A left ventricular assist device multi-source data time sequence integration method and system

Abstract

The application discloses a left ventricular assist device multi-source data time sequence integration method, comprising the following steps: acquiring anticoagulation state monitoring data, hemolysis index monitoring data, device operation parameter data and clinical event record data recorded by a medical equipment system to obtain a time sequence correlation data set; obtaining an anticoagulation deviation interval feature vector and a pump operation parameter residual error feature vector from the time sequence correlation data set; obtaining an abnormal concurrent correlation degree score from the anticoagulation deviation interval feature vector and the pump operation parameter residual error feature vector; and obtaining a time sequence data integration output file from the abnormal concurrent correlation degree score. The method effectively alleviates the misjudgment problem caused by single index processing, significantly improves the early discovery efficiency of data abnormalities, and assists in distinguishing the real causes of equipment parameter abnormalities.

Description

Technical Field

[0001] This invention relates to the field of medical device data processing, specifically to a method and system for time-series integration of multi-source data from a left ventricular assist device. Background Technology

[0002] Patients with left ventricular assist devices (LVAP) are concerned about both thrombosis and bleeding, making anticoagulation management a long-term core issue. Looking at the INR (see Table 5) once is insufficient, as patients may experience periods of low or high INR; this cumulative exposure time is also important. Pump thrombosis does not necessarily manifest as only increased pump power; it often occurs alongside elevated hemolysis markers, abnormal pump current or flow, and inadequate anticoagulation. Changes in pump power do not always indicate thrombosis; changes in blood pressure, volume status, pump rate, outflow tract problems, inflow location problems, or controller events can also cause abnormalities. Because there is no single absolute diagnostic indicator for pump thrombosis, elevated LDH can result from hemolysis during blood collection, transfusion, abnormal liver function, infection, and inflammation, while increased pump power can also be caused by elevated blood pressure, pump rate adjustments, changes in volume status, and inflow or outflow tract problems. Clinically, there is a lack of systematic tools to comprehensively assess anticoagulation, hemolysis, and pump parameters sequentially over time, which can easily lead to misinterpretations of abnormal pump parameters or anticoagulation data. Existing technologies lack a method for integrating multi-source data from left ventricular assist devices that does not directly output diagnostic conclusions, but instead integrates and extracts features from multi-source heterogeneous medical data in a temporal manner to provide pure quantitative data input for subsequent external diagnostic procedures. Summary of the Invention

[0003] This invention proposes a method for time-series integration of multi-source data from a left ventricular assist device, comprising: S1. Obtain anticoagulation status monitoring data, hemolysis index monitoring data, device operating parameter data, and clinical event record data recorded by the medical device system. Obtain a time-series associated dataset from the anticoagulation status monitoring data, the hemolysis index monitoring data, the device operating parameter data, and the clinical event record data. The time-series associated dataset is a logical set generated by integrating the anticoagulation status monitoring data, the hemolysis index monitoring data, the device operating parameter data, and the clinical event record data in chronological order. S2. Obtain the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector from the time-series correlation dataset. The anticoagulation deviation interval feature vector is the result of calculation representing the deviation of the data from the set interval, generated based on the difference and duration between the INR value and the INR target interval in the time-series correlation dataset. The pump operating parameter residual feature vector is the result of calculation representing the equipment operating parameters exceeding the baseline threshold, generated based on the changes in pump power, pump current, and hemolysis index in the time-series correlation dataset. An abnormal concurrency correlation score is obtained by combining the anticoagulation deviation interval feature vector with the pump operating parameter residual feature vector. The abnormal concurrency correlation score is calculated using the formula... Calculations show that This represents the score for the abnormal concurrency correlation. The weight representing the duration during which the INR is below the target interval. This represents the duration during which the INR is below the target range. Represents the weight of the pump power residual. Represents the pump power residual value. The weighting of changes in hemolysis indicators This represents the change in hemolysis indicators.

[0004] Furthermore, the INR target range is derived from a preset fixed safety range or a dynamically adjusted range; the anticoagulation status monitoring data includes long-term warfarin anticoagulation data and unfractionated heparin bridging data; obtaining the anticoagulation deviation range feature vector from the time-series correlation dataset specifically includes: generating the anticoagulation deviation range feature vector based on the duration for which the INR in the long-term warfarin anticoagulation data in the time-series correlation dataset is lower than the INR target range, or based on the heparin infusion rate and anti-Xa results in the unfractionated heparin bridging data.

