Intelligent hemostatic device self-adaptive control method and system based on pressure feedback

Abstract

The invention provides an intelligent hemostasis device self-adaptive control method and system based on pressure feedback, and relates to the technical field of self-adaptive control, and the method comprises the steps: collecting the real-time pressure and physiological parameters of a contact interface, constructing a pressure-physiological coupling response model to determine a safe pressure interval, and generating an initial pressure application scheme of partition cooperation according to the safe pressure interval. And after an optimal target scheme is screened out through virtual rehearsal simulation, a pressure applying unit is driven to execute, and meanwhile, model parameters are updated in real time based on execution feedback. According to the method, the hemostasis efficiency is accurately improved, and the tissue damage risk is effectively reduced.

Description

Technical Field

[0001] This invention relates to adaptive control technology, and more particularly to an adaptive control method and system for an intelligent hemostasis device based on pressure feedback. Background Technology

[0002] In the fields of trauma emergency care and surgery, effective bleeding control is a crucial aspect of saving lives. Traditional hemostasis methods typically rely on manual pressure, tourniquets, or pressure-controlled devices. These methods generally employ a static or empirical pressure application mode, where the operator applies a preset or empirically determined constant pressure based on the type and location of the injury. For devices such as tourniquets, the pressure setting is often based on population average data and remains fixed during use. The core logic of these methods is to close the severed blood vessel ends through continuous external mechanical compression. The technology is relatively simple to implement, and long-term clinical practice has demonstrated its basic effectiveness.

[0003] However, the aforementioned conventional methods have significant limitations: static pressure modes cannot adapt to the dynamically changing physiological state of the wound site. The hemodynamic state, tissue edema, and local metabolic environment of human tissues are constantly changing at different times. Fixed pressure is insufficient for effective hemostasis at some moments, while at others it can lead to distal tissue ischemia, nerve damage, or pressure necrosis due to excessive pressure. Current technologies lack a real-time, quantitative assessment and feedback mechanism to correlate the effect of pressure application with physiological state. Operators cannot accurately know whether the applied pressure is within the "safe window" that effectively stops hemostasis while minimizing tissue damage, nor can they make fine adjustments based on individual differences and real-time physiological responses. This unidirectional, open-loop control method creates an irreconcilable contradiction between safety and effectiveness in the hemostasis process, relies excessively on operator experience, and lacks overall control precision and adaptability. Summary of the Invention

[0004] The present invention provides an adaptive control method and system for an intelligent hemostasis device based on pressure feedback, which can solve the problems in the prior art.

[0005] A first aspect of the present invention provides an adaptive control method for an intelligent hemostasis device based on pressure feedback, comprising: Real-time pressure distribution data and physiological characteristics of the interface between the hemostatic device and the wound site were collected. Based on the real-time pressure distribution data and the physiological state characteristic parameters, a pressure-physiology coupled response model is constructed. By extracting the dynamic correlation features between the pressure change rate and the physiological response delay, a bidirectional prediction function of pressure regulation behavior on hemostatic efficacy and tissue damage is established, and the safe pressure application range under the current physiological state is output. Based on the safe pressure application range, the multiple pressure application units of the hemostasis device are driven and allocated differently. The upper and lower boundary values ​​of the safe pressure application range are used as the constraint conditions of each pressure application unit to generate a partitioned and coordinated initial pressure application scheme. The initial pressure application scheme was simulated virtually. By simulating the pressure transmission path and the physiological response chain evolution process, the hemostasis timeliness score and safety margin score of each candidate scheme were calculated, and the target pressure application scheme with the dual-score weighted comprehensive optimality was selected. According to the target pressure application scheme, each pressure application unit is driven to perform pressure regulation and the pressure deviation and physiological state drift are collected in real time after execution. The pressure deviation and physiological state drift are fed back to the pressure-physiology coupled response model to update the parameters of the dynamic correlation feature and the bidirectional prediction function.

[0006] Based on the real-time pressure distribution data and the physiological state characteristic parameters, a pressure-physiological coupled response model is constructed. By extracting the dynamic correlation features between the pressure change rate and the physiological response delay, a bidirectional prediction function of pressure regulation behavior on hemostatic efficacy and tissue damage is established, and the safe pressure application range under the current physiological state is output, including: The real-time pressure distribution data is processed by time-series difference processing to obtain the pressure change rate time series, and the physiological response state transition time markers in the physiological state characteristic parameters are extracted simultaneously. The time series of pressure change rate is correlated with the time markers of physiological response state transition. The time interval between the peak pressure application time and the physiological response state transition time is calculated, and dynamic correlation features characterizing the pressure-physiological conduction delay are extracted. A pressure-physiology coupled response model is constructed based on the dynamic correlation features, and the time interval in the dynamic correlation features is used as the time delay parameter of the model to establish a bidirectional prediction function. The minimum pressure threshold required to achieve the preset hemostatic effect is calculated by the positive prediction function branch of the bidirectional prediction function, and the maximum pressure threshold corresponding to the risk of triggering tissue damage is calculated by the negative prediction function branch of the bidirectional prediction function. The interval between the minimum pressure threshold and the maximum pressure threshold is determined as the safe pressure application interval.

[0007] The time series of pressure change rate is correlated with the time markers of physiological response state transitions. The time interval between the peak pressure application time and the physiological response state transition time is calculated, and dynamic correlation features characterizing the pressure-physiological conduction delay are extracted, including: Peak identification processing is performed on the pressure change rate time series to detect the moment when the pressure change rate reaches a local maximum in the pressure change rate time series, and the detected moment is determined as the peak moment of pressure application. The peak time of pressure application and the time of physiological response state transition are time-aligned on a unified time axis. For the peak time of pressure application, the physiological response state transition time marker with the closest timestamp after the peak time of pressure application is retrieved from the physiological response state transition time markers. A causal pairing relationship is established between the peak time of pressure application and the physiological response state transition time markers. Based on the causal pairing relationship, the time difference between the peak time of pressure application and the time marker of the paired physiological response state transition is calculated. The time difference is used as the time interval of a single pressure-physiological conduction delay. Multiple sets of time intervals corresponding to the causal pairing relationship are collected to form a time interval dataset. Statistical analysis is performed on the time interval dataset to calculate the central tendency measure and the dispersion measure of the time interval dataset, and the central tendency measure and the dispersion measure are combined to form dynamic correlation features.

[0008] Based on the safe pressure application range, the multiple pressure application units of the hemostatic device are differentially driven and allocated. The upper and lower boundary values ​​of the safe pressure application range are used as constraints for each pressure application unit, generating a zoned and coordinated initial pressure application scheme, including: Obtain hemostasis demand metric values ​​for the wound contact areas corresponding to multiple pressure units of the hemostasis device, and sort the multiple pressure units by urgency based on the hemostasis demand metric values ​​to obtain the urgency sequence number of each pressure unit. Based on the urgency number, a linear mapping is performed within the safe pressure application range. The pressure application unit with the first urgency number is mapped to the upper boundary value of the safe pressure application range, and the pressure application unit with the last urgency number is mapped to the lower boundary value of the safe pressure application range. The remaining pressure application units are proportionally interpolated between the upper boundary value and the lower boundary value according to the urgency number to form the initial pressure distribution value of each pressure application unit. The initial pressure distribution value is used as the pressure reference value for each pressure unit. The spatial location information of each pressure unit is obtained, and the difference between the spatial distance between adjacent pressure units and the pressure reference value is calculated. Based on the difference between the spatial distance and the pressure reference value, a pressure transmission influence coefficient between the pressure application units is constructed. The pressure reference value of each pressure application unit is then collaboratively corrected in the neighborhood using the pressure transmission influence coefficient to obtain the collaboratively corrected pressure value of each pressure application unit. The collaborative correction pressure value and the upper and lower boundary values ​​of the safe pressure application range are used for boundary constraint verification. The verified collaborative correction pressure value is used as the target execution pressure value of each pressure application unit. The target execution pressure value and the corresponding pressure application unit identifier are combined to form an initial pressure application scheme.

