomarker training dataset by using a biomarker training dataset as both input and target to the autoencoder; and determining whether the measured biomarker is valid based on a comparison of the measured biomarker input to the autoencoder with the output biomarker of the autoencoder.

[0013] Other features and aspects of the invention will become more fully apparent from the following detailed description, the appended claims, and the accompanying drawings. Brief Description of the Drawings

[0014] FIG1A illustrates an example flowchart of a method for training and deploying a machine learning-based automated certification system for clinical laboratory diagnostics according to embodiments provided herein.

[0015] FIG1B illustrates an example computer in which the method of FIG1A may be implemented according to one or more embodiments.

[0016] FIG2 illustrates a flowchart of a method for creating a machine learning-based automated certification system for clinical laboratory diagnostics according to one or more embodiments.

[0017] Figure 3A illustrates an example machine learning model that can be trained using a biomarker training dataset according to embodiments provided herein.

[0018] Figure 3B illustrates an example automated authentication system employing the trained ML model of Figure 3A according to embodiments provided herein.

[0019] Figure 4 illustrates a flowchart of another method for creating a machine learning-based automated authentication system for clinical laboratory diagnostics according to one or more embodiments.

[0020] Figure 5 illustrates a flowchart of an automated authentication method for clinical laboratory diagnostics according to one or more embodiments.

[0021] Figure 6A illustrates a first machine learning-based automated authentication system provided according to one or more embodiments. Specification 2 / 14 pages 7 CN 121844209 A

[0022] Figure 6B illustrates a second machine learning-based automated authentication system provided according to one or more embodiments.

[0023] Figure 7 illustrates an example graph of the mean squared error between a measured biomarker and a predicted (reconstructed) biomarker generated by an ML model (in response to the measured biomarker) according to one or more embodiments.

[0024] Figures 8A and 8B illustrate the planar distribution of the latent spatial representation of biomarkers measured before (Figure 8A) and after (Figure 8B) the application of a clustering model, according to one or more embodiments.

[0025] Figure 9 illustrates a flowchart of a method for automatically authenticating measured biomarkers according to one or more embodiments.

[0026] Figure 10 illustrates a flowchart of a method for analyzing measured biomarkers subjected to IV contamination, according to one or more embodiments.Detailed Description

[0027] Individuals with male or female identities are also included in the terminology, independent of the use of grammatical terms.

[0028] The embodiments provided herein provide systems and methods for detecting and identifying pre-analytical errors in clinical laboratories more efficiently, accurately, and cost-effectively. In some embodiments, a deep neural network (DNN) architecture based on one or more autoencoders in an outlier detection pattern can be utilized within one or more machine learning-based automated authentication systems. Other machine learning models can be employed for outlier detection (e.g., other dimensionality reduction algorithms). In one or more embodiments, an autoencoder model can be trained on a set of valid laboratory results (e.g., biomarkers) from clinical laboratory instruments, middleware, or laboratory information systems (LIS). The use of an autoencoder can also be viewed as a dimensionality reduction method, whereby the autoencoder learns an internal representation of the data in a "latent space." Only important features relevant to the dataset (e.g., multiple biomarkers) are encoded in the latent space, while noise and outliers are suppressed. In a prediction (outlier detection) scenario, a reconstruction error (e.g., the biomarker reconstructed by the autoencoder from the measured biomarker) is computed. A high reconstruction error is an indication of one or more outliers (e.g., invalid samples). In some embodiments, once outliers are identified, a second step can be performed to attempt to cluster them for the root cause of the outliers using machine learning clustering methods.

[0029] In one or more embodiments, an autoencoder-based automated authentication system or other machine learning-based automated authentication system for a hospital or laboratory can be trained on a dataset from a hospital or laboratory using an automated authentication system. Such a dataset represents the target population of the hospital or laboratory (e.g., to eliminate cohort risk for machine learning models).

[0030] In some embodiments, multiple models (e.g., multiple autoencoders) can be trained for different types of blood or other test or biomarker panels. For example, different autoencoders can be trained for each of the lipid panel, complete blood cell count (CBC) panel, basal metabolic panel (BMP) panel, comprehensive metabolic panel (CMP) panel, thyroid panel, immunoassay panel, urine panel, combinations of multiple panels, etc.

[0031] In one or more embodiments, a general model (e.g., utilizing all available input biomarkers) and multiple specific models for different failure models can be trained (e.g., utilizing only a subset of relevant biomarkers for each failure model). As an example, failure models can be developed to detect IV fluid contamination or other sources of error of interest.

[0032] Note that, in general, the systems and methods provided herein are generalizable and can be extended to any clinical laboratory.That is, in some embodiments, the ML model (e.g., an autoencoder) may be trained on biomarkers used for multiple tests and / or on general biomarker data (e.g., from multiple hospitals, laboratories, healthcare data providers, etc.).

[0033] In some embodiments, by creating individual models (e.g., ML models, such as autoencoders trained on a particular set of biomarkers, page 3 / 14 of the specification, 8 CN 121844209 A), fact-based model outcome reports (e.g., the prevalence of disease states x, y, and pre-analysis error z observed for a particular biomarker pattern) can be generated. Traditionally, model outcomes are reported as absence / presence predictions with some confidence bounds, such as a 70% chance that a patient will have colon cancer. The applicability of these confidence bounds depends heavily on a factor called “cohort risk,” which is a theoretical measure of the difference between the model training population and the model prediction population (especially regarding the mathematical relationship between the dependent variable and each independent variable). Cohort risk is often extremely high due to the high variability of biological and ecosystem data, leading to poor models. Fact-based results reporting highlights the most relevant information, allowing practitioners to cross-reference additional information not directly captured in biomarkers, such as symptoms.

[0034] As described above, in some embodiments provided herein, a two-step machine learning approach (e.g., error detection followed by error classification) can be employed. Two-step machine learning approaches can allow for advanced analysis of the possible root causes of errors before error detection and classification. General approaches (e.g., training models on a common biomarker dataset from multiple hospitals or laboratories) can be deployed in any clinical laboratory. Currently, in vitro diagnostics (IVD) and laboratory information systems (LIS) manufacturers offer software solutions that allow laboratories to establish automated certification rules; however, those rules must be developed, monitored, and updated individually by each laboratory. Designing, implementing, and validating such rules is both complex and time-consuming. The machine learning-based automated certification systems and methods described herein are customizable, easily implemented, and superior to rule-based approaches.

[0035] These and other embodiments of the invention are described below with reference to Figures 1A-10.