[0005] Furthermore, the pump power residual value is the difference obtained by subtracting the historical average pump power at the same pump speed from the current pump power; the step of obtaining the anticoagulation deviation interval feature vector from the time-series associated dataset also includes: based on the duration of INR being higher than the target INR interval in the time-series associated dataset, merging records of hemoglobin decrease, blood transfusion, or bleeding to generate the anticoagulation deviation interval feature vector.

[0006] Furthermore, the step of obtaining the abnormal concurrency correlation score from the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector specifically includes: when the anticoagulation deviation interval feature vector indicates low anticoagulation data and the pump operating parameter residual feature vector indicates elevated LDH and free hemoglobin, accompanied by a continuous increase in pump power or pump current residuals, by adjusting the formula... , and The weight values ​​are used to generate the abnormal concurrency correlation score.

[0007] Furthermore, the step of obtaining the abnormal concurrency correlation score from the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector also includes: when the pump operating parameter residual feature vector indicates abnormal pump power but the change in hemolysis index does not reach the threshold, and the blood pressure, volume status, pump speed adjustment, outflow channel or inflow channel data in the time-series correlation dataset can explain the abnormal pump power, a baseline calibration identifier is generated and incorporated into the time-series data integration output file.

[0008] Furthermore, the hemolysis index monitoring data includes LDH values, free hemoglobin values, and haptoglobin values; before obtaining the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector from the time-series correlation dataset, the method further includes: based on the time-series correlation dataset, removing falsely elevated data of the LDH values, free hemoglobin values, or haptoglobin values ​​caused by hemolysis of blood samples, blood transfusion, or abnormal liver function, to obtain purified hemolysis index data.

[0009] Furthermore, the clinical event record data includes infection and fever records and inflammatory marker records; before obtaining the abnormal concurrency correlation score from the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector, the method further includes: based on the infection and fever records and the inflammatory marker records in the time-series correlation dataset, identifying interference items caused by inflammatory state, such as D-dimer elevation or coagulation disorder, and performing weight downgrading processing on the interference items when calculating the abnormal concurrency correlation score.

[0010] Furthermore, the time-series data integration output file is also associated with hidden concurrent data feature identifiers; the process of obtaining the time-series data integration output file from the abnormal concurrency correlation score specifically includes: when the abnormal concurrency correlation score exceeds a preset threshold, generating the time-series data integration output file associated with data feature identifiers of acute kidney injury, stroke, or peripheral arterial embolism.

[0011] Furthermore, this application also proposes a non-volatile storage medium storing a computer program, which, when executed by a processor, implements the multi-source data timing integration method for the left ventricular assist device.

[0012] Furthermore, this application also proposes a multi-source data time-series integration system for left ventricular assist devices, comprising: an acquisition module for processing anticoagulation status monitoring data, hemolysis index monitoring data, device operating parameter data, and clinical event record data into a time-series associated dataset; a first generation module for processing the time-series associated dataset into an anticoagulation deviation interval feature vector and a pump operating parameter residual feature vector; a second generation module for processing the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector into an abnormal concurrency correlation score; and an output module for processing the abnormal concurrency correlation score into a time-series data integration output file.

[0013] This invention proposes a time-series integration method for multi-source data from left ventricular assist devices. It integrates anticoagulation status, hemolysis indicators, device operating parameters, and clinical events in chronological order and provides structured data for subsequent diagnostic procedures. This effectively alleviates the misjudgment problem caused by processing single indicators, significantly improves the early detection efficiency of data anomalies, and helps distinguish the true causes of device parameter anomalies. Attached Figure Description

[0014] Figure 1 This is a flowchart of the multi-source data timing integration method for the left ventricular assist device of the present invention; Figure 2 This is a schematic diagram of the multi-source data timing integration system of the left ventricular assist device of the present invention; Figure 3 This is a logical schematic diagram of the multi-source heterogeneous data temporal alignment and feature vector generation of the present invention; Figure 4 This is a diagram showing the parameter relationship between the abnormal concurrency correlation score calculation and weight adjustment in this invention. Figure 5 This is a diagram showing the encapsulation and feature identification association of the time-series data integration output file of this invention. Detailed Implementation

