Artificial Intelligence-Based Injury Control and Resuscitation Decision Support System and Method

Abstract

This invention relates to the field of healthcare informatics technology, specifically to an artificial intelligence-based injury control and resuscitation decision support system and method. By analyzing the temporal correlation characteristics of multimodal vital sign data in real-time multimodal physiological monitoring data of a patient, it determines whether the patient is experiencing a period of significant energy depletion compensation. If such a period is observed, it analyzes tissue metabolic side effects based on multimodal blood gas analysis data from the patient's multimodal physiological monitoring data during this period, determining the manifestation of physiological metabolic side effects. Furthermore, combining this data with the patient's multimodal physiological monitoring data during the significant energy depletion compensation period, a risk prediction model is used to automatically generate a real-time injury risk warning. This invention effectively improves the accuracy of patient risk prediction, thereby enhancing the accuracy of injury control and resuscitation decisions.

Description

Technical Field

[0001] This invention relates to the field of healthcare informatics technology, specifically to an artificial intelligence-based injury control and resuscitation decision support system and method. Background Technology

[0002] Damage control resuscitation is a key medical strategy for patients with severe trauma and hemorrhagic shock. It aims to control bleeding, restore circulating blood volume, and prevent complications through phased physiological status assessment and intervention, thus preventing irreversible physiological exhaustion. Real-time, accurate, and quantitative assessment of the patient's physiological status is a prerequisite for effective intervention.

[0003] However, traditional assessment methods often rely on doctors' clinical experience and the interpretation of single moments and single indicators (such as instantaneous blood pressure and heart rate). This experience-based assessment model has significant limitations when facing complex trauma: on the one hand, the human body has a strong neuroendocrine compensatory mechanism in the early stages of trauma, and even if blood loss reaches a certain level, the patient's blood pressure and heart rate may still remain within a relatively normal "pseudo-stable" range, thus masking the internal circulatory crisis (i.e., occult shock); on the other hand, although existing monitoring equipment can generate massive amounts of data, it lacks in-depth mining and quantitative analysis of the temporal linkages between heart rate, blood pressure, heart rate variability (HRV), and blood gas indicators, making it difficult to capture the dynamic evolution of physiological compensatory mechanisms from activation to depletion, and then to tissue metabolic disorders.

[0004] Therefore, how to utilize multimodal physiological data to construct a system that can penetrate the surface of physiological compensation and dynamically and accurately quantify the risk of injury to patients, thereby providing objective and early warning signals of critical conditions for clinical practice, is a technical problem that urgently needs to be solved in the field of trauma emergency care. Summary of the Invention

[0005] To address the aforementioned technical problems, the present invention aims to provide an artificial intelligence-based injury control and resuscitation decision support system and method, the specific technical solution of which is as follows: In a first aspect, the present invention provides an artificial intelligence-based injury control and resuscitation decision support method, comprising the following steps: Acquire real-time multimodal physiological monitoring data of patients, wherein the multimodal physiological monitoring data includes at least multimodal vital sign data and multimodal blood gas analysis data; Analyze the temporal linkage characteristics of the multimodal vital signs data to determine whether the patient experiences a period of significant energy consumption compensation in real time; If a real-time patient experiences a period of significant energy consumption compensation, the side effects at the tissue metabolic level are analyzed based on the multimodal blood gas analysis data of the real-time patient during the period of significant energy consumption compensation, and the manifestation of physiological metabolic side effects of the real-time patient is determined. Based on the visibility of the physiological metabolic side effects and the real-time multimodal physiological monitoring data of the patient during the period of significant energy consumption compensation, a risk prediction model is used to automatically generate real-time damage risk warning information for the patient.

[0006] In conjunction with the first aspect above, in some possible implementations, the multimodal vital signs data includes at least heart rate data, blood pressure data, and electrocardiogram (ECG) signal data; analyzing the temporal linkage characteristics of the multimodal vital signs data to determine whether the patient experiences a period of significant energy consumption compensation in real time includes: Determine the difference between each maximum heart rate point in the heart rate data and its preceding adjacent heart rate value, and determine the heart rate surge rate corresponding to each maximum heart rate point in the heart rate data; Using the time corresponding to each maximum heart rate point in the heart rate data as the marked time, several blood pressure data intervals are constructed based on the blood pressure values ​​of the blood pressure data at the marked time and subsequent times, and several HRV data intervals are constructed based on the HRV values ​​of the electrocardiogram signal data at the marked time and subsequent times. Based on the heart rate spike rate, and by analyzing the blood pressure stability characteristics in the blood pressure data interval and the HRV significant reduction characteristics in the HRV data interval, it is determined whether the patient experiences a period of significant energy consumption compensation in real time.

[0007] In conjunction with the first aspect above, in some possible implementations, based on the heart rate spike rate and by analyzing the blood pressure stability characteristics within the blood pressure data interval and the significant HRV decrease characteristics within the HRV data interval, it is determined whether the patient experiences a period of significant energy consumption compensation in real time, including: Analyze the blood pressure stability characteristics of the blood pressure data interval to determine the blood pressure stability of the blood pressure data interval, and analyze the HRV significant reduction characteristics of the HRV data interval to determine the significant reduction of the HRV data interval; The heart rate spike rate, blood pressure stability, and significant reduction rate are combined to obtain several candidate combinations; Based on the heart rate spike rate, blood pressure stability, and significant reduction rate of the candidate combinations, the compensatory linkage characteristic performance of the candidate combinations is determined. Determine the maximum value among all the compensatory linkage characteristic performance values. If the maximum value is greater than the set compensatory linkage characteristic performance value threshold, then determine that the real-time patient has a period of significant energy consumption compensation. Based on the candidate combination corresponding to the maximum value, the blood pressure data interval, and the time corresponding to the blood pressure data interval, determine the period of significant energy consumption compensation.

[0008] In conjunction with the first aspect mentioned above, in some possible implementations, analyzing the blood pressure stability characteristics of the blood pressure data interval and determining the blood pressure stability of the blood pressure data interval includes: Determine the standard deviation of all blood pressure values ​​within the blood pressure data interval; The standard deviation is negatively correlated and normalized to obtain the blood pressure stability of the blood pressure data interval.

[0009] In conjunction with the first aspect above, in some possible implementations, analyzing the significant decrease characteristics of HRV in the HRV data interval and determining the degree of significant decrease in the HRV data interval includes: Determine the HRV difference between every two adjacent HRV values ​​within the HRV data interval; The average of all the HRV differences is then normalized with positive correlation to obtain the degree of significant reduction in the HRV data interval.

