An abnormal sample identification and early warning system for laboratory testing

Abstract

This invention discloses an abnormal sample identification and early warning system for laboratory testing. Through a rule engine that logically judges test data according to clear rules, it ensures the consistency of abnormal identification results. It complements a machine learning model, which can uncover potential data features and patterns, assisting the rule engine in reducing false positives and false negatives, and improving the accuracy of abnormal identification. Simultaneously, the machine learning model can automatically discover patterns by learning from historical data. When faced with unknown data patterns, it can identify abnormalities based on its learned knowledge, adapting to new environments and sample types, and maintaining effective identification of unknown anomalies. Furthermore, the early warning mechanism module performs early warning operations based on the abnormal identification results according to a preset hierarchical strategy. Hierarchical early warning can provide accurate warnings based on actual conditions, avoiding over- or under-warning, improving the efficiency and effectiveness of responding to anomalies, and ensuring the smooth operation of laboratory testing.

Description

Technical Field

[0001] This invention relates to the field of abnormal sample identification and early warning technology, and in particular to an abnormal sample identification and early warning system for laboratory testing. Background Technology

[0002] In the field of laboratory testing, ensuring the accuracy and reliability of test results is of paramount importance. With the continuous increase in laboratory testing tasks and the increasing complexity and diversity of test samples, timely identification and effective early warning of abnormalities in the sample testing process have become increasingly critical.

[0003] Traditional laboratory anomaly identification methods mainly rely on human experience and fixed threshold judgments. Testers analyze test data based on their own experience, and when the data exceeds the preset fixed threshold range, it is judged as an anomaly. However, this method has many limitations: on the one hand, human experience is subjective, and different testers may have different judgments on anomalies, making it difficult to ensure the consistency and accuracy of anomaly identification; on the other hand, fixed thresholds are difficult to adapt to complex and ever-changing test environments and sample characteristics, and often cannot effectively identify some potential and unknown anomaly patterns.

[0004] Furthermore, after identifying abnormal samples, it is impossible to carry out targeted early warning operations according to different levels of abnormality, resulting in poor early warning effects and failure to notify relevant personnel to take countermeasures in a timely and effective manner. Summary of the Invention

[0005] In view of this, the present invention proposes an abnormal sample identification and early warning system for laboratory testing, which can effectively solve the shortcomings of the existing technology, such as difficulty in ensuring the consistency and accuracy of abnormal identification, difficulty in identifying unknown abnormal patterns, and inability to perform targeted early warning operations according to different abnormal situations.

[0006] The technical solution of this invention is implemented as follows:

[0007] An abnormal sample identification and early warning system for laboratory testing includes:

[0008] An anomaly detection module is used to identify anomalous samples by combining a rule engine and a machine learning model; wherein, the rule engine is used to make logical judgments on obvious anomalies, and the machine learning model is used to detect unknown data pattern anomalies.

[0009] The early warning mechanism module is used to perform early warning operations based on the anomaly identification results and in combination with the preset early warning classification strategy.

[0010] As a further optional solution to the aforementioned abnormal sample identification and early warning system for laboratory testing, the rule engine includes:

[0011] The rule definition unit is used to define gender mismatch rules and detection anomaly rules. The gender mismatch rules are used to compare whether the gender entered into the laboratory information system is consistent with the biological sex of the sample. The detection anomaly rules are used to detect whether the quality control data exceeds the normal threshold range.

[0012] The feature selection unit is used to select basic sample information, detection process metadata, and quality control data as features.

[0013] The logical judgment unit is used to identify anomalies in selected features based on defined gender mismatch rules and anomaly detection rules.

[0014] As a further optional solution to the abnormal sample identification and early warning system for laboratory testing, the basic sample information includes sample ID, patient age, and submitting department; the testing process metadata includes machine access time, operator ID, equipment ID, and reagent batch; and the quality control data includes internal standard strength, quality control sample measurement value, and positive and negative control results.

[0015] As a further optional solution to the aforementioned abnormal sample identification and early warning system for laboratory testing, the machine learning model includes:

[0016] The feature extraction unit is used to extract statistical features, derived features, and temporal features of the sample. The statistical features include the mean, variance, kurtosis, and skewness of the sample. The derived features include the ratio of the target peak to the internal standard peak of the sample, the signal ratio of different detection channels, and the similarity between samples. The temporal features include the device running time and the trend of detection results of similar samples recently.

[0017] The model recognition unit is used to identify anomalies in the statistical features, derived features, and temporal features of samples based on a pre-trained recognition model.

