A modeling device and a modeling method for modeling an isolated heart perfusion simulating myocardial ischemia-reperfusion injury

Abstract

This invention provides a modeling device and method for simulating myocardial ischemia-reperfusion injury in an isolated heart, applicable to the field of data processing. This application aims to model ischemia-reperfusion injury in an isolated heart. First, it collects multi-source core data, including samples, perfusion data, environmental data, and physiological monitoring. Utilizing an isothermal perfusion and oxygen partial pressure regulation system, it completes cardiac rewarming, cannulation, and stable perfusion, simultaneously monitoring key indicators such as heart rate, coronary blood flow, and ventricular pressure. Ischemia and reperfusion parameters are integrated to form a basic perfusion dataset. The dataset is hierarchically structured according to physiological indicators and environmental parameters, constructing ternary association nodes for samples, indicators, and parameters. A dynamic physiological association graph is generated through temporal matching and aggregation. Core features are extracted through injury assessment and temporal prediction, generating high-dimensional vectors. These vectors are imported into an adaptive engine for iterative optimization, combined with deviation visualization to form a basic modeling model, outputting modeling information applicable to simulating ischemia-reperfusion injury in an isolated heart.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a modeling device and method for simulating myocardial ischemia-reperfusion injury in an isolated heart. Background Technology

[0002] Myocardial ischemia-reperfusion injury is a key fundamental model in cardiovascular disease research and drug screening. Ex vivo heart perfusion technology has become a primary method for conducting this type of basic research because it can eliminate neurohumoral interference and facilitates precise control of experimental conditions. Currently, most conventional ex vivo heart perfusion devices and modeling methods in the industry rely on manual operation to complete steps such as heart rewarming, aortic cannulation, and perfusion parameter adjustment. They also rely on manual interpretation of changes in physiological indicators such as heart rate, coronary blood flow, and ventricular pressure, which suffers from problems such as low standardization of perfusion procedures, poor consistency of physiological data acquisition, and ineffective fusion of multi-source data.

[0003] In terms of data processing, existing devices can only achieve simple acquisition and recording of raw physiological signals, and cannot perform hierarchical management and correlation analysis of perfusion parameters, environmental parameters, and physiological indicators, making it difficult to construct a complete dataset that is time-series unified and traceable. Although some devices have basic signal display functions, they lack feature extraction mechanisms for ischemia-reperfusion injury, and cannot quantitatively identify and standardize key injury features such as ischemic pressure, myocardial enzyme release, and ST segment changes on electrocardiogram.

[0004] In terms of modeling methods, traditional approaches rely on a single indicator to determine the degree of injury, failing to establish a correlation model between injury characteristics, perfusion parameters, and physiological responses. This results in a highly subjective modeling process with poor repeatability. Furthermore, existing systems lack adaptive iterative optimization capabilities and cannot automatically correct key parameters such as perfusion flow rate, oxygen concentration, and ischemia duration based on real-time physiological deviations, making it difficult to stably reproduce standardized ischemia-reperfusion injury models.

[0005] Furthermore, problems such as asynchronous multi-source data acquisition, time mismatch, and unclear stratification are common, leading to poor comparability of experimental results and insufficient model accuracy. This makes it difficult to meet the needs of basic research, drug evaluation, and damage mechanism analysis for standardized, quantifiable, and reproducible modeling. Existing technologies have significant limitations in integrated data processing, dynamic correlation analysis, adaptive modeling, and accurate damage simulation, which restricts the accuracy and reliability of isolated cardiac ischemia-reperfusion injury research.

[0006] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0007] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part by practice of the invention.

[0008] According to one aspect of this application, a modeling method for an isolated heart perfusion modeling device simulating myocardial ischemia-reperfusion injury is provided, comprising: acquiring multi-source core modeling data; completing rapid rewarming, cannulation adaptation, and baseline perfusion stabilization of the isolated heart based on an isothermal perfusion system and a dynamic oxygen partial pressure regulation module; identifying core physiological indicators including heart rate, coronary blood flow, and left ventricular end-diastolic pressure through a myocardial electrophysiological monitoring model; integrating key variables such as ischemia duration and reperfusion velocity into the modeling process to generate a basic perfusion dataset; layering the basic perfusion dataset according to a main table of core physiological indicators and a detailed table of environmental parameters; constructing a ternary association node including sample identifier, monitoring indicator, and regulation parameter in each layer; aggregating perfusion process and physiological response information through a temporal feature matching algorithm to generate dynamic physiological associations for ischemia-reperfusion injury modeling. Figure 1 shows the modeling of a dynamic physiological correlation graph based on a damage severity assessment engine and a time-series parameter prediction model. Core features, including ischemic pressure, reperfusion oxygen concentration, myocardial enzyme release, and ST segment changes on electrocardiogram, are extracted to generate a high-dimensional damage feature embedding vector. This vector is then imported into a perfusion parameter adaptive adjustment engine, and a multi-dimensional damage index dynamic matching mechanism is used for iterative optimization. Simultaneously, a correction decision chain is generated through parameter deviation visualization, forming a basic modeling model for ischemia-reperfusion injury in ex vivo hearts that integrates multi-source data. Based on this basic modeling model, accuracy verification and dynamic correction are performed using actual damage assessment data and dynamic adjustment data of perfusion parameters. A multi-index comprehensive evaluation algorithm is used to balance model stability, damage simulation similarity, and parameter adjustability requirements, generating modeling information for ischemia-reperfusion injury simulation in ex vivo hearts.

[0009] Another aspect of this application discloses a modeling device for isolated heart perfusion simulating myocardial ischemia-reperfusion injury, comprising: a multi-source modeling data acquisition module for acquiring multi-source core modeling data such as isolated heart samples, perfusion fluid components, temperature, oxygen partial pressure, flow rate, and physiological signals to form a comprehensive modeling basic dataset; a perfusion data processing module for achieving heart rewarming, cannulation, and stable perfusion based on isothermal perfusion and dynamic oxygen partial pressure regulation, monitoring core indicators such as heart rate, coronary blood flow, and left ventricular end-diastolic pressure, incorporating key variables such as ischemia duration and reperfusion flow rate, and generating a standardized perfusion basic dataset; and a physiological feature construction module for hierarchically stratifying the perfusion basic dataset according to a main table of physiological indicators and a detailed table of environmental parameters, constructing a ternary association node of sample identifier-monitoring indicator-regulation parameter, and aggregating information through temporal feature matching to generate a dataset for ischemia-reperfusion injury. The model consists of a dynamic physiological correlation graph; a damage feature extraction module, which models the dynamic physiological correlation graph based on the damage assessment engine and time-series parameter model, extracting core features such as ischemic pressure, reperfusion oxygen concentration, myocardial enzyme release, and ST segment changes on electrocardiogram, and generating a high-dimensional damage feature embedding vector; an adaptive modeling module, which imports the damage feature embedding vector into the perfusion parameter adaptive adjustment engine, dynamically matches and iteratively optimizes the multi-dimensional damage indicators, generates a correction decision chain through parameter deviation visualization, and forms a basic modeling model of ischemia-reperfusion injury of isolated heart that integrates multi-source data; and a modeling verification output module, which performs accuracy verification and dynamic correction based on the basic modeling model combined with measured damage data and perfusion parameter adjustment data, balancing model stability, simulation similarity, and parameter adaptability, and generating complete modeling information for ischemia-reperfusion injury simulation of isolated heart.

[0010] According to another aspect of this application, an electronic device is provided, on which a computer program is stored, which, when executed by a first processor, implements the modeling method of the above-described modeling device for simulating myocardial ischemia-reperfusion injury in an isolated heart.

[0011] According to another aspect of this application, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a second processor, implements the modeling method of the above-described modeling apparatus for simulating myocardial ischemia-reperfusion injury in an isolated heart.

[0012] The beneficial effects of this application are as follows: Standardized and stable perfusion of isolated hearts is achieved through multi-source data acquisition, isothermal perfusion, and dynamic control of oxygen partial pressure. Real-time monitoring and baseline calibration of core physiological indicators such as heart rate, coronary blood flow, and left ventricular end-diastolic pressure are performed, constructing a hierarchical structured perfusion dataset. By establishing ternary association nodes for sample identification, monitoring indicators, and control parameters, and combining this with a time-series matching algorithm to generate a dynamic physiological association graph, four core features—ischemic pressure, reperfusion oxygen concentration, myocardial enzyme release, and ST segment changes on electrocardiogram—are extracted using a damage assessment engine and a time-series prediction model, generating a high-dimensional damage feature embedding vector. Through iterative optimization using an adaptive perfusion parameter adjustment engine and deviation visualization correction, a dynamically calibrable basic modeling model is formed. Accuracy verification and closed-loop optimization are then performed using measured data, ultimately outputting reproducible and quantifiable ischemia-reperfusion injury modeling information, achieving the integration of perfusion control, physiological monitoring, data fusion, and injury modeling.

[0013] This application aims to standardize the perfusion process and data acquisition, reduce sample differences and human error, and improve experimental consistency and reproducibility. In addition, it constructs a multi-dimensional data association and dynamic physiological association graph to achieve full traceability of perfusion parameters and physiological responses. Through an adaptive engine and bias correction mechanism, it automatically adjusts perfusion parameters to improve the accuracy of injury simulation and model stability. It integrates data stratification, feature extraction, iterative optimization and verification correction to form a complete modeling system that meets the needs of standardized injury models for drug screening and mechanism research.

[0014] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0015] Figure 1 A flowchart illustrating a modeling method for an ex vivo heart perfusion modeling apparatus for simulating myocardial ischemia-reperfusion injury, according to an embodiment of this application, is shown. Figure 2 A schematic diagram of the structure of an ex vivo heart perfusion modeling device for simulating myocardial ischemia-reperfusion injury is shown in one embodiment of this application. Detailed Implementation

[0016] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0017] The following is combined Figure 1 To describe the modeling method of the modeling apparatus for simulating myocardial ischemia-reperfusion injury in an isolated heart according to an exemplary embodiment of this application: S101, acquire multi-source core modeling data.