[0009] Based on the difference between the spatial distance and the pressure reference value, a pressure transmission influence coefficient is constructed between the pressure application units. This pressure transmission influence coefficient is then used to perform neighborhood-based collaborative correction of the pressure reference value for each pressure application unit, resulting in the following collaboratively corrected pressure values ​​for each unit: For a target pressure unit among multiple pressure units, obtain the neighboring pressure units that are spatially adjacent to the target pressure unit, calculate the spatial distance between the target pressure unit and each neighboring pressure unit, and simultaneously calculate the pressure reference value difference between the target pressure unit and each neighboring pressure unit. The spatial distance value and the pressure reference value difference are coupled and calculated. The pressure transmission influence coefficient of each neighboring pressure unit on the target pressure unit is obtained by multiplying the pressure reference value difference with the reciprocal of the spatial distance value. The pressure transmission influence coefficient is normalized so that the sum of the pressure transmission influence coefficients corresponding to all neighboring pressure units is a preset unit value, thus obtaining the normalized pressure transmission influence coefficient. The normalized pressure transmission influence coefficient is used as a weighting coefficient to sum the pressure reference values ​​of each neighboring pressure unit to obtain the neighborhood pressure weighted average value. The deviation between the neighborhood pressure weighted average value and the pressure reference value of the target pressure unit is then calculated. The deviation value is scaled according to a preset correction ratio to obtain the pressure correction increment. The pressure reference value of the target pressure unit is superimposed with the pressure correction increment to obtain the cooperative correction pressure value of the target pressure unit.

[0010] The initial pressure application scheme was simulated virtually. By simulating the pressure transmission path and the physiological response chain evolution process, the hemostasis timeliness score and safety margin score of each candidate scheme were calculated. The target pressure application scheme with the best weighted comprehensive score was selected, including: Obtain the target execution pressure value of each pressure application unit in the initial pressure application scheme, and construct a simulation environment model that includes the elastic modulus distribution of traumatic tissue and the vascular network topology. In the simulation environment model, the stress tensor propagation process of pressure is calculated based on the tissue elastic modulus distribution, the diffusion trajectory of pressure from each pressure-applying unit to the deep tissue is tracked, and pressure field evolution data containing the conduction time series and spatial distribution gradient are generated. The pressure field evolution data is spatially mapped to the vascular network topology. The degree of vascular closure is calculated based on the pressure value at each vascular node. The dynamic process of blood flow occlusion is constructed based on the time change trend of the degree of vascular closure, and a physiological response chain evolution sequence is obtained. By analyzing the time evolution curve of the closure state of vascular nodes in the physiological response chain evolution sequence, the critical moment when all vascular nodes reach the complete closure state is located. The time deviation between the preset hemostasis time target value and the critical moment is calculated as the hemostasis timeliness score. By monitoring the stress accumulation process of tissue units in the physiological response chain evolution sequence, the stress peak breakthrough point is captured. The safety redundancy between the preset safe stress threshold and the stress peak breakthrough point is calculated as the safety margin score. After assigning preset weight coefficients to the hemostasis timeliness score and the safety margin score respectively, the comprehensive score is obtained by summing them. The simulation process is executed on multiple candidate initial pressure application schemes, and the candidate scheme with the largest comprehensive score is selected as the target pressure application scheme.

[0011] In the simulation environment model, the stress tensor propagation process of pressure is calculated based on the tissue elastic modulus distribution. The diffusion trajectory of pressure from each pressure-applying unit to deep tissues is tracked, generating pressure field evolution data including conduction time series and spatial distribution gradients, including: Acquire the spatial mesh division data of the traumatic tissue in the simulation environment model, and map the target stress value of each stress unit to the corresponding mesh unit as the initial stress state; Based on the initial stress state and the elastic modulus of each grid cell, a stress balance equation between grid cells is established. By solving the stress balance equation, the stress transfer relationship between adjacent grid cells is calculated. The stress tensor evolution value of each grid cell at continuous time steps is iteratively calculated to form the spatiotemporal evolution matrix of the stress tensor. The pressure components of each grid cell at each time step are extracted from the stress tensor spatiotemporal evolution matrix. The pressure components are then statistically analyzed according to the depth coordinates of the grid cells. The attenuation ratio of pressure from the surface layer of the pressure-applying cell to each layer of the deep tissue is calculated, and a pressure depth penetration curve is generated. The pressure values ​​of each depth layer in the pressure depth penetration curve are correlated with the lateral spatial coordinates of the corresponding grid cells in the layer. The pressure difference distribution at different lateral positions within the same depth layer is calculated, and a two-dimensional pressure diffusion trajectory containing both depth and lateral dimensions is constructed. The pressure change rate between adjacent grid cells in the two-dimensional pressure diffusion trajectory is calculated, and the pressure change rate is used as the spatial distribution gradient value. The spatial distribution gradient value is bound to the time step information in the spatiotemporal evolution matrix of the stress tensor to generate pressure field evolution data containing the conduction time series and spatial distribution gradient.

[0012] A second aspect of the present invention provides an adaptive control system for an intelligent hemostasis device based on pressure feedback, comprising: The data acquisition unit is used to collect real-time pressure distribution data and physiological characteristic parameters at the interface between the hemostatic device and the wound site. The model building unit is used to construct a pressure-physiological coupled response model based on the real-time pressure distribution data and the physiological state characteristic parameters. By extracting the dynamic correlation features between the pressure change rate and the physiological response delay, it establishes a bidirectional prediction function of pressure regulation behavior on hemostatic efficacy and tissue damage, and outputs the safe pressure application range under the current physiological state. The scheme generation unit is used to perform differentiated drive allocation of multiple pressure application units of the hemostasis device according to the safe pressure application range, and use the upper and lower boundary values ​​of the safe pressure application range as the constraint conditions of each pressure application unit to generate an initial pressure application scheme of partitioned coordination. The simulation optimization unit is used to perform virtual pre-simulation of the initial pressure application scheme. By simulating the evolution process of pressure transmission path and physiological response chain, it calculates the hemostasis timeliness score and safety margin score of each candidate scheme and selects the target pressure application scheme with dual-score weighted comprehensive optimality. The feedback update unit is used to drive each pressure application unit to perform pressure regulation according to the target pressure application scheme and collect the pressure deviation and physiological state drift after execution in real time. The pressure deviation and physiological state drift are then fed back to the pressure-physiology coupled response model to update the parameters of the dynamic correlation feature and the bidirectional prediction function.

[0013] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0014] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0015] The beneficial effects of this application are as follows: This method enables real-time and accurate monitoring of pressure distribution at the wound site contact interface, and simultaneously acquires physiological characteristic parameters reflecting the body's state, providing a comprehensive and dynamic data foundation for subsequent intelligent regulation. The constructed pressure-physiological coupled response model can deeply analyze the intrinsic dynamic relationship between the rate of pressure change and the lag in the physiological system's response, thereby establishing a two-way predictive mechanism for pressure regulation behavior on hemostasis and tissue safety. It accurately calculates the safe pressure operating range that conforms to the current physiological state, effectively avoiding hemostasis failure due to insufficient pressure or secondary tissue damage caused by excessive pressure.

[0016] By differentiating the driving allocation of each pressure application unit according to the safe pressure range, the theoretical safety boundary is transformed into a specific constraint on each execution unit, forming an initial pressure application strategy that takes into account both the overall situation and the local situation and emphasizes synergy. Through virtual pre-simulation of the initial scheme, the pressure transmission process in biological tissue and the series of physiological responses it triggers can be simulated. This allows for the quantitative assessment of the performance of different schemes in terms of hemostasis speed and operational safety in advance. Based on a comprehensive score, the optimal target scheme is selected, significantly improving the scientific rigor and reliability of the decision-making process. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the adaptive control method of the intelligent hemostasis device based on pressure feedback according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating the generation process of pressure field evolution data based on tissue elastic modulus in an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0020] Figure 1 This is a flowchart illustrating the adaptive control method of the intelligent hemostasis device based on pressure feedback according to an embodiment of the present invention. Figure 1 As shown, the method includes: Real-time pressure distribution data and physiological characteristics of the interface between the hemostatic device and the wound site were collected. Based on the real-time pressure distribution data and the physiological state characteristic parameters, a pressure-physiology coupled response model is constructed. By extracting the dynamic correlation features between the pressure change rate and the physiological response delay, a bidirectional prediction function of pressure regulation behavior on hemostatic efficacy and tissue damage is established, and the safe pressure application range under the current physiological state is output. Based on the safe pressure application range, the multiple pressure application units of the hemostasis device are driven and allocated differently. The upper and lower boundary values ​​of the safe pressure application range are used as the constraint conditions of each pressure application unit to generate a partitioned and coordinated initial pressure application scheme. The initial pressure application scheme was simulated virtually. By simulating the pressure transmission path and the physiological response chain evolution process, the hemostasis timeliness score and safety margin score of each candidate scheme were calculated, and the target pressure application scheme with the dual-score weighted comprehensive optimality was selected. According to the target pressure application scheme, each pressure application unit is driven to perform pressure regulation and the pressure deviation and physiological state drift are collected in real time after execution. The pressure deviation and physiological state drift are fed back to the pressure-physiology coupled response model to update the parameters of the dynamic correlation feature and the bidirectional prediction function.

