A cloud-based intelligent analysis system for clinical data in acute infarction

Abstract

This invention relates to the field of clinical data analysis technology, specifically to a cloud-based intelligent analysis system for clinical data in acute infarction. The system includes a time-point extraction module, a segment integration module, a sudden segment identification module, a trend capture module, and a graded early warning module. In this invention, key time points are extracted and integrated from continuous monitoring data during hospitalization. A time series is constructed by combining the slope change rate and numerical mutation rate of multiple physiological indicators. Based on this, related behavioral time points in various data sources are simultaneously retrieved, enabling collaborative mutation identification between different data sources. Curve fitting and residual analysis of indicator change trends enhance the accuracy of trend deviation time point identification. Combining the current value of the indicator with historical change amplitudes, conditional judgments are made to achieve early warning of disease mutation events and dynamic risk level classification, improving the efficiency of comprehensive judgment on the evolution of acute disease.

Description

Technical Field

[0001] This invention relates to the field of clinical data analysis technology, and in particular to a cloud-based intelligent analysis system for clinical data of acute infarction. Background Technology

[0002] The field of clinical data analysis technology involves a series of processing operations, including the collection, cleaning, integration, analysis, and interpretation of structured or unstructured data from clinical diagnosis and treatment processes, to support tasks such as clinical decision support, disease prognosis assessment, and treatment plan optimization. Core aspects of this technology include electronic medical record text parsing, laboratory test data normalization, time-series analysis of vital sign data, clinical indicator modeling, multi-source heterogeneous data fusion, and clinical pattern recognition and reasoning methods based on statistical learning or machine learning models. Clinical data analysis typically targets specific applications such as disease diagnosis, treatment response assessment, disease risk prediction, and clinical pathway optimization, covering the entire process from data perception and processing to intelligent analysis. Technically, it integrates natural language processing, medical terminology ontology construction, time-series modeling algorithms, supervised learning, and unsupervised learning methods.

[0003] Traditional cloud-based intelligent analysis systems for clinical data in acute infarction refer to systems that upload multi-modal clinical data related to acute infarction to a remote cloud server and then perform centralized processing, archiving, and analysis of this data through computational programs based on specific analytical logic. The technical issues addressed are the automated management and intelligent analysis needs of clinical data for acute infarction. In traditional systems, patients' electrocardiogram (ECG) waveform data, laboratory test indicators, medical imaging results, and medical records are typically uploaded to the cloud platform manually or semi-automatically. Pre-defined statistical rules, decision tree models, or recognition algorithms based on ECG feature comparison rule sets are then used to perform sub-analysis and risk assessment on each type of data. Furthermore, some solutions use benchmark indicators and threshold conditions to screen and label abnormal indicators to assist physicians in their judgment. The overall processing flow often relies on a predefined path execution sequence and lacks the ability to model and uniformly analyze the correlations between multi-source heterogeneous data. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a cloud-based intelligent analysis system for clinical data in acute infarction.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a cloud-based intelligent analysis system for clinical data of acute infarction, the system comprising:

[0006] The time point extraction module receives continuous monitoring data records from patients during hospitalization via the cloud, extracts the QRS waveform amplitude and interval sequence of every 30-second segment in the electrocardiogram, selects nodes where both the slope change rate and the index value mutation rate increase, identifies and records overlapping time point nodes, and generates an acute infarction-related overlapping time point set.

[0007] The segment integration module calls the set of overlapping time points associated with acute infarction to determine whether there is a consistent upward trend in the target indicators. If there is no consistency, the nodes are shifted forward or backward by a fixed time according to the direction of trend deviation to generate an adjusted disease course time series.

[0008] The mutation segment identification module constructs a heart rate change velocity vector and an ST segment slope vector within a continuous time window based on the adjusted disease course time series, calculates the point-to-point difference sequence of the two vectors, marks the mutation segment, and generates a multi-index synergistic mutation time period.

[0009] The trend capture module calls the detection sequences of N-terminal pro-brain natriuretic peptide and troponin within the multi-indicator synergistic mutation time period, fits the change values ​​and establishes a first-order nonlinear curve, calculates the trend of the residual value change between the fitted curve and the actual sequence, marks the trend offset time point, and generates the indicator trend mutation time location set.

[0010] As a further aspect of the present invention, the acute infarction-related overlapping time point set includes waveform structure overlapping nodes, numerical mutation synchronization nodes, and diagnosis and treatment behavior change nodes; the adjusted disease course time series includes node ranking results, trend consistency markers, and time fine-tuning labels; the multi-indicator collaborative mutation time period includes heart rate fluctuation segments, waveform tilt segments, and difference amplification segments; and the indicator trend mutation time location set includes residual increase inflection points, curve offset time points, and trend turning point nodes.

[0011] As a further aspect of the present invention, the time point extraction module includes:

[0012] The data receiving submodule receives continuous monitoring data records of patients during hospitalization via the cloud, collects QRS waveform amplitude and QRS interval sequences in every 30-second segment of the electrocardiogram record, extracts continuous sampling data of creatine kinase isoenzyme and troponin, performs absolute value calculation on the difference between adjacent sampling values, constructs an index mutation intensity sequence, aligns each sequence with the time axis and performs joint processing to generate a continuous physiological parameter time mapping set;

[0013] The event extraction submodule calculates the tilt change rate of each ECG segment in the ST segment waveform based on the continuous physiological parameter time mapping set, determines whether the tilt angle increases synchronously in two consecutive time periods, and simultaneously screens nodes in the creatine kinase isoenzyme and troponin difference sequence where the mutation rate increases simultaneously. Time points that meet both conditions are marked as event candidate nodes, and a set of joint indicator mutation time nodes is generated.

[0014] The node matching submodule calls the time points in the joint indicator mutation time node set, compares them with the operation type number sequence recorded by time in the diagnosis and treatment behavior record, marks the record time points where the operation type value changes, and determines whether the time point coincides with the candidate node on the time axis or is within the fixed time tolerance range. If the condition is met, the time point is registered as an overlapping time point, and an acute infarction associated overlapping time point set is generated.

[0015] As a further aspect of the present invention, the segmentation integration module includes:

[0016] The trend extraction submodule calls the nodes in the set of overlapping time points associated with acute infarction, extracts the severity sequence of diagnosis and treatment behavior, the ST segment tilt change direction sequence, and the creatine kinase isoenzyme rise rate direction sequence within a set time range before and after each node, performs directional difference statistics on the directional changes between adjacent records in each type of sequence, and outputs the directional trend group of three parameters for each node, generating a multi-parameter trend direction group.

[0017] The consistency determination submodule, based on the multi-parameter trend direction group, combines the pairwise direction values ​​of the three direction parameters corresponding to each node, accumulates the number of consistent directions in the combination result, and if there is a number of inconsistent directions greater than or equal to one, the node is marked as a trend offset node and a trend offset node identifier set is established.

[0018] The sequence update submodule adjusts the fixed time step of the time label of each offset node according to the order indicated by the main direction based on the trend offset node identifier set, and reorders each node according to the adjusted time to generate the adjusted disease course time series.