[0036] Figure 1A illustrates an example flowchart 100a of a method for training and deploying a machine learning-based automated certification system for clinical laboratory diagnostics according to embodiments provided herein. Referring to Figure 1A, and as described in further detail below, the method of flowchart 100a includes obtaining a biomarker dataset 102 (e.g., from one or more hospitals, one or more laboratories, healthcare data providers, etc.). The biomarker dataset 102 is then used to generate a training dataset (biomarker training dataset 104).In some embodiments, biomarker dataset 102 is examined, and biomarkers of interest are identified. For example, if automated authentication is to be performed on blood groups such as lipid groups, CBC groups, BMP, CMP, thyroid groups, or immunoassay groups, biomarkers associated with one or more of these groups can be selected from biomarker dataset 102 to be included in biomarker training dataset 104. Multiple biomarker values ​​(e.g., 3 or more, 10 or more, etc.) can be adopted for each biomarker within biomarker training dataset 104. Biomarker training dataset 104 can then be used to train machine learning (ML) model 106. As described below, in some embodiments, ML model 106 may include one or more autoencoders, although other ML models may be employed. In some embodiments, the entire biomarker dataset 102 can be used as biomarker training dataset 104.

[0037] The trained ML model 106 can be used as a deployed ML model 108 within a clinical laboratory setting to predict whether test results for patient samples are valid (e.g., automatically certified) or whether the results should be manually reviewed, whether the patient samples should be repeated, and / or whether new patient samples should be obtained. As further described below, predictive models such as clustering algorithms can be employed to facilitate the identification of error sources associated with the test (e.g., identifying pre-analysis errors, such as errors during sample collection, disposal, and / or storage, and inadequate labeling or documentation).

[0038] FIG1B illustrates an example computer 120 in which the method of FIG1A can be implemented according to one or more embodiments. Referring to FIG1B, computer 120 includes a processor 122 coupled to memory 124. Memory 124 may include a biomarker dataset 102, a biomarker training dataset 104, and an ML model 106. Memory 124 may also include one or more programs 126 for implementing the methods described herein when executed by processor 122, such as examining a biomarker dataset 102 of biomarkers of interest, creating a biomarker training dataset 104 based on biomarker dataset 102, and so on. In some embodiments, processor 122 executing one or more of the programs 126 may train an ML model 106 based on the biomarker training dataset 104 (e.g., generating a trained ML model 108 (FIG. 1A) that can be deployed in hospitals, laboratories, etc.). Memory 124 may include multiple memory units and / or various types of memory. In some embodiments, all or part of memory 124 may be external to and / or remote from computer 120.Furthermore, in some embodiments, multiple processors may be employed.

[0039] Figure 2 illustrates a flowchart of a method 200 for creating a machine learning-based automated certification system for clinical laboratory diagnostics according to one or more embodiments. Referring to Figure 2, method 200 includes: in block 202, obtaining a biomarker training dataset comprising biomarker values ​​of a plurality of biomarkers. In some embodiments, the biomarker training dataset may be obtained from hospitals, laboratories, healthcare data providers (such as Dandelion Health, Inc., New York (NY), Prognos Health, Inc., New York (NY), etc.), or created based on a biomarker dataset obtained from hospitals, laboratories, healthcare data providers, etc. For example, the biomarker training dataset may be formed from a subset of biomarkers in a larger biomarker dataset (e.g., biomarkers related to the CBC group, BMP, CMP, lipid group, immunoassay group, urine group, another biomarker group, combination of biomarker groups, etc., may be selected from a larger dataset and used to form the biomarker training dataset). Figures 1A and 1B illustrate an example biomarker training dataset 104 formed from biomarker dataset 102.

[0040] In some embodiments, the biomarker training dataset may contain at least one identified biomarker from a plurality of hospitals and a plurality of laboratories. In other embodiments, the biomarker training dataset may contain only biomarkers identified for a particular hospital (e.g., at least one identified biomarker from an emergency room, inpatient clinic, outpatient clinic, physician’s office, and / or laboratory attached to a particular hospital).

[0041] In instances where the biomarker training dataset includes hospital or laboratory data, in some embodiments, biomarkers present in the biomarker training dataset (e.g., validated biomarkers, such as biomarkers validated by a rule-based method) may be obtained from a laboratory information system. In other embodiments, the biomarker training dataset may include (e.g., from middleware) unvalidated biomarkers.

[0042] In one or more embodiments, obtaining the biomarker training dataset may include obtaining a list of biomarkers to be automatically certified (e.g., from hospitals or laboratories) and creating the biomarker training dataset based on the list of biomarkers. For example, a hospital or laboratory might want to automatically certify biomarkers associated with one or more blood subgroups. The hospital or laboratory can provide a list of specific biomarkers it plans to automatically certify. Based on the provided list of biomarkers, a biomarker training dataset can be generated from a more general biomarker dataset.For example, a biomarker dataset can be obtained from a hospital or laboratory, and a biomarker training dataset can be created based on the biomarker dataset and the biomarker list. Alternatively, the biomarker dataset can be provided by another hospital or laboratory, multiple hospitals or laboratories, etc.

[0043] In some embodiments, the biomarker list may include a list of biomarkers in a biomarker group, such as biomarkers from at least one of a lipid group, a CBC group, BMP, CMP, a thyroid group, a urine group, and an immunoassay group. In another example, the biomarker group may include routine diagnostic tests with common biomarkers (e.g., biomarkers from the lipid group, CBC group, and CMP) measured during annual physical examinations, routine health checkups, etc.

[0044] In some embodiments, the biomarker training dataset may include three or more biomarker values ​​for each biomarker in the training dataset. Other numbers of biomarker values ​​may be used (e.g., 4, 5, 10, 20, 50, or more). Instruction manual 5 / 14 pages 10 CN 121844209 A

[0045] After obtaining the biomarker training dataset (e.g., biomarker training dataset 104 of Figures 1A and 1B), in box 204, method 200 includes training the autoencoder (thereby creating a trained autoencoder) using the biomarker training dataset as both input and target to the autoencoder. It will be understood that other ML-based models may be used, such as alternative dimensionality reduction algorithms (e.g., principal component analysis, isomap, t-distributed random neighborhood embedding, linear discriminant analysis, uniform manifold approximation and projection, etc.).

[0046] Figure 3A illustrates an example machine learning model, referred to as initial ML model 300, according to an embodiment provided herein, which can be trained using the biomarker training dataset (e.g., biomarker training dataset 104 of Figures 1A and 1B). In the example embodiment of Figure 3A, the initial ML model 300 is an autoencoder that maps biomarkers within the biomarker training dataset to a reduced-dimensional latent space representation (described further below). Any suitable autoencoder can be employed, such as a plain autoencoder, an undercomplete autoencoder, a denoising autoencoder, a sparse autoencoder, a shrinking autoencoder, a convolutional autoencoder, a recurrent autoencoder, a variational autoencoder, etc. Other dimensionality reduction algorithms can also be employed.