[0018] In one implementation, the data from the ex vivo heart sample is used to provide a unique identifier and basic attribute information for modeling. This includes animal species, age, weight, heart sampling time, post-ex vivo pretreatment duration, and standardized aortic cannulation location. Healthy adult SD rats are selected as experimental animals, with an age controlled between 10 and 12 weeks and a weight controlled between 220 and 250 grams to ensure that individual differences are within a controllable range. Heart sampling is strictly controlled to be completed within five minutes after anesthesia to reduce the interference of ischemic pretreatment on experimental results. Immediately after extraction, the heart is immersed in KH perfusion fluid at 4 degrees Celsius for a pretreatment time of 10 minutes to maintain myocardial viability. The aortic cannulation location is uniformly set at 2 millimeters from the valve orifice at the aortic root to ensure consistency with the fixed position. These standardized settings provide comparable basic conditions between different samples, offering a unified sample basis for subsequent data processing.

[0019] The perfusion fluid formulation data was used to provide a stable in vitro physiological environment for isolated hearts, which is crucial for maintaining cardiac activity and simulating normal physiological states. The data obtained included solute types, solute concentrations, solvent ratios, perfusion fluid pH, dissolved oxygen concentration, and isothermal control temperature. KH perfusion fluid was used as the base formulation, with specific components including sodium chloride (118 mmol / L), potassium chloride (4.7 mmol / L), calcium chloride (2.5 mmol / L), potassium dihydrogen phosphate (1.2 mmol / L), magnesium sulfate (1.2 mmol / L), sodium bicarbonate (25 mmol / L), and glucose (11 mmol / L). Ultrapure water was used as the solvent, and the perfusion fluid pH was maintained between 7.35 and 7.45. Before use, a mixture of 95% oxygen and 5% carbon dioxide was bubbled into the perfusion fluid for 30 minutes to saturate the dissolved oxygen. Throughout the process, the perfusion fluid temperature was maintained at 37 degrees Celsius using a thermostatic circulation device, consistent with the physiological body temperature of rats.

[0020] The physiological monitoring equipment parameters are used to standardize data acquisition, ensuring the accuracy and comparability of indicators such as heart rate, coronary blood flow, and left ventricular end-diastolic pressure. Acquired data includes the device sampling frequency, signal gain factor, baseline calibration value, filter threshold, and monitoring channel number. The ECG signal sampling frequency is set to 1 kHz, and the signal gain factor is set to 500. The left ventricular pressure signal sampling frequency is set to 1 kHz, and the gain factor is set to 200. The coronary blood flow monitoring module performs baseline zeroing calibration before the experiment to ensure accurate flow measurement starting point. Both ECG and pressure signals use bandpass filtering from 0.5 Hz to 100 Hz to remove noise interference. The first monitoring channel is assigned to heart rate, the second to coronary blood flow, and the third to left ventricular end-diastolic pressure, enabling simultaneous and independent acquisition of multiple indicators and avoiding signal crosstalk.

[0021] The ischemia-reperfusion timing protocol standardizes the injury simulation process, ensuring model reproducibility and comparability. Data acquired includes the equilibrium perfusion duration, the duration of ischemia-cessation perfusion, the reperfusion recovery flow rate, the total reperfusion duration, and the oxygen concentration switching point. The equilibrium perfusion duration is set to 20 minutes to allow the isolated heart to recover normal beating and contractile function under stable perfusion conditions. The ischemia duration after perfusion cessation is set to 30 minutes to simulate acute myocardial ischemia. During the reperfusion phase, the initial perfusion flow rate is restored to 10 ml / min, and the total reperfusion monitoring duration is set to 60 minutes to fully record the injury and recovery process. The oxygen concentration switching point is set at the onset of ischemia; oxygenation is stopped during the ischemia phase, and at the start of reperfusion, a mixture of 95% oxygen and 5% carbon dioxide is introduced to achieve precise switching between hypoxic and normoxic states.

[0022] Data from isolated heart samples, perfusion fluid formulations, physiological monitoring equipment parameters, and ischemia-reperfusion time-series protocols were uniformly collected and aligned temporally using a global timeline. Each rat's unique identifier was used as the sole sample identifier, linking sample attributes, perfusion parameters, monitoring settings, and time-series procedures to the same identifier, forming a structured and standardized dataset for global modeling. This dataset provides complete input for subsequent perfusion baseline dataset generation, dynamic physiological correlation graph construction, injury feature extraction, and basic model training, ensuring consistent, traceable, and reproducible data throughout the entire process.

[0023] S102, based on a constant temperature perfusion system and a dynamic oxygen partial pressure regulation module, completes rapid rewarming, cannulation adaptation, and basic perfusion stabilization of isolated hearts. It identifies core physiological indicators, including heart rate, coronary blood flow, and left ventricular end-diastolic pressure, through a myocardial electrophysiological monitoring model. It integrates key variables such as ischemia duration and reperfusion rate into the modeling process to generate a basic perfusion dataset.

[0024] In one implementation, based on multi-source physiological and environmental data such as isolated heart samples, perfusion fluid components, oxygen partial pressure, temperature, and flow rate, processing is carried out according to three rules: homeostasis maintenance, indicator monitoring, and damage induction. The homeostasis maintenance rule ensures stable physiological activity of the isolated heart throughout the experiment; the indicator monitoring rule standardizes the acquisition and analysis of key signals such as heart rate, coronary blood flow, and left ventricular end-diastolic pressure; and the damage induction rule precisely switches between ischemia and reperfusion states according to a preset time sequence. A real-time bidirectional communication connection is established between the isothermal perfusion unit and the oxygen partial pressure control unit. The isothermal perfusion unit outputs a perfusion fluid with a stable temperature and maintains a constant flow rate according to set parameters. The oxygen partial pressure control unit adjusts the oxygen and carbon dioxide ratio in the perfusion fluid in real time according to the needs of each experimental stage. The two units collaboratively confirm operating parameters such as perfusion temperature, flow rate, and oxygen concentration, and simultaneously determine data acquisition strategies such as physiological signal acquisition frequency, data storage interval, and anomaly detection threshold. Raw signals such as electrocardiogram (ECG), ventricular pressure, and coronary blood flow were continuously acquired through a real-time physiological signal acquisition interface. A global time-series synchronization mechanism was used to unify all acquired data to the same time reference. Heart rate, coronary blood flow, and left ventricular end-diastolic pressure were identified as the three core monitoring dimensions, and the filtering frequency band, amplification factor, and recording format of the raw signals were set. For rat heart samples, the perfusion temperature was stably controlled at 37 degrees Celsius, the initial oxygen partial pressure was set to the saturation concentration of a mixture of 95% oxygen and 5% carbon dioxide by volume, and the baseline perfusion flow rate was set to 10 ml per minute. All monitoring data were aligned using a unified timestamp to ensure that the perfusion parameter control and physiological indicator monitoring were completely synchronized in time, providing a time-consistent and parameter-stable foundation of data for subsequent data processing and model construction.

[0025] A four-stage linkage mechanism—rewarming, cannulation, perfusion, and stabilization—is employed to integrate the processing of multi-source baseline data. Each stage is executed sequentially, with the completion of one stage automatically triggering the start of the next, achieving uninterrupted operation throughout the entire process. The first stage involves rapid rewarming of the isolated heart using a constant-temperature perfusion system. A pre-treated heart specimen at 4°C is placed in the perfusion device, and the temperature is gradually increased at a rate of 1°C per minute until it reaches the physiological temperature of 37°C, allowing the heart tissue to quickly regain its physiological activity. The second stage involves aortic cannulation adaptation, precisely inserting the perfusion cannula into the aortic root two millimeters from the valve orifice. After fixation, the tubing is sealed, establishing a complete perfusion fluid delivery pathway to ensure no leakage or backflow. The third stage initiates steady-state perfusion. The constant-temperature perfusion unit continuously delivers 37°C saturated oxygen perfusion fluid at a fixed flow rate of 10 ml per minute, maintaining a flow rate fluctuation of no more than 5% throughout the process to maintain a stable perfusion state. The fourth stage involves monitoring cardiac status, continuously observing the amplitude and rhythm of heartbeats to confirm that ventricular contraction and coronary blood supply have entered a stable range. The physiological monitoring module establishes real-time data interaction with the isothermal perfusion system, simultaneously collecting and analyzing three core indicators—heart rate, coronary blood flow, and left ventricular end-diastolic pressure—during the stabilization phase. The core criterion for entering a stable state is that all three indicators reach their stable range, thus outputting standardized raw perfusion data. During the rewarming phase, the warming rate is strictly controlled to avoid sudden temperature increases that could damage the myocardium. During cannulation, the cannulation depth and position are standardized to ensure consistent pathway resistance across different samples. During perfusion, the flow rate and temperature are kept constant to eliminate environmental interference. During the stabilization phase, continuous observation for ten minutes is performed. When the heart rate remains between 240 and 320 beats per minute, the coronary blood flow stabilizes between 8 and 12 ml per minute, and the left ventricular end-diastolic pressure waveform is intact with no drastic amplitude fluctuations, the heart is considered to have entered a stable state. All data collected during this period is saved as standardized raw perfusion data for subsequent processing and analysis.