[0021] In one optional implementation, a pressure-physiological coupled response model is constructed based on the real-time pressure distribution data and the physiological state characteristic parameters. By extracting the dynamic correlation features between the pressure change rate and the physiological response delay, a bidirectional prediction function of pressure regulation behavior on hemostatic efficacy and tissue damage is established, and the safe pressure application range under the current physiological state is output, including: The real-time pressure distribution data is processed by time-series difference processing to obtain the pressure change rate time series, and the physiological response state transition time markers in the physiological state characteristic parameters are extracted simultaneously. The time series of pressure change rate is correlated with the time markers of physiological response state transition. The time interval between the peak pressure application time and the physiological response state transition time is calculated, and dynamic correlation features characterizing the pressure-physiological conduction delay are extracted. A pressure-physiology coupled response model is constructed based on the dynamic correlation features, and the time interval in the dynamic correlation features is used as the time delay parameter of the model to establish a bidirectional prediction function. The minimum pressure threshold required to achieve the preset hemostatic effect is calculated by the positive prediction function branch of the bidirectional prediction function, and the maximum pressure threshold corresponding to the risk of triggering tissue damage is calculated by the negative prediction function branch of the bidirectional prediction function. The interval between the minimum pressure threshold and the maximum pressure threshold is determined as the safe pressure application interval.

[0022] In constructing the pressure-physiological coupled response model, real-time pressure distribution data collected from the contact interface of the hemostatic device underwent time-series differential processing. Specifically, the pressure change rate time series was calculated by subtracting pressure values ​​from adjacent time points and dividing by the sampling time interval. The calculation time window for the pressure change rate was set to 0.1 to 0.5 seconds to ensure that rapid pressure fluctuations were captured while avoiding noise interference. Simultaneously, physiological response state transition time markers were extracted from physiological state characteristic parameters. These markers were obtained by monitoring abrupt changes in physiological indicators such as blood oxygen saturation, local tissue temperature, and microcirculatory resistance. When the change in any of these physiological indicators exceeded 15% of its baseline value, that moment was recorded as the physiological response state transition time.

[0023] In time-axis correlation analysis, the peak times in the pressure change rate time series are paired with the times of physiological response state transitions. A sliding time window method is used to search for the correspondence between these two times, calculating the time interval from the peak pressure application time to the corresponding physiological response state transition time. This time interval reflects the conduction delay characteristics of pressure stimulation in biological tissues. Depending on the injury site and individual differences, the time interval typically ranges from 0.3 seconds to 2 seconds. By collecting time interval data from multiple sets of pressure application events and physiological responses, their statistical distribution characteristics, mean, and variance are extracted as dynamic correlation features characterizing the pressure-physiological conduction delay pattern.

[0024] When constructing the pressure-physiological coupled response model, a time-delay differential equation is used as the model framework, and the extracted mean time interval is used as the time-delay parameter of the model. The establishment form is The coupling equation is defined as follows: P(t) represents the pressure state variable, and S(t) represents the physiological state variable. The bidirectional prediction function contains two computational paths: a positive branch and a negative branch. The positive prediction function branch is used to evaluate the hemostatic efficacy under a given pressure input. By establishing a mapping relationship between pressure intensity and the rate of reduction in bleeding, it calculates the minimum pressure threshold required to achieve the preset hemostatic efficacy. The preset hemostatic efficacy is generally set as a reduction in bleeding to less than 10% of the initial state. The negative prediction function branch is used to evaluate the risk of tissue damage caused by pressure application. By establishing a relationship between pressure duration, pressure intensity, and the degree of tissue hypoxia, it calculates the maximum pressure threshold corresponding to the decrease in tissue oxygen partial pressure to a critical value.

[0025] The safe pressure application range is defined by the minimum pressure threshold output from the positive prediction function branch and the maximum pressure threshold output from the negative prediction function branch. This range must simultaneously satisfy two boundary conditions: the lower boundary ensures adequate hemostasis, and the upper boundary ensures manageable tissue damage risk. For forearm trauma in adult patients, the typical safe pressure application range is 120 mmHg to 180 mmHg. The model also incorporates a dynamic adjustment mechanism, which adjusts the time lag parameters and prediction function coefficients online based on real-time physiological state drift data, enabling the safe pressure application range to adapt to individual physiological response differences and time-varying characteristics.

[0026] In one optional implementation, a time-axis correlation analysis is performed between the pressure change rate time series and the physiological response state transition time markers to calculate the time interval between the pressure application peak time and the physiological response state transition time, and to extract dynamic correlation features characterizing the pressure-physiological conduction delay law, including: Peak identification processing is performed on the pressure change rate time series to detect the moment when the pressure change rate reaches a local maximum in the pressure change rate time series, and the detected moment is determined as the peak moment of pressure application. The peak time of pressure application and the time of physiological response state transition are time-aligned on a unified time axis. For the peak time of pressure application, the physiological response state transition time marker with the closest timestamp after the peak time of pressure application is retrieved from the physiological response state transition time markers. A causal pairing relationship is established between the peak time of pressure application and the physiological response state transition time markers. Based on the causal pairing relationship, the time difference between the peak time of pressure application and the time marker of the paired physiological response state transition is calculated. The time difference is used as the time interval of a single pressure-physiological conduction delay. Multiple sets of time intervals corresponding to the causal pairing relationship are collected to form a time interval dataset. Statistical analysis is performed on the time interval dataset to calculate the central tendency measure and the dispersion measure of the time interval dataset, and the central tendency measure and the dispersion measure are combined to form dynamic correlation features.

[0027] In the correlation analysis between the pressure change rate time series and the time markers of physiological response state transitions, peak identification processing was first performed on the collected pressure change rate time series. This processing employed a sliding window algorithm, setting the time window width to 0.5 to 2 seconds, and determining local extrema of the pressure change rate values ​​within the window. When the pressure change rate at a certain moment was detected to be greater than the values ​​of the three sampling points before and after it simultaneously, and this value exceeded 1.5 times the standard deviation of the overall series, that moment was marked as the pressure application peak moment. To avoid noise interference, the identified candidate peak moments were further verified by calculating the integral value of the pressure change rate within 200 milliseconds before and after the peak point. When the integral value exceeded a preset threshold of 15 kPa·s, the peak was confirmed as valid.

[0028] After peak identification, all pressure application peak times and physiological response state transition times are projected onto a unified time axis. The time axis uses the device startup time as zero and records the timestamps of each event with millisecond-level precision. For each pressure application peak time, a time series retrieval is performed in the set of physiological response state transition time markers. The retrieval strategy is to traverse all physiological response markers, filter out candidate markers with timestamps greater than the current pressure peak time, and select the marker with the smallest timestamp difference as the pairing object. When multiple physiological response markers with the same time interval exist after the same pressure peak time, the marker with the higher physiological response intensity value is preferentially selected to establish a causal pairing relationship. This pairing mechanism ensures that a clear causal chain is established between pressure stimulation and physiological response.

[0029] Based on the established causal pairings, the time difference for each pair is calculated. This time difference is obtained by subtracting the peak pressure application time from the time of physiological response transition in the paired pair. This difference reflects the delay in the transmission of pressure stimulus through tissues to trigger the physiological state transition. During the continuous monitoring period, the system continuously collects these pairings, accumulating them into a time interval dataset. To ensure data quality, outliers are removed from the time interval dataset. Data intervals exceeding three times the median or less than the physical conduction limit of 0.1 seconds are considered invalid and deleted.