[0010] In conjunction with the first aspect above, in some possible implementations, the multimodal blood gas analysis data includes lactate data and base excess data; based on the multimodal blood gas analysis data of the real-time patient during the period of significant energy consumption compensation, the side effects at the tissue metabolic level are analyzed to determine the manifestation of physiological metabolic side effects in the real-time patient, including: The characteristics of the increase in lactate data of real-time patients during the period of significant energy consumption compensation were analyzed to determine the trend of lactate increase in real-time patients. The characteristics of the decrease in base excess data of real-time patients during the period of significant energy consumption compensation are analyzed to determine the degree of decrease in base excess of real-time patients. Based on the increasing trend of lactate and the decreasing performance of base excess, the prominence of changes in base excess over changes in lactate in real-time patients is determined. Based on the prominence of the change in base excess compared to the change in lactate, the trend of lactate increase is weighted and corrected to obtain the real-time manifestation of physiological metabolic side effects in patients.

[0011] In conjunction with the first aspect above, in some possible implementations, the rising characteristics of lactate data in real-time patients during the period of significant energy consumption compensation are analyzed to determine the trend of lactate increase in real-time patients, including: Based on the differences between each lactate value and the maximum normal lactate value in the lactate data of the real-time patient during the period of significant energy consumption compensation, and the differences between each lactate value and the next lactate value, the amount of lactate increase in each lactate value in the lactate data is determined. The lactate increase of all the stated values ​​was summed and normalized with positive correlation to obtain the real-time lactate trend of the patient.

[0012] In conjunction with the first aspect above, in some possible implementations, the characteristics of the decrease in base excess data of real-time patients during the period of significant energy consumption compensation are analyzed to determine the degree of base excess decrease in real-time patients, including: The standard alkali surplus data during the period of significant energy consumption compensation is corrected using the standard alkali surplus method to obtain standard alkali surplus data; Based on the difference between each alkali surplus value in the standard alkali surplus data and its next alkali surplus value, the decrease in alkali surplus for each alkali surplus value in the standard alkali surplus data is determined. All the base excess decreases were summed and positively correlated and normalized to obtain the real-time base excess decrease performance of the patients.

[0013] In conjunction with the first aspect above, in some possible implementations, the risk prediction model is composed of a neural network, which automatically generates real-time injury risk warning information for patients, including: A training dataset was constructed by acquiring the physiological metabolic side effects of several historical patients and the multimodal physiological monitoring data of these patients during periods of significant energy consumption compensation. Label the training dataset with damage risk warning values, and use the training dataset and its damage risk warning value labels to train the neural network to obtain the trained neural network. The real-time physiological metabolic side effects of the patient and the multimodal physiological monitoring data of the patient during the period of significant energy consumption compensation are input into the trained neural network, and the trained neural network is used to automatically generate the damage risk warning value of the real-time patient. Based on the value of the damage risk warning, damage control and resuscitation decision-making suggestions are automatically generated.

[0014] Secondly, the present invention also provides an artificial intelligence-based injury control and resuscitation decision support system, including a memory and a processor. The memory is used to store executable computer program code, and the processor is used to call and run the executable computer program code from the memory, causing the system to execute an artificial intelligence-based injury control and resuscitation decision support method according to the first aspect or any possible implementation thereof.

[0015] Thirdly, the present invention also provides a computer program product comprising: computer program code, which, when run on a computer, causes the computer to execute an artificial intelligence-based injury control and resuscitation decision support method in one of the first aspects or any possible implementations of the first aspect.

[0016] Fourthly, the present invention also provides a computer-readable storage medium storing computer program code that, when executed on a computer, causes the computer to perform an artificial intelligence-based injury control and resuscitation decision support method according to the first aspect or any possible implementation thereof.

[0017] This invention offers the following advantages: By acquiring real-time multimodal physiological monitoring data of patients, including at least multimodal vital sign data and multimodal blood gas analysis data, a data foundation for patient risk prediction is obtained. The temporal correlation of the multimodal vital sign data is analyzed to determine whether the patient experiences a period of significant energy consumption compensation, thereby accurately identifying the critical time interval of high energy consumption and low elasticity compensation caused by excessive sympathetic nerve excitation. Subsequently, by analyzing the tissue metabolic side effects based on the multimodal blood gas analysis data of the real-time patient during the period of significant energy consumption compensation, the manifestation of physiological metabolic side effects in the real-time patient is determined, quantifying the severity of metabolic side effects. Finally, the manifestation of physiological metabolic side effects and the multimodal physiological monitoring data of the real-time patient during the period of significant energy consumption compensation are input into a risk prediction model. The risk prediction model automatically generates early warning information on the injury risk of the real-time patient, improving the accuracy of patient risk prediction and thus improving the accuracy of injury control and resuscitation decisions. Attached Figure Description

[0018] To more clearly illustrate the technical solutions and advantages of the embodiments of the present invention, the accompanying drawings used in this embodiment will be briefly introduced below.

[0019] Figure 1 This is a flowchart illustrating the steps of an artificial intelligence-based injury control and resuscitation decision support method according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating the steps of determining whether a patient is experiencing a period of significant energy consumption compensation in real time, according to an embodiment of the present invention. Figure 3 This is a flowchart illustrating the steps for determining the real-time manifestation of physiological metabolic side effects in a patient according to an embodiment of the present invention. Figure 4 This is a flowchart illustrating the steps of automatically generating real-time patient injury risk warning information according to an embodiment of the present invention. Figure 5 This is a schematic diagram of the structure of an artificial intelligence-based injury control and resuscitation decision support system according to an embodiment of the present invention. Detailed Implementation

[0020] To clearly illustrate the technical features of the embodiments of the present invention, the present invention will be described in detail below through specific implementation methods and in conjunction with the accompanying drawings.

[0021] The following will provide a detailed description of the artificial intelligence-based injury control and resuscitation decision support system and method provided in the embodiments of the present invention, with reference to the accompanying drawings.

[0022] Figure 1 This diagram illustrates the basic flowchart of an artificial intelligence-based injury control and resuscitation decision support method provided by an embodiment of the present invention. Figure 1 As shown, the method specifically includes the following steps: Step S100: Acquire real-time multimodal physiological monitoring data of the patient. This multimodal physiological monitoring data includes at least multimodal vital sign data and multimodal blood gas analysis data.

[0023] For trauma patients in the emergency room, due to the varying types of trauma, severity of injuries, and patients' physical conditions, not all patients require large-scale blood transfusions or damage control surgery. For some patients with milder conditions who are still in the compensatory phase, a gentler monitoring and resuscitation strategy should be adopted to avoid unnecessary intervention and resource waste. However, for some critically ill patients who have already entered the compensatory phase, damage control resuscitation (DCR) should be initiated rapidly, and they should be transferred to the operating room as soon as possible for damage control surgery (DCS) to maximize survival and reduce complications. Therefore, accurately identifying those critically ill patients who truly require aggressive intervention is a crucial task in emergency care.

[0024] To accurately identify the risk categories of trauma patients, multiple monitoring devices are used to monitor various physiological data in real time or collect samples periodically, acquiring multimodal physiological monitoring data. This multimodal physiological monitoring data includes at least multimodal vital signs data and multimodal blood gas analysis data. Furthermore, based on the multimodal physiological monitoring data acquired by these devices, the patient's injury risk warning information is further assessed and provided to physicians for reference in determining whether to initiate an injury control and resuscitation program.