[0018] As a further optional solution to the aforementioned abnormal sample identification and early warning system for laboratory testing, the early warning mechanism module includes:

[0019] The identification result judgment unit is used to judge the degree of abnormality based on the abnormality identification result, wherein the degree of abnormality includes severe, suspicious and slight;

[0020] The early warning operation execution unit is used to execute different levels of early warning operations based on the degree of abnormality and in conjunction with a preset early warning classification strategy;

[0021] The early warning push unit is used to deliver information to operators using different push methods based on different levels of early warning operations.

[0022] As a further optional solution to the aforementioned abnormal sample identification and early warning system for laboratory testing, the early warning grading strategy includes:

[0023] The Level 1 warning strategy is used to interrupt the current sample testing process or prevent report generation when the anomaly level is severe.

[0024] The Level 2 early warning strategy is used to continue the detection process and mark suspicious information on the report as a reminder when the anomaly level is suspicious.

[0025] The three-level early warning strategy is used to allow the detection process to continue and generate logs when the anomaly level is minor.

[0026] As a further optional solution to the abnormal sample identification and early warning system for laboratory testing, the early warning push unit includes:

[0027] The interface pop-up sub-unit is used to pop up an interface pop-up window on the laboratory information system client when performing a first-level or second-level early warning operation. The interface pop-up window displays the abnormal sample number, abnormality type and handling suggestions.

[0028] The mobile push subunit is used to push real-time information to the mobile device when performing a Level 3 early warning operation. The real-time information includes the anomaly type and severity level.

[0029] A method for identifying and issuing early warnings of abnormal samples in laboratory testing, specifically including:

[0030] Anomaly sample identification is performed by combining a rule engine and a machine learning model; wherein, the rule engine is used to perform hard-coded logical judgments on obvious anomalies, and the machine learning model is used to detect unknown data pattern anomalies.

[0031] Based on the anomaly identification results, and in conjunction with the preset early warning classification strategy, early warning operations are carried out.

[0032] A computing device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for identifying and warning of abnormal samples for laboratory testing.

[0033] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method for identifying and issuing early warnings of abnormal samples for laboratory testing.

[0034] The beneficial effects of this invention are as follows: Firstly, the rule engine performs logical judgments on the detection data based on predefined, explicit rules. Unlike judgments relying on human experience, the rule engine consistently follows the same rules when processing the same data, thus ensuring the consistency of anomaly identification results. Secondly, the deterministic judgment of the rule engine complements the flexibility of the machine learning model. The machine learning model can learn from a large amount of data to discover potential data features and patterns that are difficult to describe with explicit rules, further assisting the rule engine in making more accurate judgments. The combination of the two can comprehensively consider multiple factors, reducing false positives and false negatives, and improving the accuracy of anomaly identification. Thirdly, the machine learning model can learn from a large amount of historical detection data... Based on its learning, the machine learning model automatically discovers potential patterns and rules in the data. When faced with unknown data patterns, the model can analyze and judge based on its learned knowledge, identifying samples that differ significantly from normal patterns as anomalies. This allows it to better adapt to new data patterns and anomalies, enabling the system to maintain its ability to effectively identify unknown anomalies even when facing constantly changing detection environments and sample types. In addition, the early warning mechanism module performs early warning operations based on the anomaly identification results and a preset early warning grading strategy. Through this grading early warning approach, the system can provide accurate early warnings based on the actual situation of the anomaly, avoiding over-warning or under-warning issues and improving the efficiency and effectiveness of responding to anomalies. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Fig. 1 This is a schematic diagram of the composition of an abnormal sample identification and early warning system for laboratory testing according to the present invention;

[0037] Fig. 2 This is a flowchart illustrating an abnormal sample identification and early warning method for laboratory testing according to the present invention.

[0038] Fig. 3 This is a schematic diagram of the composition of a computing device according to the present invention. Detailed Implementation

[0039] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] refer to Figs. 1 to 3 An abnormal sample identification and early warning system for laboratory testing includes an abnormal identification module and an early warning mechanism module, wherein:

[0041] An anomaly detection module is used to identify anomalous samples by combining a rule engine and a machine learning model; wherein, the rule engine is used to perform logical judgments on clearly defined anomalies, and the machine learning model is used to detect unknown data pattern anomalies; in some embodiments, the rule engine includes:

[0042] The rule definition unit is used to define gender mismatch rules and detection anomaly rules. The gender mismatch rules are used to compare whether the gender entered into the laboratory information system is consistent with the biological sex of the sample. The detection anomaly rules are used to detect whether the quality control data exceeds the normal threshold range.