[0030] Statistical analysis was performed on the cleaned time interval dataset to calculate measures of central tendency, including the mean and median. The mean reflects the overall level of delay time, while the median characterizes typical delay characteristics and is insensitive to extreme values. Measures of dispersion, including the standard deviation and interquartile range, were also calculated. The standard deviation quantifies the fluctuation range of delay time, and the interquartile range describes the robust dispersion characteristics of the data distribution. The calculated mean, median, standard deviation, and interquartile range were combined in a fixed order to form a four-dimensional dynamic correlation feature vector. This feature vector fully characterizes the delay pattern of pressure-physiological transmission under the current traumatic state, where the measure of central tendency indicates the baseline time of the delayed response, and the measure of dispersion reveals the stability of the physiological response. When the standard deviation is less than 20% of the mean, it indicates that the pressure transmission path is stable, and a more aggressive pressure control strategy can be adopted; when the standard deviation exceeds 50% of the mean, it suggests a high degree of uncertainty in the physiological response, requiring an expansion of the margin of the safe pressure range to ensure a balance between hemostatic efficacy and tissue safety even under fluctuating delay characteristics.

[0031] In one optional implementation, the multiple pressure application units of the hemostatic device are differentially driven and allocated according to the safe pressure application range, and the upper and lower boundary values ​​of the safe pressure application range are used as constraints for each pressure application unit to generate a zoned and coordinated initial pressure application scheme, including: Obtain hemostasis demand metric values ​​for the wound contact areas corresponding to multiple pressure units of the hemostasis device, and sort the multiple pressure units by urgency based on the hemostasis demand metric values ​​to obtain the urgency sequence number of each pressure unit. Based on the urgency number, a linear mapping is performed within the safe pressure application range. The pressure application unit with the first urgency number is mapped to the upper boundary value of the safe pressure application range, and the pressure application unit with the last urgency number is mapped to the lower boundary value of the safe pressure application range. The remaining pressure application units are proportionally interpolated between the upper boundary value and the lower boundary value according to the urgency number to form the initial pressure distribution value of each pressure application unit. The initial pressure distribution value is used as the pressure reference value for each pressure unit. The spatial location information of each pressure unit is obtained, and the difference between the spatial distance between adjacent pressure units and the pressure reference value is calculated. Based on the difference between the spatial distance and the pressure reference value, a pressure transmission influence coefficient between the pressure application units is constructed. The pressure reference value of each pressure application unit is then collaboratively corrected in the neighborhood using the pressure transmission influence coefficient to obtain the collaboratively corrected pressure value of each pressure application unit. The collaborative correction pressure value and the upper and lower boundary values ​​of the safe pressure application range are used for boundary constraint verification. The verified collaborative correction pressure value is used as the target execution pressure value of each pressure application unit. The target execution pressure value and the corresponding pressure application unit identifier are combined to form an initial pressure application scheme.

[0032] The hemostasis demand metric values ​​for the wound contact areas corresponding to multiple pressure application units of the hemostasis device are obtained. These hemostasis demand metric values ​​are quantified by comprehensively considering the bleeding rate, vascular damage degree, and tissue type characteristics of the wound area. Specifically, pressure sensors and optical blood flow sensors installed at the bottom of each pressure application unit acquire dynamic microscopic blood flow information of the contact area. Combined with temperature gradient distribution data collected by an infrared thermal imaging module, a multimodal fusion algorithm is used to calculate the hemostasis demand metric value for each area. The hemostasis demand metric values ​​corresponding to each pressure application unit are sorted in descending order of numerical value and assigned an urgency number corresponding to the sorting position. Number 1 indicates the most urgent area, and the increasing numbers indicate decreasing urgency.

[0033] When performing linear mapping allocation within the safe pressure application range, assume the safe pressure application range is from P_min to P_max, and the total number of pressure application units is N. Pressure application units with urgency index 1 are directly mapped to the upper boundary value P_max, and pressure application units with urgency index N are mapped to the lower boundary value P_min. For pressure application units with urgency index i (where...) For each pressure application unit, the initial pressure distribution value is determined proportionally. The calculation formula is: the initial pressure of the pressure application unit equals the lower boundary value plus the difference between the upper and lower boundary values ​​multiplied by the sequence number normalization factor. This factor is the maximum sequence number minus the current sequence number and then divided by the maximum sequence number minus 1. This mapping method ensures that emergency areas receive higher pressure values ​​while all distribution values ​​fall within the safe range.

[0034] The spatial location information of each pressure application unit on the wound surface is obtained, and the two-dimensional coordinates of each unit are recorded through a built-in position encoding module. The Euclidean distance between adjacent pressure application units is calculated as a spatial distance metric, and the pressure reference value difference between these adjacent units is extracted. When constructing the pressure transmission influence coefficient, this coefficient is negatively correlated with the spatial distance; the closer the distance, the stronger the pressure transmission influence. In the specific implementation, the influence coefficient adopts a Gaussian decay function, with spatial distance as the independent variable, and the decay rate is determined by the elastic modulus and pressure diffusion characteristics of the wound tissue. When the pressure reference value difference between adjacent units is too large and the spatial distance is small, a large pressure gradient will be generated, which will lead to local tissue stress concentration.

[0035] The pressure reference value of each pressure-applying unit is adjusted using a neighborhood collaborative correction based on the pressure transmission influence coefficient. For a target pressure-applying unit, all its neighboring units are traversed, and the pressure reference value of each neighboring unit is multiplied by its corresponding pressure transmission influence coefficient. The influence contribution values ​​of all neighboring units are accumulated, and then weighted and fused with the target unit's own pressure reference value. The weighting coefficient is dynamically adjusted according to the urgency index of the target unit. Units with high urgency retain a larger proportion of the original pressure reference value, while units with low urgency adopt more neighborhood collaborative correction, thereby balancing hemostatic efficacy and pressure distribution uniformity.

[0036] Boundary constraint verification is performed on the collaboratively corrected pressure values, checking whether the collaboratively corrected pressure value of each pressure application unit exceeds the safe pressure application range. If the corrected pressure value of a unit is greater than the upper boundary value, it is truncated to the upper boundary value; if it is less than the lower boundary value, it is raised to the lower boundary value. After verification, the final target execution pressure value of each pressure application unit is combined with the corresponding physical identifier of the pressure application unit to form structured data containing the pressure application unit number, target execution pressure value, and execution timing mark. This data is the initial pressure application scheme, which can be directly used for instruction parsing and execution of the drive control module.

[0037] In one optional implementation, a pressure transmission influence coefficient between the pressure application units is constructed based on the difference between the spatial distance and the pressure reference value. This pressure transmission influence coefficient is then used to perform neighborhood-based collaborative correction of the pressure reference value for each pressure application unit, resulting in a collaboratively corrected pressure value for each unit, including: For a target pressure unit among multiple pressure units, obtain the neighboring pressure units that are spatially adjacent to the target pressure unit, calculate the spatial distance between the target pressure unit and each neighboring pressure unit, and simultaneously calculate the pressure reference value difference between the target pressure unit and each neighboring pressure unit. The spatial distance value and the pressure reference value difference are coupled and calculated. The pressure transmission influence coefficient of each neighboring pressure unit on the target pressure unit is obtained by multiplying the pressure reference value difference with the reciprocal of the spatial distance value. The pressure transmission influence coefficient is normalized so that the sum of the pressure transmission influence coefficients corresponding to all neighboring pressure units is a preset unit value, thus obtaining the normalized pressure transmission influence coefficient. The normalized pressure transmission influence coefficient is used as a weighting coefficient to sum the pressure reference values ​​of each neighboring pressure unit to obtain the neighborhood pressure weighted average value. The deviation between the neighborhood pressure weighted average value and the pressure reference value of the target pressure unit is then calculated. The deviation value is scaled according to a preset correction ratio to obtain the pressure correction increment. The pressure reference value of the target pressure unit is superimposed with the pressure correction increment to obtain the cooperative correction pressure value of the target pressure unit.

[0038] In actual hemostasis, the pressure distribution at the wound site is not isolated; there are mechanical coupling effects and physiological response transmission phenomena between adjacent pressure-applying units. To achieve coordinated work among the pressure-applying units, it is necessary to establish a pressure transmission mechanism between them.

[0039] For any target pressure application unit on the hemostatic device, its physical coordinates are determined using a spatial positioning module. These coordinates can be two-dimensional planar coordinates or three-dimensional spatial coordinates. A spatial search radius is set around the target pressure application unit, and all neighboring pressure application units falling within this radius are identified. The search radius needs to consider the size of the wound area and the density of the pressure application units, typically ranging from 10 to 30 millimeters. The spatial distance between the target pressure application unit and each neighboring pressure application unit is calculated using the Euclidean distance formula and recorded as d_i. Simultaneously, the pressure reference value of the target pressure application unit is extracted. Calculate the pressure reference value difference between the pressure reference value P_i of each neighboring pressure unit and the pressure reference value P_i. .