[0047] In some embodiments, the latent space representation can encode data across different scales. For example, normalized flows, autoregressive models, variational autoencoders (VAEs), and deep energy-based models are all examples of deep generative learning.Furthermore, the latent space can be constrained to a known parametric distribution (e.g., Gaussian or Gaussian mixture) or a nonparametric distribution (such as using a vector quantization variational autoencoder (VQ-VAE)). It will be understood that other ML models can be employed.

[0048] Referring to FIG3A, ML model 300 includes an encoder 302 configured to map input biomarkers to a latent space L and a decoder 304 configured to map the latent space L to output biomarkers (e.g., reconstructed biomarkers). A biomarker training dataset can be input to encoder 302 to generate latent features in the latent variable LA in a multidimensional latent space (e.g., generating latent features in a low-dimensional space). The features of the latent variable LA can be segmented into multiple sets of features, such as W, X, Y, and Z as shown in FIG3A. Each of these features may correspond to a specific attribute in the properties of the biomarker in the biomarker training dataset input to encoder 302. In some embodiments, at least one of these features (e.g., feature Z) may retain intrinsic properties for the biomarker dataset, where variations cannot be controlled by setting.

[0049] During training, decoder 304 can use features of latent variable LA to reconstruct the biomarker input to encoder 302 (e.g., as similar as possible to the input biomarker). Reconstruction loss module 306 can compare the input biomarker with the output (reconstructed) biomarker to determine if the input and output biomarkers are close to each other. For example, reconstruction loss module 306 can analyze the properties of the input and output biomarkers to determine if latent variable LA is correct and if the encoder model (encoder 302) and decoder model (decoder 304) are correctly trained.

[0050] Once the ML model (e.g., the autoencoder in the embodiment of FIG2) has been trained using the biomarker training dataset as both input and target, method 200 includes, in block 206, deploying the trained autoencoder in a hospital or laboratory. In some embodiments, the trained autoencoder may be deployed in the same hospital or laboratory that supplies the biomarker dataset used to generate the biomarker training dataset.

[0051] FIG3B illustrates an example automated authentication system 350 employing the trained ML model 300 of FIG3A according to an embodiment provided herein. Referring to FIG3B, the trained ML model 300 (e.g., the autoencoder in the embodiment of FIG3B) can be employed for automated authentication by inputting a measured biomarker 352 into the trained model 300 to generate an output (reconstructed) biomarker 354.Error calculation module 356 can be used to determine the amount of reconstruction error between the measured biomarker 352 input to the trained ML model 300 and the output (reconstructed) biomarker 354 output from the trained ML model 300. In some embodiments, reconstruction error can represent one or more mean square error values ​​of the biomarker. High reconstruction error can identify a patient sample as invalid, while low reconstruction error can indicate that the patient sample is valid.

[0052] As further described below, in some embodiments, if a sample is determined to be invalid, a classification model can be used to predict the source of error associated with the patient sample. In the embodiment of FIG3B, a latent space representation (e.g., latent variable LA) can be used in conjunction with clustering model 358 to classify outlier latent space features into clusters associated with pre-analysis errors (e.g., clinical condition clusters, contamination clusters, or any other relevant clusters). Clustering model 358 may include one or more clustering algorithms. For example, clustering algorithms such as k-means clustering or other dimensionality reduction and / or grouping algorithms can be used. Other example clustering or grouping algorithms that can be used include t-distributed random neighborhood embedding, isometric feature map embedding, locally linear embedding, spectral embedding, multidimensional scaling, latent Dirichlet assignment, etc.

[0053] In some embodiments, it may be desirable for hospitals or laboratories to employ automated authentication models based on multiple machine learning trained on different biomarker sets. For example, separate autoencoders can be trained for different combinations of CBC groups, BMP, CMP, lipid groups, immunoassay groups, biomarker groups, etc.

[0054] Figure 4 illustrates a flowchart of another method 400 for creating a machine learning-based automated authentication system for clinical laboratory diagnostics according to one or more embodiments. Referring to Figure 4, method 400 includes: in box 402, obtaining a first biomarker list for automatic authentication; in box 404, obtaining a first training dataset for the first biomarker list, the first training dataset including the biomarker value of each biomarker in the first biomarker list; and in box 406, training a first autoencoder using the first training dataset as both input and target to create a first trained autoencoder. Method 400 further includes: in box 408, obtaining a second biomarker list for automatic authentication; in box 410, obtaining a second training dataset for the second biomarker list, the second training dataset including the biomarker value of each biomarker in the second biomarker list; and in box 412, training a second autoencoder using the second training dataset as both input and target to create a second trained autoencoder.Subsequently, method 400 includes, in block 414, deploying a first trained autoencoder and a second trained autoencoder.

[0055] In some embodiments, the first biomarker list and the second biomarker list may be obtained from a hospital or laboratory (e.g., a hospital or laboratory that plans to use the trained autoencoder). The first training dataset and the second training dataset may be created from one or more biomarker datasets obtained from a hospital or laboratory that plans to use the trained autoencoder, or from a different source (such as another hospital or laboratory, a healthcare data provider, etc.).

[0056] In one or more embodiments, the first biomarker list may include a list of biomarkers in a first biomarker group, and the second biomarker list may include a list of biomarkers in a second biomarker group. For example, the first list and the second list may include different sets of biomarkers from at least one of the lipid group, CBC group, BMP, CMP, thyroid group, urine group, immunoassay group, etc.

[0057] FIG5 illustrates a flowchart of a method 500 for automated certification of clinical laboratory diagnostics according to one or more embodiments. Referring to Figure 5, method 500 includes: in block 502, receiving one or more biomarkers measured for a patient. For example, patient samples can be collected from patients at hospitals, inpatient clinics, outpatient clinics, walk-in laboratories, etc. Tests can then be performed to measure one or more biomarkers of the patient sample, such as CBC group, BMP, CMP, thyroid group, urine group, immunoassay group, another biomarker group, etc. Subsequently, method 500 includes: in block 504, inputting the measured biomarkers into an autoencoder to generate output biomarkers, wherein the autoencoder is trained on the biomarker training dataset using both the input and target of the autoencoder. As described above, in some embodiments, the biomarker training set used to train the autoencoder may come from a hospital or laboratory in which the autoencoder is deployed. In other embodiments, the biomarker training set may come from one or more different hospitals and / or laboratories.

[0058] Method 500 further includes, in block 506, determining whether the measured biomarker is valid based on a comparison between the measured biomarker input to the autoencoder and the output biomarker of the autoencoder.In some embodiments, determining whether a measured biomarker is valid may include determining the reconstruction error between the measured biomarker and the output biomarker (e.g., at least one mean squared error value or another error measure based on the measured biomarker from the patient sample and the output (reconstructed) biomarker from the autoencoder in response to the measured biomarker).