[0026] Real-time baseline verification was performed on standardized raw perfusion data, sequentially executing a three-layer processing logic of baseline calibration, signal compensation, and state determination to optimize data and screen for anomalies layer by layer. The baseline calibration stage used the no-load signal acquired before the experiment as a benchmark, performing zero-point correction on all monitoring data to uniformly calibrate the baseline value of coronary blood flow to a fixed reference value of 10 ml / min, eliminating systematic errors caused by sensor drift or differences in tubing resistance. The signal compensation stage employed a fixed-parameter filtering compensation mechanism to address electromagnetic interference and mechanical vibration noise introduced during acquisition. Bandpass filtering of 0.5 Hz to 100 Hz was used for ECG and ventricular pressure signals to filter out high-frequency noise and low-frequency baseline drift, improving signal smoothness and numerical reliability. The state determination stage judged real-time data according to preset physiological intervals. Abnormal physiological states were identified when the heart rate exceeded the normal range of 240 to 320 beats / min, the coronary blood flow decreased by more than 50% within one minute, or the left ventricular end-diastolic pressure remained above 15 mmHg or below 5 mmHg. When the above conditions occur, the corresponding data segment is immediately marked, invalid data recording is paused, and an abnormal status alert is sent to the control system until all cardiac indicators return to reasonable ranges before data acquisition resumes. To address individual differences in activity and contractile function among different ex vivo heart samples, the system automatically initiates adaptive parameter fine-tuning, slightly adjusting perfusion flow rate and oxygen concentration to quickly bring heart rate, coronary blood flow, and ventricular pressure back to stable ranges. After three layers of processing and optimization correction, complete optimized perfusion data is finally generated, including baseline calibration results, filter compensation parameters, and abnormal status determination records, providing a high-quality, highly reliable data source for subsequent data integration and modeling.

[0027] Optimized perfusion data were recorded and integrated in a closed loop. A synchronous acquisition link was used to bind and correlate three types of information throughout the entire process: perfusion control parameters, physiological monitoring indicators, and system operating status. The system synchronously reads time-series changes in ischemia duration, reperfusion rate, heart rate, coronary blood flow, and left ventricular end-diastolic pressure, as well as the operating status information of the isothermal perfusion unit, oxygen partial pressure regulation unit, and physiological monitoring module. All data are aligned using a unified global timeline to ensure a one-to-one correspondence between parameters and indicators at the same moment. Using the unique ID of each rat as the primary key for sample identification, all data corresponding to the same sample are linked and stored, constructing a complete, time-consistent, traceable, and reusable basic perfusion dataset. The dataset comprehensively records key process information, including 20 minutes of balanced perfusion, 30 minutes of ischemia-induced cessation of perfusion, and a reperfusion flow rate of 10 ml / min. It also saves the entire process's heart rate, coronary blood flow, and left ventricular end-diastolic pressure curves, as well as the adjustment records of oxygen partial pressure during the ischemia-induced cessation of oxygen supply and the restoration of a 95% oxygen / 5% carbon dioxide mixture during the reperfusion phase. All information is integrated into a unified dataset, ensuring that each physiological indicator value accurately corresponds to specific perfusion control parameters and time points. This provides standardized, high-quality input support for subsequent dynamic physiological correlation map construction, injury feature extraction, and basic model training.

[0028] S103, the perfusion basic dataset is layered according to the dimensions of the main table of core physiological indicators and the detailed table of environmental parameters. Each layer constructs a ternary association node including sample identifier, monitoring indicator and regulation parameter. The perfusion process and physiological response information are aggregated by the time series feature matching algorithm to generate a dynamic physiological association graph for ischemia-reperfusion injury modeling.

[0029] In one implementation, based on the requirements of completeness, temporal consistency, traceability, and feature extraction for ischemia-reperfusion modeling of the perfusion baseline dataset, a systematic dimensional hierarchical and structured analysis of the dataset is conducted. All data is divided into two independent but related data levels: a main table of core physiological indicators and a detailed table of environmental parameters. The field composition, primary key generation rules, and storage structure constraints of the two data levels are defined respectively to ensure unified data format, standardized calling, and traceability and reproducibility.

[0030] The core physiological indicators table stores key physiological indicators that directly reflect the real-time working state of the myocardium and is the core basis for assessing cardiac activity and the degree of damage. The table includes specific fields such as real-time heart rate, real-time coronary blood flow, left ventricular end-diastolic pressure, and data acquisition timestamp. The normal range for heart rate is 240 to 320 beats per minute, the stable range for coronary blood flow is 8 to 12 ml per minute, and the normal range for left ventricular end-diastolic pressure is 5 to 15 mmHg. All indicators are marked with timestamps with one millisecond precision.

[0031] The Environmental Parameter Details table stores environmental and operational parameters related to perfusion control, providing standardized condition records for injury simulation. Specific fields in this table include real-time perfusion temperature, real-time oxygen partial pressure concentration, real-time perfusion flow rate, ischemia duration, and reperfusion set flow rate. The perfusion temperature is fixed at 37 degrees Celsius, the oxygen partial pressure concentration is 95% oxygen and 5% carbon dioxide by volume during equilibrium perfusion and reperfusion phases, and zero oxygen concentration during the ischemia phase. Both the baseline perfusion flow rate and reperfusion flow rate are set to 10 ml / min, and the ischemia duration is uniformly set to 30 minutes.

[0032] The sample identifier serves as the unified primary key for the entire dataset. The sample identifier is generated using a unique combination of the experimental animal's ID and the experimental date, for example, Rat20260509001. This unified primary key establishes a fixed association between the main table of core physiological indicators and the detailed table of environmental parameters, ensuring that physiological indicators and corresponding environmental parameters are accurately matched whenever data is queried. All data is stored and retrieved using the sample identifier as a unique index, guaranteeing a one-to-one correspondence between physiological indicators and perfusion control parameters for the same sample throughout the entire process of balanced perfusion, ischemia, and reperfusion, without misalignment, omissions, or duplications. This provides a reliable data structure foundation for the subsequent construction of ternary association nodes and the generation of dynamic physiological association graphs.

[0033] Based on the requirements of hierarchical architecture and topological association construction, ternary association nodes for sample identifiers, monitoring indicators, and regulatory parameters are established in the main table of core physiological indicators and the detailed table of environmental parameters, respectively. Each node has an independent index, a clear data source, and fixed association logic. The sample identifier serves as the unique index of the node, generated by combining the experiment number and date, to distinguish data from different experimental subjects and ensure clear attribution of each data set. The monitoring indicators correspond to real-time physiological data such as heart rate, coronary blood flow, and left ventricular end-diastolic pressure in the main table of core physiological indicators, and environmental data such as perfusion temperature, oxygen partial pressure, and real-time flow rate in the detailed table of environmental parameters. The regulatory parameters correspond to control parameters that are pre-set or dynamically adjusted during the experiment, including baseline perfusion flow rate, ischemia duration, reperfusion flow rate, and oxygen concentration setpoint.

[0034] A fixed mapping relationship and data matching logic are established between the three nodes. Sample identifiers, monitoring indicators, and control parameters under the same timestamp are automatically bound, and cross-layer data achieves bidirectional interaction through sample identifiers and timestamps. Any physiological data in the main table of core physiological indicators can find the corresponding perfusion control parameter in the environmental parameter details table through sample identifiers and timestamps, making the data have the characteristics of being associative, queryable, and traceable.

[0035] In the main table of core physiological indicators, heart rate monitoring indicators are directly linked to baseline perfusion flow rate control parameters using sample identifiers as indexes. For example, the heart rate value corresponding to sample identifier Rat20260509001 is 280 beats per minute, forming a node association with a baseline perfusion flow rate of 10 ml per minute, with corresponding timestamps recorded synchronously. In the environmental parameter details table, oxygen partial pressure monitoring indicators are linked to reperfusion oxygen concentration control parameters using the same sample identifier as an index. For example, the oxygen partial pressure monitoring value for the same sample number is 95% (volume fraction), consistent with the oxygen concentration control value during reperfusion. All nodes are precisely matched through a unified global timestamp, ensuring complete temporal synchronization of physiological indicators and perfusion parameters without deviation or misalignment, ultimately forming a data standardization organization model with clear hierarchical structure, associative nodes, and temporally aligned data.

[0036] Combining the requirements of continuous myocardial physiological response, controllable perfusion process, and identifiable injury characteristics, this study uses a globally unified time series as the axis and sets three constraints—time series matching window, feature similarity, and topological connectivity—for each of the three experimental stages: equilibrium perfusion, ischemia, and reperfusion. This achieves precise aggregation and association of data at each stage. The time series matching window is set according to the duration and physiological rate of change of different perfusion stages to limit the effective time range of data association and avoid cross-stage data interference. The equilibrium perfusion stage lasts 20 minutes, and a 60-second time series matching window is set to match steady-state physiological characteristics. The ischemia stage lasts 30 minutes, during which myocardial injury changes rapidly and significantly; a 30-second time series matching window is set to improve the accuracy of injury feature capture. The reperfusion stage lasts 60 minutes, during which the injury recovery process is relatively gradual; a 45-second time series matching window is set.

[0037] Feature similarity thresholds are used to determine the reliability of the association between monitored indicators and regulatory parameters within the same stage. In the equilibrium perfusion stage, aiming to maintain steady state, the feature similarity threshold is set at 90%. In the ischemia stage, focusing on the strong correlation of injury characteristics, the feature similarity threshold between myocardial enzyme release and ischemic pressure is set at 85%. In the reperfusion stage, based on the consistency of recovery trends, the feature similarity threshold is set at 88%. Topological connectivity thresholds are used to ensure the integrity of the generated dynamic association graph structure and the coherence of node connections, avoiding broken connections or isolated nodes. A unified topological connectivity threshold of 90% is set for all three stages to ensure that physiological indicators at any given time can form effective connections with their corresponding perfusion parameters.

[0038] Based on the three constraints mentioned above, the system automatically determines the matching method for ternary nodes, the multi-dimensional feature aggregation process, and the overall topology generation rules. During the ischemic phase, the data is segmented in 30-second increments, and features such as ventricular pressure, oxygen partial pressure, myocardial enzyme release, and electrocardiogram signals are aggregated and calculated within each window. When the feature similarity within a window reaches 85% or more and the topological connectivity reaches 90% or more, the node is deemed valid and the association aggregation is completed. According to this rule, the continuously changing physiological responses during the ischemic phase can be completely correlated with the perfusion parameter regulation process, ensuring accurate and reliable integrated data and providing a standardized execution basis for the subsequent generation of dynamic physiological correlation graphs.