[0025] In a specific example, multimodal physiological monitoring data of the patient is acquired in real time or periodically through monitoring devices or sampling tests directly connected to the patient. This multimodal physiological monitoring data includes: Multimodal vital signs data: heart rate, non-invasive / invasive blood pressure, blood oxygen saturation, respiratory rate, body temperature, electrocardiogram signal, heart rate variability (HRV), etc. Multimodal advanced hemodynamic data: stroke volume variability, cardiac output, pulse pressure variability, etc.; Multimodal ventilator parameter data: tidal volume, airway pressure, end-expiratory carbon dioxide, etc. Multimodal blood gas analysis data: pH, lactate, base excess, potassium ions, etc.

[0026] Among these methods, the patient's electrocardiogram (ECG) signal can be collected using an ECG monitor, and the heart rate variability (HRV) data corresponding to the ECG signal can be continuously obtained using existing HRV calculation algorithms.

[0027] It should be understood that the aforementioned multimodal vital signs data, multimodal advanced hemodynamic data, and multimodal ventilator parameter data can be acquired through continuous monitoring by equipment. However, multimodal blood gas analysis data such as pH, lactate, base excess, and potassium ions require manual sample collection and timely analysis using a blood gas analyzer. The acquired multimodal physiological monitoring data undergoes preprocessing, including outlier removal, missing value imputation, and time-label alignment.

[0028] Step S200: Analyze the temporal linkage characteristics of multimodal vital signs data to determine whether the patient experiences a period of significant energy consumption compensation in real time.

[0029] During the compensatory phase and early decompensation stages, the body can still maintain macroscopic hemodynamic stability through rapid neuroendocrine mobilization, but the internal environment is already in a fragile balance of high power consumption and low reserves. It is necessary to capture "compensatory load signals" in real time, predict the risk of collapse, and intervene proactively in accordance with physiological rhythms.

[0030] The transition from compensated to decompensated states is a continuous and progressively deepening pathophysiological process, involving abnormal excitation of the neuroendocrine system, abnormal tissue metabolism, and microcirculatory paralysis. Specifically, trauma and bleeding reduce circulating blood volume, activating baroreceptors and triggering explosive activation of the SAMS (Syndrome-Associated Muscle Mass System); large amounts of catecholamines (adrenaline and noradrenaline) are released into the blood; heart rate increases, myocardial contractility strengthens (attempting to increase cardiac output), and systemic vasoconstriction (especially in the skin, internal organs, and muscles) strongly constricts, creating a compensatory effect. During this process, blood pressure is maintained due to vasoconstriction and increased heart rate, with a slightly delayed decrease. However, the heart rate velocity (HRV) calculated based on continuous electrocardiogram signals is significantly reduced due to sympathetic dominance and parasympathetic inhibition. Since a significant reduction in HRV requires several heartbeats to calculate, it is the most delayed factor. When the patient is in a compensatory state, they are in a state of strong sympathetic excitation. Although blood pressure and heart rate are still within the compensatory range, the significant decrease in HRV is rapidly depleting the patient's physiological reserves, and the body is quietly approaching the edge of decompensation.

[0031] Therefore, this embodiment of the invention analyzes the temporal linkage characteristics of multimodal vital sign data to determine whether a patient is experiencing a period of significant energy consumption compensation in real time. This period of significant energy consumption compensation refers to the time interval during which the patient is in a state of high energy consumption compensation. For example, within a certain time period, when the patient's heart rate spikes, given that blood pressure is affected first, greater attention needs to be paid to the stability of blood pressure data within that time period. Conversely, given that HRV is affected later, greater attention needs to be paid to the degree of significant reduction in HRV within that time period. Since blood pressure remains relatively stable during compensation, while HRV decreases significantly, the higher the stability of blood pressure data and the higher the degree of significant reduction in HRV, the more likely that the time period is a period of significant energy consumption compensation for the patient.

[0032] In one possible implementation, multimodal vital sign data includes at least heart rate data, blood pressure data, and electrocardiogram (ECG) signal data; such as Figure 2 As shown, step S200, analyzing the temporal linkage characteristics of multimodal vital sign data to determine whether the patient experiences a period of significant energy consumption compensation includes: Step S201: Determine the difference between each maximum heart rate point in the heart rate data and its preceding adjacent heart rate value, and determine the heart rate surge rate corresponding to each maximum heart rate point in the heart rate data.

[0033] Mark the maximum values ​​in the heart rate data as individual heart rate maxima. Compare each heart rate maxima with its preceding neighboring heart rate values ​​to determine the heart rate surge rate corresponding to each maxima. This heart rate surge rate reflects the degree of surge of each heart rate maxima relative to its preceding neighboring heart rate values.

[0034] In a specific example, the difference between each heart rate maxima and its preceding heart rate minima (the nearest preceding heart rate minima) is calculated, and the time interval between each heart rate maxima and its preceding heart rate minima is determined. The ratio of this difference to the time interval is calculated, and this ratio is normalized to a maximum and minimum value. That is, based on medical common sense or historical big data, an upper limit for extremely high surge rate and a lower limit for significantly pathological surge rate are determined. When the ratio is greater than or equal to the lower limit for significantly pathological surge rate and less than or equal to the upper limit for extremely high surge rate, the ratio is positively normalized using the upper limit for extremely high surge rate and the lower limit for significantly pathological surge rate. When the ratio is less than the lower limit for significantly pathological surge rate, the normalization result is set to 0. When the ratio is greater than the upper limit for extremely high surge rate, the normalization result is set to 1. The final normalized result is taken as the heart rate surge rate corresponding to each heart rate maxima. The upper limit of the extremely high surge rate and the lower limit of the significant pathological surge rate can be pre-defined by experts based on historical big data. For example, the upper limit of the extremely high surge rate and the lower limit of the significant pathological surge rate can be taken as the 95th percentile and 75th percentile of the historical heart rate pathological surge rate distribution, respectively.

[0035] Step S202: Using the time corresponding to each maximum heart rate point in the heart rate data as the marked time, construct several blood pressure data intervals based on the blood pressure values ​​at the marked time and subsequent times, and construct several HRV data intervals based on the HRV values ​​of the electrocardiogram signal data at the marked time and subsequent times.

[0036] In a specific example, the time corresponding to each maximum heart rate point in the heart rate data. Marking is performed separately to obtain multiple marking times. For the current marked time Set an effective physiological response lag window ,in, The preset maximum delay time for the compensatory response, for example, can be set to 5-10 minutes, covering the physiological lag period of blood pressure adjustment and HRV changes. (Based on the current marked time) The starting time and based on the subsequent time within the physiological response hysteresis window. Multiple blood pressure data intervals are constructed from continuous time-series data. The starting time of each blood pressure data interval is the current labeled time. The termination time is the lag window of this physiological response. In addition to the marked time At some point outside of [the specified time]. Simultaneously, using the currently marked time... The starting time and based on the subsequent time within the physiological response hysteresis window. The HRV values ​​at consecutive time points are used to construct multiple HRV data intervals.