[0059] Because the autoencoder is trained using the same biomarker as both the input and target, the autoencoder will result in a small reconstruction error if the biomarker measured from the patient sample to the autoencoder is similar to a biomarker in the biomarker training dataset. Furthermore, if the biomarker training dataset used to train the autoencoder includes validated biomarkers (e.g., from one or more laboratory information systems), a low reconstruction error between the biomarker measured from the patient sample and the (reconstructed) biomarker output from the autoencoder in response to the measured biomarker indicates that the measured biomarker is valid. Similarly, a high reconstruction error between a biomarker measured from a patient sample and a biomarker output from an autoencoder in response to the measured biomarker indicates that the measured biomarker is invalid (e.g., manual inspection, re-acquiring the patient sample, etc. may be required). In this way, a trained autoencoder can be used for automated authentication of patient biomarkers.

[0060] FIG6A illustrates a first machine learning-based automated authentication system 600a provided according to one or more embodiments. Referring to FIG6A, the first machine learning-based automated authentication system 600a can communicate with a first analyzer 602a configured to determine one or more biomarkers (e.g., creatinine, BUN, BUN / creatinine ratio, chloride, red blood cell width (RDW), sodium, albumin, hematocrit, platelet count, red blood cell or white blood cell count, neutrophils, lymphocytes, eosinophils, etc.) of a patient blood sample. The resulting biomarkers are fed to processor 604 and passed through a deployed ML model 606 (e.g., a trained autoencoder, such as trained ML model 300 of FIG. 3B) stored in memory 608. ML model 606 takes the measured biomarkers as input and provides predicted (reconstructed) biomarkers as output.

[0061] Error calculation module 610 (labeled as error calculator in FIG. 6A) can be used to determine the amount of reconstruction error (e.g., the mean squared error value of the biomarker) between the measured biomarkers input to ML model 606 and the reconstructed biomarkers output from ML model 606. High reconstruction error can identify an invalid patient sample, while low reconstruction error can indicate that the patient sample is valid.Figure 7 illustrates an example graph of the mean squared error between a measured biomarker and a predicted (reconstructed) biomarker generated by ML model 606 (in response to the measured biomarker) according to one or more embodiments. As shown in Figure 7, the mean squared error values ​​for most normal biomarkers are below a predefined threshold (e.g., approximately 0.013 in Figure 7, although any suitable threshold may be used). Most abnormal biomarkers have mean squared error values ​​above the predefined threshold.

[0062] Error calculation module 610 (within memory 608) may include computer program instructions that, when executed by processor 604, cause processor 604 to determine the reconstruction error between the measured biomarker and the output (reconstructed) biomarker of the trained ML model 606 (e.g., determining at least one mean squared error value based on the measured biomarker and the output biomarker of the trained autoencoder).

[0063] Clustering model 612 may be used to predict the source of error for the measured biomarker of a patient. For example, in some embodiments described on page 8 / 14 of a specification, CN 121844209 A, the latent spatial representation of the measured biomarker (e.g., the latent variable LA in Figure 3B) may be employed, together with clustering model 612 (e.g., similar to clustering model 358 in Figure 3B), to classify the latent spatial features of outliers into clusters associated with pre-analytical errors (e.g., clinical condition clusters, pollution clusters, and / or any other relevant clusters). Clustering model 612 may include one or more clustering algorithms (as described above with respect to clustering model 358).

[0064] In one or more embodiments, memory 608 may include one or more programs 614 having computer program instructions that, when executed by processor 604, cause processor 604 to employ clustering model 612 to predict the cause of the difference between the measured biomarker input to ML model 606 (e.g., a trained autoencoder) and the output biomarker obtained from the result of ML model 606. For example, one or more procedures 614 may use clustering model 612 to determine the classification of the representation of the measured biomarker (such as the latent spatial representation of the measured biomarker generated by ML model 606). In some embodiments, clustering model 612 may cluster the latent spatial representation of the measured biomarker into one of a plurality of clusters that classify (e.g., predict) sources of reconstruction error between the measured biomarker and the output (reconstructed) biomarker. Example clusters include clinical condition clusters, contamination clusters, and / or any other relevant clusters that identify sources of error.

[0065] Figures 8A and 8B illustrate the planar distribution of the latent spatial representations (e.g., features W, X, Y, and Z of Figure 3B) of biomarkers measured before (Figure 8A) and after (Figure 8B) the application of clustering model 612 (Figure 6A) according to one or more embodiments. As shown in Figures 8A and 8B, clustering model 612 spatially divides latent features 800 (Figure 8A) into multiple clusters 802, 804, 806, and 808 (Figure 8B). (More generally, the clusters are separated in the feature space.) Each cluster may represent a specific source of pre-analysis error (e.g., clinical condition, contamination, erroneous samples, etc.). For example, such clusters can be determined by training clustering model 612 on a training dataset that includes known pre-analysis errors.

[0066] Therefore, in some embodiments, the program 614(one or more) stored in memory 608 may include computer program instructions that cause processor 604 to employ a clustering algorithm to cluster the representation of the measured biomarker into one of a plurality of clusters, which may be used to classify (e.g., predict the cause) differences (e.g., reconstruction errors) between the measured biomarker and the output biomarker of a trained ML model in response to the measured biomarker. For example, after using clustering model 612, a prediction of the source of error for the measured biomarker can be determined by identifying the cluster(s) in which the potential representation of the measured biomarker exists.

[0067] FIG6B illustrates a second machine learning-based automated authentication system 600b provided according to one or more embodiments. Referring to FIG6B, the second machine learning-based automated authentication system 600b may communicate with a first analyzer 602a (and / or a second analyzer 602b or other analyzers not shown) configured to determine one or more biomarkers of a patient sample. The resulting biomarkers are transmitted to processor 604 and processed by a first ML model 606 stored in memory 608. For example, the first ML model 606 may be a first autoencoder trained on a biomarker dataset of a first blood group. The first ML model 606 may be used to automatically authenticate biomarkers measured by a first analyzer 602a (or another analyzer), such as by generating output (reconstructed) biomarkers in response to the measured biomarkers input to the first ML model 606.For example, one or more programs 614 executed by processor 604 may receive biomarkers from first analyzer 602a (e.g., directly, from medical personnel, or via a central computer system), use first ML model 606 to generate reconstructed biomarkers based on the biomarkers received from first analyzer 602a, determine the validity of the biomarkers (e.g., using error calculation module 610 as described above), predict the cause if the biomarkers are invalid (e.g., using clustering model 612 as described above), and / or make recommendations.

[0068] Multiple uniquely trained ML models 606a-606n (e.g., uniquely trained autoencoders) may be employed to automatically authenticate different biomarker groups (e.g., using error calculation model 610) and identify sources of error in such biomarker groups (e.g., using clustering model 612 or other clustering models).