[0039] A comprehensive integration process was implemented, encompassing the hierarchical data structure, the construction method of ternary association nodes, and the temporal feature matching algorithm. This involved unifying and streamlining the core physiological indicator master table and environmental parameter detail table, the rule-constructed ternary association nodes for sample identifiers, monitoring indicators, and control parameters, and valid data validated through temporal matching windows, feature similarity, and topological connectivity constraints. The integration process used a globally unified timestamp as a benchmark, aligning perfusion control parameters, physiological monitoring indicators, temporal change data, and experimental stage status information item by item to achieve end-to-end association of all data elements. Information aggregation was performed using the temporal feature matching algorithm, sequentially connecting all valid ternary nodes across the three stages: the equilibrium perfusion period, the ischemia period, and the reperfusion period. Data aggregation was completed with a 60-second temporal matching window during the equilibrium perfusion period, a 30-second time-resolution aggregation during the ischemia period, and a 45-second trend aggregation during the reperfusion period. Only data nodes meeting both feature similarity and topological connectivity criteria were retained within each stage to ensure the authenticity and reliability of the associations.

[0040] Once aggregated, a complete dynamic correlation is formed, which can intuitively reflect the physiological response patterns of the myocardium during the entire ischemia-reperfusion process and the corresponding relationship with the regulation of perfusion parameters. For example, during the ischemic phase, the decrease in ventricular pressure, the drop in oxygen partial pressure to zero, the increase in myocardial enzymes, and the ST segment shift can all be correlated one-to-one with the regulation parameters for stopping perfusion and shutting off oxygen supply. During the reperfusion phase, changes such as the recovery of flow rate, the increase in oxygen concentration, and the gradual recovery of myocardial indicators can also be fully traced.

[0041] The final result is a dynamic physiological correlation graph for ischemia-reperfusion injury modeling. This graph, with time on the horizontal axis and physiological indicators and perfusion parameters on the vertical axis, comprehensively displays the interconnected changes throughout the entire process, from the equilibrium perfusion steady state to the occurrence of ischemic injury and then to reperfusion recovery. This dynamic physiological correlation graph can be directly input into the injury severity assessment engine and the time-series parameter prediction model, providing standardized, traceable, and reproducible input data for subsequent core feature extraction and high-dimensional injury feature embedding vector construction.

[0042] S104 models the dynamic physiological correlation graph based on the damage severity assessment engine and the time-series parameter prediction model, extracts core features including ischemic pressure, reperfusion oxygen concentration, myocardial enzyme release, and ST segment changes in electrocardiogram, and generates a high-dimensional damage feature embedding vector.

[0043] In one implementation, the damage assessment engine and the time-series parameter prediction model are first loaded, and the previously generated dynamic physiological correlation graph is read and parsed as the sole input data. The dynamic physiological correlation graph uses the sample identifier as a unique index and fully contains ternary correlation information consisting of the sample identifier, monitoring indicators, and regulatory parameters. It also covers time-series perfusion data and myocardial physiological response data throughout the entire process of balanced perfusion, ischemia, and reperfusion, providing a unified and traceable input foundation for joint modeling. The damage assessment engine primarily handles the quantitative determination, grading, and reliability verification of damage-related indicators, while the time-series parameter prediction model is used to fit, extrapolate, and extract patterns from the time-series changes throughout the ischemia-reperfusion process. The two models establish real-time data-level linkage; the damage assessment results output by the damage assessment engine are transmitted to the time-series parameter prediction model in real time, forming a collaborative computing and time-synchronized joint modeling architecture. The dynamic physiological correlation graph, containing complete stage information for 20 minutes of balanced perfusion, 30 minutes of ischemia, and 60 minutes of reperfusion, is imported into the model system. The engine and model start running synchronously, achieving parallel execution and collaborative output of damage assessment and time-series prediction.

[0044] Joint modeling operations are performed based on a dynamic physiological correlation graph. The damage assessment engine traverses the physiological indicators in the correlation graph point by point, judges the damage correlation of each data point, and screens out key data dimensions directly related to myocardial ischemia-reperfusion injury. A time-series parameter prediction model performs curve fitting on the entire time-series changes, extracting the time-series change patterns of the three stages of injury occurrence, injury development, and injury recovery. The two models collaboratively locate and lock in four core features: ischemic pressure, reperfusion oxygen concentration, myocardial enzyme release, and ST segment changes on electrocardiogram. Ischemia pressure reflects the myocardial mechanical load state caused by the decrease in ventricular pressure during the ischemic stage; an example value is the lowest pressure during the ischemic period dropping to 5 mmHg to 8 mmHg. Reperfusion oxygen concentration reflects the degree of oxygen supply recovery during the reperfusion stage; an example value is the recovery from zero oxygen concentration during the ischemic stage to 95% by volume fraction. Myocardial enzyme release reflects the degree of myocardial cell damage; an example value is an increase to more than three times the baseline value ten minutes after reperfusion. ST segment changes on an electrocardiogram (ECG) are used to reflect myocardial electrophysiological damage. Example values ​​are ST segment deviations of 0.1 to 0.3 mV lasting for more than 30 seconds. During the ischemic phase, the damage assessment engine identifies ischemic pressure characteristics resulting from a rapid drop in ventricular pressure, and the time-series parameter prediction model simultaneously fits the pressure's decreasing trend over time. During the reperfusion phase, the engine identifies the oxygen partial pressure recovery value, the model predicts the oxygen concentration recovery rate, and simultaneously locks onto the magnitude of myocardial enzyme elevation and the morphology of ST segment deviation.

[0045] The four core features obtained from the co-localization were sequentially extracted and standardized dimensionally to unify the dimensions and numerical ranges. Ischemia pressure features were extracted from continuous ventricular pressure monitoring data, specifically the lowest pressure value and the amplitude of fluctuation throughout the ischemic period. Reperfusion oxygen concentration features were extracted from oxygen partial pressure monitoring data, specifically the initial oxygen concentration and the oxygen concentration after reperfusion stabilization. Myocardial enzyme release features were extracted from perfusion fluid detection data, specifically the increase in enzyme concentration per minute. ST segment change features were extracted from ECG signal data, specifically the ST segment offset amplitude and duration. All extracted feature values ​​were normalized to eliminate calculation bias caused by unit differences. Ischemia pressure values ​​were normalized to the zero-to-one range, reperfusion oxygen concentration was standardized according to its proportion to saturated oxygen concentration, myocardial enzyme release was ratio-based with reference to the baseline value during the equilibrium perfusion period, and ST segment offset was uniformly quantified to the zero-to-one range in millivolts.

[0046] Four standardized features—ischemic pressure, reperfusion oxygen concentration, myocardial enzyme release, and ST segment shift on electrocardiogram—are combined in a fixed dimensional order to construct a four-dimensional high-dimensional injury feature embedding vector. The vector dimensions are ordered as follows: ischemic pressure, reperfusion oxygen concentration, myocardial enzyme release, and ST segment shift on electrocardiogram. Each dimension maintains a strict temporal correspondence, comprehensively characterizing the degree of injury, its trend, and temporal patterns from ischemia to reperfusion recovery in a single experiment. This vector retains all output information from the joint modeling, with no data loss, and can be directly imported into the adaptive perfusion parameter adjustment engine. Example vectors can be represented as 0.8, 0.95, 2.3, and 0.2, corresponding to the standardized ischemic pressure, reperfusion oxygen concentration, myocardial enzyme release, and ST segment shift, respectively.

[0047] The constructed high-dimensional damage feature embedding vector is output and saved, serving as standard input data for subsequent adaptive adjustment, iterative optimization, and correction decision chain generation of perfusion parameters. This completes the entire joint modeling process of the damage severity assessment engine and the time-series parameter prediction model, achieving a complete transformation from dynamic physiological correlation diagrams to standardized high-dimensional damage feature embedding vectors. This provides crucial data support and feature input for the subsequent construction of basic modeling models for ischemia-reperfusion injury in ex vivo hearts.

[0048] S105 embeds the damage features into a vector and imports it into the perfusion parameter adaptive adjustment engine. It uses a multi-dimensional damage index dynamic matching mechanism for iterative optimization and generates a correction decision chain through parameter deviation visualization, forming a basic modeling model of ischemia-reperfusion injury in ex vivo heart that integrates multi-source data.

[0049] In one implementation, based on the physiological characteristics, parameter control range, injury response patterns, and multi-source data fusion requirements of isolated cardiac ischemia-reperfusion injury simulation, the generated injury feature embedding vector, preset perfusion control parameters, and temporal iteration rules are individually adapted, analyzed, and feature extracted to form four types of core input features for constructing the basic model. First, injury characterization feature extraction is performed. Four standardized values—ischemic pressure, reperfusion oxygen concentration, myocardial enzyme release, and ST segment shift on electrocardiogram—are directly read from the high-dimensional injury feature embedding vector and transformed into injury characterization features that intuitively reflect the degree and trend of myocardial injury. For example, ischemic pressure of 0.7, myocardial enzyme release ratio of 1.9, and ST segment shift amplitude of 0.2 are extracted from the vector. These values ​​collectively characterize the injury level and development status of the current sample, serving as a direct basis for the model to judge the degree of injury.

[0050] Next, parameter adaptation features are generated, transforming the adjustable physical parameter ranges of the irrigation system into parameter constraint features recognizable by the model. The adjustable range for irrigation temperature is set to 36°C to 38°C, with a default stable point of 37°C; the oxygen concentration control range is set to 0% to 95% (volume fraction); and the adjustable range for irrigation flow rate is set to 8 ml / min to 12 ml / min, with a standard flow rate of 10 ml / min. These ranges, together with the default values, constitute the parameter adaptation features, ensuring that the model does not exceed the safe and effective operating range of the physical equipment during iteration.

[0051] Next, iterative optimization features are extracted, transforming the model's internal computational rules into executable iterative optimization features. The maximum number of iterations is set to fifty, and the convergence threshold is set to 0.5%. That is, when the output change is less than this threshold for three consecutive iterations, the model is considered to have reached convergence and the iteration stops. Simultaneously, the computational step size for each iteration is set to 0.1 to ensure smooth and rapid convergence during the iteration process, avoiding oscillations or non-convergence.