[0037] Step S203: Based on the heart rate spike rate, and by analyzing the blood pressure stability characteristics in the blood pressure data interval and the HRV significant reduction characteristics in the HRV data interval, determine whether the patient experiences a period of significant energy consumption compensation in real time.

[0038] By analyzing the temporal correlation between heart rate spike rate, blood pressure stability characteristics within blood pressure data intervals, and significant HRV reduction characteristics within HRV data intervals, we can accurately identify the key time interval in which patients experience a high-energy-consumption, low-elasticity compensatory state caused by excessive sympathetic nerve excitation, i.e., the period of significant energy consumption compensation.

[0039] In one possible implementation, based on the heart rate spike rate and by analyzing blood pressure stability characteristics within blood pressure data intervals and significant HRV reduction characteristics within HRV data intervals, it is determined whether the patient experiences a period of significant energy compensation in real time, including: First, analyze the blood pressure stability characteristics of the blood pressure data intervals to determine the blood pressure stability of the blood pressure data intervals, and analyze the significant reduction characteristics of HRV in the HRV data intervals to determine the significant reduction of HRV in the HRV data intervals.

[0040] In a specific example, the blood pressure stability characteristics of a blood pressure data interval are analyzed to determine the blood pressure stability of the blood pressure data interval. This includes: determining the standard deviation of all blood pressure values ​​in the blood pressure data interval; and performing negative correlation normalization on the standard deviation to obtain the blood pressure stability of the blood pressure data interval.

[0041] Specifically, for all blood pressure values ​​within each blood pressure data interval, the standard deviation of these values ​​is calculated and negative correlation normalization is performed. Negative correlation normalization means that the values ​​before and after normalization are negatively correlated; that is, the larger the value before normalization, the smaller the value after normalization. The process of negative correlation normalization includes setting an extremely stable (rigid) threshold. (e.g., 2 mmHg, measurement error limit) and an instability threshold. (e.g., 15 mmHg, significant fluctuations), when the standard deviation is greater than or equal to the extremely stable (rigid) threshold. And less than or equal to the instability threshold Then, this extremely stable (rigid) threshold is utilized. and instability threshold The standard deviation is subjected to min-max normalization, and the difference between the value of 1 and the min-max normalization result is taken as the negative correlation normalization result. This standard deviation is less than the extremely stable (rigid) threshold. When the standard deviation is greater than the instability threshold, the negative correlation normalization result is set to 1. If the negative correlation normalization result is set to 0, the final negative correlation normalization result will be used as the blood pressure stability of the blood pressure data interval.

[0042] In a specific example, the significant reduction characteristics of HRV in the HRV data interval are analyzed to determine the degree of significant reduction of the HRV data interval, including: determining the HRV difference value between every two adjacent HRV values ​​in the HRV data interval; averaging all HRV difference values ​​and performing positive correlation normalization to obtain the degree of significant reduction of the HRV data interval.

[0043] Specifically, for each HRV data interval, the absolute value of the difference between all two adjacent HRV values ​​is taken as the HRV difference value. The average of all HRV difference values ​​is calculated as the average HRV difference value, and then positive correlation normalization is performed. Positive correlation normalization means that the values ​​before and after normalization are positively correlated; that is, the larger the value before normalization, the larger the value after normalization. In other words, a lower limit for HRV difference is determined. (e.g., 5ms, extremely regular heart rate with almost no fluctuation) and an upper limit of HRV difference. (e.g., 50ms, severe heart rhythm irregularity), when the mean HRV difference is greater than or equal to the lower limit of HRV difference. And less than or equal to the upper limit of healthy HRV difference Then use the lower limit of HRV difference. Upper limit of difference from healthy HRV The mean HRV difference is subjected to maximum-min normalization, and the difference between the maximum-min normalization results is taken as the positive correlation normalization result. This mean HRV difference is less than the lower limit of HRV difference. When the mean HRV difference is greater than the upper limit of the healthy HRV difference, the positive correlation normalization result is set to 0. If the positive correlation normalization result is set to 1, the final positive correlation normalization result will be used as the degree of significant reduction in the HRV data interval.

[0044] Secondly, a full permutation and combination of heart rate spike rate, blood pressure stability, and significant reduction rate were performed to obtain several candidate combinations.

[0045] Specifically, based on the heart rate surge rate at each heart rate maxima, the blood pressure stability in each blood pressure data interval, and the significant reduction in each HRV data interval, a full permutation and combination of the heart rate surge rate, blood pressure stability, and significant reduction is performed to obtain several candidate combinations. Each candidate combination includes one heart rate surge rate, one blood pressure stability, and one significant reduction.

[0046] Next, based on the heart rate spike rate, blood pressure stability, and significant reduction rate in the candidate combinations, the compensatory linkage characteristics of the candidate combinations are determined.

[0047] Specifically, for candidate combinations whose heart rate spike rate is less than the set spike rate threshold, to avoid false positives when the heart rate does not spike, the compensatory linkage characteristic performance of the corresponding candidate combination is directly set to 0. The set spike rate threshold can be determined based on a physiological baseline and the statistical characteristics of abnormal fluctuations. For example, heart rate data segments are obtained from the patient in a non-traumatic state (such as previous physical examination data) or during a relatively stable phase in the early stages of trauma (such as the first 5-10 minutes after connecting to the monitor, before any strenuous activity). Based on the coefficient of variation of the heart rate in this baseline segment, multiple extreme points of heart rate spike rate are determined using the same method as described above. The maximum value among all extreme points is then determined, and the product of this maximum value and a safety factor (set empirically, such as 1.2) is used as the set spike rate threshold.

[0048] For any i-th candidate combination where the heart rate spike rate is greater than or equal to a set spike rate threshold, the compensatory linkage characteristic performance of the i-th candidate combination is calculated using the following formula based on the heart rate spike rate, blood pressure stability, and significant reduction in the i-th candidate combination: In the formula: This indicates the degree of compensatory linkage performance of the i-th candidate combination; This represents the heart rate spike rate of the i-th candidate combination; This represents the blood pressure stability of the i-th candidate combination; This indicates the degree of significant reduction for the i-th candidate combination.