[0069] FIG9 illustrates a flowchart of a method 900 for automatically authenticating a measured biomarker according to one or more embodiments. Referring to FIG9, in block 902, method 900 includes obtaining a reconstruction error between the measured biomarker and an output (reconstructed) biomarker generated by an ML model in response to the measured biomarker. For example, the measured biomarker may be input into a trained autoencoder (such as a generally trained or specially trained autoencoder) to generate an output (reconstructed) biomarker. The amount of reconstruction error between the input biomarker and the output biomarker may then be determined.

[0070] In block 904, method 900 includes determining whether the reconstruction error is high (e.g., above a predetermined threshold). If the reconstruction error is not high, the sample is determined to be valid (block 906); otherwise, in block 908, clustering is performed based on the measured biomarker. For example, clustering algorithms (as described above) may be used to cluster a dimensionality-reduced representation of the measured biomarker (such as a latent space representation generated by an autoencoder).

[0071] Method 900 includes: in block 910, determining whether clustering results in a distinct pattern (e.g., whether at least one defined cluster is generated). If no distinct pattern is generated, no determination is made regarding the source of error in the sample (block 912); otherwise, if a distinct pattern is observed after clustering, a determination can be made regarding which cluster(s) the representation of the measured biomarker exists in, such as a first outlier category (e.g., cluster) representing sample error due to clinical condition (block 914), an nth outlier category representing sample contamination (block 916), and / or the like. Any number of outlier categories may be employed.

[0072] FIG10 illustrates a flowchart of a method 1000 for analyzing measured biomarkers contaminated with IV according to one or more embodiments. Referring to FIG10, in block 1002, method 1000 includes obtaining a reconstruction error between the measured biomarker and an output (reconstructed) biomarker generated by an ML model in response to the measured biomarker. For example, the measured biomarker may be input into a trained autoencoder (such as a generally trained or specially trained autoencoder) to generate an output (reconstructed) biomarker. The amount of reconstruction error between the input biomarker and the output biomarker may then be determined.

[0073] In block 1004, method 1000 includes determining whether the reconstruction error is high (e.g., above a predetermined threshold). If the reconstruction error is not high, the sample is determined to be valid (block 1006); otherwise, in block 1008, clustering is performed based on the measured biomarker. For example, clustering algorithms (as described above) can be used to cluster the dimensionality-reduced representation of the measured biomarker (such as a latent spatial representation generated by an autoencoder).

[0074] Method 1000 includes: in block 1010, determining whether clustering results in a distinguishing pattern (e.g., whether at least one defined cluster is generated). If no distinguishing pattern is generated, no determination is made regarding the source of error in the samples (block 1012); otherwise, if a distinguishing pattern is observed after clustering, a determination can be made regarding which (which) clusters the representation of the measured biomarker exists in. In some embodiments, a first outlier category (e.g., a cluster) may represent a first level or type of IV contamination (block 1014), a second outlier category may represent a second level or type of IV contamination (block 1016), and a third outlier category may represent a third level or type of IV contamination (block 1018). Outlier categories of any number and / or type can be employed (e.g., clustering for additional levels of IV contamination, additional types of IV contamination, specific biomarkers, and / or biomarker relationships as outliers, etc.).

[0075] In some embodiments, clustering (e.g., outlier categories) can be developed to identify diseases and / or other clinical conditions, such as acute kidney injury (AKI), myocardial infarction, segmental elevation myocardial infarction (STEMI), non-ST elevation myocardial infarction (NSTEMI), unstable angina, acute coronary syndrome, heart failure, atrial fibrillation, acute liver failure, etc.

[0076] In some embodiments, dimensionality reduction models and clustering models can initially be trained to identify only a few clinical conditions.The dimensionality reduction model and / or clustering model can be retrained to identify additional clinical conditions (e.g., from the use of the trained model in a real-world, clinical setting) when training data becomes available. Additionally, the dimensionality reduction model and / or clustering model deployed in a hospital or laboratory can initially be trained using training data from one or more different hospitals or laboratories (or healthcare data providers). Subsequently, the dimensionality reduction model and / or clustering model can be retrained on data from the location where the model was deployed (when training data becomes available).

[0077] In some embodiments, the dimensionality reduction model and clustering model can be trained to aid in the microscopic analysis of patient samples. For example, the dimensionality reduction model and clustering model can be trained to identify, quantify, label, or otherwise facilitate the analysis of one or more features of cells, morphological elements, bacteria, parasites, etc., such as differential leukocyte identification, red blood cell morphology, body fluid identification, microscopic urine analysis, Gram staining, parasite smears, pathological smears, etc.

[0078] While some embodiments of autoencoders have been described herein, it will be understood that other dimensionality reduction models, such as normalized flow, autoregressive models, deep energy-based models, generative adversarial networks (GANs), other deep generative learning models, etc., may be employed.

[0079] Non-limiting illustrative embodiments Illustrative embodiment 1. A method for creating a machine learning-based automated authentication system for clinical laboratory diagnostics, comprising: obtaining a biomarker training dataset, the biomarker training dataset including biomarker values ​​of a plurality of biomarkers; training an autoencoder using the biomarker training dataset as both input and target of an autoencoder to create a trained autoencoder; and deploying the trained autoencoder in a hospital or laboratory.

[0080] Illustrative embodiment 2. The method according to illustrative embodiment 1, wherein obtaining the biomarker training dataset includes obtaining a list of biomarkers to be automatically authenticated from a hospital or laboratory, and creating the biomarker training dataset based on the list of biomarkers.

[0081] Illustrative embodiment 3. According to the method of one of the foregoing embodiments, creating a biomarker training dataset includes obtaining a biomarker dataset from a hospital or laboratory, and creating the biomarker training dataset based on the biomarker dataset and a biomarker list.

[0082] Illustrative Embodiment 4. According to the method of one of the foregoing embodiments, the biomarker training dataset includes at least three biomarker values ​​for each biomarker in the biomarker list.

[0083] Illustrative Embodiment 5. According to the method of one of the foregoing embodiments, the biomarker list includes a list of biomarkers in a biomarker group.

[0084] Illustrative Example 6. The method according to one of the foregoing embodiments, wherein the biomarker group comprises biomarkers from at least one of the lipid group, the complete blood cell count (CBC) group, the basal metabolic group (BMP), the comprehensive metabolic group (CMP), the thyroid group, the urine group, and the immunoassay group.

[0085] Illustrative Example 7. The method according to one of the foregoing embodiments, wherein the biomarker group comprises routine diagnostic tests.

[0086] Illustrative Example 8. The method according to one of the foregoing embodiments, wherein the routine diagnostic tests comprise biomarkers from the lipid group, the complete blood cell count (CBC) group, and the comprehensive metabolic group (CMP).

