Patient-Type Adaptive Intelligent Negative Pressure Therapy System and Control Method

By using an intelligent negative pressure therapy system that adapts to patient type, wound signals are collected in real time and optimized in a rolling closed loop. This solves the problem of dynamic optimization of negative pressure therapy equipment when the wound condition changes, achieving a balance between safety and treatment efficiency, and significantly improving treatment outcomes.

CN123120243APending Publication Date: 2026-11-17THE FIRST AFFILIATED HOSPITAL OF ARMY MEDICAL UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611335413.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-31
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing negative pressure therapy equipment cannot dynamically optimize treatment parameters when the physiological state of the wound is constantly changing, making it difficult to balance treatment safety and efficacy. It also suffers from problems such as large individual differences and poor controllability of safety risks.

Method used

An intelligent negative pressure therapy system based on patient type adaptation is adopted. By acquiring patients' clinical characteristic data and collecting wound microenvironment signals in real time, a reduced-order model of wound physiological state is constructed, and rolling closed-loop optimization control is carried out. Combined with risk value constraints and multi-objective optimization, dynamic adjustment of treatment parameters is achieved.

Benefits of technology

It achieves a balance between safety and treatment efficiency when the wound condition changes dynamically, shortens the average healing time by 27%, and reduces the incidence of adverse events by 63%.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123120243A_ABST
    Figure CN123120243A_ABST
Patent Text Reader

Abstract

This invention relates to the field of simulation testing technology, and discloses an intelligent negative pressure wound therapy system and control method based on patient type adaptation. The system includes: S1, automatically matching the corresponding maximum acceptable risk value threshold according to the patient's risk level; S2, real-time acquisition of multidimensional physiological signals of the wound microenvironment, updating the current wound state estimate, and constructing a reduced-order model of the wound physiological state to predict the evolution trend of the wound physiological state over the next N control cycles; S3, calculating the probability of adverse events occurring in the future under the current strategy, and obtaining the corresponding risk value; S4, solving a multi-objective optimization problem in a finite time domain under hard constraints to obtain the optimal control sequence for the current cycle; S5, executing the optimal control command for the current cycle, and repeating S2~S4 in the next cycle until the operation ends. This application addresses the dynamic optimization problem of wound treatment by constructing a fully closed-loop intelligent control system, effectively improving treatment efficiency while ensuring safety.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of negative pressure therapy equipment technology, and more specifically to an intelligent negative pressure therapy system and control method that is patient type adaptive. Background Technology

[0002] Currently, parameter settings for clinical negative pressure therapy devices primarily rely on the experience and judgment of medical staff. They need to comprehensively assess multiple dimensions of information, including the patient's medical history, wound type, coagulation function, and age, to set or adjust treatment parameters. While some devices have introduced closed-loop control mechanisms, their control objective is limited to the single task of maintaining actual pressure near a preset value. Further attempts at intelligent control are limited to initial parameter matching based on static patient characteristics, i.e., selecting a fixed prescription according to preset classification rules at the start of treatment. However, these technologies still present the following problems in practical application.

[0003] First, the control architecture of existing technologies is inherently open-loop, making it difficult to cope with the time-varying characteristics of wound conditions. A wound is not a static, controlled object; as treatment progresses, key physiological indicators such as the degree of local tissue edema, the nature of exudate, and the state of microcirculation continuously and dynamically evolve. Existing equipment is essentially an open-loop or semi-open-loop system; once parameters are set, the system remains unchanged until the next adjustment, making it lack the ability to proactively sense and predict.

[0004] Second, the existing safety protection mechanisms generally adopt a threshold triggering method, and a response will only be triggered when the threshold is reached. However, for patients with mild coagulation dysfunction, reducing the negative pressure from -125 mmHg to -80 mmHg can reduce the risk of bleeding and ensure their safety, but at the cost of significantly delaying the healing speed. This means that in order to ensure the patient's safety, treatment efficiency must be sacrificed, and it is difficult to achieve a dynamic balance between the two.

[0005] Third, when existing negative pressure therapy devices are used in conjunction with auxiliary physical therapy, there can be physical coupling and risk superposition effects between different treatment modalities. For example, negative pressure leakage may cause the cold compress to adhere tightly to the skin, and the superposition of low temperature and negative pressure may cause frostbite and mechanical damage, making it difficult for existing devices to be used synergistically.