[0019] As a further aspect of the present invention, the protrusion identification module includes:

[0020] The vector generation submodule collects heart rate records and ST segment potential sequences within each time period based on the interval between adjacent time nodes in the adjusted disease course time series. It calculates the difference between adjacent record values ​​for the heart rate records and divides it by the corresponding time interval to obtain the heart rate change rate sequence. It fits a straight line to the ST segment potential points and calculates the corresponding slope sequence. These are then organized into two time-aligned vectors to generate a continuous segment feature vector set.

[0021] The difference calculation submodule calls the heart rate change rate vector and ST segment slope vector in the continuous segment feature vector set, performs difference calculation on each corresponding time point in the vector, obtains the point-to-point difference sequence between heart rate and waveform slope, performs amplitude value filtering within a sliding window on the difference sequence, extracts the intervals where the difference amplitude exceeds the difference mean offset threshold, and establishes a fluctuation difference interval distribution set.

[0022] The ST segment extraction submodule calculates the difference between the maximum and minimum values ​​within a given time period by calling the original heart rate record based on the start and end times of the differential value interval distribution. Simultaneously, it extracts the ST segment slope change trend label and performs a structured combination of the segment start and end times, heart rate fluctuation amplitude, and slope change direction to generate a multi-indicator collaborative mutation time period.

[0023] As a further aspect of the present invention, the mean deviation threshold of the difference is the sum of the mean and standard deviation of the difference sequence within the sliding window, wherein the mean is the arithmetic mean of all point-to-point differences within the sliding window, and the standard deviation is the root mean square of the deviations of all point-to-point differences from the mean within the sliding window.

[0024] The interval where the extracted difference exceeds the mean difference offset threshold is a time period in which the difference between several consecutive points is greater than the mean difference offset threshold. The start time of the time period is the time label of the first point that meets the threshold condition, and the end time is the time label of the last point that meets the threshold condition in the continuous segment.

[0025] As a further aspect of the present invention, the trend capture module includes:

[0026] The sequence extraction submodule calls the multi-index co-mutation time period to obtain the N-terminal brain natriuretic peptide precursor and troponin detection value sequences in each segment, constructs a fixed time window according to the segment length, and smooths the change sequence of the detection items within the window to generate an index sliding change sequence set.

[0027] The residual calculation submodule performs first-order polynomial fitting on each detection value sequence based on the index sliding change sequence set, calculates the corresponding point residual value between the fitted sequence and the original sequence, establishes a sequence of residual values ​​changing over time, calculates the incremental difference within continuous point segments, calculates the trend offset, and generates an index trend offset sequence.

[0028] The inflection point calibration submodule calls the indicator trend offset sequence, calculates the offset increase for the trend offset of adjacent time windows, compares the offset increase with the stable offset fluctuation threshold, determines whether the offset increase in two adjacent time windows continuously exceeds the stable offset fluctuation threshold, registers the time position that meets the condition as the trend change time point, and generates the indicator trend change time location set.

[0029] As a further aspect of the present invention, the offset increase is the change in trend offset between two adjacent time windows; the stable offset fluctuation threshold refers to setting the standard deviation of all trend offsets in the trend offset sequence as the stable offset fluctuation threshold within the time period in which no trend change occurs.

[0030] As a further aspect of the present invention, the system further includes:

[0031] The graded early warning module determines the risk level of acute infarction based on each time point in the time location set of the trend change of the indicators, according to the combined situation of creatine kinase isoenzyme reaching the upper limit of clinical diagnosis and the rapid increase trend of N-terminal brain natriuretic peptide precursor. If both conditions are met, the point is marked as an early warning node, and the stratification standard corresponding to the current value of creatine kinase isoenzyme is called to perform graded labeling, generating an acute infarction risk level labeling table.

[0032] The acute infarction risk level labeling table includes the risk identification time point, the joint indicator label value, and the level classification reference item.

[0033] As a further aspect of the present invention, the graded early warning module includes:

[0034] The parameter extraction submodule calls the time points in the time location set of the indicator trend change time, and extracts the current detection value, maximum detection value and detection value change difference of the corresponding creatine kinase isoenzyme and N-terminal brain natriuretic peptide precursor within the set time interval before and after, respectively, and constructs the time window feature vector set of the two indicators to generate the dual indicator window feature sequence.

[0035] The joint judgment submodule determines, based on each time point in the dual-indicator window feature sequence, whether the current detection value of creatine kinase isoenzyme has reached the upper limit of clinical diagnostic reference, and whether the difference between the detection value of N-terminal brain natriuretic peptide precursor at the current time point and the minimum detection value within the time window is greater than twice the standard deviation of all detection values ​​within the time window. If both conditions are met, the time point is registered as an early warning node, and the set of joint indicator trigger nodes is obtained.

[0036] The risk labeling submodule maps the creatine kinase isoenzyme detection value corresponding to each time point in the set of joint indicator trigger nodes to the corresponding interval in the risk grading standard, retrieves the stratification segment number to which the current value belongs, and appends the grading number to the corresponding time point to establish an acute infarction risk level labeling table.

[0037] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0038] In this invention, key time points are extracted and integrated from continuous monitoring data during hospitalization. A time series is constructed by combining the slope change rate and numerical mutation rate of multiple physiological indicators. Based on this, related behavioral time points in various data sources are retrieved simultaneously to achieve collaborative mutation identification between different data sources. By curve fitting and residual analysis of indicator change trends, the accuracy of trend deviation time point identification is enhanced. By combining the current value of the indicator with the historical change range, condition combination judgment is implemented to achieve early warning of disease mutation events and dynamic classification of risk levels. This enhances the ability to capture mutation events under multi-indicator fusion and improves the efficiency of comprehensive judgment on the evolution of acute diseases. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0040] Figure 1 This is a system flowchart of the present invention;

[0041] Figure 2 This is a schematic diagram of the system framework of the present invention;

[0042] Figure 3 This is a flowchart of the time point extraction module of the present invention;

[0043] Figure 4 This is a flowchart of the sequence integration module of the present invention;

[0044] Figure 5 This is a flowchart of the protrusion recognition module of the present invention;

[0045] Figure 6 This is a flowchart of the trend capture module of the present invention;

[0046] Figure 7 This is a flowchart of the graded early warning module of the present invention. Detailed Implementation

[0047] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0048] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0049] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0050] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0051] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0052] Please see Figure 1 A cloud-based intelligent analysis system for clinical data of acute infarction disease, comprising a time point extraction module, a segment integration module, a protrusion identification module, a trend capture module, and a graded early warning module;

[0053] The time point extraction module receives continuous monitoring data records from patients during hospitalization via the cloud, extracts the QRS waveform amplitude and interval sequence of every 30-second segment in the electrocardiogram, retrieves the ST segment waveform slope change, and the absolute value of the difference between the creatine kinase isoenzyme and troponin values ​​for each sampling. It selects the node time points where both the slope change rate and the index value mutation rate increase in chronological order as the initial positioning points, performs continuous screening on the operation type number in the diagnosis and treatment behavior records, extracts the time nodes where the operation type changes and integrates the source nodes, and matches each node with other source nodes within a certain time range before and after on the time axis. If multiple matches exist, the node is recorded as an overlapping time point node, generating an acute infarction associated overlapping time point set.