[0049] In the above formula, based on the characteristic that blood pressure is affected first during bodily compensation, when the patient's heart rate spike rate is higher, more attention should be paid to the blood pressure stability within the blood pressure data interval of the candidate combination; conversely, based on the characteristic that HRV is affected by lag, when the patient's heart rate spike rate is lower, more attention should be paid to the significant reduction in the patient's HRV data interval within the candidate combination. Thus, by using the heart rate spike rate to fuse the blood pressure stability and significant reduction in the HRV data interval within the candidate combination, the compensatory linkage characteristic performance of the i-th candidate combination is determined. When the blood pressure stability and significant reduction are higher, the corresponding compensatory linkage characteristic performance is higher, which can be used to reflect the comprehensive stress state of the patient's physiological compensatory efficacy and the degree of neural exhaustion.

[0050] Finally, the maximum value among all compensatory linkage characteristic performance values ​​is determined. If the maximum value is greater than the set compensatory linkage characteristic performance value threshold, it is determined that the patient has a period of significant energy consumption compensation in real time. Based on the candidate combination corresponding to the maximum value, the heart rate maximum point, the blood pressure data interval, and the time corresponding to the blood pressure data interval, the period of significant energy consumption compensation is determined.

[0051] Specifically, the maximum value among all compensatory linkage characteristic manifestations is determined. To avoid misclassifying a normal state as a compensatory state, it is checked whether this maximum value is greater than a set threshold for the compensatory linkage characteristic manifestation. If it is greater than the set threshold, it indicates that the patient has entered a period of significant energy consumption compensation; otherwise, it indicates that the patient has not entered a period of significant energy consumption compensation. If the patient has not entered a period of significant energy consumption compensation, it means that the current patient is not exhibiting physiological compensation, and the manifestation of physiological metabolic side effects for that patient can be directly set to 0, without needing to perform the subsequent calculation process to determine the manifestation of physiological metabolic side effects for the real-time patient.

[0052] When a patient enters a period of significant energy consumption compensation, the earliest and latest time points among the candidate combinations corresponding to the maximum heart rate, blood pressure data intervals, and the corresponding times of the blood pressure data intervals are determined. The time period formed by these earliest and latest time points is defined as the period of significant energy consumption compensation. A compensation linkage characteristic performance threshold is set as a compensation initiation threshold. In the first few minutes after the patient is connected to the monitor (the stable period / baseline window), a series of compensation linkage characteristic performance values ​​are calculated using the same method as described above. The maximum value among these performance values ​​is determined, representing the "background noise" of the patient in a non-stress state. Then, an empirical multiple K (usually between 1.5 and 2.0) is set. This empirical multiple represents the degree of "significance." The product of this empirical multiple K and the maximum value is calculated, and this product is used as the threshold for setting the compensation linkage characteristic performance value.

[0053] Based on the above scheme, by capturing the heart rate surge rate corresponding to each maximum heart rate point in the heart rate data and constructing several HRV data intervals, and based on the heart rate surge rate, analyzing the blood pressure stability characteristics in the blood pressure data intervals and the significant HRV reduction characteristics in the HRV data intervals, it is possible to accurately identify whether a patient is experiencing a period of significant energy consumption compensation through the specific physiological combination of heart rate surge (trigger) + blood pressure stabilization (result) + HRV reduction (cost), thereby significantly reducing the missed diagnosis rate.

[0054] Step S300: If the real-time patient experiences a period of significant energy consumption compensation, analyze the side effects at the tissue metabolic level based on the multimodal blood gas analysis data of the real-time patient during the period of significant energy consumption compensation, and determine the manifestation of physiological metabolic side effects of the real-time patient.

[0055] If a patient experiences a period of significant energy depletion compensation, the body maintains central blood pressure through strong systemic vasoconstriction during this period. However, this compensatory mechanism incurs a significant "physiological cost" at the peripheral microcirculation level, namely, metabolic abnormalities caused by insufficient tissue perfusion. The specific pathological evolution is as follows: vasoconstriction leads to a sharp reduction in microcirculatory blood flow, resulting in insufficient oxygen supply to tissue cells to support mitochondrial aerobic metabolism. This forces the activation of the inefficient anaerobic glycolysis pathway for energy, thereby producing large amounts of acidic metabolites such as lactic acid. Subsequently, the body's buffer system (mainly the bicarbonate system) is consumed to neutralize the acidic substances, leading to significant changes in physiological and biochemical indicators.

[0056] Therefore, if a patient experiences a period of significant energy compensation in real time, this embodiment analyzes the side effects at the tissue metabolic level based on multimodal blood gas analysis data during this period to determine the manifestation of physiological metabolic side effects in real time. For example, by monitoring and analyzing lactate levels, which reflect the degree of anaerobic metabolism, and base excess (BE), which reflects the degree of buffer base consumption, the patient's side effects (tissue hypoxia and metabolic acidosis) can be analyzed to determine the manifestation of physiological metabolic side effects. This reflects the severity of latent metabolic debt exhibited in deep tissues under the compensatory state of maintaining stable surface vital signs.

[0057] In one possible implementation, multimodal blood gas analysis data includes lactate data and base excess data, such as Figure 3 As shown, step S300 analyzes tissue metabolic side effects based on multimodal blood gas analysis data of the patient during the period of significant energy consumption compensation in real time, and determines the manifestation of physiological metabolic side effects of the patient in real time, including: Step S301: Analyze the characteristics of the increase in lactate data of real-time patients during the period of significant energy consumption compensation, and determine the trend of lactate increase in real-time patients.

[0058] The characteristics of lactate increases in patients during periods of significant energy compensation were analyzed to determine the degree of lactate increase, which reflects the significance of the upward trend in lactate levels within the patient's body. For example, the higher the patient's current baseline lactate level (the more it exceeds the normal range) and the greater the net increase in lactate over time, the greater the degree of lactate increase, reflecting a higher degree of lactate deterioration in the patient's body.

[0059] In one possible implementation, step S301 analyzes the characteristics of the increase in lactate data of the real-time patient during the period of significant energy consumption compensation to determine the trend of lactate increase in the real-time patient, including: First, based on the differences between each lactate value and the maximum normal lactate value in the real-time lactate data of patients during the period of significant energy consumption compensation, as well as the differences between each lactate value and the next lactate value, the amount of lactate increase in each lactate value in the lactate data is determined.

[0060] Specifically, regarding the normal range for obtaining lactate data, in clinical medical testing, the normal reference range for arterial blood lactate is usually defined as 0.5-2.2 mmol / L. The maximum value within this normal range is taken as the normal maximum value for lactate. Calculate the lactate values ​​and their normal maximum values ​​in the lactate data during periods of significant energy consumption compensation. The ratio of lactate values ​​to the previous lactate value is calculated, and the difference between each lactate value and the previous lactate value is calculated. Based on this ratio and difference, the increase in lactate value for each lactate value in the lactate data is determined.

[0061] Secondly, all increases in lactate were summed and normalized with positive correlation to obtain the real-time trend of lactate increase in patients.