[0006] Fourth, while existing equipment is described as having intelligent functions, its core remains a lookup-based matching method, comparing parameter characteristics with pre-set prescriptions and selecting the closest preset treatment plan. This method is adequate for most standard patients, but it inevitably fails to handle rare symptoms or multiple complications, leading to treatment failure or even dangerous situations. Furthermore, even for standard patients, the continuous changes in their physiological state during treatment can cause the initial plan to gradually become mismatched, affecting the treatment outcome. Summary of the Invention

[0007] The present invention aims to provide an intelligent negative pressure therapy system and control method based on patient type adaptation, which solves the problems of existing negative pressure trauma therapy equipment being unable to dynamically optimize treatment parameters when the physiological state of the wound is constantly changing, and also being difficult to balance treatment safety and treatment effect, resulting in large individual differences in treatment and poor controllability of safety risks.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: a patient-type adaptive intelligent negative pressure therapy system control method, comprising the following steps: S1. Acquire patients' clinical characteristic data and classify them into different risk levels based on preset classification rules; automatically match the corresponding maximum acceptable risk value threshold according to the risk level. ; S2 collects multidimensional physiological signals of the wound microenvironment in real time, updates the current wound state estimate, and constructs a wound physiological state reduction model based on the current wound state estimate and historical sequence to predict the evolution trend of the wound physiological state in the next N control cycles. S3, based on the predicted future state and candidate control strategies, calculate the probability of adverse events occurring in the future period under the current strategy, and obtain the corresponding Value at Risk (VaR); S4, in Under the hard constraints, a multi-objective optimization problem in a finite time domain is solved to obtain the optimal control sequence for the current cycle. The control parameters include the target negative pressure value, the pressure change rate, the intermittent cycle parameter, and the auxiliary equipment command. S5 executes the optimal control command for the current cycle, and repeats S2~S4 in the next cycle to form a rolling closed-loop optimization control until the end of the operation.

[0009] Meanwhile, this solution also provides a patient-type adaptive intelligent negative pressure therapy system, applied to the aforementioned patient-type adaptive intelligent negative pressure therapy system control method, including: The physiological state sensing module, through a multimodal flexible sensor array integrated within the dressing, collects pressure, temperature, humidity, and drainage fluid optical signals of the wound microenvironment in real time, and outputs an estimate of the current wound state. The patient type classification engine receives input patient clinical characteristic data, including age, wound type, coagulation indicators, and comorbidity information, and classifies patients into different risk level categories. The Risk Value Constraint Engine is used to adaptively set the maximum acceptable risk threshold VaR_max according to the patient's risk level, and dynamically calculate the risk value VaR corresponding to the strategy based on the current wound status and candidate control strategies. The model predictive control optimizer is used to solve a finite-time multi-objective optimization problem in each control cycle based on the predicted future wound state and VaR hard constraints, and outputs the optimal control sequence including negative pressure parameters and auxiliary treatment parameters. The multimodal collaborative execution module is used to receive unified control commands output by the MPC optimizer and drive the negative pressure pump, electric valve, cold compress device, and hemostatic microcapsule release mechanism to perform corresponding operations, thereby realizing the collaborative control of multiple treatment modalities.

[0010] The principles and advantages of this scheme are: This solution addresses the dynamic optimization of wound treatment by constructing a fully closed-loop intelligent control system encompassing perception, prediction, evaluation, optimization, and execution, thereby achieving a balance between safety and treatment efficiency.

[0011] This solution models the entire treatment process as an infinite-time-domain constrained optimal control problem. Unlike traditional open-loop control systems that involve one-time parameter setting and fixed execution, this approach employs a rolling time-domain optimization mechanism. Based on the latest acquired wound state data, it re-solves for the optimal control sequence for the next N control cycles, executing only the control instructions for the current cycle, and iteratively optimizing again for subsequent cycles based on the latest state. This allows the system to continuously track the time-varying characteristics of the wound, always approaching the optimal treatment trajectory under safe constraints, thus solving the problem that traditional open-loop control cannot adapt to the dynamic evolution of the wound.

[0012] Meanwhile, based on the identification results of a clinical big data database, this solution abstracts the complex physiological processes of wounds into a discrete-time linear time-invariant state-space model, which serves as the prediction core for MPC (Multimodal Clinical Processing). Using currently acquired multimodal sensor data as input, it can accurately predict the evolution trend of the wound state over the next N steps. The prediction accuracy can be continuously optimized based on real-time data during treatment to ensure that the prediction results closely match the individual characteristics of the patient.

[0013] The concept of patient safety is further quantified into VaR (VaR) index. Based on the patient's clinical characteristics and the distribution of historical adverse event data, the probability distribution of unacceptable adverse events such as tissue ischemia and necrosis and active bleeding in the next N steps under the candidate control strategy is calculated. The maximum possible tissue damage loss (i.e., VaR value) at a 95% confidence level is obtained and set as a hard constraint in the MPC optimization process, thereby achieving predictive risk avoidance and improving efficiency while ensuring safety. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating the control method of the intelligent negative pressure therapy system based on patient type adaptation according to the present invention. Figure 2 This is a schematic diagram of the intelligent negative pressure therapy system based on patient type adaptation according to the present invention. Detailed Implementation

[0015] The following detailed description illustrates the specific implementation method: The patient-type adaptive intelligent negative pressure therapy system and control method in this embodiment model the dynamic optimization problem of negative pressure therapy parameters as a rolling time-domain optimal control problem with hard risk constraints. It realizes the advanced prediction and active intervention of wound status through the MPC framework, and realizes the quantitative management and hard guarantee of patient safety risks through VaR constraints.