[0054] The segment integration module calls each node in the acute infarction associated overlapping time point set, and retrieves three items within a certain time range before and after the node: the severity level of diagnosis and treatment behavior, the direction of change of ST segment slope, and the direction of change of creatine kinase isoenzyme rise rate. It determines whether there is a consistent upward trend among the three. If not, the node is shifted forward or backward by a fixed time according to the direction of trend deviation, and all nodes are reordered to form a complete time series, generating an adjusted disease course time series.

[0055] The abrupt segment identification module constructs a heart rate change velocity vector and an ST segment slope vector within a continuous time window based on the interval between adjacent time nodes in the adjusted disease course time series. It calculates the point-to-point difference sequence between the two vectors, marks the time periods with prominent changes in the difference sequence as abrupt segments, records the start and end times of each segment and the corresponding heart rate fluctuation amplitude and waveform slope change trend, and generates multi-indicator collaborative abrupt segment time periods.

[0056] The heart rate change velocity vector is a sequence of heart rate changes per unit time; the ST segment slope vector is a sequence of ST segment amplitude changes over time, which can be calculated by linear fitting of the slope from the QRS endpoint to the highest point of the ST segment; the difference sequence is the numerical difference between the two vectors at the same moment, and the criterion is that the increase in the difference exceeds the historical average within the time period.

[0057] The trend capture module calls the detection sequences of N-terminal pro-brain natriuretic peptide and troponin within the time period of multi-indicator synergistic mutation. Using the length of each interval as a window, it extracts the value change sequence of each indicator within the window, fits the change value and establishes a first-order nonlinear curve, calculates the trend of the residual value change between the fitted curve and the actual sequence, and if the residual value continuously increases beyond the stable fluctuation range, it is marked as the trend offset time point, generating the indicator trend mutation time location set.

[0058] N-terminal pro-brain natriuretic peptide (BNP) and troponin are widely used biomarkers for monitoring acute cardiovascular events. BNP primarily reflects changes in ventricular load, while troponin is a direct indicator of myocardial injury. Trend deviation points are identified through the residual growth trend. The residual is the difference between the actual value and the fitted value, reflecting the degree to which the current indicator deviates from stable changes.

[0059] The graded early warning module locates each time point in the trend change time set of the indicators, extracts the current value, historical maximum value and growth rate of the corresponding creatine kinase isoenzyme and N-terminal brain natriuretic peptide precursor within the time interval before and after, and judges based on the combined situation of creatine kinase isoenzyme reaching the upper limit of clinical diagnosis and N-terminal brain natriuretic peptide precursor showing a rapid growth trend. If both conditions are met, the point is marked as an early warning node, and the stratification standard corresponding to the current value of creatine kinase isoenzyme is called to perform graded labeling, generating an acute infarction risk level labeling table.

[0060] The upper limit of clinical reference for creatine kinase isoenzymes and the increasing trend of N-terminal pro-brain natriuretic peptide are both defined by publicly available medical standards. Creatine kinase isoenzymes are usually used to determine whether an acute myocardial infarction has occurred, while N-terminal pro-brain natriuretic peptide is used to assess cardiac load. The grading standards are based on the clinically accepted range of reference values ​​and are not original settings.

[0061] The acute infarction-related overlapping time point set includes waveform structure overlap nodes, numerical mutation synchronization nodes, and diagnosis and treatment behavior change nodes. The adjusted disease course time series includes node ranking results, trend consistency markers, and time fine-tuning labels. The multi-indicator synergistic mutation time period includes heart rate fluctuation segments, waveform tilt segments, and difference amplification segments. The indicator trend mutation time location set includes residual increase inflection points, curve offset time points, and trend turning point nodes. The acute infarction risk level labeling table includes risk identification time points, joint indicator label values, and level classification reference items.

[0062] Please see Figure 2 and Figure 3 The time point extraction module includes:

[0063] The data receiving submodule receives continuous monitoring data records of patients during hospitalization via the cloud, collects QRS waveform amplitude and QRS interval sequences in every 30-second segment of the electrocardiogram record, extracts continuous sampling data of creatine kinase isoenzyme and troponin, performs absolute value calculation on the difference between adjacent sampling values, constructs an index mutation intensity sequence, aligns each sequence with the time axis and performs joint processing to generate a continuous physiological parameter time mapping set;

[0064] Taking a patient admitted with a suspected acute myocardial infarction as an example, data was collected from their electrocardiogram monitor between the 10th and 12th minute after admission. Specifically, each 30-second segment of the electrocardiogram recording was marked with a time label. , , , Within each segment, all QRS groups are identified, and the amplitude value of the R-wave peak of each group is recorded and the average value within the segment is calculated to obtain the QRS waveform amplitude sequence, for example... Simultaneously, the time difference between each QRS group and the vertex of its next group, i.e., the RR interval, is calculated, and the average RR interval within the segment is also calculated to obtain the QRS interval sequence, for example... Simultaneously, synchronized laboratory test data were retrieved from the patient's electronic medical record system, extracting continuous sampling data of creatine kinase isoenzyme (CK-MB) and troponin I (cTnI) collected around the same time point. For example, CK-MB was... to The sampled value corresponding to the time point The sampled value corresponding to cTnI is Next, the absolute value of the difference between adjacent sample values ​​of these two myocardial enzymes is calculated to construct an index mutation intensity sequence. Taking CK-MB as an example, the first value of its mutation intensity sequence is... The second value is The third value is , to obtain the sequence Performing the same calculation on cTnI yields the sequence. Finally, the QRS waveform amplitude sequence, QRS interval sequence, CK-MB mutation intensity sequence, and cTnI mutation intensity sequence were arranged along a common time axis. Alignment and joint processing are performed to generate a continuous physiological parameter time mapping set.

[0065] The event extraction submodule calculates the tilt change rate of each ECG segment in the ST segment waveform based on the continuous physiological parameter time mapping set, determines whether the tilt angle increases synchronously in two consecutive time periods, and simultaneously screens nodes in the creatine kinase isoenzyme and troponin difference sequence where the mutation rate increases simultaneously. Time points that meet the two conditions are marked as event candidate nodes, and a set of joint indicator mutation time nodes is generated.