[0040] The pressure transmission influence coefficient is constructed using a coupling method of physical distance and pressure difference. For the i-th neighboring pressure-applying unit, its pressure transmission influence coefficient on the target pressure-applying unit is calculated as follows: This calculation method reflects two core physical mechanisms: the larger the pressure difference, the stronger the traction effect of the pressure applied by the neighboring unit on the target unit, while the reciprocal of the spatial distance reflects the attenuation law of mechanical transmission, that is, the closer the distance, the more significant the effect.

[0041] To eliminate the influence of dimensions and ensure a reasonable allocation of influence weights for each neighboring unit, the pressure transmission influence coefficients of all neighboring pressure-applying units are normalized. The sum of all influence coefficients is then calculated. Where n is the number of neighboring pressure-applying units, then each influence coefficient is divided by this sum to obtain the normalized pressure transmission influence coefficient w_i, which satisfies This normalization operation keeps the sum of the influence contributions of each neighboring unit constant at a unit value.

[0042] Using the normalized pressure transmission influence coefficient as a weighting coefficient, the pressure reference values ​​of each neighboring pressure unit are weighted and summed to obtain the weighted average neighborhood pressure. This weighted average comprehensively reflects the overall pressure expectation of the surrounding neighborhood on the target cell. The deviation between the weighted average neighborhood pressure and the current pressure baseline value of the target pressure cell is calculated. .

[0043] A preset correction scaling factor α is introduced to scale the deviation value. This factor typically ranges from 0.1 to 0.4 to avoid excessive correction amplitude that could cause pressure oscillations. The pressure correction increment is then calculated. The original pressure reference value of the target pressure unit is superimposed with this correction increment to obtain the coordinated correction pressure value. This collaboratively modified pressure value retains the characteristics of the target unit while incorporating collaborative control information from neighboring units, enabling the entire hemostasis device to form a distributed and coordinated pressure field. This avoids local pressure abrupt changes or discontinuities, improving hemostasis efficiency while reducing the risk of secondary tissue damage.

[0044] In one optional implementation, a virtual simulation is performed on the initial pressure application scheme. By simulating the pressure transmission path and the physiological response chain evolution process, the hemostasis timeliness score and safety margin score of each candidate scheme are calculated, and the target pressure application scheme with the dual-score weighted comprehensive optimality is selected, including: Obtain the target execution pressure value of each pressure application unit in the initial pressure application scheme, and construct a simulation environment model that includes the elastic modulus distribution of traumatic tissue and the vascular network topology. In the simulation environment model, the stress tensor propagation process of pressure is calculated based on the tissue elastic modulus distribution, the diffusion trajectory of pressure from each pressure-applying unit to the deep tissue is tracked, and pressure field evolution data containing the conduction time series and spatial distribution gradient are generated. The pressure field evolution data is spatially mapped to the vascular network topology. The degree of vascular closure is calculated based on the pressure value at each vascular node. The dynamic process of blood flow occlusion is constructed based on the time change trend of the degree of vascular closure, and a physiological response chain evolution sequence is obtained. By analyzing the time evolution curve of the closure state of vascular nodes in the physiological response chain evolution sequence, the critical moment when all vascular nodes reach the complete closure state is located. The time deviation between the preset hemostasis time target value and the critical moment is calculated as the hemostasis timeliness score. By monitoring the stress accumulation process of tissue units in the physiological response chain evolution sequence, the stress peak breakthrough point is captured. The safety redundancy between the preset safe stress threshold and the stress peak breakthrough point is calculated as the safety margin score. After assigning preset weight coefficients to the hemostasis timeliness score and the safety margin score respectively, the comprehensive score is obtained by summing them. The simulation process is executed on multiple candidate initial pressure application schemes, and the candidate scheme with the largest comprehensive score is selected as the target pressure application scheme.

[0045] After obtaining the initial pressure application scheme, the target execution pressure value corresponding to each pressure application unit is extracted, and a simulation environment model including the elastic modulus distribution of the traumatic tissue and the vascular network topology is constructed. The elastic modulus distribution of the traumatic tissue is obtained by dividing the trauma site into several voxel units, and each voxel unit is assigned an elastic modulus value corresponding to the tissue type. For example, the elastic modulus of the skin layer is set to... The fat layer is set to The muscle layer is set to The vascular network topology was extracted from pre-acquired ultrasound Doppler image data, and the spatial coordinates, diameter parameters, and connection relationships of each vascular node were marked.

[0046] In the simulation environment model, the stress tensor propagation process under pressure is calculated based on the elastic modulus distribution of each voxel element. The target applied pressure value of each pressure-bearing element is applied as the initial boundary condition, and the stress transfer between adjacent voxel elements is calculated according to the constitutive equation of elasticity, tracing the diffusion trajectory of pressure from the contact interface to the deeper tissue. The simulation step size is set to... At each time step, the pressure value and stress tensor components of each voxel element are recorded, generating pressure field evolution data that includes time series and spatial distribution gradients. This data is stored in a three-dimensional matrix, with the matrix dimensions corresponding to the voxel space coordinates and the time axis.

[0047] The pressure field evolution data is spatially mapped to the vascular network topology. Coordinate matching is used to determine the voxel unit where each vascular node is located, and the pressure time series at that location is extracted. The degree of vascular closure at each node is calculated based on a vascular closure mechanics model, which defines the degree of closure as a function of the ratio of external pressure to the compressive strength of the vascular wall. Closure reaches 100% when the external pressure exceeds the critical closure pressure of the vascular wall. The temporal trend of the closure degree at each vascular node is continuously tracked to construct a dynamic process of blood flow occlusion. When an upstream vascular node closes, the blood flow at downstream nodes decreases accordingly, forming a chain response effect, generating a physiological response chain evolution sequence containing the evolution of the closure state of each vascular node.

[0048] The time evolution curves of vascular node closure states in the physiological response chain evolution sequence are analyzed to identify the critical moment when all vascular nodes reach complete closure. The difference between the preset hemostasis target value and the critical moment is calculated, and the time deviation is obtained through normalization. This deviation is then mapped to a hemostasis timeliness score from 0 to 100, with a higher score for a smaller deviation. The stress accumulation process of each tissue unit in the physiological response chain evolution sequence is monitored simultaneously to capture the stress peak breakthrough point, i.e., the time and location of the maximum stress value. The difference between the preset safe stress threshold and the stress peak is calculated. This difference is used as a safety redundancy and normalized to map a safety margin score from 0 to 100, with a higher score for a larger redundancy.

[0049] Preset weighting coefficients are assigned to both the hemostasis timeliness score and the safety margin score. These weighting coefficients are set according to clinical needs; for example, in emergency trauma scenarios, the timeliness weight is set to 0.7, and the safety margin weight is set to 0.3. The comprehensive score is obtained by multiplying the two scores by their corresponding weighting coefficients and summing the results. The above simulation process is executed sequentially for multiple candidate initial pressure application schemes, generating a comprehensive score list for each candidate scheme. The candidate scheme with the highest comprehensive score is selected as the target pressure application scheme, which effectively controls the risk of tissue damage while ensuring hemostasis efficiency.

[0050] In one optional implementation, the stress tensor propagation process of pressure is calculated based on the tissue elastic modulus distribution in the simulation environment model, tracing the diffusion trajectory of pressure from each pressure-applying unit to deeper tissues, and generating pressure field evolution data including conduction time series and spatial distribution gradients, including: Acquire the spatial mesh division data of the traumatic tissue in the simulation environment model, and map the target stress value of each stress unit to the corresponding mesh unit as the initial stress state; Based on the initial stress state and the elastic modulus of each grid cell, a stress balance equation between grid cells is established. By solving the stress balance equation, the stress transfer relationship between adjacent grid cells is calculated. The stress tensor evolution value of each grid cell at continuous time steps is iteratively calculated to form the spatiotemporal evolution matrix of the stress tensor. The pressure components of each grid cell at each time step are extracted from the stress tensor spatiotemporal evolution matrix. The pressure components are then statistically analyzed according to the depth coordinates of the grid cells. The attenuation ratio of pressure from the surface layer of the pressure-applying cell to each layer of the deep tissue is calculated, and a pressure depth penetration curve is generated. The pressure values ​​of each depth layer in the pressure depth penetration curve are correlated with the lateral spatial coordinates of the corresponding grid cells in the layer. The pressure difference distribution at different lateral positions within the same depth layer is calculated, and a two-dimensional pressure diffusion trajectory containing both depth and lateral dimensions is constructed. The pressure change rate between adjacent grid cells in the two-dimensional pressure diffusion trajectory is calculated, and the pressure change rate is used as the spatial distribution gradient value. The spatial distribution gradient value is bound to the time step information in the spatiotemporal evolution matrix of the stress tensor to generate pressure field evolution data containing the conduction time series and spatial distribution gradient.