[0087] Illustrative Example 9. The method according to one of the foregoing embodiments further comprises deploying an error calculation module in a hospital or laboratory, the error calculation module being configured to determine a reconstruction error between a measured biomarker input to a trained autoencoder and a reconstructed biomarker output from the autoencoder in response to the measured biomarker.

[0088] Illustrative Example 10. The method according to one of the foregoing embodiments further includes employing a deployed autoencoder on the biomarker measured for the patient sample.

[0089] Illustrative Example 11. The method according to one of the foregoing embodiments, wherein employing the deployed autoencoder includes: obtaining a patient sample from a patient; performing at least one test on the patient sample to generate a measured biomarker from the patient sample; inputting the measured biomarker into the autoencoder to generate an output biomarker; and determining whether the measured biomarker is valid based on a comparison of the measured biomarker with the output biomarker.

[0090] Illustrative Example 12. The method according to one of the foregoing embodiments, wherein the measured biomarker includes at least one of the lipid group, the complete blood cell count (CBC) group, the basal metabolic group (BMP), the comprehensive metabolic group (CMP), the thyroid group, the urine group, and the immunoassay group.

[0091] Illustrative Example 13. According to the method of one of the foregoing embodiments, the comparison includes determining a reconstruction error between the measured biomarker and the output biomarker.

[0092] Illustrative Example 14. According to the method of one of the foregoing embodiments, determining the reconstruction error includes determining at least one mean squared error value.

[0093] Illustrative Example 15. According to the method of one of the foregoing embodiments, the method further includes determining the cause of the difference between the measured biomarker and the output biomarker.

[0094] Illustrative Example 16.According to the method of one of the foregoing embodiments, the reason for determining the difference between the measured biomarker and the output biomarker includes determining the classification of the representation of the measured biomarker.

[0095] Illustrative Example 17. According to the method of one of the foregoing embodiments, the representation of the measured biomarker includes a latent spatial representation of the measured biomarker generated by an autoencoder.

[0096] Illustrative Example 18. According to the method of one of the foregoing embodiments, determining the classification of the representation of the measured biomarker includes: clustering the representation of the measured biomarker into one of a plurality of clusters using a clustering algorithm; and classifying the difference between the measured biomarker and the output biomarker based on the plurality of clusters.

[0097] Illustrative Example 19. According to the method of one of the foregoing embodiments, the plurality of clusters includes a clinical condition cluster and a pollution cluster.

[0098] Illustrative Example 20. A method for creating a machine learning-based automated authentication system for clinical laboratory diagnostics includes: obtaining a first biomarker list for automated authentication; obtaining a first training dataset for the first biomarker list, the first training dataset including biomarker values ​​for each biomarker in the first biomarker list; training a first autoencoder using the first training dataset as both input and target to create a first trained autoencoder; obtaining a second biomarker list for automated authentication; obtaining a second training dataset for the second biomarker list, the second training dataset including biomarker values ​​for each biomarker in the second biomarker list; training a second autoencoder using the second training dataset as both input and target to create a second trained autoencoder; and deploying the first trained autoencoder and the second trained autoencoder.

[0099] Illustrative Example 21. According to the method of illustrative embodiment 20, wherein: obtaining the first biomarker list and the second biomarker list includes obtaining the first biomarker list and the second biomarker list from a hospital or laboratory; obtaining the first training dataset and the second training dataset includes obtaining at least one biomarker dataset from a hospital or laboratory, and creating the first training dataset and the second training dataset based on the at least one biomarker dataset; and deploying the first trained autoencoder and the second trained autoencoder includes deploying the first trained autoencoder and the second trained autoencoder in a hospital or laboratory. Specification 12 / 14 pages 17 CN 121844209 A

[0100] Illustrative embodiment 22.According to the method of one of the foregoing embodiments, the first biomarker list includes a list of biomarkers in a first biomarker group, and the second biomarker list includes a list of biomarkers in a second biomarker group.

[0101] Illustrative Example 23. According to the method of one of the foregoing embodiments, the first biomarker group and the second biomarker group include different sets of biomarkers from at least one of the lipid group, the complete blood cell count (CBC) group, the basal metabolic group (BMP), the comprehensive metabolic group (CMP), the thyroid group, the urine group, and the immunoassay group.

[0102] Illustrative Example 24. An automated authentication system based on machine learning for clinical laboratory diagnostics includes: a processor; a memory coupled to the processor, the memory including an autoencoder trained on the biomarker training dataset by using a biomarker training dataset as both input and target of the autoencoder; and computer program instructions stored in the memory, which, when executed by the processor, cause the processor to: receive one or more biomarkers measured for a patient; input the measured biomarkers into the autoencoder to generate an output biomarker; and determine whether the measured biomarkers are valid based on a comparison of the measured biomarkers input into the autoencoder with the output biomarkers of the autoencoder.

[0103] Illustrative Example 25. An automated authentication system according to Illustrative Example 24, wherein the biomarker training set includes a biomarker dataset from a hospital or laboratory.

[0104] Illustrative Example 26. An automated authentication system according to one of the foregoing embodiments, wherein the autoencoder is deployed at a hospital or laboratory providing the biomarker dataset.

[0105] Illustrative Example 27. According to one of the preceding embodiments, the automated authentication system wherein the autoencoder is deployed at a different hospital or laboratory than the hospital or laboratory providing the biomarker dataset.

[0106] Illustrative Example 28. According to one of the preceding embodiments, the automated authentication system wherein the biomarker training set comprises biomarkers from a biomarker group.

[0107] Illustrative Example 29. According to one of the preceding embodiments, the automated authentication system wherein the biomarker group comprises biomarkers from at least one of a lipid group, a complete blood cell count (CBC) group, a basal metabolic group (BMP), a comprehensive metabolic group (CMP), a thyroid group, a urine group, and an immunoassay group.

[0108] Illustrative Example 30. According to one of the preceding embodiments, the automated authentication system wherein the biomarker training set comprises biomarkers from routine diagnostic tests.

[0109] Illustrative Example 31.The automated authentication system according to one of the foregoing embodiments further includes, when the routine diagnostic tests include biomarkers from the lipid group, the complete blood cell count (CBC) group, and the comprehensive metabolic group (CMP).

[0110] Illustrative Example 32. The automated authentication system according to one of the foregoing embodiments further includes computer program instructions stored in memory, which, when executed by a processor, cause the processor to determine a reconstruction error between the measured biomarker and the output biomarker.

[0111] Illustrative Example 33. The automated authentication system according to one of the foregoing embodiments further includes computer program instructions stored in memory, which, when executed by a processor, cause the processor to determine at least one mean square error based on the measured biomarker and the output biomarker of the autoencoder.