[0015] This invention proposes a method for time-series integration of multi-source data from a left ventricular assist device, comprising: S1. Obtain anticoagulation status monitoring data, hemolysis index monitoring data, device operating parameter data, and clinical event record data recorded by the medical device system. Obtain a time-series associated dataset from the anticoagulation status monitoring data, the hemolysis index monitoring data, the device operating parameter data, and the clinical event record data. The time-series associated dataset is a logical set generated by integrating the anticoagulation status monitoring data, the hemolysis index monitoring data, the device operating parameter data, and the clinical event record data in chronological order.

[0016] The data includes: anticoagulation status monitoring data (reflecting the patient's long-term warfarin anticoagulation and unfractionated heparin bridging process); hemolysis index monitoring data (reflecting intravascular hemolysis and blood compatibility); device operating parameter data (recorded by the left ventricular assist device controller as pump operating parameters); and clinical event record data (recorded clinical symptoms and confounding factors affecting coagulation or hemolysis assessment). The time-series association dataset originates from the anticoagulation status monitoring data, hemolysis index monitoring data, device operating parameter data, and clinical event record data. Its functional boundary is to perform time axis alignment and logical integration of multi-source heterogeneous data, without performing risk assessment operations. The time-series integration method for multi-source data of the left ventricular assist device implemented in this invention is a data processing process. Each step itself does not involve diagnostic purposes but only provides a data foundation for subsequent diagnostic procedures. The legal characterization of this step conforms to the regulations for non-diagnostic time-series integration methods for multi-source data of the left ventricular assist device. By focusing on data time-series integration, the direct causal link between the steps of the time-series integration method for multi-source data of the left ventricular assist device and disease diagnosis is severed, thus meeting the subject matter requirements of the patent application.

[0017] Specifically, the system records warfarin dosage and INR daily, automatically generating an INR trend chart; during the bridging period, it records unfractionated heparin pump rate, dosage, and anti-Xa results (see Table 5); it synchronously imports the controller's pump rate, flow rate, power, current, pulsatility index (see Table 5), alarm events, and controller events, placing these data in the same system in chronological order. The classification of multi-source data and the correspondence between specific parameters are shown in Table 1 below: Table 1 shows the classification of multi-source monitoring data and the corresponding specific parameters.

[0018] In one specific implementation scenario, the patient starts warfarin on the first postoperative day, with unfractionated heparin used as a bridging anticoagulant. Once warfarin takes effect and the INR reaches the target range, unfractionated heparin is discontinued. The system aligns the daily warfarin dose, INR value, target INR range, heparin infusion rate, and discontinuation time along the time axis to form a time-series correlated dataset. For example... Figure 1 As shown, the overall data flow of this invention starts with data acquisition and then gradually performs integration and feature extraction operations.

[0019] S2. Obtain the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector from the time-series correlation dataset. The anticoagulation deviation interval feature vector is the result of calculation representing the deviation of the data from the set interval, generated based on the difference and duration between the INR value and the INR target interval in the time-series correlation dataset. The pump operating parameter residual feature vector is the result of calculation representing the equipment operating parameters exceeding the baseline threshold, generated based on the changes in pump power, pump current, and hemolysis index in the time-series correlation dataset.

[0020] The anticoagulation deviation interval feature vector is derived from the difference and duration between the INR value and the target INR interval in the time-series correlation dataset. Its functional boundary is to identify the calculation result of anticoagulation data deviating from the set interval, and it does not directly diagnose pump thrombosis. The pump operating parameter residual feature vector is derived from the changes in pump power, pump current, and hemolysis index in the time-series correlation dataset. Its functional boundary is to identify the equipment status calculation result of abnormal pump operating resistance or load. The INR target interval is derived from the preset fixed safety interval or dynamic adjustment interval, and its function is to provide a clear boundary benchmark for data calculation. The pump power residual value (refer to Table 5) is obtained by subtracting the historical average pump power at the same pump speed from the current pump power. Its physical meaning is to characterize the degree of deviation of equipment operating parameters. The specific construction dimensions and discrimination logic of the feature vectors are shown in Table 2 below: Table 2 is the feature vector construction and discrimination logic mapping table.