[0051] like Figure 2 As shown, the method includes: In the virtual pre-simulation phase, it is necessary to accurately simulate the propagation of pressure in the tissue. This involves reading the three-dimensional spatial mesh data of the traumatic tissue from the simulation environment model. This meshing uses the finite element method to discretize the tissue space into several cubic or tetrahedral elements. Each mesh element carries spatial coordinate information, volume parameters, and a tissue type identifier. After obtaining the target applied pressure value for each pressure-applying element in the initial pressure application scheme, the pressure value is mapped to the corresponding mesh element on the surface of the simulation model based on the physical location of the pressure-applying element. The mapping process uses a projection algorithm; when a pressure-applying element covers multiple mesh elements, the pressure load is distributed according to the contact area ratio. The distributed pressure value is then converted into an initial stress state and stored in the stress tensor matrix of the mesh element.

[0052] Based on the elastic modulus values ​​of each grid cell, a stress balance equation between grid cells is established. The elastic modulus values ​​are read from a pre-established database of tissue mechanics parameters. Different types of tissues have different elastic moduli; the elastic modulus of muscle tissue at the injury site is typically in the range of 5-50 kPa, subcutaneous adipose tissue is approximately 0.5-5 kPa, and vascular wall tissue is approximately 100-500 kPa. The stress balance equation is based on the theory of continuum mechanics and is expressed for any grid cell as follows: the product of the resultant stress acting on the cell surface and the deformation strain inside the cell must remain in balance. The partial differential equations are transformed into a system of linear equations through discretization, and the stress transfer coefficient matrix between adjacent grid cells is solved. Iterative calculations are performed using a time-stepping method, with each time step set to 0.01 seconds. The stress components transferred from adjacent cells to each grid cell at the current moment are calculated, superimposed, and the stress tensor of that cell is updated. The iteration process continues until the pressure propagation reaches a steady state or a preset duration, forming a spatiotemporal evolution matrix containing the stress tensor values ​​of each grid cell at each time step.

[0053] Pressure components are extracted from the spatiotemporal evolution matrix of the stress tensor, a three-dimensional matrix whose average diagonal elements represent hydrostatic pressure. After extracting the pressure components of each grid cell at each time step, the grid cells are layered according to their depth coordinates. The depth coordinates are set to zero at the contact surface of the pressure-applying cell and increase towards the interior of the tissue. The depth range is divided into several layered intervals, each with a thickness of 2 mm. The average pressure of all grid cells within each layer is statistically analyzed, and the attenuation ratio is calculated by comparing the average pressure of the surface layer with that of each deeper layer, forming a pressure-depth penetration curve with depth on the x-axis and pressure attenuation ratio on the y-axis. This curve reflects the attenuation law of pressure with tissue depth, typically exhibiting an exponential decay characteristic.

[0054] Further analysis of the lateral pressure distribution within the same depth layer was conducted. Pressure values ​​for each depth layer were extracted from the pressure-depth penetration curve, and the lateral spatial coordinates of the grid cells within that layer were correlated. The lateral coordinates were implemented using a two-dimensional plane coordinate system with the center of the pressure-exerting cell as the origin. Pressure differences between grid cells at different lateral positions within the same depth layer were calculated, and a lateral pressure distribution contour map was plotted. The depth and lateral dimensions were combined to construct a two-dimensional pressure diffusion trajectory matrix. The rows of this matrix represent depth layers, the columns represent lateral positions, and the matrix elements are the pressure values ​​at the corresponding positions.

[0055] The rate of pressure change between adjacent grid cells is calculated as the spatial gradient. For the depth direction, the difference in pressure values ​​between adjacent depth layers is calculated and divided by the layer spacing; for the lateral direction, the pressure difference between adjacent lateral positions is calculated and divided by the spatial distance. The gradient value characterizes the spatial non-uniformity of the pressure field, and high gradient regions indicate the risk of stress concentration. The spatial gradient value is correlated with the time step information in the spatiotemporal evolution matrix of the stress tensor to establish a four-dimensional data structure: three spatial dimensions plus one time dimension. This data structure fully describes the evolution of the pressure field over time in three-dimensional space, forming a pressure field evolution data package for subsequent calculations of hemostasis timeliness and safety margin scoring.

[0056] A second aspect of the present invention provides an adaptive control system for an intelligent hemostasis device based on pressure feedback, comprising: The data acquisition unit is used to collect real-time pressure distribution data and physiological characteristic parameters at the interface between the hemostatic device and the wound site. The model building unit is used to construct a pressure-physiological coupled response model based on the real-time pressure distribution data and the physiological state characteristic parameters. By extracting the dynamic correlation features between the pressure change rate and the physiological response delay, it establishes a bidirectional prediction function of pressure regulation behavior on hemostatic efficacy and tissue damage, and outputs the safe pressure application range under the current physiological state. The scheme generation unit is used to perform differentiated drive allocation of multiple pressure application units of the hemostasis device according to the safe pressure application range, and use the upper and lower boundary values ​​of the safe pressure application range as the constraint conditions of each pressure application unit to generate an initial pressure application scheme of partitioned coordination. The simulation optimization unit is used to perform virtual pre-simulation of the initial pressure application scheme. By simulating the evolution process of pressure transmission path and physiological response chain, it calculates the hemostasis timeliness score and safety margin score of each candidate scheme and selects the target pressure application scheme with dual-score weighted comprehensive optimality. The feedback update unit is used to drive each pressure application unit to perform pressure regulation according to the target pressure application scheme and collect the pressure deviation and physiological state drift after execution in real time. The pressure deviation and physiological state drift are then fed back to the pressure-physiology coupled response model to update the parameters of the dynamic correlation feature and the bidirectional prediction function.

[0057] The method further includes: When implementing the adaptive control method of the intelligent hemostasis device based on pressure feedback, the hemostasis device includes multiple pressure sensor arrays and corresponding pressure application unit arrays distributed at the contact interface. The pressure sensor arrays use resistive pressure sensors with a measurement range of 0 to 500 mmHg and a sampling frequency of 100 Hz. The sensor arrays are arranged in a two-dimensional grid with a grid spacing of 5 to 10 mm. The sensor output signals are filtered by a low-pass filter to remove high-frequency noise above 50 Hz before entering the data acquisition module. The data acquisition module organizes the pressure values ​​of all sensor nodes at the same time into a two-dimensional pressure distribution matrix, which is the real-time pressure distribution data. At the same time, the physiological state characteristic parameter acquisition module acquires heart rate data through a photoplethysmography pulse wave sensor with a sampling frequency of 200 Hz, acquires blood oxygen concentration percentage through a blood oxygen saturation sensor, and acquires local temperature of the contact interface through a skin temperature sensor with a measurement accuracy of 0.1 degrees Celsius. All physiological state characteristic parameters are unified to a 100 Hz time base through linear interpolation. The aligned pressure distribution data and physiological state characteristic parameters are transmitted to the control processing unit through a data bus.

[0058] After receiving data, the control processing unit constructs a pressure-physiological coupled response model. This model maintains a historical data buffer with a time window length of 10 seconds. Data within a continuous time period is extracted from the buffer to calculate the pressure change rate. The pressure change rate is obtained by subtracting the corresponding elements of the pressure distribution matrix at the current time from those at the previous time and dividing by the time interval, which is 0.01 seconds. Simultaneously, the heart rate change sequence is extracted from the physiological state characteristic parameters. Moments with a relative heart rate change exceeding 5% are marked as physiological response moments. The time difference between each physiological response moment and the moment when the peak occurs in the previous pressure change rate matrix is ​​calculated. This time difference is the physiological response delay. All physiological response delay values ​​are statistically analyzed to construct a delay distribution histogram. The delay time interval with the highest frequency of occurrence is extracted from the histogram as the typical physiological response delay range, which is usually between 0.5 and 3 seconds. Each element in the pressure change rate matrix is ​​correlated with the heart rate change value after the corresponding delay time, and the linear correlation coefficient between the two is calculated. Pressure change rate elements with an absolute correlation coefficient greater than 0.6 and heart rate change pairs constitute dynamic correlation features.