[0052] Finally, data fusion features were extracted, and the correlation logic between sample data, physiological monitoring data, environmental parameter data, and perfusion regulation data was extracted as data fusion features. Using sample identifiers as a unified primary key and global timestamps as a synchronization benchmark, a one-to-one correspondence between physiological indicators and perfusion parameters was established, stipulating that all data at the same time point must be paired and cannot be accessed individually. This rule ensures that the model will not experience temporal misalignment or parameter mismatch when accessing data, guaranteeing the accuracy and consistency of multi-source data fusion. After the above analysis and extraction, the four types of information—damage characterization features, parameter adaptation features, iterative optimization features, and data fusion features—together constitute the complete core input for building the basic modeling model, providing a stable, standardized, and quantifiable data foundation for subsequent state constraint setting, engine iterative optimization, and decision chain generation.

[0053] Five dynamic change factors—ischemic pressure, reperfusion oxygen concentration, myocardial enzyme release, ST segment shift, and perfusion velocity—were logically transformed and thresholded one by one. This transformed continuously changing physical signals into quantitative constraint parameters that the model could directly recognize, ultimately forming five sets of state adaptation constraint information for the modeling process. First, the parameters for determining the degree of injury were transformed. The numerical ranges of ischemic pressure, myocardial enzyme release, and ST segment shift were divided into three levels of injury criteria: mild, moderate, and severe. A myocardial enzyme release exceeding 2.5 times the baseline value was defined as severe injury; ischemic pressure below 5 mmHg for more than 15 seconds was defined as severe ischemia; and an ST segment shift amplitude exceeding 0.2 mV for more than 30 seconds was defined as severe electrophysiological injury. These thresholds together constitute the injury degree determination parameters, used by the model to identify the injury level in real time.

[0054] Secondly, deviation calibration parameters are generated. The difference between real-time monitored indicators and expected target values ​​is converted into deviation calibration criteria. The allowable deviation range for ST segment offset is set to ±0.05 mV; deviation calibration is initiated immediately if this range is exceeded. Automatic calibration is triggered when the coronary flow rate deviates from the set value by more than 1 ml / min, ensuring that physiological indicators remain consistent with the target state. Thirdly, control step size limits are constructed. The single adjustment amplitude of perfusion flow rate and oxygen concentration is converted into a safe control step size. The single adjustment amplitude of perfusion flow rate is set not to exceed ±1 ml / min, and the single adjustment amplitude of oxygen concentration is set not to exceed ±10%, avoiding myocardial stress injury caused by parameter mutations and forming a stable and controllable controllability constraint.

[0055] Fourth, establish time-series matching constraints. All data are strictly aligned using a globally unified timestamp, stipulating that physiological indicators, perfusion parameters, and stage states must correspond and match at the same time point, allowing no time misalignment. Data must be updated synchronously at the switching points between the balancing perfusion, ischemia, and reperfusion stages to ensure accurate time-series relationships. Fifth, establish physiological stability protection parameters. Safe operating ranges for heart rate, coronary blood flow, and left ventricular end-diastolic pressure are set as hard constraints. The heart rate is set to be maintained at 240–320 beats / min, coronary blood flow at 8–12 ml / min, and left ventricular end-diastolic pressure at 5–15 mmHg. Exceeding these ranges is immediately considered abnormal, and regulation is suspended to ensure the stability of the heart's physiological state.

[0056] The objective function of the adaptive adjustment engine for perfusion parameters is vector-imported, matched, calibrated, and iteratively optimized by combining damage characteristics and state constraints. This generates multi-dimensional damage index dynamic matching parameters, visual deviation comparison parameters, correction decision chain generation parameters, and multi-source data fusion parameters, forming optimization information for the modeling rules. First, the objective function of the adaptive adjustment engine for perfusion parameters is loaded. This objective function is a weighted superposition type objective optimization function, used to ensure that the simulated damage effect is highly consistent with the expected damage state while satisfying physiological safety and parameter constraints. The specific expression of the objective function is: F = 0.5 × F1 + 0.3 × F2 + 0.2 × F3, where F is the overall optimization target value, and the optimization process is directed towards minimizing the F value; 0.5 is the damage deviation weight, 0.3 is the physiological stability weight, and 0.2 is the parameter smoothing weight. The weight values ​​are consistent with the logic of prioritizing damage simulation, followed by physiological stability, and then parameter smoothing adjustment in ex vivo heart experiments.

[0057] Damage bias subfunction F1: Used to calculate the sum of squared differences between the current damage feature and the target damage feature, specifically F1 = (Pischemictischemic)2 = 1 / 2 * ... P ischemia 0)²+(O reperfusion t O refill 0)²+(E enzyme t E enzyme 0)²+(ST segment t In the formula ST segment 0)², P ischemia t is the characteristic value of the ischemic pressure in the current iteration, and P ischemia 0 is the characteristic value of the target ischemic pressure; O reperfusion t is the characteristic value of the reperfusion oxygen concentration in the current iteration, and O reperfusion 0 is the characteristic value of the target reperfusion oxygen concentration; E enzyme t is the characteristic value of the myocardial enzyme release in the current iteration, and E enzyme 0 is the characteristic value of the target myocardial enzyme release; ST segment t is the characteristic value of the ST segment offset in the current iteration, and ST segment 0 is the characteristic value of the target ST segment offset.

[0058] The physiological stabilization subfunction F2 is used to penalize situations where heart rate, coronary blood flow, and left ventricular end-diastolic pressure exceed the safe range. Specifically, F2 = |HRt| HR standard |+|CFt CF standard |+|LVEDPt In the formula LVEDP, HRt is the real-time heart rate, and HR is the standard heart rate of 280 beats / minute; CFt is the real-time coronary flow, and CF is the standard flow of 10 ml / minute; LVEDPt is the real-time left ventricular end-diastolic pressure, and LVEDP is the standard pressure of 10 mmHg.

[0059] The parameter smoothing sub-function F3 is used to limit the adjustment range of irrigation flow rate and oxygen concentration to avoid abrupt changes. Specifically, F3 = |ΔV| + |ΔO| where ΔV is the single adjustment change of irrigation flow rate and ΔO is the single adjustment change of oxygen concentration.

[0060] High-dimensional damage features are embedded into the vector input engine. The vectors are consistent with those mentioned earlier, namely [P ischemia, O reperfusion, E enzyme, ST segment]. An example input vector is [0.7, 0.95, 1.9, 0.2], and the target damage feature vector is [0.6, 0.95, 1.5, 0.15]. Simultaneously, physiological constraint parameters are loaded: heart rate 240–320 bpm, coronary blood flow 8–12 mL / min, left ventricular end-diastolic pressure 5–15 mmHg; and regulatory constraint parameters are: flow rate ≤ ±1 mL / min per adjustment, oxygen concentration ≤ ±10% per adjustment.

[0061] The engine substitutes the vector into the objective function to calculate the initial F-value, and then begins iterative calculation according to the iteration rules, with a maximum of 50 iterations and an iteration step size of 0.1. The convergence threshold is that the F-value change is ≤0.5% after three consecutive iterations. In each iteration, parameters such as reperfusion oxygen concentration, perfusion flow rate, and ischemia duration are adjusted based on the F-value change to continuously reduce the F-value. After each iteration, the deviation data between the current indicator and the target value is output, and control instructions are generated based on the deviation. After iteration convergence, the engine generates four types of parameters: dynamic matching parameters for multi-dimensional damage indicators, visualization deviation comparison parameters, correction decision chain generation parameters, and multi-source data fusion parameters.

[0062] Integrating three key elements—core input information, state adaptation constraint information, and model rule optimization information—generated in the early stages, this paper describes a comprehensive collaborative design and unified scheduling of four stages: damage feature import, dynamic matching iteration, parameter deviation visualization, and decision chain correction. This results in a logically coherent, stable, and reproducible basic modeling system for ischemia-reperfusion injury in ex vivo hearts. First, a closed-loop execution process is established, comprising five stages: feature input, constraint limitation, engine calculation, deviation visualization, and decision correction. In the feature input stage, high-dimensional damage features are embedded into a vector input system. These vectors contain four standardized features: ischemic pressure, reperfusion oxygen concentration, myocardial enzyme release, and ST segment changes on electrocardiogram. Example input vectors are 0.7, 0.95, 1.9, and 0.2. The constraint limiting stage loads parameters for damage assessment, index deviation calibration, adjustment step size limitation, temporal matching constraint, and physiological stability assurance. Heart rate is limited to 240-320 beats per minute, coronary blood flow to 8-12 ml per minute, left ventricular end-diastolic pressure to 5-15 mmHg, perfusion flow rate adjustment not exceeding 1 ml per minute per instance, and oxygen concentration adjustment not exceeding 10%. The engine calculation stage activates the adaptive adjustment engine for perfusion parameters, performing iterative calculations according to the objective function F = 0.5 × F1 + 0.3 × F2 + 0.2 × F3, with a maximum of 50 iterations and a convergence threshold of 0.5%. The deviation visualization stage plots real-time comparison curves between damage indicators and perfusion parameters. The decision correction stage automatically adjusts parameters based on the deviation.

[0063] In the parameter deviation visualization stage, the system displays in real-time comparisons of the deviations of ischemic pressure, reperfusion oxygen concentration, myocardial enzyme release, and ST segment shift from the target values. For example, it shows that the deviation between the current ischemic pressure (0.7) and the target value (0.6) is 0.1; the current reperfusion oxygen concentration (0.95) has no deviation from the target value (0.95); the current myocardial enzyme release (1.9) has a deviation of 0.4 from the target value (1.5); and the current ST segment shift (0.2) has a deviation of 0.05 from the target value (0.15). All deviations are presented simultaneously in both numerical and graphical forms, intuitively reflecting the difference between the model output and the expected state.