[0112] Illustrative Example 34. The automated authentication system according to one of the foregoing embodiments further includes computer program instructions stored in memory, which, when executed by a processor, cause the processor to predict the cause of the difference between the measured biomarker and the output biomarker of the autoencoder.

[0113] Illustrative Example 35. The automated authentication system according to one of the foregoing embodiments further includes computer program instructions stored in memory on pages 13 / 14 of the specification (CN 121844209 A), which, when executed by a processor, cause the processor to determine the classification of the representation of the measured biomarker.

[0114] Illustrative Embodiment 36. The automated authentication system according to one of the foregoing embodiments, wherein the representation of the measured biomarker includes a latent spatial representation of the measured biomarker, wherein the latent spatial representation is generated by an autoencoder.

[0115] Illustrative Embodiment 37. The automated authentication system according to one of the foregoing embodiments, further includes computer program instructions stored in memory, which, when executed by a processor, cause the processor to cluster the representation of the measured biomarker into one of a plurality of clusters using a clustering algorithm; and to classify the differences between the measured biomarker and the output biomarker based on the plurality of clusters.

[0116] Illustrative Embodiment 38. The automated authentication system according to one of the foregoing embodiments, wherein the plurality of clusters includes a clinical condition cluster and a contamination cluster.

[0117] Illustrative Example 39. The automated authentication system according to one of the foregoing embodiments further includes computer program instructions stored in memory, which, when executed by a processor, cause the processor to recommend manual processing to obtain a test for the one or more biomarkers to be measured for the patient based on the difference between the measured biomarker and the output biomarker of the autoencoder.

[0118] Illustrative Example 40. A method for automatically certifying a clinical laboratory diagnosis, comprising: receiving one or more biomarkers measured for a patient; inputting the measured biomarkers into an autoencoder to generate an output biomarker, wherein the autoencoder is trained on a biomarker training dataset using both the biomarker training dataset as input and target to the autoencoder; and determining whether the measured biomarker is valid based on a comparison of the measured biomarker input to the autoencoder with the output biomarker of the autoencoder.

[0119] Illustrative Example 41. The method according to Illustrative Example 40, wherein the biomarker training dataset comprises a biomarker dataset from a hospital or laboratory, and wherein the autoencoder is deployed at the hospital or laboratory providing the biomarker dataset.

[0120] Illustrative Example 42. The method according to one of the preceding embodiments, wherein the biomarker training dataset comprises a biomarker dataset from a first hospital or laboratory, and wherein the autoencoder is deployed at a second hospital or laboratory different from the first hospital or laboratory providing the biomarker dataset.

[0121] Illustrative Example 43. According to the method described in one of the foregoing embodiments, the autoencoder is a variational autoencoder.

[0122] The foregoing description discloses only exemplary embodiments of the invention. Modifications to the apparatus and methods disclosed above that fall within the scope of the invention will be readily apparent to those skilled in the art.

[0123] Therefore, although the invention has been disclosed in conjunction with exemplary embodiments thereof, it should be understood that other embodiments may fall within the spirit and scope of the invention as defined by the following claims.Instruction manual 14 / 14 page 19 CN 121844209 A Figure 1A Figure 1B Instruction manual drawing 1 / 12 page 20 CN 121844209 A Figure 2 Instruction manual drawing 2 / 12 page 21 CN 121844209 A Figure 3A Instruction manual drawing 3 / 12 page 22 CN 121844209 A Figure 3B Instruction manual drawing 4 / 12 page 23 CN 121844209 A Figure 4 Instruction manual drawing 5 / 12 page 24 CN 121844209 A Figure 5 Instruction manual drawing 6 / 12 page 25 CN 121844209 A Figure 6A Figure 6B Instruction manual drawing 7 / 12 page 26 CN 121844209 A Figure 7 Instruction manual drawing 8 / 12 page 27 CN 121844209 A Figure 8A Instruction manual drawing 9 / 12 page 28 CN 121844209 A Figure 8B Instruction Manual Drawings 10 / 12 Page 29 CN 121844209 A Figure 9 Instruction Manual Drawings 11 / 12 Page 30 CN 121844209 A Figure 10 Instruction Manual Drawings 12 / 12 Page 31 CN 121844209 A.

Claims

1. A method for creating a machine learning-based automated authentication system for clinical laboratory diagnostics, comprising: Obtain the biomarker training dataset, which includes biomarker values ​​for multiple biomarkers; The autoencoder is trained by using a biomarker training dataset as both input and target data. as well as Deploy trained autoencoders in hospitals or laboratories.

2. The method of claim 1, wherein obtaining the biomarker training dataset comprises obtaining a list of biomarkers to be automatically certified from a hospital or laboratory, and creating the biomarker training dataset based on the list of biomarkers.

3. The method of claim 2, wherein creating the biomarker training dataset comprises obtaining a biomarker dataset from a hospital or laboratory, and creating the biomarker training dataset based on the biomarker dataset and the biomarker list.

4. The method of claim 3, wherein the biomarker training dataset comprises at least three biomarker values ​​for each biomarker in the biomarker list.

5. The method of claim 2, wherein the biomarker list comprises a list of biomarkers in a biomarker group.

6. The method of claim 5, wherein the biomarker group comprises biomarkers from at least one of the lipid group, the complete blood cell count (CBC) group, the basal metabolic group (BMP) group, the comprehensive metabolic group (CMP) group, the thyroid group, the urine group, and the immunoassay group.

7. The method of claim 5, wherein the biomarker group comprises routine diagnostic tests.

8. The method of claim 7, wherein the routine diagnostic test comprises biomarkers from the lipid group, the complete blood count (CBC) group, and the comprehensive metabolic group (CMP).

9. The method of claim 1, further comprising deploying an error calculation module in a hospital or laboratory, the error calculation module being configured to determine a reconstruction error between a measured biomarker input to a trained autoencoder and a reconstructed biomarker output from the autoencoder in response to the measured biomarker.

10. The method of claim 1, further comprising employing the deployed autoencoder on the biomarkers measured for the patient sample.

11. The method of claim 10, wherein employing the deployed autoencoder comprises: Obtain patient samples from patients; Perform at least one test on the patient sample to generate the measured biomarker from the patient sample; The measured biomarkers are input into an autoencoder to generate output biomarkers; as well as The effectiveness of the measured biomarker is determined by comparing it with the output biomarker.

12. The method of claim 11, wherein the biomarker measured comprises at least one of the lipid group, complete blood cell count (CBC) group, basal metabolic group (BMP) group, comprehensive metabolic group (CMP) group, thyroid group, urine group, and immunoassay group.

13. The method of claim 11, wherein the comparison includes determining the reconstruction error between the measured biomarker and the output biomarker.