[0021] Specifically, when the INR remains below the target range, a feature vector representing low anticoagulation data is generated; when the INR is above the target range and bleeding records are included, a feature vector representing high anticoagulation data is generated. Similarly, when pump power or current continuously increases at the same pump rate, and LDH or free hemoglobin levels are also elevated, a feature vector representing abnormal equipment operating parameters is generated. For example... Figure 3 As shown, the time-series correlation dataset performs transformation operations on the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector, respectively.

[0022] In one specific implementation scenario, patients mainly rely on warfarin for long-term anticoagulation after discharge. The system observed that the INR was below the target range for a long time, and at the same time, it captured the continuous increase of pump power residual and LDH. It generated the anticoagulation deviation range feature vector and the pump operating parameter residual feature vector, respectively.

[0023] S3. An abnormal concurrency correlation score is obtained from the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector, wherein the abnormal concurrency correlation score is obtained through the formula... Calculations show that This represents the score for the abnormal concurrency correlation. The weight representing the duration during which the INR is below the target interval. This represents the duration during which the INR is below the target range. Represents the weight of the pump power residual. Represents the pump power residual value. The weighting of changes in hemolysis indicators This represents the change in hemolysis markers. Specifically, the stated... The physical meaning of represents the calculation result of hemolysis indicators exceeding the baseline threshold. The calculation method is as follows: First, acquire hemolysis indicator monitoring data (including LDH values, free hemoglobin values, and haptoglobin values) from the time-series correlation dataset. Based on the clinical event record data in the time-series correlation dataset, remove falsely elevated data caused by hemolysis in blood samples, blood transfusion, or abnormal liver function to obtain purified hemolysis indicator data and corresponding baseline thresholds. Then, calculate the deviation of each indicator: calculate the difference between the current LDH value and the LDH baseline threshold to obtain the LDH deviation (recorded as 0 when below the baseline); calculate the difference between the current free hemoglobin value and the free hemoglobin baseline threshold to obtain the free hemoglobin deviation; calculate the difference between the haptoglobin baseline threshold and the current haptoglobin value to obtain the haptoglobin decrease. Finally, due to the different dimensions of each indicator, normalize the above deviations (e.g., convert them to multiples exceeding the normal upper limit) and perform a weighted summation. The calculation formula can be expressed as: ;in, These are the internal weighting coefficients for each hemolysis indicator in the calculation of its change value. This is a normalized mapping function. When the anticoagulation deviation interval feature vector indicates low anticoagulation data and the pump operating parameter residual feature vector indicates increased LDH and free hemoglobin, accompanied by a continuous increase in pump power or pump current residuals, the function calculated based on the above method... The numerical value will be substituted into the aforementioned abnormal concurrency correlation scoring formula, triggering an adjustment in the formula. The weight values ​​are determined to generate a higher score for abnormal concurrency correlation.

[0024] The abnormal concurrency correlation score is derived from the logical cross-calculation of the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector. Its dynamic feature is based on the time sequence of low anticoagulation exposure, followed by hemolysis and abnormal pump power. Its functional boundary is to output the concurrency correlation value; it does not directly adjust the anticoagulation dose or output a diagnostic conclusion. This step is essentially a weighted numerical calculation of multiple feature dimensions, producing only quantitative values ​​for subsequent use and not involving any clinical medical judgment logic. The non-diagnostic attribute of this data processing is reflected in the purely mathematical nature of the formula calculation; each variable only represents the degree of deviation of physical parameters and does not establish a direct correspondence with the diagnostic results of a specific disease. The parameter definition and weight adjustment logic of the abnormal concurrency correlation score are shown in Table 3 below: Table 3 shows the definition and weight adjustment of the scoring parameters for abnormal concurrency correlation.

[0025] Specifically, first determine if there is low anticoagulation exposure in the anticoagulation background, then look for evidence of hemolysis, then check for abnormal pump operation, then consider the time sequence, and finally rule out other causes such as elevated blood pressure, volume changes, infection, and hemolysis during blood collection. Substitute these factors into a formula to generate an abnormal complication correlation score. For example... Figure 4 As shown, the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector are used together as input parameters in the calculation process of the scoring formula.