[0062] Specifically, the lactate increases corresponding to all lactate data during periods of significant energy compensation are summed and normalized with a positive correlation. The real-time lactate trend increase of patients is then obtained using the following formula: In the formula: This indicates the real-time trend of lactate elevation in patients; This represents the m-th lactate value in the lactate data during the period of significant energy consumption compensation. This represents the (m+1)th lactate value in the lactate data during a period of significant energy consumption compensation. This indicates the normal maximum value of lactic acid. This represents the total number of lactate values ​​in lactate data during periods of significant energy consumption compensation. Represents the corrected linear unit, when When, output ,when When, output ; This represents a normalization function, such as a max-min normalization function, used to sum the values ​​of each lactate increase. Normalize to the range [0,1].

[0063] In the above formula, This represents the increase in lactic acid at adjacent sampling times. When the increase in lactic acid is positive and larger, it indicates that the current rate of lactic acid deterioration is faster. This ratio represents the level of concern based on the current baseline lactate level. The higher the ratio, the closer the patient's current baseline lactate level is to or exceeds the critical threshold, and the smaller their physiological compensation space. By multiplying this level of concern by the increase in lactate, a non-linear amplification of disease deterioration in high-risk states is achieved: that is, the same increase in lactate has a smaller weight when it occurs at low lactate levels, but a much larger weight when it occurs at high lactate levels, thus enabling a more accurate assessment of the trend of lactate increase at different baselines.

[0064] Step S302: Analyze the characteristics of the decrease in base excess data of real-time patients during the period of significant energy consumption compensation, and determine the degree of decrease in base excess of real-time patients.

[0065] During periods of significant energy depletion compensation, if patients experience side effects, their base excess data will decrease significantly. This reflects both the severity of systemic metabolic acidosis and buffer base depletion, and its changes are often more sensitive than those of lactate. Therefore, analyzing the characteristics of the decrease in base excess data during periods of significant energy depletion compensation and determining the real-time rate of decrease in base excess is crucial to reflecting the significance of the decrease in base excess within the patient's body.

[0066] In one possible implementation, step S302 analyzes the characteristics of the decrease in base excess data of real-time patients during periods of significant energy compensation, and determines the degree of base excess decrease in real-time patients, including: First, the alkali surplus data during periods of significant energy consumption compensation is corrected using the standard alkali surplus method to obtain standard alkali surplus data.

[0067] Specifically, regarding base excess (ALE) data from patients, trauma patients often experience hyperventilation due to pain, anxiety, or brain injury, leading to a decrease in PaCO2 and causing respiratory alkalosis. This affects the true extent of the ALE decrease, making it appear relatively normal. Therefore, for ALE data during periods of significant energy compensation, the existing Standard Base Excess (SBE) method is used for correction. This corrected standard ALE data eliminates the influence of respiratory factors (such as PaCO2 reduction due to hyperventilation) on acid-base balance, thus reflecting the degree of metabolic acidosis directly. Since the process of correcting ALE data using the existing SBE method is prior art, it will not be elaborated here.

[0068] Secondly, based on the difference between each alkali surplus value in the standard alkali surplus data and its subsequent alkali surplus value, the decrease in alkali surplus for each alkali surplus value in the standard alkali surplus data is determined.

[0069] Specifically, the difference between each alkali surplus value in the standard alkali surplus data and the next alkali surplus value is calculated. If the difference is less than or equal to 0, the difference is set to 0; otherwise, the difference is taken as the actual difference result. This determines several difference result values ​​of the standard alkali surplus data, and these difference result values ​​are used as the alkali surplus reduction of each alkali surplus value in the standard alkali surplus data.

[0070] Finally, all base excess decreases were summed and normalized with positive correlation to obtain the real-time base excess decrease performance of patients.

[0071] Specifically, the decreases in all base excess data from the standard base excess data are summed and normalized with positive correlation. The real-time performance of the patient's base excess decrease is then obtained using the following formula: In the formula: This indicates the degree of decrease in base excess in patients in real time; This represents the kth alkali surplus value in the standard alkali surplus data during the period of significant energy consumption compensation. This represents the (k+1)th alkali surplus value in the standard alkali surplus data during the period of significant energy consumption compensation. This represents the total number of alkali residue values ​​in the standard alkali residue data during periods of significant energy consumption compensation. Represents the corrected linear unit, when When, output ,when When, output ; This represents a normalization function, such as a maximum-minimum normalization function, used to sum the values ​​of each decrease in alkali residue. Normalize to the range [0,1].

[0072] Step S303: Based on the trend of lactate increase and the degree of base excess decrease, determine the prominence of real-time changes in base excess over changes in lactate.

[0073] When the consumption (decrease) of base excess is more dramatic than the increase in lactate, it conforms to the characteristics of 'implicit liability' during the compensation period. Therefore, by analyzing the trend of lactate increase and the degree of base excess decrease, the prominence of real-time changes in base excess compared to changes in lactate can be determined, which reflects the degree of prominence of changes in base excess decrease compared to changes in lactate increase.

[0074] Specifically, calculate the upward trend of lactic acid. The sum of the values ​​of the parameter ϵ (which takes the smallest value greater than 0 to prevent the denominator from being 0) and the value of the sum of the values ​​of the parameter ϵ. And calculate the degree of performance of the decrease in alkali residue. Added value ratio And use a normalization function, such as a maximum-minimum normalization function, to normalize this ratio. Normalized to the range of [0,1], the normalized value is used as the salient representation of real-time changes in base excess compared to changes in lactate.

[0075] Step S304: Based on the prominence of changes in base excess compared to changes in lactate, a weighted correction is applied to the trend of lactate increase to obtain the real-time manifestation of physiological metabolic side effects in patients.

[0076] The tendency of increasing lactate levels is weighted and multiplied by the salientity of changes in base excess compared to changes in lactate levels, and the product is used as the real-time visibility of physiological metabolic side effects in patients. By using the salientity of changes in base excess compared to changes in lactate levels to weight the tendency of increasing lactate levels, the risk score can be prospectively increased even when a sharp decrease in base excess is detected before a significant increase in lactate is observed. This allows for early identification before lactate levels actually spike, and ensures a higher risk score when both lactate and base excess deteriorate simultaneously, thus achieving a reliable assessment of the visibility of physiological metabolic side effects in patients.

[0077] Based on the above scheme, by analyzing the characteristics of the increase in lactate data and the decrease in base excess data during the period of significant energy consumption compensation, the degree of increase in lactate and the degree of decrease in base excess are determined. Then, the degree of prominence of changes in base excess over changes in lactate is determined, and the degree of prominence is used to weight and correct the degree of increase in lactate, so as to obtain the real-time physiological metabolic side effects of patients. This can accurately reflect the true process of microcirculatory failure in patients and provide a highly discriminative core feature input for early warning of patient damage risks.

[0078] Step S400: Based on the manifestation of physiological metabolic side effects and real-time multimodal physiological monitoring data of patients during periods of significant energy consumption compensation, a risk prediction model is used to automatically generate real-time damage risk warning information for patients.