[0043] The feature selection unit is used to select basic sample information, detection process metadata, and quality control data as features.

[0044] The logical judgment unit is used to identify anomalies in selected features based on defined gender mismatch rules and anomaly detection rules.

[0045] Specifically, in a blood sample testing scenario in a laboratory, the following rules apply: Gender mismatch rule: The laboratory information system records the patient's gender information. However, during blood sample testing, certain indicators may differ between genders. For example, the normal range for hemoglobin in men differs from that in women. The rule definition unit defines a rule that compares the patient's gender recorded in the laboratory information system with the biological sex corresponding to the sample (determined through information such as sample origin). If a male sample corresponds to female information, or vice versa, it is considered a gender mismatch. Detection anomaly rule: For the quality control data of blood samples, such as the quality control values ​​of the testing instrument, a normal threshold range is set. When the obtained quality control data exceeds this range, the detection anomaly rule is triggered.

[0046] Features are selected from sample-related information, which includes basic sample information, detection process metadata, and quality control data.

[0047] Based on gender mismatch rules and anomaly detection rules, the features selected by the feature selection unit are evaluated. For example, if the gender information in the laboratory information system corresponding to a sample does not match the actual biological sex, or if the quality control data exceeds the normal threshold range, the sample is determined to be an abnormal sample. It should be noted that when gender mismatch or anomaly detection is triggered, basic sample information (such as sample ID, collection time, submitting department, and sample type) can help operators quickly locate the source and background of the problematic sample. For example, if a department continuously produces gender mismatched samples, it may indicate a systemic error in the sample collection or entry process of that department; sample type (such as blood or saliva) may affect the interpretation of test results (for example, the detection success rate of certain loci in saliva is low). Metadata from the testing process (such as machine access time, operator ID, equipment ID used, and reagent batch) can help analyze whether anomalies are related to specific operations, equipment, or reagents. For example, if a device frequently triggers "internal standard Ct value anomaly" in its test samples, it may indicate that the device needs calibration or that the reagent has a quality problem. If an operator consistently handles samples with gender mismatch, it may indicate an oversight in their operation process (such as sample confusion). When a gender mismatch rule is triggered, the system will combine basic sample information (such as patient age) and testing process metadata (such as operator ID) to generate more detailed warning information to help operators quickly troubleshoot problems. After a testing anomaly rule is triggered, the system can trace possible causes through the equipment ID and reagent batch, rather than simply reporting "signal anomaly".

[0048] In this way, the rule engine judges gender mismatch and quality control data anomalies through explicit rules, ensuring accurate identification in known anomaly situations. The machine learning model can discover unknown data pattern anomalies, making up for the shortcomings of the rule engine. The combination of the two enables the system to comprehensively cover various anomaly situations, greatly improving the accuracy of anomaly identification. For example, in the blood sample detection mentioned above, it not only avoids misjudgment of test results due to incorrect gender information, but also promptly detects potential unknown abnormal samples.

[0049] In some embodiments, the basic sample information includes sample ID, patient age, and department to which the sample is sent; the metadata of the testing process includes the time of use, operator ID, equipment ID, and reagent batch; and the quality control data includes internal standard strength, quality control sample measurement value, and positive and negative control results.

[0050] Specifically, by comprehensively considering multi-dimensional features such as basic sample information, detection process metadata, and quality control data, the rule engine can accurately judge known gender mismatches and quality control data anomalies based on clear rules. The machine learning model, on the other hand, uses patterns learned from a large amount of historical data to discover unknown data pattern anomalies. The combination of the two enables the system to comprehensively and accurately identify abnormal samples from multiple perspectives, greatly reducing the probability of misjudgment and missed judgment. For example, in blood sample testing, it avoids erroneous test result reports caused by incorrect gender information or failure to detect quality control anomalies in a timely manner, while also capturing some potential and complex anomalies. Secondly, the recording of detection process metadata, such as the time of use, operator ID, equipment ID, and reagent batch, provides detailed traceability clues for the detection process. When an anomaly occurs, the specific detection link, operator, and equipment can be quickly located, facilitating the analysis of the cause of the anomaly and the implementation of targeted improvement measures. For example, if it is found that samples tested by a certain batch of reagents frequently show anomalies, the quality of that batch of reagents can be checked and replaced in a timely manner to ensure the stability of the detection quality.