[0026] In a specific implementation scenario, if the system determines that there is a low INR first, followed by an increase in LDH and free hemoglobin, accompanied by a continuous increase in pump power, then the values ​​of each parameter in the formula will increase, generating a higher abnormal concurrency correlation score; if the pump power is abnormal but the increase in blood pressure or pump speed adjustment can explain it, then a lower abnormal concurrency correlation score will be generated.

[0027] S4. Obtain a time-series data integration output file from the abnormal concurrency correlation score, wherein the time-series data integration output file is a structured data set derived from the abnormal concurrency correlation score, used for subsequent diagnostic procedures to read, and does not contain diagnostic conclusions itself.

[0028] The time-series data integration output file originates from the abnormal concurrency correlation score. Its functional boundary is to provide a structured intermediate data set to subsequent external diagnostic procedures or databases. It does not contain disease diagnostic conclusions, nor does it perform drug dosage adjustments or automatic diagnostic operations. The final output object of this invention is limited to a structured data file without diagnostic conclusions. This severs the direct causal link between the steps of the left ventricular assist device multi-source data time-series integration method and the diagnosis of diseases in living human beings, thus defining the left ventricular assist device multi-source data time-series integration method as a data processing process that provides intermediate data. The data encapsulation format and field definitions of the output file are shown in Table 4 below: Table 4 defines the fields for encapsulation in the time-series data integration output file.

[0029] Specifically, the system packages the generated abnormal concurrency correlation score and associated feature vectors into a structured data file in JSON or XML format for external medical information systems to read. For example... Figure 5 As shown, the final encapsulated output file not only contains the score value, but also embeds the associated feature identifiers.

[0030] In one specific implementation scenario, the system identifies a high concurrent correlation score for low anticoagulation data and abnormal pump power, and outputs a structured data set containing the score value and related timestamps to help external systems detect risk trends as early as possible.

[0031] Furthermore, the INR target range is derived from a preset fixed safety range or a dynamically adjusted range; the anticoagulation status monitoring data includes long-term warfarin anticoagulation data and unfractionated heparin bridging data; obtaining the anticoagulation deviation range feature vector from the time-series correlation dataset specifically includes: generating the anticoagulation deviation range feature vector based on the duration for which the INR in the long-term warfarin anticoagulation data in the time-series correlation dataset is lower than the INR target range, or based on the heparin infusion rate and anti-Xa results in the unfractionated heparin bridging data.

[0032] Among them, the INR target range serves as the baseline boundary for data calculation, and its fixed or dynamically adjusted source clearly defines the standard for difference calculation; long-term warfarin anticoagulation data refers to the daily warfarin dose and INR records for long-term anticoagulation maintenance after patient discharge; unfractionated heparin bridging data refers to the transitional heparin records during the early postoperative period before warfarin takes effect. The working principle of this additional feature is to determine the deviation status of anticoagulation data by accumulating low anticoagulation exposure time or identifying insufficient bridging.

[0033] Specifically, the system records the daily warfarin dose, whether a dose was missed or discontinued, the INR value, and the duration of INR below the target. It also records the heparin infusion rate and anti-Xa results. When the above data deviates from the set safe range, it triggers the generation of an anticoagulation deviation interval feature vector.

[0034] In one specific implementation scenario, warfarin did not take full effect in the early postoperative period, and the aPTT during the unfractionated heparin bridging period (see Table 5) did not meet the target. The system generates an anticoagulation deviation interval feature vector based on the data during this time period.

[0035] Furthermore, the pump power residual value is the difference obtained by subtracting the historical average pump power at the same pump speed from the current pump power; the step of obtaining the anticoagulation deviation interval feature vector from the time-series associated dataset also includes: based on the duration of INR being higher than the target INR interval in the time-series associated dataset, merging records of hemoglobin decrease, blood transfusion, or bleeding to generate the anticoagulation deviation interval feature vector.

[0036] Among them, the calculation logic of the pump power residual value difference clarifies the source of its value and ensures the clarity of the operation boundary; the anticoagulation deviation interval feature vector comes from the duration of INR higher than the target interval combined with clinical bleeding events, and its working principle is to identify the correlation between high anticoagulation exposure and bleeding complications.