[0066] The slope change rate of each ECG segment in the ST segment waveform is calculated. This calculation retrieves ST segment data acquired in the same time segment as the QRS waveform amplitude. Linear regression analysis is performed on the ST segment waveform data points within each 30-second segment, and the slope of the resulting straight line is taken as the slope change rate of that segment. For example, at time point... The obtained ST segment tilt change rate sequence is Then, it is determined whether a synchronous increasing trend in tilt angle occurs in two consecutive time periods. This determination process involves comparing the rate of change of tilt at three adjacent time points, i.e., determining whether the conditions are met. and In this example, comparison The slope, and If the conditions are met, compare The slope, and The conditions are met, therefore the time period and All showed an increasing trend. Next, we simultaneously screened for nodes in the creatine kinase isoenzyme and troponin difference sequences calculated in the previous steps where the mutation rate increased simultaneously. This mutation rate is the mutation intensity value calculated above. The screening process is to determine whether there is a certain time point. At this point, the mutation intensity of CK-MB is compared with that of cTnI at the previous calculation node. All showed an increase, using the aforementioned CK-MB mutation intensity sequence. and cTnI mutation intensity sequence Their time tags are respectively ,exist At time 1, the CK-MB mutation intensity was 1. greater than Moment However, cTnI in Temporal change intensity Greater than Moment , while CK-MB in Moment Less than Moment Therefore, there are no synchronously rising nodes. For a complete demonstration, the cTnI mutation intensity sequence is now set as follows: Then in At time 1, the CK-MB mutation intensity increased from 10 to 13, and the cTnI mutation intensity increased from 0.48 to 1.0, satisfying the synchronous increase condition. Finally, the time points that met both the conditions of a continuous increase in ST segment tilt angle and a synchronous increase in the mutation rate of myocardial enzyme indicators were marked as candidate event nodes. In the adjusted settings, the time period... The slope of the ST segment increases continuously, and in The mutation rate of myocardial enzymes increases synchronously at any given time, therefore... Mark them as event candidate nodes and generate a set of joint indicator mutation time nodes.

[0067] The node matching submodule calls the time points in the joint indicator mutation time node set, compares them with the operation type number sequence recorded by time in the diagnosis and treatment behavior record, marks the record time points where the operation type value changes, and determines whether the time point coincides with the candidate node on the time axis or is within the fixed time tolerance range. If the condition is met, the time point is registered as an overlapping time point, and an acute infarction associated overlapping time point set is generated.

[0068] For candidate nodes The system retrieves the patient's medical records from the hospital information system. These records contain a sequence of operation type numbers recorded by time. For example, after the patient's admission, the sequence of operation type numbers would be: (Hospital admission registration) (Establishment of intravenous access) (Administer nitroglycerin) (Maintenance dose), this module then marks the recorded time points where the operation type value changes; in this example, the operation type value is at... The time changes from 101 to 201. The time point changes from 201 to 302, therefore the time point is marked. and Next, it is determined whether the marked time point coincides with the candidate node on the time axis or falls within a fixed time tolerance range. This fixed time tolerance range is set to 120 seconds. This value is based on the average response time from the discovery of abnormal indicators to intervention in clinical practice. This average time was obtained through a retrospective analysis of the electronic medical records of 200 diagnosed acute myocardial infarction patients, with a median response time of 115 seconds, thus rounded down to 120 seconds. The determination process involves calculating the absolute difference between the candidate node time and the marked time point and comparing it to 120 seconds. In this example, the calculation... ,because The conditions are met, and It also meets the conditions, but selects the operation marker point that is closest to the time of the candidate node, i.e. If the conditions are met, then the candidate node time point will be selected. Register as overlapping time points to generate a set of overlapping time points associated with acute infarction.

[0069] Please see Figure 2 and Figure 4 The segmentation integration module includes:

[0070] The trend extraction submodule calls the nodes in the set of overlapping time points associated with acute infarction, extracts the severity sequence of diagnosis and treatment behavior, the ST segment tilt change direction sequence, and the creatine kinase isoenzyme rise rate direction sequence within a set time range before and after each node, performs directional difference statistics on the directional changes between adjacent records in each type of sequence, and outputs the directional trend group of three parameters for each node, generating a multi-parameter trend direction group.

[0071] For overlapping time points Three types of sequences were extracted within a set time range before and after the event point. This time range was set at 60 minutes before and after the event point, based on the analysis of historical AMI patient data, which found that key information on critical disease progression is usually concentrated within one hour around the event point. The specific sequences extracted were: the severity level sequence of treatment behavior, the ST segment tilt change direction sequence, and the creatine kinase isoenzyme rise rate direction sequence. First, the severity level sequence of treatment behavior was obtained by mapping the operation type number to a preset severity level table. For example, level 1 corresponds to observation (numbers 100-199), level 2 corresponds to routine treatment (numbers 200-299), and level 3 corresponds to emergency intervention (numbers 300-399). The ST segment tilt change direction sequence was obtained by judging the sign of the ST segment tilt change rate: a positive value is an increase (marked as 1), a negative value is a decrease (marked as -1), and no change is 0. The slope sequence in the aforementioned example... All values ​​are positive. The directional sequence of the rate of increase of creatine kinase isoenzymes is obtained by calculating the second difference of CK-MB concentration values. Positive values ​​represent accelerated increase (labeled as 1), and negative values ​​represent decelerated increase (labeled as -1). For example, the CK-MB concentration sequence... Its first-order difference (rate) is The second-order difference (acceleration) is The corresponding direction sequence is Subsequently, directional difference statistics were performed on the directional changes between adjacent records in each sequence type, based on the severity level sequence of the diagnostic and treatment behavior. For example, the directional difference is ST segment tilt direction sequence ,exist Take two points nearby, and the difference is CK-MB rising rate direction sequence exist The difference in the vicinity is Finally, the output shows the directional trend groups of three types of parameters for each node, such as the node... The corresponding directional trend group is Generate multi-parameter trend direction groups.

[0072] The consistency determination submodule is based on the multi-parameter trend direction group. It combines the pairwise direction values ​​of the three direction parameters corresponding to each node, accumulates the number of consistent directions in the combination result, and marks the node as a trend deviation node if the number of inconsistent directions is greater than or equal to one, and establishes a trend deviation node identifier set.

[0073] For example, nodes Obtaining directional trend groups Here, 1 indicates an increase in the severity of the treatment behavior, 0 indicates no change in the ST segment tilt trend, and -2 indicates that the CK-MB rise rate has changed from acceleration to deceleration. The three directional parameters corresponding to each node are combined pairwise. The directional values ​​here are classified according to their positive or negative values: positive values ​​represent an enhancing or upward trend (denoted as P), negative values ​​represent a weakening or downward trend (denoted as N), and zero represents stability (denoted as S). The trend group... The corresponding directional classification is Then, by combining them in pairs, we obtain... , , Three combinations were then considered, and the number of directions with consistent orientation was accumulated in the combination results. The criterion for determining direction consistency was that both directions were either P or N. For example... or Recorded as consistent, and , , All of these are inconsistent. In this example, the number of inconsistent combinations is 3. Next, we check if the number of inconsistent directions is greater than or equal to one item. In this example, the number of inconsistent directions is 3. If the condition is met, then the node will be... Mark them as trend offset nodes and establish a trend offset node identifier set.

[0074] The sequence update submodule adjusts the fixed time step of the time label of each offset node according to the order indicated by the main direction based on the trend offset node identifier set, and re-sorts each node according to the adjusted time to generate the adjusted disease course time series.