[0059] A bidirectional prediction function is established based on the extracted dynamic correlation features. The bidirectional prediction function includes a hemostasis efficacy prediction branch and a tissue damage prediction branch. The hemostasis efficacy prediction branch calculates the proportion of pressure-covered vascular nodes to the total number of bleeding vascular nodes based on the spatial mapping relationship between the values ​​of each element in the pressure distribution matrix and the distribution of blood vessels at the wound site. This proportion is used as the hemostasis coverage rate indicator. When the hemostasis coverage rate exceeds 85%, it is considered effective hemostasis. The pressure application rate threshold is set to no more than 50 mmHg per second. The tissue damage prediction branch judges the damage risk based on the comparison between the local pressure peak value in the pressure distribution matrix and the tissue tolerance pressure threshold. The baseline tolerance pressure threshold is set to 300 mmHg. The threshold decreases by 3% for every 1 degree Celsius increase in temperature. The bidirectional prediction function outputs the safe pressure application range under the current physiological state. The lower boundary value is usually between 120 and 180 mmHg, and the upper boundary value is set to 90% of the tissue tolerance pressure threshold.

[0060] The differentiated drive allocation module configures the pressure application units according to the safe pressure application range. The drive configuration process first divides the wound site into multiple sub-regions based on the spatial distribution characteristics of the pressure gradient in the pressure distribution matrix. Each sub-region contains at least 4 sensor nodes. The mean and variance of the pressure distribution within each sub-region are calculated. When the variance exceeds 30% of the mean, the pressure distribution in that region is considered uneven. The drive intensity is adjusted by the amplitude of the actuator control signal of the pressure application unit. The amplitude of the control signal ranges from 0 to 5 volts, with each volt corresponding to a pressure output of 100 mmHg. For sub-regions where the pressure is below the lower boundary value, the amplitude of the control signal is increased to the voltage value corresponding to the lower boundary value. For sub-regions where the pressure is above the upper boundary value, the amplitude of the control signal is decreased to the voltage value corresponding to the upper boundary value. The drive configuration results of all pressure application units are organized into the initial pressure application scheme.

[0061] After loading the initial pressure application scheme, the virtual pre-simulation module constructs a simulation environment that includes the mechanical properties of traumatic tissue and the topology of the vascular network. The elastic modulus of muscle tissue is set to 10 to 30 kPa, and the elastic modulus of adipose tissue is set to 2 to 5 kPa. The vascular network topology is represented by a graph structure of nodes and edges. Based on the expected output pressure values ​​of each pressure application unit in the initial pressure application scheme, the simulation environment calculates the pressure attenuation process of pressure transmission from the tissue to the vascular nodes. The pressure decreases by 5% to 15% for every 1 mm increase in transmission distance. When the pressure transmitted to the vascular node exceeds the intravascular pressure at that node, the blood vessel closes. The module counts the proportion of completely closed vascular nodes to the total number of bleeding vascular nodes. When this proportion reaches 95%, hemostasis is considered complete. The time from the start of the simulation to the completion of hemostasis is recorded as the hemostasis timeliness index. At the same time, the stress accumulation value of the tissue unit is monitored. When the stress accumulation value exceeds the tissue damage threshold, it is marked as a potential damage area. The proportion of the potential damage area to the total contact area is calculated as a safety margin index. The weight of hemostasis timeliness is set to 0.6, and the weight of safety margin is set to 0.4. The two indexes are weighted and summed to obtain a comprehensive score. The candidate scheme with the highest comprehensive score is selected as the target pressure application scheme.

[0062] The execution control module sends drive commands to each pressure application unit according to the target pressure application plan. The execution timing adopts a phased loading strategy, dividing the target pressure value into 3 to 5 steps. The pressure increment of each step is 20% to 30% of the total pressure change, and the time interval between steps is 0.5 to 1 second. During the execution of the drive commands by the pressure application unit, the pressure sensor array continuously collects the pressure distribution data after execution and calculates the pressure deviation between the actual measured pressure value and the expected pressure value. At the same time, the physiological state characteristic parameter acquisition module continuously monitors heart rate, blood oxygen concentration and skin temperature, and calculates the physiological state drift amount by calculating the percentage change in heart rate, percentage change in blood oxygen concentration and absolute value of temperature change. The pressure deviation amount and the physiological state drift amount are fed back to the pressure-physiology coupled response model. The model updates the dynamic correlation feature library according to the fed-back data and corrects the parameters in the bidirectional prediction function.

[0063] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0064] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0065] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An adaptive control method for an intelligent hemostasis device based on pressure feedback, characterized in that, include: Real-time pressure distribution data and physiological characteristics of the interface between the hemostatic device and the wound site were collected. Based on the real-time pressure distribution data and the physiological state characteristic parameters, a pressure-physiology coupled response model is constructed. By extracting the dynamic correlation features between the pressure change rate and the physiological response delay, a bidirectional prediction function of pressure regulation behavior on hemostatic efficacy and tissue damage is established, and the safe pressure application range under the current physiological state is output. Based on the safe pressure application range, the multiple pressure application units of the hemostasis device are driven and allocated differently. The upper and lower boundary values ​​of the safe pressure application range are used as the constraint conditions of each pressure application unit to generate a partitioned and coordinated initial pressure application scheme. The initial pressure application scheme was simulated virtually. By simulating the pressure transmission path and the physiological response chain evolution process, the hemostasis timeliness score and safety margin score of each candidate scheme were calculated, and the target pressure application scheme with the dual-score weighted comprehensive optimality was selected. According to the target pressure application scheme, each pressure application unit is driven to perform pressure regulation and the pressure deviation and physiological state drift are collected in real time after execution. The pressure deviation and physiological state drift are fed back to the pressure-physiology coupled response model to update the parameters of the dynamic correlation feature and the bidirectional prediction function.

2. The method according to claim 1, characterized in that, Based on the real-time pressure distribution data and the physiological state characteristic parameters, a pressure-physiological coupled response model is constructed. By extracting the dynamic correlation features between the pressure change rate and the physiological response delay, a bidirectional prediction function of pressure regulation behavior on hemostatic efficacy and tissue damage is established, and the safe pressure application range under the current physiological state is output, including: The real-time pressure distribution data is processed by time-series difference processing to obtain the pressure change rate time series, and the physiological response state transition time markers in the physiological state characteristic parameters are extracted simultaneously. The time series of pressure change rate is correlated with the time markers of physiological response state transition. The time interval between the peak pressure application time and the physiological response state transition time is calculated, and dynamic correlation features characterizing the pressure-physiological conduction delay are extracted. A pressure-physiology coupled response model is constructed based on the dynamic correlation features, and the time interval in the dynamic correlation features is used as the time delay parameter of the model to establish a bidirectional prediction function. The minimum pressure threshold required to achieve the preset hemostatic effect is calculated by the positive prediction function branch of the bidirectional prediction function, and the maximum pressure threshold corresponding to the risk of triggering tissue damage is calculated by the negative prediction function branch of the bidirectional prediction function. The interval between the minimum pressure threshold and the maximum pressure threshold is determined as the safe pressure application interval.

3. The method according to claim 2, characterized in that, The time series of pressure change rate is correlated with the time markers of physiological response state transitions. The time interval between the peak pressure application time and the physiological response state transition time is calculated, and dynamic correlation features characterizing the pressure-physiological conduction delay are extracted, including: Peak identification processing is performed on the pressure change rate time series to detect the moment when the pressure change rate reaches a local maximum in the pressure change rate time series, and the detected moment is determined as the peak moment of pressure application. The peak time of pressure application and the time of physiological response state transition are time-aligned on a unified time axis. For the peak time of pressure application, the physiological response state transition time marker with the closest timestamp after the peak time of pressure application is retrieved from the physiological response state transition time markers. A causal pairing relationship is established between the peak time of pressure application and the physiological response state transition time markers. Based on the causal pairing relationship, the time difference between the peak time of pressure application and the time marker of the paired physiological response state transition is calculated. The time difference is used as the time interval of a single pressure-physiological conduction delay. Multiple sets of time intervals corresponding to the causal pairing relationship are collected to form a time interval dataset. Statistical analysis is performed on the time interval dataset to calculate the central tendency measure and the dispersion measure of the time interval dataset, and the central tendency measure and the dispersion measure are combined to form dynamic correlation features.