[0037] Specifically, the system continuously monitors the duration of INR above the target range. When the system simultaneously receives records of decreased hemoglobin or blood transfusion, it generates a feature vector characterizing the elevated anticoagulation data.

[0038] In one specific implementation scenario, a patient's INR exceeds the standard and presents with gastrointestinal bleeding. The system integrates the INR trend and bleeding records to generate a feature vector of the anticoagulation deviation interval.

[0039] Furthermore, the abnormal concurrency correlation score obtained from the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector specifically includes: when the anticoagulation deviation interval feature vector indicates low anticoagulation data and the pump operating parameter residual feature vector indicates elevated LDH (refer to Table 5), elevated free hemoglobin (refer to Table 5), and accompanied by a continuous increase in pump power or pump current residual, by adjusting the formula... , and The weight values ​​are used to generate the abnormal concurrency correlation score.

[0040] The adjusted abnormal concurrency correlation score is derived from the temporal concurrency relationship between low anticoagulation exposure and hemolysis and abnormal pump parameters. Its working principle is to lock the data concurrency correlation based on the logical chain that first the anticoagulation data is low and then the hemolysis and pump load increase. This process only performs numerical adjustment of the weights and does not output diagnostic qualitative results.

[0041] Specifically, the system determines that LDH is significantly elevated from baseline or reaches 2-2.5 times or more of the upper limit of normal. Combined with elevated free hemoglobin and a sustained increase in pump power residual, this improves... , and The weight values ​​are used to generate a higher abnormal concurrency correlation score.

[0042] In one specific implementation scenario, a patient's INR remained below the target value for an extended period, subsequently resulting in tea-colored urine and elevated LDH levels. The controller recorded persistently abnormal pump power, and the system generated a high abnormal concurrency correlation score.

[0043] Furthermore, the step of obtaining the abnormal concurrency correlation score from the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector also includes: when the pump operating parameter residual feature vector indicates abnormal pump power but the change in hemolysis index does not reach the threshold, and the blood pressure, volume status, pump speed adjustment, outflow channel or inflow channel data in the time-series correlation dataset can explain the abnormal pump power, a baseline calibration identifier is generated and incorporated into the time-series data integration output file.

[0044] The baseline calibration identifier is derived from the interference of elevated blood pressure, changes in volume status, or mechanical problems of the equipment on pump parameters. Its working principle is to distinguish non-thrombotic abnormalities by excluding evidence of hemolysis and matching other explainable factors.

[0045] Specifically, the system detects abnormal pump power or flow rate, but LDH and free hemoglobin are normal. At the same time, it monitors the increase in mean arterial pressure or pump speed adjustment records and generates a baseline calibration indicator.

[0046] In one specific implementation scenario, a patient's blood pressure rises, causing a change in pump power. The system identifies the abnormal pump power but has no evidence of hemolysis, and the blood pressure data can explain the abnormality. It then generates a baseline calibration flag and incorporates it into the time-series data integration output file.

[0047] Furthermore, the hemolysis index monitoring data includes LDH values, free hemoglobin values, and haptoglobin values ​​(refer to Table 5); before obtaining the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector from the time-series correlation dataset, the method further includes: based on the time-series correlation dataset, removing falsely elevated data of the LDH values, free hemoglobin values, or haptoglobin values ​​caused by hemolysis of blood samples, blood transfusion, or abnormal liver function, to obtain purified hemolysis index data.

[0048] The purified hemolysis index data comes from the dataset after removing confounding and interfering terms from the original hemolysis index monitoring data. Its working principle is to eliminate non-thrombotic hemolysis interference to ensure the accuracy of subsequent data processing. This step is a typical data preprocessing operation, simply providing clean data input for subsequent calculations.

[0049] Specifically, the system identifies blood hemolysis or recent transfusion records, marks and removes abnormal changes in LDH and haptoglobin levels as false increases, and outputs purified hemolysis index data.

[0050] In one specific implementation scenario, a patient's abnormal liver function leads to a decrease in haptoglobin and an increase in indirect bilirubin. The system identifies the abnormal liver function record and removes related false positive data on hemolysis, thus obtaining purified hemolysis index data.