[0075] For example, nodes The time stamps are adjusted to a fixed time step in the order indicated by the principal direction. The principal direction is determined based on the different weights of three parameters (treatment behavior, ST segment, and CK-MB) on disease progression. According to a confirmatory study involving 300 AMI patients, logistic regression analysis was used to derive the weight coefficients of each parameter on the prediction of major adverse cardiovascular events (MACE), which were: treatment behavior (0, 5), ST segment (0, 3), and CK-MB (0, 2). The principal direction is defined as the direction of change of the parameter with the highest weight. In this case, treatment behavior has the highest weight, and its direction is P (increase), indicating disease progression. Therefore, the time adjustment direction is forward (time stamp decreases). The fixed time step is set to 300 seconds. This value is based on the average delay time observed clinically after key interventions in physiological indicators. This average delay time was obtained through analysis of continuous monitoring data from 150 patients after PCI, and the median of the delay time distribution was 280 seconds, so it was rounded to 300 seconds. The adjustment process is: New time = Original time - Time step, i.e. If the dominant direction is N (weakening), the time is adjusted backward, i.e., new time = original time + time step. If the dominant direction is S (stable), the time is not adjusted. After adjusting the time labels of all offset nodes, all nodes (including nodes not marked as offset) are reordered according to the adjusted time. For example, if there are nodes in the original sequence... After adjustment, it becomes , and generate the adjusted disease course time series.

[0076] Please see Figure 2 and Figure 5 The protrusion recognition module includes:

[0077] The vector generation submodule collects heart rate records and ST segment potential sequences within each time period based on the interval between adjacent time nodes in the adjusted disease course time series. It calculates the difference between adjacent record values ​​for the heart rate records and divides it by the corresponding time interval to obtain the heart rate change rate sequence. It fits a straight line to the ST segment potential points and calculates the corresponding slope sequence, which are then organized into two time-aligned vectors to generate a continuous segment feature vector set.

[0078] For example, the first time interval in the sequence, i.e. and Between, the time period is Collect all heart rate recordings and ST segment potential sequences within this time period, for example, the first rate value is... (times / minute) / second, the second is (beats / minute) / second, and so on, to obtain a complete rate sequence. Simultaneously, the corresponding ST segment potential sequence within this time period is acquired. A straight line is fitted to each ST segment potential point within a 30-second interval, and its slope is calculated, resulting in an ST segment slope sequence aligned with the heart rate change rate sequence. For example... Finally, the heart rate change rate sequence and the ST segment slope sequence are sorted into two time-aligned vectors to generate a continuous segment feature vector set.

[0079] The difference calculation submodule calls the heart rate change rate vector and ST segment slope vector in the continuous segment feature vector set, performs difference calculation on each corresponding time point in the vector, and obtains the point-to-point difference sequence between heart rate and waveform slope. It then performs amplitude value filtering within a sliding window on the difference sequence, extracts the intervals where the difference amplitude exceeds the difference mean offset threshold, and establishes a fluctuation difference interval distribution set.

[0080] The difference mean offset threshold is the sum of the mean and standard deviation of the difference sequence within the sliding window, where the mean is the arithmetic mean of all point-to-point differences within the sliding window, and the standard deviation is the root mean of the squared deviations of all point-to-point differences from the mean within the sliding window.

[0081] The intervals where the difference magnitude exceeds the mean difference offset threshold are the time periods of several consecutive point-to-point differences that are greater than the mean difference offset threshold. The start time of the time period is the time label of the first point that meets the threshold condition, and the end time is the time label of the last point that meets the threshold condition in the continuous segment.

[0082] At the first time point, the difference is... The difference at the second time point is This process is repeated to obtain a complete difference sequence. Then, a sliding window is used to filter the amplitude values ​​within this sequence. The sliding window size is set to 5 consecutive data points, a conclusion drawn from the analysis of data from 200 myocardial infarction patients. This window size ensures sufficient data points for statistical calculation while being sensitive to short-term fluctuations. Within each window, the intervals where the difference amplitude exceeds the mean deviation threshold are extracted. This mean deviation threshold is dynamically calculated and is the sum of the arithmetic mean and standard deviation of all point-to-point differences within the sliding window, using a window containing 5 difference points. For example, its mean is Its standard deviation is Therefore, the mean offset threshold of the difference in this window is Subsequently, several consecutive time periods with point-to-point differences greater than the threshold are selected. In this example window, the difference 0.25 is greater than the threshold 0.234. The next value is set to 0.26, which is also greater than the updated threshold. The values ​​after that are less than the threshold. The start time of this interval is the time label of the first point that meets the condition (the point with a difference of 0.25), and the end time is the time label of the last point that meets the condition in the continuous segment (the point with a difference of 0.26). This establishes a distribution set of fluctuation difference intervals.

[0083] The ST segment extraction submodule calculates the difference between the maximum and minimum values ​​within a time period by calling the original heart rate record based on the start and end times of the concentrated difference interval distribution of fluctuation difference intervals. At the same time, it extracts the ST segment slope change trend label and performs a structured combination of the segment start and end times, heart rate fluctuation amplitude, and slope change direction to generate a multi-indicator collaborative mutation time period.

[0084] For example, a identified mutation segment is This involves retrieving the original heart rate records for that time period and calculating the difference between the maximum and minimum heart rate during that period, i.e., the heart rate fluctuation range. For example, it can be calculated as follows: The algorithm executes the following steps per minute, simultaneously extracting the trend label of the ST segment slope change for that segment. This label is obtained by judging the overall trend of the ST segment slope sequence within the segment. For example, if the slope sequence of that segment is... Its overall trend is calculated by the difference between the first and last values. If the change is positive, it is marked as "increasing" (U); if it is negative, it is marked as "decreasing" (D); if the absolute value of the change is less than a preset small change amount (e.g., 0.01 mV / s), it is marked as "stable" (S). Finally, the start and end times of the segment, the calculated heart rate fluctuation amplitude, and the extracted slope change direction are structurally combined to form a record, for example: {start time: 120s, end time: 180s, heart rate fluctuation amplitude: 5 times / minute, slope change direction: "U"}, generating a multi-indicator collaborative mutation time period.

[0085] Please see Figure 2 and Figure 6 The trend capture module includes:

[0086] The sequence extraction submodule calls the multi-index co-mutation time period to obtain the N-terminal brain natriuretic peptide precursor and troponin detection value sequences in each segment. A fixed time window is constructed according to the segment length, and the change sequence of the detection item within the window is smoothed to generate an index sliding change sequence set.