[0051] In some embodiments, the machine learning model includes:

[0052] The feature extraction unit is used to extract statistical features, derived features, and temporal features of the sample. The statistical features include the mean, variance, kurtosis, and skewness of the sample. The derived features include the ratio of the target peak to the internal standard peak of the sample, the signal ratio of different detection channels, and the similarity between samples. The temporal features include the device running time and the trend of detection results of similar samples recently.

[0053] The model recognition unit is used to identify anomalies in the statistical features, derived features, and temporal features of samples based on a pre-trained recognition model.

[0054] Specifically, statistical feature extraction involves calculating the mean, variance, kurtosis, and skewness of the test data for each blood sample. For example, for a series of test data on hemoglobin content in a blood sample, the mean of these data is calculated to understand the average level; variance measures the dispersion of the data; kurtosis reflects the sharpness of the data distribution; and skewness reflects the asymmetry of the data distribution.

[0055] Derivative feature extraction: Calculate the ratio of the target peak to the internal standard peak. In blood testing, the ratio of the peak of certain specific components to the internal standard peak is an important analytical indicator. Calculate the signal ratio of different detection channels (such as different wavelength detection channels). The signal differences between different channels may reflect certain characteristics of the sample. Calculate the similarity between samples. By comparing the detection data of the current sample with those of other normal samples, the degree of similarity can be determined.

[0056] Temporal feature extraction: Record the running time of the device, as devices that run for a long time may experience performance drift; analyze the trend of test results of similar samples in the near future (such as blood samples or samples of the same test items), such as the average change trend of hemoglobin test results in the past week;

[0057] A pre-trained recognition model is used, which is trained on a large amount of historical blood sample data (including normal samples and data of various known abnormal samples, and labeled whether they are abnormal). The statistical features, derived features and temporal features of the samples extracted by the feature extraction unit are input into the recognition model. The model analyzes and judges the input features based on the learned patterns and rules. For example, if it is found that the variance of a sample's statistical features is much larger than the variance range of normal samples, and the ratio of the target peak to the internal standard peak in the derived features is abnormal, and combined with the fact that the detection results of similar samples in the recent time series features show an upward trend, the model will comprehensively judge that the sample may be abnormal.

[0058] Thus, the feature extraction unit extracts sample features from multiple dimensions such as statistics, derivation, and time series, comprehensively mining the information in the sample data. The model recognition unit, trained on a large amount of historical data, can learn various complex feature patterns and anomaly patterns. Compared with relying solely on rule engines, it can not only more accurately identify anomalies that conform to explicit rules, but also discover some hidden and unknown anomalies, greatly expanding the scope of anomaly recognition and improving the ability to discover anomalous samples. For example, it can identify some potential anomalous samples caused by long-term changes in equipment performance, which may not be detected in time by rule judgment. Secondly, the machine learning model can adapt to new testing projects, equipment, and sample types through continuous learning and updating. When the hospital laboratory introduces new testing equipment or carries out new testing projects, it only needs to retrain or fine-tune the model using new historical data, and the model can adapt to new data features and anomaly patterns. In contrast, rule engines often need to redefine and adjust rules according to new situations, which is relatively cumbersome. This adaptability and flexibility enable the system to better cope with changes in the laboratory testing environment.

[0059] An early warning mechanism module is used to perform early warning operations based on anomaly identification results and a preset early warning classification strategy; in some embodiments, the early warning mechanism module includes:

[0060] The identification result judgment unit is used to judge the degree of abnormality based on the abnormality identification result, wherein the degree of abnormality includes severe, suspicious and slight;

[0061] The early warning operation execution unit is used to execute different levels of early warning operations based on the degree of abnormality and in conjunction with a preset early warning classification strategy;

[0062] The early warning push unit is used to deliver information to operators using different push methods based on different levels of early warning operations.

[0063] In some embodiments, the early warning classification strategy includes:

[0064] The Level 1 warning strategy is used to interrupt the current sample testing process or prevent report generation when the anomaly level is severe.

[0065] The Level 2 early warning strategy is used to continue the detection process and mark suspicious information on the report as a reminder when the anomaly level is suspicious.

[0066] The three-level early warning strategy is used to allow the detection process to continue and generate logs when the anomaly level is minor.