14. The method of claim 13, wherein determining the reconstruction error includes determining at least one mean square error value.

15. The method of claim 11, further comprising determining the cause of the difference between the measured biomarker and the output biomarker.

16. The method of claim 15, wherein determining the cause of the difference between the measured biomarker and the output biomarker includes determining the classification of the representation of the measured biomarker.

17. The method of claim 16, wherein the representation of the measured biomarker comprises a latent spatial representation of the measured biomarker generated by an autoencoder.

18. The method of claim 16, wherein determining the classification of the representation of the measured biomarker comprises: Clustering algorithms are used to cluster the representations of the measured biomarkers into one of multiple clusters; as well as The differences between the measured biomarkers and the output biomarkers are classified based on the multiple clusters mentioned above.

19. The method of claim 18, wherein the plurality of clusters includes clinical condition clusters and contamination clusters.

20. A method for creating a machine learning-based automated authentication system for clinical laboratory diagnostics, comprising: Obtain a list of primary biomarkers for automated authentication; Obtain a first training dataset for a first biomarker list, the first training dataset including the biomarker value of each biomarker on the first biomarker list; The first training dataset is used as both the input and target of the first autoencoder to train the first autoencoder, thereby creating the first trained autoencoder; Obtain a list of second biomarkers for automated authentication; Obtain a second training dataset for the second biomarker list, the second training dataset including the biomarker value of each biomarker on the second biomarker list; The second training dataset is used as both the input and target of the second autoencoder to train the second autoencoder, thereby creating the second trained autoencoder; as well as Deploy the first trained autoencoder and the second trained autoencoder.

21. The method of claim 20, wherein: Obtaining the list of primary biomarkers and the list of secondary biomarkers includes obtaining the list of primary biomarkers and the list of secondary biomarkers from the hospital or laboratory; Obtaining the first training dataset and the second training dataset includes obtaining at least one biomarker dataset from a hospital or laboratory, and creating the first training dataset and the second training dataset based on the at least one biomarker dataset; as well as Deploying the first trained autoencoder and the second trained autoencoder includes deploying the first trained autoencoder and the second trained autoencoder in a hospital or laboratory.

22. The method of claim 21, wherein the first biomarker list comprises a list of biomarkers in a first biomarker group, and the second biomarker list comprises a list of biomarkers in a second biomarker group.

23. The method of claim 22, wherein the first biomarker group and the second biomarker group comprise a different set of biomarkers from at least one of the lipid group, the complete blood cell count (CBC) group, the basal metabolic group (BMP) group, the comprehensive metabolic group (CMP) group, the thyroid group, the urine group, and the immunoassay group.

24. A machine learning-based automated authentication system for clinical laboratory diagnostics, comprising: processor; The memory coupled to the processor includes an autoencoder trained on the biomarker training dataset by using the biomarker training dataset as both input and target of the autoencoder. as well as Computer program instructions stored in memory, when executed by a processor, cause the processor to: Receive one or more biomarkers measured for the patient; The measured biomarkers are input into an autoencoder to generate output biomarkers; as well as The validity of a measured biomarker is determined by comparing the biomarker input to the autoencoder with the biomarker output from the autoencoder.

25. The automated authentication system of claim 24, wherein the biomarker training set comprises a biomarker dataset from a hospital or laboratory.

26. The automated authentication system of claim 25, wherein the autoencoder is deployed at a hospital or laboratory that provides a biomarker dataset.

27. The automated authentication system of claim 25, wherein the autoencoder is deployed at a different hospital or laboratory than the hospital or laboratory providing the biomarker dataset.

28. The automated authentication system of claim 24, wherein the biomarker training set comprises biomarkers from a biomarker group.

29. The automated authentication system of claim 28, wherein the biomarker group comprises biomarkers from at least one of the lipid group, the complete blood cell count (CBC) group, the basal metabolic group (BMP) group, the comprehensive metabolic group (CMP) group, the thyroid group, the urine group, and the immunoassay group.

30. The automated authentication system of claim 24, wherein the biomarker training set comprises biomarkers from routine diagnostic tests.

31. The automated authentication system of claim 30, wherein the routine diagnostic tests include biomarkers from the lipid group, the complete blood count (CBC) group, and the comprehensive metabolic group (CMP).

32. The automated authentication system of claim 24, further comprising computer program instructions stored in a memory, which, when executed by a processor, cause the processor to determine a reconstruction error between the measured biomarker and the output biomarker.

33. The automatic authentication system of claim 32, further comprising computer program instructions stored in a memory, which, when executed by a processor, cause the processor to determine at least one mean square error based on the measured biomarker and the output biomarker of the autoencoder.

34. The automated authentication system of claim 24, further comprising computer program instructions stored in a memory, which, when executed by a processor, cause the processor to predict the cause of the difference between the measured biomarker and the output biomarker of the autoencoder.

35. The automated authentication system of claim 34, further comprising computer program instructions stored in a memory, which, when executed by a processor, cause the processor to determine the classification of the representation of the measured biomarker.

36. The automated authentication system of claim 35, wherein the representation of the measured biomarker includes a latent spatial representation of the measured biomarker, wherein the latent spatial representation is generated by an autoencoder.

37. The automatic authentication system of claim 36, further comprising computer program instructions stored in a memory, which, when executed by a processor, cause the processor to... Clustering algorithms are used to cluster the representations of the measured biomarkers into one of several clusters; and The differences between the measured biomarkers and the output biomarkers are classified based on the multiple clusters mentioned above.

38. The automated authentication system of claim 37, wherein the plurality of clusters includes a clinical condition cluster and a contamination cluster.

39. The automated authentication system of claim 24, further comprising computer program instructions stored in memory, which, when executed by a processor, cause the processor to recommend manual processing of a test to obtain the one or more biomarkers measured for the patient, based on the difference between the measured biomarker and the output biomarker of the autoencoder.

40. A method for automatically certifying clinical laboratory diagnostics, comprising: Receive one or more biomarkers measured for the patient; The measured biomarkers are input into an autoencoder to generate output biomarkers, wherein the autoencoder is trained on a biomarker training dataset using both the biomarker training dataset as input and target. as well as The validity of a measured biomarker is determined by comparing the biomarker input to the autoencoder with the biomarker output from the autoencoder.

41. The method of claim 40, wherein the biomarker training set comprises a biomarker dataset from a hospital or laboratory, and wherein the autoencoder is deployed at the hospital or laboratory providing the biomarker dataset.

42. The method of claim 40, wherein the biomarker training set comprises a biomarker dataset from a first hospital or laboratory, and wherein the autoencoder is deployed at a second hospital or laboratory different from the first hospital or laboratory providing the biomarker dataset.

43. The method of claim 40, wherein the autoencoder is a variational autoencoder.

Images