[0079] The visibility of physiological metabolic side effects and real-time multimodal physiological monitoring data of patients during periods of significant energy consumption compensation are used as inputs to a pre-acquired risk prediction model. This model automatically generates real-time warning information on the patient's injury risk for physicians' reference. The risk prediction model is a classification model, such as one composed of a neural network, used to classify the types of risks present in patients based on the visibility of their physiological metabolic side effects and multimodal physiological monitoring data during periods of significant energy consumption compensation.

[0080] In one possible implementation, the risk prediction model is composed of neural networks, such as... Figure 4As shown, step S400 automatically generates real-time patient injury risk warning information using a risk prediction model, including: Step S401: Obtain the physiological metabolic side effects manifestation of several historical patients and the multimodal physiological monitoring data of several historical patients during their periods of significant energy consumption compensation, and construct a training dataset.

[0081] Following the same method described above for obtaining the manifestation of physiological metabolic side effects in real-time patients and multimodal physiological monitoring data of real-time patients during periods of significant energy consumption compensation, a large number of historical patients' physiological metabolic side effect manifestations and several historical patients' multimodal physiological monitoring data during periods of significant energy consumption compensation were collected. A training dataset was then constructed using all the collected historical patients' physiological metabolic side effect manifestations and multimodal physiological monitoring data. Each training sample in the training dataset corresponds to the physiological metabolic side effect manifestations and multimodal physiological monitoring data of one historical patient.

[0082] Step S402: Label the training dataset with damage risk warning values, and use the training dataset and its damage risk warning value labels to train the neural network to obtain the trained neural network.

[0083] Each training sample in the training dataset is labeled with a damage risk warning value. This damage risk warning value is a numerical data point ranging from [0,1]. The larger the value of the label, the higher the risk level of the patient. For example, the damage risk warning value label for each training sample can be determined through a standardized manual assessment based on historical clinical outcome data of patients.

[0084] In a specific example, under authorization, a large amount of historical patient clinical outcome data is retrieved from a hospital's patient medical database. (Indicator of Massive Transfusion Protocol: Whether a mass transfusion protocol has been initiated; its value is 1, not 0.) (Indicator of Emergency Surgery: Whether an emergency damage control surgery was performed is set to 1, not 0.) (Indicator of Death: Whether there is an early death indication; a value of 1 indicates death, and a value of 0 indicates death.) ISS (Injury Severity Score: An internationally recognized trauma scoring standard, calculated based on the sum of squares of the simplified injury grades (AIS) of the three most severely injured areas out of six body regions (head and neck, face, chest, abdomen, limbs, and body surface), ranging from [1 to 75 points).) , , Assign reference weights to each item in the ISS based on their relative emphasis. Higher relative emphasis corresponds to higher reference weights. For example, setting... , , Reference weights of ISS , , and The values ​​were 0.4, 0.3, 0.2, and 0.1 respectively, and then the damage risk warning value label was calculated through weighted fusion. .in, This represents the function that takes the minimum value. This represents the maximum severity score of the injury (i.e., 75 points).

[0085] After labeling the damage risk warning value of each training sample in the training dataset, the training dataset is divided into a training set and a validation set in a 7:3 ratio. The neural network is trained using the training set, with the cross-entropy function used as the loss function during training. Gradient descent is used to train until the loss function converges. The robustness of the training results is verified using the validation set, and the trained neural network is obtained.

[0086] Step S403: Input the real-time physiological metabolic side effects manifestation of the patient and the multimodal physiological monitoring data of the patient during the period of significant energy consumption compensation into the trained neural network, and use the trained neural network to automatically generate the damage risk warning value of the real-time patient.

[0087] The real-time physiological metabolic side effects of the patient, as well as the multimodal physiological monitoring data of the patient during the period of significant energy consumption compensation, are input into the trained neural network. The trained neural network then automatically generates a warning value for the patient's damage risk in real time.

[0088] Step S404: Based on the value of the damage risk warning, automatically generate damage control and resuscitation decision-making suggestions.

[0089] Based on real-time injury risk warning values ​​automatically generated by neural networks, injury control and resuscitation decision-making suggestions are automatically generated. These suggestions, along with the injury risk warning values, constitute the patient's injury risk warning information. This information is sent to physicians for reference, enabling timely responses to injury control during the patient's compensatory period. For example, by statistically analyzing historical patient injury risk warning values, a first threshold (e.g., 0.4) and a second threshold (e.g., 0.8) can be determined. When the injury risk warning value is less than the first threshold (e.g., 0.4), it is recommended to maintain the current frequency of vital sign monitoring, adopt a restrictive fluid resuscitation strategy, and avoid excessive fluid infusion leading to coagulation dilution. When the injury risk warning value is greater than or equal to the first threshold (e.g., 0.4) and less than the second threshold (e.g., 0.8), it is recommended to immediately perform thromboelastography (TEG), notify the blood transfusion department to prepare blood (red blood cells: plasma 1:1), and prepare for tranexamic acid (TXA) administration. When the injury risk warning value is greater than or equal to the second threshold (e.g., 0.8), it is recommended to immediately initiate a massive transfusion protocol (MTP), activate the damage control surgery (DCS) team, and prepare for emergency hemostasis surgery or vascular embolization.

[0090] Based on the above technical solution, the patient's injury risk warning value is automatically generated by using a trained neural network, and injury control and resuscitation decision-making suggestions are automatically generated based on the value of the injury risk warning value, thereby ultimately realizing hierarchical decision support.

[0091] Based on the same inventive concept, embodiments of the present invention also provide an artificial intelligence-based injury control and resuscitation decision support system, such as... Figure 5 As shown, the system includes: a memory, a processor, and computer program code stored in the memory and running on the processor, wherein when the processor executes the computer program code, the system can execute any of the aforementioned artificial intelligence-based injury control and resuscitation decision support methods.

[0092] Based on the same inventive concept, embodiments of the present invention also provide a computer program product, which includes: computer program code, which, when run on a computer, causes the computer to execute any of the aforementioned artificial intelligence-based injury control and resuscitation decision support methods.

[0093] Based on the same inventive concept, embodiments of the present invention also provide a computer-readable storage medium storing computer program code, which, when executed on a computer, causes the computer to perform any of the aforementioned artificial intelligence-based injury control and resuscitation decision support methods.

[0094] It should be noted that the above embodiments are only used to explain the technical solution of the present invention and are not intended to limit it. Modifications or equivalent substitutions made by those skilled in the art to this solution, as long as they do not depart from the core technology of the present invention, are all within the protection scope of the present invention.