[0087] For example, recording {start time: 120s, end time: 180s, ...}, the system obtains the sequence of N-terminal pro-brain natriuretic peptide (NT-proBNP) and troponin I (cTnI) levels for each segment. These levels are obtained from the testing section of the electronic medical record system and matched according to their reporting time. For instance, the time interval from 120s to 180s may contain one or more levels. If no level is found within that time interval, the system searches forward or backward for the most recent level as a representative. A fixed time window is then constructed based on the segment length, with the window length set at 6 hours. This setting is based on... A statistical analysis of the half-lives of NT-proBNP and cTnI and the frequency of routine clinical re-examinations in 500 heart failure patients showed that 6 hours is an effective timescale for capturing dynamic changes in these indicators and is consistent with clinical practice. The center point of the time interval for synergistic mutations of multiple indicators was used as the center of the time window. All NT-proBNP and cTnI test values ​​within this time window were extracted to form the original sequences. For example, for NT-proBNP, the change sequence within the window was smoothed using a moving average method with a window size of 3. After processing, a smoothed sequence was obtained. ,Right now pg / mL, the smoothed sequences of the two indicators corresponding to all mutation time periods are aggregated to generate a set of indicator sliding change sequences.

[0088] The residual calculation submodule performs first-order polynomial fitting on each detection value sequence based on the sliding change sequence set of indicators, calculates the residual values ​​at corresponding points between the fitted sequence and the original sequence, establishes a sequence of residual values ​​changing over time, and calculates the incremental difference within continuous point segments using the formula:

[0089] ;

[0090] The trend offset is calculated and a sequence of indicator trend offsets is generated.

[0091] in, Indicates the first Within the first time window Trend deviation of the detection index Indicates the first Within the first time window The normalized value of the fitting residual for a class of detection indicators is obtained by dividing the difference between the original detection data and the first-order fitted sequence by the absolute maximum value of the sequence. Indicates the first Class detection indicators in all The average of the normalized residuals over each time window, calculated from all... Calculate the average value. Indicates the first Within the first time window The normalized variation of the detection index is obtained by dividing the difference between the maximum and minimum detection values ​​within the time window by the maximum value of the entire index sequence. This indicates the number of time windows included in this type of detection indicator throughout the entire monitoring period. The index in the summation symbol indicates the number of all time windows used for variance calculation;

[0092] For example, smoothed sequences of NT-proBNP pg / mL, perform first-order polynomial fitting, that is, use the least squares method to find a straight line. To best fit this set of data points, we obtain a fitted sequence. For example, we calculate the residual values ​​between corresponding points of the fitted sequence and the original smoothed sequence. The residual sequence is... ,Right now After establishing a sequence of residual values ​​changing over time, the incremental difference within consecutive point segments is calculated using the formula:

[0093] ;

[0094] The calculation yields the trend deviation, a formula used to determine the severity of a detection indicator's deviation from its local linear trend within a specific time window. This deviation is weighted by incorporating the actual fluctuation amplitude within that window. The formula's logic involves first calculating the normalized residual... Its mean The difference quantifies the degree to which a single point deviates from the average trend. Then, the denominator... It measures the dispersion of residuals across all time windows. The entire fractional structure resembles a standardization process, highlighting residuals that deviate abnormally from the overall fluctuation pattern, and is finally multiplied by the normalized magnitude of change. Its function is to give higher weight to trend deviations that occur when the indicator itself changes drastically. Here is an explanation of the parameters in the formula: Indicates the first Within the first time window Class detection indicators (e.g.) The trend offset of NT-proBNP, Indicates the first Within the first time window The normalized value of the fitting residuals for a class of detection indicators is calculated by dividing the difference between the original detection data and the first-order fitted sequence by the absolute maximum value of the residuals of that indicator across all time windows. Indicates the first Class detection indicators in all The arithmetic mean of the normalized residuals over a time window Indicates the first Within the first time window The normalized variation of a detection index is calculated by dividing the difference between the maximum and minimum detection values ​​within the time window by the maximum value of the index over the entire monitoring period. This indicates the total number of time windows encompassed by this type of detection indicator throughout the entire monitoring period. The index in the summation symbol represents the range from 1 to... Each time window number is now calculated using a specific example, setting the NT-proBNP index (i.e. During the entire monitoring period, a total of Calculate the time window. Trend offset of each time window ;

[0095] Table 1. Calculation data for NT-proBNP index

[0096] Time window number (i) Original residuals (pg / mL) Maximum value within the window (pg / mL) Minimum value within the window (pg / mL) 1 50 900 800 2 -80 1100 900 3 120 1500 1000 4 -40 1400 1300

[0097] Table 1 lists the intermediate data required for the calculation. First, the global parameters are determined, and the absolute maximum value of all residuals is... The maximum detectable value of NT-proBNP during the entire monitoring period was 1500 pg / mL;

[0098] Calculate normalized residuals :

[0099] ;

[0100] ;

[0101] ;

[0102] ;

[0103] Calculate the normalized residual average :

[0104] ;

[0105] Calculate the normalized change magnitude :

[0106] ;

[0107] Substitute into the formula to calculate :

[0108] ;

[0109] The advantage of this formula lies in its ability to more accurately identify "true" inflection points—those that not only exhibit statistically abnormal trends but also show significant changes in clinical indicator values—by combining the relative deviation of the residuals with the fluctuation range of the indicator itself. This avoids interference from spurious trend changes caused by minor data fluctuations, ultimately resulting in a more accurate calculated trend shift. The dimensionless value of 0.079 comprehensively reflects the anomaly of NT-proBNP changes within the second time window and is used to generate the indicator trend offset sequence.

[0110] The inflection point calibration submodule calls the indicator trend offset sequence, calculates the offset increase for the trend offset of adjacent time windows, compares the offset increase with the stable offset fluctuation threshold, determines whether the offset increase in two adjacent time windows continuously exceeds the stable offset fluctuation threshold, registers the time position that meets the condition as the trend change time point, and generates the indicator trend change time location set.

[0111] The offset increase is the change in trend offset between two adjacent time windows; the stable offset fluctuation threshold refers to setting the standard deviation of all trend offsets in the trend offset sequence as the stable offset fluctuation threshold within the time interval where no trend change occurs.

[0112] For example, the NT-proBNP trend offset sequence is... The increase in trend offset is calculated for adjacent time windows, which is the difference between two adjacent points. The first increase is... The second increase was The third increase was The amplification sequence is obtained. Then, the increase in offset is compared with a stable offset fluctuation threshold. This stable offset fluctuation threshold is set as the standard deviation of the trend offset sequence over time periods without trend abrupt changes. The benchmark dataset is derived from the same indicator sequences of 100 patients with stable coronary artery disease, and the standard deviation of their trend offset sequences is calculated to be 0 to 150. Therefore, the stable offset fluctuation threshold is set to 0 to 150. Next, it is determined whether the increase in offset exceeds this threshold consecutively in two adjacent time windows, i.e., whether a trend abrupt change occurs. and In this example, the increase sequence is Only the second increase value, 0.176, is greater than the threshold value, 0.150, which does not meet the condition of "two consecutive increases". For a complete demonstration, the increase sequence is now set as follows. Then the second increase The third increase If the condition of continuously exceeding the threshold is met, the starting time position of the continuous segment that meets the condition is registered as the trend change point. Under this setting, the second increase corresponds to the change from time window 2 to 3, and the third increase corresponds to the change from time window 3 to 4. The continuity occurs near time window 3, so the starting time point of time window 3 is marked as the trend change point, and the indicator trend change time location set is generated.