4. The method according to claim 1, characterized in that, Based on the safe pressure application range, the multiple pressure application units of the hemostatic device are differentially driven and allocated. The upper and lower boundary values ​​of the safe pressure application range are used as constraints for each pressure application unit to generate a zoned and coordinated initial pressure application scheme, including: Obtain hemostasis demand metric values ​​for the wound contact areas corresponding to multiple pressure units of the hemostasis device, and sort the multiple pressure units by urgency based on the hemostasis demand metric values ​​to obtain the urgency sequence number of each pressure unit. Based on the urgency number, a linear mapping is performed within the safe pressure application range. The pressure application unit with the first urgency number is mapped to the upper boundary value of the safe pressure application range, and the pressure application unit with the last urgency number is mapped to the lower boundary value of the safe pressure application range. The remaining pressure application units are proportionally interpolated between the upper boundary value and the lower boundary value according to the urgency number to form the initial pressure distribution value of each pressure application unit. The initial pressure distribution value is used as the pressure reference value for each pressure unit. The spatial location information of each pressure unit is obtained, and the difference between the spatial distance between adjacent pressure units and the pressure reference value is calculated. Based on the difference between the spatial distance and the pressure reference value, a pressure transmission influence coefficient between the pressure application units is constructed. The pressure reference value of each pressure application unit is then collaboratively corrected in the neighborhood using the pressure transmission influence coefficient to obtain the collaboratively corrected pressure value of each pressure application unit. The collaborative correction pressure value and the upper and lower boundary values ​​of the safe pressure application range are used for boundary constraint verification. The verified collaborative correction pressure value is used as the target execution pressure value of each pressure application unit. The target execution pressure value and the corresponding pressure application unit identifier are combined to form an initial pressure application scheme.

5. The method according to claim 4, characterized in that, Based on the difference between the spatial distance and the pressure reference value, a pressure transmission influence coefficient is constructed between the pressure application units. This pressure transmission influence coefficient is then used to perform neighborhood-based collaborative correction of the pressure reference value for each pressure application unit, resulting in the following collaboratively corrected pressure values ​​for each unit: For a target pressure unit among multiple pressure units, obtain the neighboring pressure units that are spatially adjacent to the target pressure unit, calculate the spatial distance between the target pressure unit and each neighboring pressure unit, and simultaneously calculate the pressure reference value difference between the target pressure unit and each neighboring pressure unit. The spatial distance value and the pressure reference value difference are coupled and calculated. The pressure transmission influence coefficient of each neighboring pressure unit on the target pressure unit is obtained by multiplying the pressure reference value difference with the reciprocal of the spatial distance value. The pressure transmission influence coefficient is normalized so that the sum of the pressure transmission influence coefficients corresponding to all neighboring pressure units is a preset unit value, thus obtaining the normalized pressure transmission influence coefficient. The normalized pressure transmission influence coefficient is used as a weighting coefficient to sum the pressure reference values ​​of each neighboring pressure unit to obtain the neighborhood pressure weighted average value. The deviation between the neighborhood pressure weighted average value and the pressure reference value of the target pressure unit is then calculated. The deviation value is scaled according to a preset correction ratio to obtain the pressure correction increment. The pressure reference value of the target pressure unit is superimposed with the pressure correction increment to obtain the cooperative correction pressure value of the target pressure unit.

6. The method according to claim 1, characterized in that, The initial pressure application scheme was simulated virtually. By simulating the pressure transmission path and the physiological response chain evolution process, the hemostasis timeliness score and safety margin score of each candidate scheme were calculated. The target pressure application scheme with the best weighted comprehensive score was selected, including: Obtain the target execution pressure value of each pressure application unit in the initial pressure application scheme, and construct a simulation environment model that includes the elastic modulus distribution of traumatic tissue and the vascular network topology. In the simulation environment model, the stress tensor propagation process of pressure is calculated based on the tissue elastic modulus distribution, the diffusion trajectory of pressure from each pressure-applying unit to the deep tissue is tracked, and pressure field evolution data containing the conduction time series and spatial distribution gradient are generated. The pressure field evolution data is spatially mapped to the vascular network topology. The degree of vascular closure is calculated based on the pressure value at each vascular node. The dynamic process of blood flow occlusion is constructed based on the time change trend of the degree of vascular closure, and a physiological response chain evolution sequence is obtained. By analyzing the time evolution curve of the closure state of vascular nodes in the physiological response chain evolution sequence, the critical moment when all vascular nodes reach the complete closure state is located. The time deviation between the preset hemostasis time target value and the critical moment is calculated as the hemostasis timeliness score. By monitoring the stress accumulation process of tissue units in the physiological response chain evolution sequence, the stress peak breakthrough point is captured. The safety redundancy between the preset safe stress threshold and the stress peak breakthrough point is calculated as the safety margin score. After assigning preset weight coefficients to the hemostasis timeliness score and the safety margin score respectively, the comprehensive score is obtained by summing them. The simulation process is executed on multiple candidate initial pressure application schemes, and the candidate scheme with the largest comprehensive score is selected as the target pressure application scheme.

7. The method according to claim 6, characterized in that, In the simulation environment model, the stress tensor propagation process of pressure is calculated based on the tissue elastic modulus distribution. The diffusion trajectory of pressure from each pressure-applying unit to deep tissues is tracked, generating pressure field evolution data including conduction time series and spatial distribution gradients, including: Acquire the spatial mesh division data of the traumatic tissue in the simulation environment model, and map the target stress value of each stress unit to the corresponding mesh unit as the initial stress state; Based on the initial stress state and the elastic modulus of each grid cell, a stress balance equation between grid cells is established. By solving the stress balance equation, the stress transfer relationship between adjacent grid cells is calculated. The stress tensor evolution value of each grid cell at continuous time steps is iteratively calculated to form the spatiotemporal evolution matrix of the stress tensor. The pressure components of each grid cell at each time step are extracted from the stress tensor spatiotemporal evolution matrix. The pressure components are then statistically analyzed according to the depth coordinates of the grid cells. The attenuation ratio of pressure from the surface layer of the pressure-applying cell to each layer of the deep tissue is calculated, and a pressure depth penetration curve is generated. The pressure values ​​of each depth layer in the pressure depth penetration curve are correlated with the lateral spatial coordinates of the corresponding grid cells in the layer. The pressure difference distribution at different lateral positions within the same depth layer is calculated, and a two-dimensional pressure diffusion trajectory containing both depth and lateral dimensions is constructed. The pressure change rate between adjacent grid cells in the two-dimensional pressure diffusion trajectory is calculated, and the pressure change rate is used as the spatial distribution gradient value. The spatial distribution gradient value is bound to the time step information in the spatiotemporal evolution matrix of the stress tensor to generate pressure field evolution data containing the conduction time series and spatial distribution gradient.

8. An adaptive control system for an intelligent hemostasis device based on pressure feedback, used to implement the method of any one of claims 1-7, characterized in that, include: The data acquisition unit is used to collect real-time pressure distribution data and physiological characteristic parameters at the interface between the hemostatic device and the wound site. The model building unit is used to construct a pressure-physiological coupled response model based on the real-time pressure distribution data and the physiological state characteristic parameters. By extracting the dynamic correlation features between the pressure change rate and the physiological response delay, it establishes a bidirectional prediction function of pressure regulation behavior on hemostatic efficacy and tissue damage, and outputs the safe pressure application range under the current physiological state. The scheme generation unit is used to perform differentiated drive allocation of multiple pressure application units of the hemostasis device according to the safe pressure application range, and use the upper and lower boundary values ​​of the safe pressure application range as the constraint conditions of each pressure application unit to generate an initial pressure application scheme of partitioned coordination. The simulation optimization unit is used to perform virtual pre-simulation of the initial pressure application scheme. By simulating the evolution process of pressure transmission path and physiological response chain, it calculates the hemostasis timeliness score and safety margin score of each candidate scheme and selects the target pressure application scheme with dual-score weighted comprehensive optimality. The feedback update unit is used to drive each pressure application unit to perform pressure regulation according to the target pressure application scheme and collect the pressure deviation and physiological state drift after execution in real time. The pressure deviation and physiological state drift are then fed back to the pressure-physiology coupled response model to update the parameters of the dynamic correlation feature and the bidirectional prediction function.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images