[0067] Specifically, once the anomaly detection module (rule engine or machine learning model) determines that a sample is abnormal, the result judgment unit begins to work. Based on preset standards, it assesses the severity of the anomaly. For example, if a sample shows a gender mismatch, it may lead to completely incorrect analysis and interpretation of all subsequent gender-based test results, which would have a significant impact on clinical diagnosis and is therefore judged as a serious anomaly. If the quality control data exceeds the normal threshold range, but the exceedance is small, and the impact on the test results is relatively controllable through comprehensive analysis with other relevant features, it is judged as a minor anomaly. For unknown data pattern anomalies identified by the machine learning model, if the anomaly involves multiple key detection indicators and the feature expression is relatively complex, it may have a significant uncertainty impact on the test results and is judged as a suspicious anomaly.

[0068] For severe anomalies (such as gender mismatch), the highest level of alert is executed, immediately halting subsequent testing of the current sample (if still in progress) to prevent further spread of erroneous results. Simultaneously, a comprehensive review mechanism is activated, requiring relevant testing personnel to re-verify sample information, including patient identity and sample origin, and to conduct a detailed review and inspection of the testing process. For suspicious anomalies, a medium-level alert is executed, marking the sample as a key focus and notifying testing personnel to re-examine and analyze the sample's testing data. For minor anomalies, a lower-level alert is executed, reminding testing personnel to routinely review the sample's testing process and data, focusing on testing indicators potentially affected by quality control data anomalies to ensure the accuracy of test results. Furthermore, the alert push unit employs different push methods based on the alert level to ensure that important information is quickly and accurately delivered to relevant personnel, avoiding delays and omissions in information transmission. This allows operators to promptly understand the anomaly and take appropriate action. For example, multi-channel push notifications for severe anomalies ensure that key personnel receive information immediately, buying time for timely error correction.

[0069] In some embodiments, the early warning push unit includes:

[0070] The interface pop-up sub-unit is used to pop up an interface pop-up window on the laboratory information system client when performing a first-level or second-level early warning operation. The interface pop-up window displays the abnormal sample number, abnormality type and handling suggestions.

[0071] The mobile push subunit is used to push real-time information to the mobile device when performing a Level 3 early warning operation. The real-time information includes the anomaly type and severity level.

[0072] Specifically, when the early warning operation execution unit determines that it is a Level 1 early warning (such as a serious abnormality of gender mismatch in the sample) or a Level 2 early warning (such as a relatively complex suspicious abnormality identified by the machine learning model that requires special attention), the interface pop-up sub-unit responds quickly. On the client interface of the laboratory information system, an eye-catching interface pop-up will immediately pop up. This pop-up will attract the operator's attention with bright colors (such as a red background with a yellow warning icon), display the abnormal sample number in detail, such as "XY20241015001 - A" (where "A" may represent an abnormality identifier), clarify the abnormality type, such as "gender mismatch abnormality" or "suspicious data pattern abnormality", and provide handling suggestions, such as "immediately stop the testing process and verify the sample information" or "arrange expert consultation and re-examine the testing data".

[0073] If the warning level reaches level three (e.g., quality control data slightly exceeds the threshold range, although the impact is relatively small, it still needs attention), after the warning operation execution unit performs the level three warning operation, the mobile terminal push subunit starts working. Through integration with the hospital's internal mobile information platform, it pushes real-time information to relevant testing personnel, laboratory managers, and other mobile devices (such as mobile phones and tablets). The information content is concise and clear, including the type of abnormality, such as "minor abnormality in quality control data," and the severity level "level three warning." It may also include brief handling instructions, such as "Please review the relevant quality control data after completing the current work."

[0074] Thus, the intuitive display of the interface pop-up sub-unit on the laboratory information system client allows testing personnel operating in front of the computer to obtain key anomaly information immediately without switching systems or searching for other channels. The eye-catching design and detailed content presentation ensure that operators can quickly understand the anomaly and take action according to the recommendations. The mobile push sub-unit breaks spatial limitations, allowing relevant personnel to receive early warning information in a timely manner through mobile devices even if they are not at their workstations, ensuring timely information delivery and avoiding processing delays caused by personnel absence. Secondly, different push methods are designed for different warning levels and scenarios, ensuring that relevant information is accurately delivered to the right personnel. For level one and two warnings, the interface pop-up ensures that the direct operators and on-site supervisors can handle the situation in a timely manner. For level three warnings, the mobile push allows relevant personnel to handle the situation at the appropriate time, avoiding excessive interference. This hierarchical and categorized push method promotes collaborative work among personnel in different positions within the laboratory and improves the overall efficiency of handling anomalies.