[0113] Please see Figure 2 and Figure 7 The tiered early warning module includes:

[0114] The parameter extraction submodule calls the time point of the index trend change time location set, and extracts the current detection value, maximum detection value and detection value change difference of the corresponding creatine kinase isoenzyme and N-terminal brain natriuretic peptide precursor within the set time interval before and after. It constructs the time window feature vector set of the two indicators respectively, and generates the dual indicator window feature sequence.

[0115] The starting point of the calibrated time window 3 was used to extract the corresponding creatine kinase isoenzyme (CK-MB) and N-terminal pro-brain natriuretic peptide (NT-proBNP) values ​​within a set time interval before and after this point. This time interval was set to 2 hours, based on clinical observations that myocardial markers show significant changes within 2 hours after a critical event. The extracted parameters included: the current detection value, the maximum detection value within the interval, and the difference in detection values ​​(maximum value minus minimum value). Taking CK-MB as an example, within a 2-hour interval near the starting point of time window 3, the current value was 28 ng / mL, the maximum value within the interval was 42 ng / mL, and the difference in values ​​was... The same operation was performed on NT-proBNP at ng / mL to construct time window feature vector sets for the two indicators, generating dual-indicator window feature sequences.

[0116] The joint judgment submodule determines whether the current detection value of creatine kinase isoenzyme reaches the upper limit of clinical diagnosis based on each time point in the dual-indicator window feature sequence, and whether the difference between the detection value of N-terminal brain natriuretic peptide precursor at the current time point and the minimum detection value in the time window is greater than twice the standard deviation of all detection values ​​in the time window. If both conditions are met, the time point is registered as an early warning node, and the set of joint indicator trigger nodes is obtained.

[0117] The CK-MB and NT-proBNP feature vectors generated in time window 3 are used to perform two judgments. The first judgment is to determine whether the current creatine kinase isoenzyme detection value reaches the upper limit of the clinical diagnostic reference. This upper limit is set at 25 ng / mL based on cardiovascular disease reports. This is a highly specific diagnostic threshold that has been validated through large-scale epidemiological surveys and clinical studies. In this case, the current CK-MB detection value is 28 ng / mL. If the condition is met, the second step is to determine whether the difference between the current detection value of the N-terminal pro-brain natriuretic peptide (NT-proBNP) precursor and the minimum detection value within the time window is greater than twice the standard deviation of all detection values ​​within the time window. First, calculate the standard deviation of the NT-proBNP sequence in pg / mL, setting its mean to 1300 and its standard deviation to approximately 187. Twice the standard deviation is 374. Then, calculate the difference between the current detection value and the minimum value within the window. The current value is the first value of the sequence, 1000, and the minimum value within the window is also 1000; therefore, the difference is 0. If the first condition is not met, the second condition is not met. If both conditions are met, the time point is registered as an early warning node. In this example, the second condition is not met, so the time point is not registered. For a complete demonstration, the current value of the NT-proBNP sequence is set to 1500, the minimum value to 1000, the difference to 500, and twice the standard deviation to 374. If the second condition is met, and both conditions are met, then the time point is registered as an early warning node, and the set of joint indicator trigger nodes is obtained.

[0118] The risk labeling submodule maps the creatine kinase isoenzyme detection value corresponding to each time point in the joint indicator trigger node set to the corresponding interval in the risk grading standard, retrieves the stratification segment number to which the current value belongs, and appends the grading number to the corresponding time point to establish an acute infarction risk level labeling table.

[0119] The risk stratification criteria were established based on a prognostic analysis of 1000 AMI patients. The study showed that peak CK-MB was significantly associated with the risk of major adverse cardiovascular events within one year. The specific stratification criteria are as follows:

[0120] Table 2. Risk Classification Criteria for Acute Infarction

[0121]

[0122] As shown in Table 2, this standard divides CK-MB detection values ​​into four intervals, corresponding to four risk classification numbers. In the demonstration case in this section, the current CK-MB value that triggers the warning node is 28 ng / mL, which falls within the interval... Within the current time frame, the stratification segment number to which the current value belongs is retrieved, and the stratification number is obtained as 2. Finally, this stratification number is appended to the corresponding time point to form a record, such as {Time point: Start of time window 3, CK-MB value: 28ng / mL, Risk stratification: 2}, to establish an acute infarction risk level labeling table.

[0123] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0124] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0125] It should be understood that, in various embodiments of the present invention, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0126] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0127] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0128] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0129] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0130] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0131] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0132] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A cloud-based intelligent analysis system for clinical data in acute infarction, characterized in that, The system includes: The time point extraction module receives continuous monitoring data records from patients during hospitalization via the cloud, extracts the QRS waveform amplitude and interval sequence of every 30-second segment in the electrocardiogram, selects nodes where both the slope change rate and the index value mutation rate increase, identifies and records overlapping time point nodes, and generates an acute infarction-related overlapping time point set. The segment integration module calls the set of overlapping time points associated with acute infarction to determine whether there is a consistent upward trend in the target indicators. If there is no consistency, the nodes are shifted forward or backward by a fixed time according to the direction of trend deviation to generate an adjusted disease course time series. The mutation segment identification module constructs a heart rate change velocity vector and an ST segment slope vector within a continuous time window based on the adjusted disease course time series, calculates the point-to-point difference sequence of the two vectors, marks the mutation segment, and generates a multi-index synergistic mutation time period. The protrusion recognition module includes: The vector generation submodule collects heart rate records and ST segment potential sequences within each time period based on the interval between adjacent time nodes in the adjusted disease course time series. It calculates the difference between adjacent record values ​​for the heart rate records and divides it by the corresponding time interval to obtain the heart rate change rate sequence. It fits a straight line to the ST segment potential points and calculates the corresponding slope sequence. These are then organized into two time-aligned vectors to generate a continuous segment feature vector set. The difference calculation submodule calls the heart rate change rate vector and ST segment slope vector in the continuous segment feature vector set, performs difference calculation on each corresponding time point in the vector, obtains the point-to-point difference sequence between heart rate and waveform slope, performs amplitude value filtering within a sliding window on the difference sequence, extracts the intervals where the difference amplitude exceeds the difference mean offset threshold, and establishes a fluctuation difference interval distribution set. The ST segment extraction submodule calculates the difference between the maximum and minimum values ​​within the time period by calling the original heart rate record based on the start and end times of the difference abrupt change segments in the distribution of the fluctuation difference interval. At the same time, it extracts the ST segment slope change trend label and performs a structured combination of the segment start and end times, heart rate fluctuation amplitude, and slope change direction to generate a multi-indicator collaborative mutation time period. The trend capture module calls the detection sequences of N-terminal pro-brain natriuretic peptide and troponin within the multi-indicator synergistic mutation time period, fits the change values ​​and establishes a first-order nonlinear curve, calculates the trend of the residual value change between the fitted curve and the actual sequence, marks the trend offset time point, and generates the index trend mutation time location set.