[0064] In the decision chain generation and correction phase, the system automatically generates continuous parameter adjustment instructions based on the magnitude and direction of the deviation. When myocardial enzyme release is too high and ST segment deviation is too large, it automatically generates a correction instruction to prolong ischemia duration by five minutes or increase the reperfusion flow rate from 10 ml / min to 11 ml / min. When ischemic pressure deviation is too large, it automatically generates a correction instruction to reduce the initial oxygen concentration of reperfusion by 5%. All instructions follow the control step size limit, and a single parameter adjustment does not exceed the safety range, forming an orderly execution of the correction decision chain.

[0065] In the data fusion and storage phase, sample data, physiological monitoring data, environmental parameter data, and perfusion control data generated throughout the process are uniformly aligned and associated for storage according to unique sample identifiers and global timestamps. Every set of damage characteristics, every iteration result, every parameter correction, and every physiological indicator is fully recorded to ensure that the data is traceable, verifiable, and reusable.

[0066] S106, based on the basic modeling model combined with actual damage assessment data and dynamic adjustment data of perfusion parameters, performs accuracy verification and dynamic correction. Through a multi-index comprehensive evaluation algorithm, it balances the requirements of model stability, damage simulation similarity and parameter adjustability, and generates modeling information for ischemia-reperfusion injury simulation of isolated heart.

[0067] In one implementation, a systematic pattern mining and standardized feature extraction are carried out on the intrinsic calculation relationship between the basic model of ischemia-reperfusion in isolated heart, actual damage assessment data, dynamic adjustment data of perfusion parameters, and physiological monitoring data throughout the process. The model operation logic, experimental measurement results, parameter control process, and physiological dynamic changes are correlated and analyzed to finally form four types of core features: model representation features, damage measurement features, parameter control features, and physiological response features.

[0068] First, model representation features are extracted, focusing on the basic modeling model. Structural attributes, input-output rules, and computational logic directly related to model operation are extracted. These features include high-dimensional damage feature embedding vector input rules, iterative calculation logic of the adaptive adjustment engine for perfusion parameters, objective function calculation methods, convergence criteria, and data storage association methods. Specifically, the extracted content includes four fixed features in the vector input dimension: ischemic pressure, reperfusion oxygen concentration, myocardial enzyme release, and ECG ST segment changes. Iterative calculations use a maximum of fifty iterations and a convergence threshold of 0.5%. The objective function employs a weighted calculation method considering damage bias, physiological stability, and parameter smoothing. Data is stored using a joint index of sample identifiers and global timestamps. The example is to extract a four-dimensional structure from the model with the vector input format of 0.7, 0.95, 1.9, and 0.2. The iteration termination condition is that the change in the objective function value is less than 0.5 percent for three consecutive iterations. The objective function operation adopts the weighted calculation logic of F=0.5×F1+0.3×F2+0.2×F3. The above contents together constitute the complete model representation features.

[0069] Next, the measured features of the injury are extracted. Based on actual injury assessment data, quantitative indicators and stage division criteria that can truly reflect the degree of myocardial injury are extracted. The measured features of the injury include measured values ​​of myocardial enzyme release, measured amplitude of ST segment shift on electrocardiogram, injury grade determination results, and the division of injury status in each stage of balanced perfusion, ischemia, and reperfusion. Specifically, the extracted content includes a measured value of myocardial enzyme release of 2.3 times the baseline value, a measured amplitude of ST segment shift of 0.2 mV lasting more than 30 seconds, an injury grade of moderate injury, an ischemic stage of 20 to 50 minutes (30 minutes in total), and a reperfusion stage of 50 to 110 minutes (60 minutes in total). For example, the enzyme concentration increased to 2.3 times the baseline value after 10 minutes of reperfusion from myocardial enzyme detection data, and the ST segment shift amplitude of 0.2 mV was extracted from the electrocardiogram record. Based on these values, the sample injury degree was determined to be moderate, forming measured features of the injury that can be directly used for model comparison.

[0070] Third, the extraction of parameter regulation characteristics is carried out. Using dynamic adjustment data of perfusion parameters as the source, information on temporal changes, adjustment amplitudes, and switching nodes related to perfusion operations is extracted. Parameter regulation characteristics include perfusion flow rate adjustment records, oxygen concentration switching records, ischemia duration adjustment values, reperfusion flow rate setpoints, and parameter switching time points at each stage. Specifically, the extracted content includes a base perfusion flow rate of 10 ml / min, a flow rate that drops to 0 ml / min during the ischemia stage, and a return to 10 ml / min during the reperfusion stage. The oxygen concentration switches from 95% to 0 at the onset of ischemia and returns to 95% at the onset of reperfusion. The ischemia duration is fixed at 30 minutes. For example, the perfusion flow rate is extracted from the regulation records as follows: 10 ml / min from 0 to 20 minutes, 0 ml / min from 20 to 50 minutes, and returns to 10 ml / min from 50 to 110 minutes. The oxygen concentration switches to 0 at 20 minutes and returns to 95% at 50 minutes. These temporal and numerical values ​​together constitute the parameter regulation characteristics.

[0071] Finally, physiological response characteristics were extracted based on the full-process physiological monitoring curve, revealing trends and fluctuations that reflect dynamic changes in the myocardium. These physiological response characteristics include the range of heart rate changes, the coronary blood flow fluctuation range, the amplitude of left ventricular end-diastolic pressure response, and the trends of these indicators with each perfusion phase. Specifically, the extracted content includes a heart rate maintained between 260 and 300 beats per minute throughout the process; coronary blood flow remaining stable at 10 ml / min during the equilibrium perfusion phase, rapidly decreasing to near zero ml / min during the ischemic phase, and gradually recovering to 9 ml / min during the reperfusion phase; and left ventricular end-diastolic pressure decreasing from 10 mmHg to 6 mmHg during the ischemic phase. For example, the extracted data shows a heart rate stable at 280 beats per minute during the equilibrium perfusion phase, slightly increasing to 300 beats per minute during the ischemic phase, and decreasing back to 270 beats per minute during the reperfusion phase; and coronary blood flow exhibiting a three-stage trend of stability, a sharp drop, and a recovery with the perfusion status. These trends and numerical values ​​together constitute the physiological response characteristics.

[0072] The extracted model representation features, measured injury features, parameter regulation features, and physiological response features were integrated and cross-dimensionally correlated and verified using a unified dimension. These four types of features were aligned, correlated, and verified within the same data framework, ultimately forming a multi-dimensional model accuracy verification fusion feature set that can be directly used for model evaluation. First, a unified dimension integration operation was performed, converting all four types of features into standardized data with the same dimension, time series, and sample identifier. Using a globally unified timeline set in the experiment as the benchmark, the timeline covered 20 minutes of equilibrium perfusion, 30 minutes of ischemia, and 60 minutes of reperfusion, for a total duration of 110 minutes. All features were arranged using timestamps with one millisecond precision. A unique sample identifier was used as the unified primary key, such as Rat20260509001, assigning all model representation features, measured injury features, parameter regulation features, and physiological response features corresponding to the same sample to the same identifier, ensuring a unique correspondence among data subjects.

[0073] Subsequently, cross-dimensional correlation matching was performed to align the damage features calculated by the model with the measured damage features obtained from actual detection at the same time point. The model-output ischemic pressure value of 0.7, myocardial enzyme release ratio of 1.9, and ST segment offset amplitude of 0.2 were matched item by item with the measured ischemic pressure of 6 mmHg, myocardial enzyme release of 2.3 times the baseline value, and ST segment offset of 0.2 mV. Simultaneously, the adjustment time points of perfusion parameters were precisely matched with the time points of physiological indicator mutations. For example, the parameter switching moment at the 20th minute when ischemia begins and the flow rate drops to zero ml was aligned with the physiological mutation moments when heart rate begins to rise, coronary blood flow begins to decrease, and ventricular pressure begins to decrease, ensuring that the timing of regulatory actions is consistent with the physiological response.

[0074] Next, a three-level correlation verification is performed. The first level verifies the consistency between the model output and the measured data, calculating the deviation between the damage value calculated by the model and the actual detected damage value. The allowable deviation range is no more than 10%. In the example, the model output for myocardial enzyme release is 1.9 times, while the measured value is 2.3 times. The deviation is within the allowable range, and the data is considered consistent. The second level verifies the rationality of parameter adjustments and physiological responses, verifying whether physiological laws such as the inevitable decrease in coronary blood flow when the flow rate decreases and the gradual recovery of the ST segment when the oxygen concentration increases hold true, ensuring that parameter regulation is consistent with the logic of physiological changes. The third level verifies the adaptability of the model logic to actual physiological laws, checking whether the number of iterations, convergence threshold, and objective function calculation conform to the actual changing trends in the isolated heart experiment, and confirming that the model calculation has no reverse or abnormal output.

[0075] After all validations are passed, multi-dimensional information fusion is completed, integrating four types of information—model attributes, measured results, regulatory records, and physiological changes—into a unified whole, forming a multi-dimensional model accuracy validation fusion feature set. This feature set includes all information such as sample identifiers, timestamps, model input vectors, iteration parameters, measured damage values, perfusion flow rate changes, oxygen concentration switching records, heart rate change curves, coronary blood flow fluctuations, and left ventricular end-diastolic pressure changes. All features are arranged in a unified temporal sequence, with each time point corresponding to a complete set of model data, measured data, regulatory data, and physiological data, forming a complete, timely, and clearly correlated fusion feature set that can be directly used to assess model stability, damage simulation similarity, and parameter adjustability.

[0076] Based on the established multi-dimensional model accuracy verification fusion feature set, a multi-index comprehensive evaluation algorithm is used to perform weighted calculations, balance constraints, and dynamic corrections on three core indicators: model stability, damage simulation similarity, and parameter adjustability. The final output is standardized verification data including damage degree, perfusion parameters, physiological indicators, and iteration errors. First, the weight allocation of the three core indicators is set. According to the experimental objective of isolated cardiac ischemia-reperfusion injury simulation, the weight of damage simulation similarity is set to 0.5, the weight of model stability is set to 0.3, and the weight of parameter adjustability is set to 0.2. This weight allocation conforms to the experimental design logic of prioritizing damage restoration, followed by operational stability, and then parameter adaptation.