[0075] A method for identifying and issuing early warnings of abnormal samples in laboratory testing, specifically including:

[0076] Anomaly sample identification is performed by combining a rule engine and a machine learning model; wherein, the rule engine is used to perform hard-coded logical judgments on obvious anomalies, and the machine learning model is used to detect unknown data pattern anomalies.

[0077] Based on the anomaly identification results, and in conjunction with the preset early warning classification strategy, early warning operations are carried out.

[0078] A computing device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for identifying and warning of abnormal samples for laboratory testing.

[0079] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method for identifying and issuing early warnings of abnormal samples for laboratory testing.

[0080] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An abnormal sample identification and early warning system for laboratory testing, characterized in that, include: An anomaly detection module is used to identify anomalous samples by combining a rule engine and a machine learning model; wherein, the rule engine is used to make logical judgments on obvious anomalies, and the machine learning model is used to detect unknown data pattern anomalies. The early warning mechanism module is used to perform early warning operations based on the anomaly identification results and in combination with the preset early warning classification strategy.

2. The abnormal sample identification and early warning system for laboratory testing according to claim 1, characterized in that, The rule engine includes: The rule definition unit is used to define gender mismatch rules and detection anomaly rules. The gender mismatch rules are used to compare whether the gender entered into the laboratory information system is consistent with the biological sex of the sample. The detection anomaly rules are used to detect whether the quality control data exceeds the normal threshold range. The feature selection unit is used to select basic sample information, detection process metadata, and quality control data as features. The logical judgment unit is used to identify anomalies in selected features based on defined gender mismatch rules and anomaly detection rules.

3. The abnormal sample identification and early warning system for laboratory testing according to claim 2, characterized in that, The basic information of the sample includes sample ID, patient age, and department of submission; the metadata of the testing process includes machine time, operator ID, equipment ID, and reagent batch; the quality control data includes internal standard strength, quality control sample test value, and positive and negative control results.

4. The abnormal sample identification and early warning system for laboratory testing according to claim 1, characterized in that, The machine learning model includes: The feature extraction unit is used to extract statistical features, derived features, and temporal features of the sample. The statistical features include the mean, variance, kurtosis, and skewness of the sample. The derived features include the ratio of the target peak to the internal standard peak of the sample, the signal ratio of different detection channels, and the similarity between samples. The temporal features include the device running time and the trend of detection results of similar samples recently. The model recognition unit is used to identify anomalies in the statistical features, derived features, and temporal features of samples based on a pre-trained recognition model.

5. The abnormal sample identification and early warning system for laboratory testing according to claim 1, characterized in that, The early warning mechanism module includes: The identification result judgment unit is used to judge the degree of abnormality based on the abnormality identification result, wherein the degree of abnormality includes severe, suspicious and slight; The early warning operation execution unit is used to execute different levels of early warning operations based on the degree of abnormality and in conjunction with a preset early warning classification strategy; The early warning push unit is used to deliver information to operators using different push methods based on different levels of early warning operations.

6. The abnormal sample identification and early warning system for laboratory testing according to claim 5, characterized in that, The early warning classification strategy includes: The Level 1 warning strategy is used to interrupt the current sample testing process or prevent report generation when the anomaly level is severe. The Level 2 early warning strategy is used to continue the detection process and mark suspicious information on the report as a reminder when the anomaly level is suspicious. The three-level early warning strategy is used to allow the detection process to continue and generate logs when the anomaly level is minor.

7. The abnormal sample identification and early warning system for laboratory testing according to claim 5, characterized in that, The early warning push unit includes: The interface pop-up sub-unit is used to pop up an interface pop-up window on the laboratory information system client when performing a first-level or second-level early warning operation. The interface pop-up window displays the abnormal sample number, abnormality type and handling suggestions. The mobile push subunit is used to push real-time information to the mobile device when performing a Level 3 early warning operation. The real-time information includes the anomaly type and severity level.

8. A method for identifying and issuing early warnings of abnormal samples in laboratory testing, characterized in that, Specifically, it includes: Anomaly sample identification is performed by combining a rule engine and a machine learning model; wherein, the rule engine is used to perform hard-coded logical judgments on obvious anomalies, and the machine learning model is used to detect unknown data pattern anomalies. Based on the anomaly identification results, and in conjunction with the preset early warning classification strategy, early warning operations are carried out.

9. A computing device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the abnormal sample identification and early warning method for laboratory testing as described in claim 8.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the steps of the abnormal sample identification and early warning method for laboratory testing as described in claim 8.

Images