2. The cloud-based intelligent analysis system for clinical data of acute infarction according to claim 1, characterized in that, The acute infarction-related overlapping time point set includes waveform structure overlapping nodes, numerical mutation synchronization nodes, and diagnosis and treatment behavior change nodes. The adjusted disease course time series includes node ranking results, trend consistency markers, and time fine-tuning labels. The multi-indicator collaborative mutation time period includes heart rate fluctuation segments, waveform tilt segments, and difference amplification segments. The indicator trend mutation time location set includes residual increase inflection points, curve offset time points, and trend turning point nodes.

3. The cloud-based intelligent analysis system for clinical data of acute infarction according to claim 2, characterized in that, The time point extraction module includes: The data receiving submodule receives continuous monitoring data records of patients during hospitalization via the cloud, collects QRS waveform amplitude and QRS interval sequences in every 30-second segment of the electrocardiogram record, extracts continuous sampling data of creatine kinase isoenzyme and troponin, performs absolute value calculation on the difference between adjacent sampling values, constructs an index mutation intensity sequence, aligns each sequence with the time axis and performs joint processing to generate a continuous physiological parameter time mapping set; The event extraction submodule calculates the tilt change rate of each ECG segment in the ST segment waveform based on the continuous physiological parameter time mapping set, determines whether the tilt angle increases synchronously in two consecutive time periods, and simultaneously screens nodes in the creatine kinase isoenzyme and troponin difference sequence where the mutation rate increases simultaneously. Time points that meet both conditions are marked as event candidate nodes, and a set of joint indicator mutation time nodes is generated. The node matching submodule calls the time points in the joint indicator mutation time node set, compares them with the operation type number sequence recorded by time in the diagnosis and treatment behavior record, marks the record time points where the operation type value changes, and determines whether the time point coincides with the candidate node on the time axis or is within the fixed time tolerance range. If the condition is met, the time point is registered as an overlapping time point, and an acute infarction associated overlapping time point set is generated.

4. The cloud-based intelligent analysis system for clinical data of acute infarction according to claim 3, characterized in that, The segmentation integration module includes: The trend extraction submodule calls the nodes in the set of overlapping time points associated with acute infarction, extracts the severity sequence of diagnosis and treatment behavior, the ST segment tilt change direction sequence, and the creatine kinase isoenzyme rise rate direction sequence within a set time range before and after each node, performs directional difference statistics on the directional changes between adjacent records in each type of sequence, and outputs the directional trend group of three parameters for each node, generating a multi-parameter trend direction group. The consistency determination submodule, based on the multi-parameter trend direction group, combines the pairwise direction values ​​of the three direction parameters corresponding to each node, accumulates the number of consistent directions in the combination result, and if there is a number of inconsistent directions greater than or equal to one, the node is marked as a trend offset node and a trend offset node identifier set is established. The sequence update submodule adjusts the fixed time step of the time label of each offset node according to the order indicated by the main direction based on the trend offset node identifier set, and reorders each node according to the adjusted time to generate the adjusted disease course time series.

5. The cloud-based intelligent analysis system for clinical data of acute infarction according to claim 4, characterized in that, The difference mean offset threshold is the sum of the mean and standard deviation of the difference sequence within the sliding window, where the mean is the arithmetic mean of all point-to-point differences within the sliding window, and the standard deviation is the root mean of the squared deviations of all point-to-point differences from the mean within the sliding window. The interval where the extracted difference magnitude exceeds the difference mean offset threshold is a time period in which the difference between several consecutive points is greater than the difference mean offset threshold. The start time of the time period is the time label of the first point that meets the threshold condition, and the end time is the time label of the last point that meets the threshold condition in the continuous segment.

6. The cloud-based intelligent analysis system for clinical data of acute infarction according to claim 5, characterized in that, The trend capture module includes: The sequence extraction submodule calls the multi-index co-mutation time period to obtain the N-terminal brain natriuretic peptide precursor and troponin detection value sequences in each segment, constructs a fixed time window according to the segment length, and smooths the change sequence of the detection items within the window to generate an index sliding change sequence set. The residual calculation submodule performs first-order polynomial fitting on each detection value sequence based on the set of sliding change sequences of the indicators, calculates the residual value of the corresponding point between the fitted sequence and the original sequence, establishes a sequence of residual values ​​changing with time, calculates the incremental difference within continuous point segments, calculates the trend offset, and generates an indicator trend offset sequence. The inflection point calibration submodule calls the indicator trend offset sequence, calculates the offset increase for the trend offset of adjacent time windows, compares the offset increase with the stable offset fluctuation threshold, determines whether the offset increase in two adjacent time windows continuously exceeds the stable offset fluctuation threshold, registers the time position that meets the condition as the trend change time point, and generates the indicator trend change time location set.

7. The cloud-based intelligent analysis system for clinical data of acute infarction according to claim 6, characterized in that, The offset increase is the change in trend offset between two adjacent time windows; the stable offset fluctuation threshold refers to setting the standard deviation of all trend offsets in the trend offset sequence as the stable offset fluctuation threshold within the time period in which no trend change occurs.

8. The cloud-based intelligent analysis system for clinical data of acute infarction according to claim 7, characterized in that, The system also includes: The graded early warning module determines the risk level of acute infarction based on each time point in the time location set of the trend change of the indicators, according to the combined situation of creatine kinase isoenzyme reaching the upper limit of clinical diagnosis and the rapid increase trend of N-terminal brain natriuretic peptide precursor. If both conditions are met, the point is marked as an early warning node, and the stratification standard corresponding to the current value of creatine kinase isoenzyme is called to perform graded labeling, generating an acute infarction risk level labeling table. The acute infarction risk level labeling table includes the risk identification time point, the joint indicator label value, and the level classification reference item.

9. The cloud-based intelligent analysis system for clinical data of acute infarction according to claim 8, characterized in that, The tiered early warning module includes: The parameter extraction submodule calls the time points in the time location set of the indicator trend change time, and extracts the current detection value, maximum detection value and detection value change difference of the corresponding creatine kinase isoenzyme and N-terminal brain natriuretic peptide precursor within the set time interval before and after, respectively constructs the time window feature vector set of the two indicators, and generates the dual indicator window feature sequence. The joint judgment submodule determines, based on each time point in the dual-indicator window feature sequence, whether the current detection value of creatine kinase isoenzyme has reached the upper limit of clinical diagnostic reference, and whether the difference between the detection value of N-terminal brain natriuretic peptide precursor at the current time point and the minimum detection value within the time window is greater than twice the standard deviation of all detection values ​​within the time window. If both conditions are met, the time point is registered as an early warning node, and the set of joint indicator trigger nodes is obtained. The risk labeling submodule maps the creatine kinase isoenzyme detection value corresponding to each time point in the set of joint indicator trigger nodes to the corresponding interval in the risk grading standard, retrieves the stratification segment number to which the current value belongs, and appends the grading number to the corresponding time point to establish an acute infarction risk level labeling table.

Images