[0077] In the model stability assessment, the stability is determined by calculating the fluctuation range and the number of anomalies in the results of multiple repeated model runs. The system extracts the output data of the model during fifty iterations and statistically analyzes the fluctuation amplitude of three indicators: heart rate, coronary blood flow, and left ventricular end-diastolic pressure. The model is considered stable if the overall fluctuation amplitude of these three indicators is less than 10%, 15%, and 10% respectively, and no abnormal states exceeding the physiological safety range are observed. In the example, the maximum fluctuation of heart rate in multiple model runs was 8%, the maximum fluctuation of coronary blood flow was 12%, and the maximum fluctuation of left ventricular end-diastolic pressure was 9%, with no abnormal trigger records. Therefore, the model stability score is set to 0.92.

[0078] In the injury simulation similarity assessment, the deviation between the model's output injury features and the measured injury data is used for judgment. The three injury indicators output by the model—ischemic pressure, myocardial enzyme release, and ST segment deviation—are compared with their corresponding actual measured indicators, and the average deviation is used as the similarity criterion. The maximum allowable average deviation is 15%. When the actual average deviation is less than this value, the similarity is considered to meet the requirements. In the example, the average deviation between the model output and the measured data is 9.6%, therefore the injury simulation similarity score is set to 0.88.

[0079] In the parameter adjustability assessment stage, the model's parameter adjustment range and response performance under different samples and perfusion conditions are analyzed for evaluation. The model is tested to ensure stable control and effective damage simulation across perfusion flow rates of 8-12 ml / min, oxygen concentrations of 0-95%, and ischemia durations of 20-40 minutes. Good adaptability is considered achieved when the model responds normally across all adjustment ranges without failure. In this example, the model operates normally across all set ranges; therefore, the parameter adjustability score is set to 0.90.

[0080] Subsequently, a weighted summation was performed according to preset weights to obtain the model's comprehensive evaluation score. The score calculation formula is: comprehensive score = 0.5 times the damage simulation similarity score + 0.3 times the model stability score + 0.2 times the parameter adjustability and adaptability score. The model is dynamically corrected based on the comprehensive score. When the score falls below 0.85, internal parameters such as the iteration step size, convergence threshold, and objective function weighting coefficients are automatically adjusted until the comprehensive score reaches the acceptable range. Finally, the evaluation and correction results are unified into standardized validation data. The data includes physiological indicators such as the damage severity level corresponding to the sample, equilibrium perfusion flow rate, reperfusion oxygen concentration, heart rate, coronary blood flow, and left ventricular end-diastolic pressure, as well as model validation information such as iteration error, mean deviation, and comprehensive score.

[0081] Standardized validation data is fused with the basic modeling model, measured damage data, and dynamic control data through closed-loop calibration. The fusion process uses standardized validation data as the correction benchmark, measured damage data as the truth constraint, and dynamic control data as the adaptive optimization basis. It systematically corrects and calibrates the internal parameters, calculation rules, mapping relationships, and convergence logic of the basic modeling model, achieving a high degree of matching between the model output and the measured results. First, full-dimensional data alignment is performed, binding the damage degree, perfusion parameters, physiological indicators, and iteration errors in the standardized validation data to the output of the basic modeling model, measured damage data, and dynamic control data one-to-one using a unique sample identifier and a global timestamp. This ensures that the model's calculated values, measured values, and control action values ​​correspond completely at the same time, providing a unified benchmark for model correction.

[0082] First, the vector input and feature mapping rules were revised. The order of the four-dimensional input dimensions—ischemic pressure, reperfusion oxygen concentration, myocardial enzyme release, and ST segment changes on ECG—remained unchanged. Based on feature biases in the validation data, the feature mapping coefficients were corrected. The original mapping coefficients were 1.0, 1.0, 1.0, and 1.0; the revised coefficients were 0.98, 1.0, 0.95, and 1.02, making the correspondence between model input features and injury output more closely resemble actual physiological patterns. Second, the weighting coefficients of the objective function were revised. Based on the original objective function F = 0.5 × F1 + 0.3 × F2 + 0.2 × F3, the weight allocation was optimized based on the model performance reflected in the validation data. The revised objective function is F = 0.52 × F1 + 0.30 × F2 + 0.18 × F3, further improving the similarity of injury simulation while maintaining physiological stability and parameter smoothing constraints.

[0083] Next, the internal parameters of the iterative calculation were modified. The maximum number of iterations was retained to be fifty, the original iteration step size was modified from 0.1 to 0.08, the original convergence threshold was modified from 0.5% to 0.4%, and a new consecutive convergence judgment number of three was added. That is, the model is considered to have reached convergence and the calculation is stopped only when the objective function value changes by less than 0.4% in three consecutive iterations. Subsequently, the damage judgment calculation rules were modified. The original myocardial enzyme release damage threshold of 2.5 times the baseline value was modified to 2.3 times the baseline value; the original ECG ST segment deviation damage threshold of 0.2 mV was modified to 0.18 mV; and the original ischemic pressure damage threshold of less than 5 mmHg was modified to less than 6 mmHg, so that the model damage judgment results were completely consistent with the measured data.

[0084] The adaptive adjustment triggering rules were further refined. The original triggering threshold for perfusion flow rate adjustment, which was set at a flow deviation of more than 1 ml per minute, was revised to a flow deviation of more than 0.8 ml per minute. Similarly, the original triggering threshold for oxygen concentration adjustment, which was set at a ST segment deviation of more than 0.05 mV, was revised to a ST segment deviation of more than 0.04 mV, thereby improving the model's sensitivity to changes in physiological indicators. Finally, the temporal matching and association rules were revised. The temporal matching window for the ischemic phase remained at 30 seconds, and the topological connectivity threshold remained at 90%. However, the feature similarity threshold was revised from 85% to 86%, ensuring complete consistency between the node association rules within the model and the construction rules of the dynamic physiological association graph.

[0085] Simultaneously, measured injury data is written into the model output layer as a truth constraint, limiting the model output results to within ±10% of the measured values ​​to avoid out-of-limit calculations. Dynamic control data is transformed into adaptive model logic, with the actual switching sequence and adjustment amplitude of perfusion flow rate, oxygen concentration, and ischemia duration written into the rule base, enabling the model to automatically reproduce the real control path. After all corrections and fusions are completed, a complete, accurate, reproducible, and scalable modeling information for isolated cardiac ischemia-reperfusion injury simulation is formed. This modeling information includes a complete experimental procedure, standard perfusion parameters, monitoring index specifications, internal model parameters, injury judgment criteria, adaptive adjustment rules, and a complete and reproducible operating method. Standard perfusion parameters include a temperature of 37 degrees Celsius, a flow rate of 10 ml / min, an ischemia duration of 30 minutes, and a reperfusion duration of 60 minutes; monitoring index specifications include maintaining a heart rate of 240-320 beats / min and coronary blood flow of 8-12 ml / min; and internal model parameters include all calibrated values ​​such as feature mapping coefficients, iteration step size, convergence threshold, and objective function weighting coefficients. The final output modeling information can be directly used to reproduce isolated cardiac ischemia-reperfusion injury simulation experiments on different experimental platforms and by different operators, ensuring a unified process, consistent parameters, comparable results, and complete reproducibility.

[0086] In one implementation, such as Figure 2 As shown, this application also provides a modeling device for simulating myocardial ischemia-reperfusion injury in an isolated heart, comprising: The multi-source data acquisition module is used to acquire structured and unstructured multi-source heterogeneous safety production data from all dimensions, including equipment operation, environmental monitoring, personnel operation, fire protection, logistics energy consumption, and business processes in medical institutions, forming a comprehensive safety data collection set. The data fusion processing module is used to clean, normalize, and extract features from multi-source heterogeneous data to achieve standardized transformation. Through weighted fusion, spatiotemporal correlation analysis, and deep learning, it achieves deep data coupling and generates a dynamic safety production assessment fusion dataset. The security feature construction module is used to layer the fused dataset according to the dimensions of risk identification, level determination, and trend prediction, construct a three-element association node of indicator-data-evaluation result, and integrate multi-dimensional security information through a security indicator association algorithm to generate a security feature association system for full-domain control. The evaluation modeling module is used to extract core features of operating status, environmental parameters, operating procedures, security level, and energy consumption indicators based on the security feature association system combined with the analytic hierarchy process and fuzzy comprehensive evaluation model, and generate security feature embedding vectors. The edge-cloud collaboration module is used to embed security features into vectors and import them into the edge-cloud collaborative computing engine for real-time updates and iterative optimization of the anomaly warning mechanism. It generates a risk correction decision chain through indicator deviation visualization, forming a dynamic assessment basic model. The dynamic evaluation output module is used to conduct accuracy verification and dynamic correction based on the evaluation base model combined with real-time monitoring and historical safety data, and output dynamic safety production evaluation results that are adapted to the refined safety management and proactive risk prevention of medical institutions.

[0087] The various embodiments in this application are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the modeling method, electronic device, electronic device, and readable storage medium embodiments for the modeling apparatus for simulating myocardial ischemia-reperfusion injury modeling of isolated heart perfusion are basically similar to the modeling method embodiments of the modeling apparatus for simulating myocardial ischemia-reperfusion injury described above, so the description is relatively simple. Relevant parts can be referred to in the description of the modeling method embodiments of the modeling apparatus for simulating myocardial ischemia-reperfusion injury described above.