[0051] Furthermore, the clinical event record data includes infection and fever records and inflammatory marker records; before obtaining the abnormal concurrency correlation score from the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector, the method further includes: based on the infection and fever records and the inflammatory marker records in the time-series correlation dataset, identifying interference items caused by inflammatory state, such as increased D-dimer (refer to Table 5) or coagulation disorders, and performing weight downgrading processing on the interference items when calculating the abnormal concurrency correlation score.

[0052] The interference identification mechanism is derived from infection and fever records and inflammatory marker records. Its working principle is to reduce the weight of related indicators in the abnormal concurrency correlation score by identifying the interference of inflammatory status on coagulation indicators.

[0053] Specifically, the system acquires records of elevated CRP (see Table 5) or procalcitonin, identifies D-dimer elevation and coagulation disorder interference items, and performs weight downgrading on these interference items when calculating the abnormal concomitant correlation score.

[0054] In one specific implementation scenario, a patient develops sepsis, leading to a hypercoagulable state and elevated D-dimer levels. After the system identifies the infection and inflammation records, it reduces the weight of D-dimer in the data processing to avoid misjudgment.

[0055] Furthermore, the time-series data integration output file is also associated with hidden concurrent data feature identifiers; the process of obtaining the time-series data integration output file from the abnormal concurrency correlation score specifically includes: when the abnormal concurrency correlation score exceeds a preset threshold, generating the time-series data integration output file associated with data feature identifiers of acute kidney injury, stroke, or peripheral arterial embolism.

[0056] The hidden concurrent data feature identifiers are derived from acute kidney injury or thromboembolic events hidden when the abnormal concurrent correlation score exceeds a preset threshold. The system outputs complication data features concurrently with the structured dataset to prompt comprehensive clinical review. The output file is essentially a dataset and does not directly output a disease diagnosis conclusion. The hidden concurrent data feature identifiers include one or more data feature identifiers for complications such as acute kidney injury, stroke, or peripheral arterial embolism. Specifically, the acute kidney injury-related data feature identifier refers to the data feature identifier for the complication of acute kidney injury, and the stroke or peripheral arterial embolism data feature identifier refers to the data feature identifier for the complication of stroke or peripheral arterial embolism. When the abnormal concurrent correlation score calculated by the system exceeds the preset threshold, the system will synchronously retrieve and generate the hidden concurrent data feature identifier and concurrently output the data features of the aforementioned complication to prompt external clinical systems or doctors to conduct a comprehensive review (e.g., reminding doctors to perform organ function assessments). However, this identifier itself only serves as a data feature prompt and does not directly output a disease diagnosis conclusion. Specifically, when the system generates an abnormal concurrent correlation score that exceeds a preset threshold, it simultaneously retrieves the feature identifiers of acute kidney injury, stroke, or peripheral arterial embolism data and binds them to generate a composite time-series data integration output file.

[0057] In one specific implementation scenario, when a patient is diagnosed with an abnormal comorbidity score exceeding a preset threshold and experiences a large release of hemoglobin, the system generates a time-series data integration output file that indicates acute kidney injury and stroke, reminding the doctor to conduct organ function assessment.