Claims

1. An artificial intelligence-based injury control and resuscitation decision support method, characterized in that, Includes the following steps: Acquire real-time multimodal physiological monitoring data of patients, wherein the multimodal physiological monitoring data includes at least multimodal vital sign data and multimodal blood gas analysis data; Analyze the temporal linkage characteristics of the multimodal vital signs data to determine whether the patient experiences a period of significant energy consumption compensation in real time; If a real-time patient experiences a period of significant energy consumption compensation, the side effects at the tissue metabolic level are analyzed based on the multimodal blood gas analysis data of the real-time patient during the period of significant energy consumption compensation, and the manifestation of physiological metabolic side effects of the real-time patient is determined. Based on the visibility of the physiological metabolic side effects and the real-time multimodal physiological monitoring data of the patient during the period of significant energy consumption compensation, a risk prediction model is used to automatically generate real-time damage risk warning information for the patient.

2. The injury control and resuscitation decision support method based on artificial intelligence according to claim 1, characterized in that, The multimodal vital signs data include at least heart rate data, blood pressure data, and electrocardiogram signal data; Analyzing the temporal correlation characteristics of the multimodal vital signs data to determine whether the patient experienced a period of significant energy depletion in real time includes: Determine the difference between each maximum heart rate point in the heart rate data and its preceding adjacent heart rate value, and determine the heart rate surge rate corresponding to each maximum heart rate point in the heart rate data; Using the time corresponding to each maximum heart rate point in the heart rate data as the marked time, several blood pressure data intervals are constructed based on the blood pressure values ​​of the blood pressure data at the marked time and subsequent times, and several HRV data intervals are constructed based on the HRV values ​​of the electrocardiogram signal data at the marked time and subsequent times. Based on the heart rate spike rate, and by analyzing the blood pressure stability characteristics in the blood pressure data interval and the HRV significant reduction characteristics in the HRV data interval, it is determined whether the patient experiences a period of significant energy consumption compensation in real time.

3. The injury control and resuscitation decision support method based on artificial intelligence according to claim 2, characterized in that, Based on the heart rate spike rate, and by analyzing the blood pressure stability characteristics within the blood pressure data interval and the significant HRV decrease characteristics within the HRV data interval, it is determined whether the patient experiences a period of significant energy compensation in real time, including: Analyze the blood pressure stability characteristics of the blood pressure data interval to determine the blood pressure stability of the blood pressure data interval, and analyze the HRV significant reduction characteristics of the HRV data interval to determine the significant reduction of the HRV data interval; The heart rate spike rate, blood pressure stability, and significant reduction rate are combined to obtain several candidate combinations; Based on the heart rate spike rate, blood pressure stability, and significant reduction rate of the candidate combinations, the compensatory linkage characteristic performance of the candidate combinations is determined. Determine the maximum value among all the compensatory linkage characteristic performance values. If the maximum value is greater than the set compensatory linkage characteristic performance value threshold, then determine that the real-time patient has a period of significant energy consumption compensation. Based on the candidate combination corresponding to the maximum value, the blood pressure data interval, and the time corresponding to the blood pressure data interval, determine the period of significant energy consumption compensation.

4. The injury control and resuscitation decision support method based on artificial intelligence according to claim 3, characterized in that, Analyze the blood pressure stability characteristics of the blood pressure data interval to determine the blood pressure stability of the blood pressure data interval, including: Determine the standard deviation of all blood pressure values ​​within the blood pressure data interval; The standard deviation is negatively correlated and normalized to obtain the blood pressure stability of the blood pressure data interval.

5. The injury control and resuscitation decision support method based on artificial intelligence according to claim 3, characterized in that, Analyze the significant decrease characteristics of HRV in the HRV data interval to determine the degree of significant decrease in the HRV data interval, including: Determine the HRV difference between every two adjacent HRV values ​​within the HRV data interval; The average of all the HRV differences is then normalized with positive correlation to obtain the degree of significant reduction in the HRV data interval.

6. The injury control and resuscitation decision support method based on artificial intelligence according to claim 1, characterized in that, The multimodal blood gas analysis data includes lactate data and base excess data; based on the multimodal blood gas analysis data of the real-time patient during the period of significant energy consumption compensation, the side effects at the tissue metabolic level are analyzed to determine the manifestation of physiological metabolic side effects of the real-time patient, including: The characteristics of the increase in lactate data of real-time patients during the period of significant energy consumption compensation were analyzed to determine the trend of lactate increase in real-time patients. The characteristics of the decrease in base excess data of real-time patients during the period of significant energy consumption compensation are analyzed to determine the degree of decrease in base excess of real-time patients. Based on the increasing trend of lactate and the decreasing performance of base excess, the prominence of changes in base excess over changes in lactate in real-time patients is determined. Based on the prominence of the change in base excess compared to the change in lactate, the trend of lactate increase is weighted and corrected to obtain the real-time manifestation of physiological metabolic side effects in patients.

7. The injury control and resuscitation decision support method based on artificial intelligence according to claim 6, characterized in that, The rise characteristics of lactate data in real-time patients during the period of significant energy consumption compensation are analyzed to determine the trend of lactate increase in real-time patients, including: Based on the differences between each lactate value and the maximum normal lactate value in the lactate data of the real-time patient during the period of significant energy consumption compensation, and the differences between each lactate value and the next lactate value, the amount of lactate increase in each lactate value in the lactate data is determined. The lactate increase of all the stated values ​​was summed and normalized with positive correlation to obtain the real-time lactate trend of the patient.

8. The injury control and resuscitation decision support method based on artificial intelligence according to claim 6, characterized in that, The characteristics of the decrease in base excess data of real-time patients during the period of significant energy consumption compensation are analyzed to determine the degree of base excess decrease in real-time patients, including: The standard alkali surplus data during the period of significant energy consumption compensation is corrected using the standard alkali surplus method to obtain standard alkali surplus data; Based on the difference between each alkali surplus value in the standard alkali surplus data and its next alkali surplus value, the decrease in alkali surplus for each alkali surplus value in the standard alkali surplus data is determined. All the base excess decreases were summed and positively correlated and normalized to obtain the real-time base excess decrease performance of the patients.

9. The injury control and resuscitation decision support method based on artificial intelligence according to claim 1, characterized in that, The risk prediction model is composed of a neural network, which automatically generates real-time early warning information on the patient's injury risk, including: A training dataset was constructed by acquiring the physiological metabolic side effects of several historical patients and the multimodal physiological monitoring data of these patients during periods of significant energy consumption compensation. Label the training dataset with damage risk warning values, and use the training dataset and its damage risk warning value labels to train the neural network to obtain the trained neural network. The physiological metabolic side effects of the real-time patient and the multimodal physiological monitoring data of the real-time patient during the period of significant energy consumption compensation are input into the trained neural network, and the damage risk warning value of the real-time patient is automatically generated using the trained neural network. Based on the value of the damage risk warning, damage control and resuscitation decision-making suggestions are automatically generated.

10. An artificial intelligence-based injury control and resuscitation decision support system, characterized in that, The device includes a memory, a processor, and executable computer program code stored in the memory and executable on the processor. When the processor executes the computer program code, it performs an artificial intelligence-based injury control and resuscitation decision support method as described in any one of claims 1 to 9.

Images