Claims

1. A modeling method of a modeling device of an isolated heart perfusion modeling device simulating myocardial ischemia reperfusion injury, characterized by, include: Acquire multi-source core modeling data; Based on the isothermal perfusion system and dynamic oxygen partial pressure regulation module, the isolated heart is rapidly rewarmed, cannulated and adapted, and basic perfusion is stabilized. The core physiological indicators, including heart rate, coronary blood flow and left ventricular end-diastolic pressure, are identified through the myocardial electrophysiological monitoring model. The key variables of ischemia duration and reperfusion rate are integrated into the modeling process to generate a basic perfusion dataset. The perfusion basic dataset is hierarchically divided into a main table of core physiological indicators and a detailed table of environmental parameters. Each layer constructs a ternary association node including sample identifier, monitoring indicator, and regulation parameter. The perfusion process and physiological response information are aggregated through a time-series feature matching algorithm to generate a dynamic physiological association graph for ischemia-reperfusion injury modeling. Based on the damage severity assessment engine and the time-series parameter prediction model, the dynamic physiological correlation graph is modeled, and core features including ischemic pressure, reperfusion oxygen concentration, myocardial enzyme release, and ST segment changes on electrocardiogram are extracted to generate a high-dimensional damage feature embedding vector. The damage feature embedding vector is imported into the perfusion parameter adaptive adjustment engine, and a multi-dimensional damage index dynamic matching mechanism is used for iterative optimization. At the same time, a correction decision chain is generated through parameter deviation visualization, forming a basic modeling model of ischemia-reperfusion injury in ex vivo heart that integrates multi-source data. Based on the basic modeling model combined with actual damage assessment data and dynamic adjustment data of perfusion parameters, the accuracy is verified and dynamically corrected. Through a multi-index comprehensive evaluation algorithm, the modeling information for ischemia-reperfusion injury simulation of isolated heart ischemia ...

2. The method of claim 1, wherein, Based on a isothermal perfusion system and a dynamic oxygen partial pressure regulation module, rapid rewarming, cannulation adaptation, and baseline perfusion stabilization of isolated hearts were achieved. A myocardial electrophysiological monitoring model identified core physiological indicators including heart rate, coronary blood flow, and left ventricular end-diastolic pressure. Key variables such as ischemia duration and reperfusion rate were incorporated into the modeling process to generate a baseline perfusion dataset, including: Based on multi-source physiological and environmental data such as isolated heart samples, perfusion fluid composition, oxygen partial pressure, temperature, and flow rate, and following the rules of homeostasis maintenance, indicator monitoring, and damage induction, the perfusion parameters and data acquisition strategy are determined collaboratively by the isothermal perfusion unit and the oxygen partial pressure regulation unit. The data processing method and indicator monitoring dimensions are determined through real-time physiological signal acquisition and global temporal synchronization. A four-stage linkage mechanism of rewarming-cannulation-perfusion-stabilization is adopted to process multi-source basic data in an integrated manner. First, the isothermal perfusion system completes the rapid rewarming of the isolated heart, aortic cannulation and steady-state perfusion. Then, the physiological monitoring module completes the collection and analysis of core indicators, and correlates the core requirements of heart rate, coronary blood flow and left ventricular end-diastolic pressure to generate standardized raw perfusion data. Real-time baseline verification is performed on standardized raw perfusion data, adaptive parameter fine-tuning is initiated to address index fluctuations caused by sample differences, a filtering compensation mechanism is triggered for monitoring signal noise, and a stability determination process is executed for abnormal physiological states, generating optimized perfusion data that includes baseline calibration, signal compensation, and state determination. The optimized perfusion data is recorded and integrated in a closed loop. The entire process is associated through a synchronous acquisition link, and ischemia duration, reperfusion flow rate, temporal changes of physiological indicators and system operation status information are read synchronously to form a basic perfusion dataset that supports ischemia-reperfusion injury modeling.

3. The method of claim 1, wherein, The perfusion baseline dataset is hierarchically structured according to a main table of core physiological indicators and a detailed table of environmental parameters. Each layer constructs a ternary association node comprising sample identifier, monitoring indicator, and regulatory parameter. A time-series feature matching algorithm is used to aggregate perfusion process and physiological response information, generating a dynamic physiological association graph for ischemia-reperfusion injury modeling, including: Based on the requirements of completeness, temporal consistency, traceability, and feature extraction for ischemia-reperfusion modeling of the perfusion basic dataset, the data is hierarchically layered and structured according to the main table of core physiological indicators and the detailed table of environmental parameters, clarifying the layered fields, primary key rules, and storage structure constraints. Based on the requirements of hierarchical architecture and topological association construction, a three-element association node of sample identifier, monitoring indicator and control parameter is established in each layer of data. The mapping relationship, matching logic and data interaction method are determined to form a standardized organizational model with clear hierarchy, associative nodes and time sequence alignment. Combining the requirements of continuous myocardial physiological response, controllable perfusion process, and identifiable injury characteristics, time-series matching windows, feature similarity, and topological connectivity constraints are set according to time series and perfusion stages to determine the execution rules for node matching, feature aggregation, and topology generation. The data layering, ternary node construction, and temporal feature matching algorithm are integrated to generate a dynamic physiological correlation graph for ischemia-reperfusion injury modeling.

4. The method of claim 1, wherein, Damage features are embedded into vectors and imported into an adaptive adjustment engine for perfusion parameters. A multi-dimensional damage index dynamic matching mechanism is used for iterative optimization. Simultaneously, a correction decision chain is generated through parameter deviation visualization, forming a fundamental modeling model for ischemia-reperfusion injury in ex vivo hearts that integrates multi-source data. This model includes: Based on the physiological characteristics, parameter regulation range, injury response law, and multi-source data fusion requirements of isolated heart ischemia-reperfusion injury simulation, the injury feature embedding vector, perfusion regulation parameters, and time-series iteration rules are adapted, analyzed, and feature extracted to generate injury characterization features, parameter adaptation features, iterative optimization features, and data fusion features, forming the core input information for building the basic model. Logical transformation and threshold definition are performed on dynamic change elements such as ischemic pressure, reperfusion oxygen concentration, myocardial enzyme release, ST segment changes, and perfusion flow rate to generate damage degree judgment parameters, index deviation calibration parameters, control step size limit parameters, time sequence matching constraint parameters, and physiological stability guarantee parameters, forming state adaptation constraint information in the modeling process. The objective function of the adaptive adjustment engine for perfusion parameters is combined with damage characteristics and state constraints for vector import, matching calibration and iterative optimization to generate multi-dimensional damage index dynamic matching parameters, visual deviation comparison parameters, correction decision chain generation parameters and multi-source data fusion parameters, forming model rule optimization information; By integrating core input information, state adaptation constraint information, and model rule optimization information, a collaborative design is carried out throughout the entire process of injury feature import, dynamic matching iteration, parameter deviation visualization, and decision chain correction, forming a basic modeling model of ischemia-reperfusion injury in ex vivo heart that integrates multi-source data.

5. The method of claim 4, wherein, Based on the fundamental modeling model combined with actual damage assessment data and dynamic adjustment data of perfusion parameters, accuracy verification and dynamic correction are performed. A multi-index comprehensive evaluation algorithm is used to balance model stability, damage simulation similarity, and parameter adjustability requirements, generating modeling information for isolated cardiac ischemia-reperfusion injury simulation, including: The algorithm mines and extracts features from the solution relationships of the basic model of ischemia-reperfusion in isolated heart, actual damage assessment data, and dynamic adjustment data of perfusion parameters, and generates model characterization features, measured damage features, parameter regulation features, and physiological response features. The model representation features, damage measurement features, parameter regulation features, and physiological response features are integrated and correlated for verification to generate a multi-dimensional model accuracy verification fusion feature set. Based on the multi-dimensional model accuracy verification fusion feature set, a multi-index comprehensive evaluation algorithm is used to perform weighted balancing and dynamic correction on model stability, damage simulation similarity, and parameter adjustability, generating standardized verification data that includes damage degree, perfusion parameters, physiological indicators, and iteration error. By integrating standardized validation data with basic modeling models, measured damage data, and dynamic control data through closed-loop calibration, the entire process of optimizing and adapting the isolated heart ischemia-reperfusion simulation model is completed, generating modeling information for isolated heart ischemia-reperfusion injury simulation.

6. A modeling device for modeling an isolated heart perfusion simulating myocardial ischemia reperfusion injury, characterized in that, The device includes: The multi-source modeling data acquisition module is used to acquire multi-source core modeling data such as isolated heart samples, perfusion fluid components, temperature, oxygen partial pressure, flow rate and physiological signals, forming a set of basic data for global modeling; The perfusion data processing module is used to achieve cardiac rewarming, intubation and stable perfusion based on isothermal perfusion and dynamic oxygen partial pressure regulation, monitor core indicators such as heart rate, coronary blood flow and left ventricular end-diastolic pressure, incorporate key variables such as ischemia duration and reperfusion rate, and generate a standardized perfusion basic dataset. The physiological feature construction module is used to hierarchically divide the perfusion basic dataset into a main table of physiological indicators and a detailed table of environmental parameters, construct a three-element association node of sample identifier, monitoring indicator and regulation parameter, and generate a dynamic physiological association graph for ischemia-reperfusion injury modeling by aggregating information through temporal feature matching. The damage feature extraction module is used to model the dynamic physiological correlation graph based on the damage assessment engine and the time-series parameter model, extract the core features of ischemic pressure, reperfusion oxygen concentration, myocardial enzyme release, and ST segment changes in electrocardiogram, and generate a high-dimensional damage feature embedding vector. The adaptive modeling module is used to embed damage features into vectors and import them into the adaptive adjustment engine of perfusion parameters. It dynamically matches and iteratively optimizes multi-dimensional damage indicators, generates a correction decision chain through parameter deviation visualization, and forms a basic modeling model of ischemia-reperfusion injury in ex vivo heart that integrates multi-source data. The modeling verification output module is used to perform accuracy verification and dynamic correction based on the basic modeling model combined with measured damage data and perfusion parameter adjustment data, balance model stability, simulation similarity and parameter adaptability, and generate complete modeling information for isolated heart ischemia-reperfusion injury simulation.

7. An electronic device, comprising: include: First processor; And a memory for storing executable instructions of the first processor; wherein the first processor is configured to execute the modeling method of the modeling apparatus for simulating myocardial ischemia-reperfusion injury of an ex vivo heart as described in any one of claims 1 to 5 by executing the executable instructions.

Images