[0058] Table 5

[0059] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for time-series integration of multi-source data from a left ventricular assist device, characterized in that, include: S1. Obtain anticoagulation status monitoring data, hemolysis index monitoring data, device operating parameter data, and clinical event record data recorded by the medical device system. Obtain a time-series associated dataset from the anticoagulation status monitoring data, the hemolysis index monitoring data, the device operating parameter data, and the clinical event record data. The time-series associated dataset is a logical set generated by integrating the anticoagulation status monitoring data, the hemolysis index monitoring data, the device operating parameter data, and the clinical event record data in chronological order. S2. Obtain the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector from the time-series correlation dataset. The anticoagulation deviation interval feature vector is the result of calculation representing the deviation of the data from the set interval, generated based on the difference and duration between the INR value and the INR target interval in the time-series correlation dataset. The pump operating parameter residual feature vector is the result of calculation representing the equipment operating parameters exceeding the baseline threshold, generated based on the changes in pump power, pump current, and hemolysis index in the time-series correlation dataset. S3. An abnormal concurrency correlation score is obtained from the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector, wherein the abnormal concurrency correlation score is obtained through the formula... Calculations show that This represents the score for the abnormal concurrency correlation. The weight representing the duration during which the INR is below the target interval. This represents the duration during which the INR is below the target range. Represents the weight of the pump power residual. Represents the pump power residual value. The weighting of changes in hemolysis indicators Represents changes in hemolysis markers; S4. Obtain a time-series data integration output file from the abnormal concurrency correlation score, wherein the time-series data integration output file is a structured data set derived from the abnormal concurrency correlation score, used for subsequent diagnostic procedures to read, and does not contain diagnostic conclusions itself. The INR target range is derived from a preset fixed safety range or a dynamically adjusted range; the anticoagulation status monitoring data includes long-term warfarin anticoagulation data and unfractionated heparin bridging data; obtaining the anticoagulation deviation range feature vector from the time-series correlation dataset specifically includes: generating the anticoagulation deviation range feature vector based on the duration for which the INR in the long-term warfarin anticoagulation data in the time-series correlation dataset is lower than the INR target range, or based on the heparin infusion rate and anti-Xa results in the unfractionated heparin bridging data; The pump power residual value is the difference obtained by subtracting the historical average pump power at the same pump speed from the current pump power; the step of obtaining the anticoagulation deviation interval feature vector from the time-series associated dataset further includes: based on the duration of INR being higher than the target INR interval in the time-series associated dataset, merging records of hemoglobin decrease, blood transfusion or bleeding, to generate the anticoagulation deviation interval feature vector.

2. The method for time-series integration of multi-source data from a left ventricular assist device according to claim 1, characterized in that, The abnormal concurrency correlation score obtained from the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector specifically includes: when the anticoagulation deviation interval feature vector indicates low anticoagulation data and the pump operating parameter residual feature vector indicates elevated LDH and free hemoglobin, accompanied by a continuous increase in pump power or pump current residuals, the score is obtained by adjusting the formula... and The weight values ​​are used to generate the abnormal concurrency correlation score.

3. The method for time-series integration of multi-source data from a left ventricular assist device according to claim 1, characterized in that, The method of obtaining the abnormal concurrency correlation score from the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector further includes: when the pump operating parameter residual feature vector indicates abnormal pump power but the change in hemolysis index does not reach the threshold, and the blood pressure, volume status, pump speed adjustment, outflow channel or inflow channel data in the time series correlation dataset can explain the abnormal pump power, a baseline calibration identifier is generated and incorporated into the time series data integration output file.

4. The method for time-series integration of multi-source data from a left ventricular assist device according to claim 1, characterized in that, The hemolysis index monitoring data includes LDH values, free hemoglobin values, and haptoglobin values. Before obtaining the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector from the time-series correlation dataset, the method further includes: based on the time-series correlation dataset, removing falsely elevated data of the LDH values, free hemoglobin values, or haptoglobin values ​​caused by hemolysis of blood samples, blood transfusion, or abnormal liver function, to obtain purified hemolysis index data.

5. The method for time-series integration of multi-source data from a left ventricular assist device according to claim 1, characterized in that, The clinical event record data includes infection and fever records and inflammatory marker records; before obtaining the abnormal concurrency correlation score from the anticoagulation deviation interval feature vector and the pump operating parameter residual feature vector, the method further includes: based on the infection and fever records and the inflammatory marker records in the time-series correlation dataset, identifying interference items caused by inflammatory state, such as D-dimer elevation or coagulation disorder, and performing weight downgrading processing on the interference items when calculating the abnormal concurrency correlation score.

6. The method for time-series integration of multi-source data from a left ventricular assist device according to claim 1, characterized in that, The time-series data integration output file is also associated with hidden concurrent data feature identifiers; the process of obtaining the time-series data integration output file from the abnormal concurrency correlation score specifically includes: when the abnormal concurrency correlation score exceeds a preset threshold, generating the time-series data integration output file associated with data feature identifiers of acute kidney injury, stroke, or peripheral arterial embolism.

7. A non-volatile storage medium, characterized in that, The non-volatile storage medium stores a computer program that, when executed by a processor, implements the multi-source data timing integration method for left ventricular assist devices as described in any one of claims 1 to 6.

Images