[0016] Option 1 A control method for an intelligent negative pressure therapy system based on patient type adaptation is provided, as shown in the attached figure. Figure 1 As shown, it includes the following steps: S1. Acquire patients' clinical characteristic data and classify them into different risk levels based on preset classification rules; automatically match the corresponding maximum acceptable risk value threshold according to the risk level. .

[0017] In this embodiment, medical staff input or select key patient clinical characteristics through a human-computer interface. These clinical characteristic data include the patient's age, wound type, coagulation function indicators (INR value), platelet count, and comorbidity information. Based on these characteristics, the built-in clinical rule engine categorizes patients into three risk levels, including... Low-risk category: Age <60 years, normal coagulation function, no history of bleeding, wound is a common trauma / postoperative wound, corresponding to =60; Medium risk level: Age 60-75 years, mild coagulation dysfunction, and underlying conditions such as diabetes / hypertension. =35 (balancing treatment efficiency and safety); High-risk level: Age > 75 years, severe coagulation dysfunction, history of active bleeding, wound is infected / malignant tumor wound, corresponding to =15 (Prioritize safety).

[0018] Then, based on the clinical rule engine, the maximum acceptable risk threshold for the patient is automatically set according to the patient's risk level. .

[0019] Specifically, the maximum acceptable value at risk threshold The calculation method is as follows: (1) In the formula, Basic contribution threshold; , , These are risk adjustment coefficients for age, coagulation function, and comorbidities, respectively, with each adjustment coefficient dynamically assigned based on the patient's clinical characteristics.

[0020] For example, for young and middle-aged patients with normal coagulation function, It can be set at a higher level to allow for stronger treatment intensity; for elderly, frail patients or those with coagulation disorders, It is then set at a lower level to prioritize safety.

[0021] S2 collects multidimensional physiological signals of the wound microenvironment in real time, updates the current wound state estimate, and constructs a wound physiological state reduction model based on the current wound state estimate and historical sequence to predict the evolution trend of the wound physiological state in the next N control cycles.

[0022] In this embodiment, during the treatment execution phase, the system collects the latest readings of the multimodal sensors at a fixed sampling period (e.g., every 30 seconds or every 1 minute) to obtain multidimensional physiological signals and update the system's estimate of the current physiological state of the wound. These multidimensional physiological signals include pressure, temperature, humidity, and drainage fluid characteristics.

[0023] Based on the updated state estimate and historical sequence, a reduced-order model of the wound's physiological state is constructed, and the evolution trend of the physiological state in the next N steps (e.g., the next 30 minutes to 2 hours) is predicted using this reduced-order model. In this embodiment, the reduced-order model of the wound's physiological state is a discrete-time linear time-invariant state-space model, represented as follows: Status Update: (2) Observation output: (3) In the formula, The physiological state of the wound at time k; This is the control input vector at time k; Let k be the uncontrollable perturbation input vector at time k; This is the process noise vector; Let be the sensor output vector at time k; To measure the noise vector; This is the state transition matrix; To control the input matrix; The perturbation input matrix; This is the output mapping matrix.

[0024] Because biological wounds have a high degree of uncertainty, three perturbation terms are added to make them more closely resemble actual clinical scenarios, ensuring that the model can maintain its predictive accuracy even in complex clinical environments.

[0025] In this embodiment, wound condition estimation It includes 6 core state variables, including local pressure on the wound. Local temperature of the wound Wound humidity Bleeding risk potential Local tissue oxygen saturation and exudation rate , can be represented as (4) In this embodiment, and Both represent the wound physiological state vector in the k-th control cycle, and they are equivalent.

[0026] The system's control input commands can be represented as: (5) Before making predictions, the MPC optimizer needs to be initialized and initial control commands generated. Based on the initial patient characteristics, the built-in wound physiological state reduction model is invoked to generate the first round of predictions for future states. The MPC (Model Predictive Control) optimizer... Under hard constraints, the optimization problem is solved to generate a complete parameter set for the first control cycle (including target negative pressure value, pressure change rate, intermittent cycle parameters, auxiliary equipment instructions, etc.), and the instructions are sent to each actuator to start treatment.

[0027] Then, based on the updated state estimate and historical sequence, the state in the time domain is recursively predicted. In each control period t, based on the current state estimate, an open-loop prediction is performed for the state in the next N steps (prediction time domain), predicting the physiological state evolution trend in the next N steps. The predicted wound physiological state evolution trend is expressed as follows: (6) (7) In the formula, This is the current moment in the control cycle; To predict the step size; To predict the length of the time domain; The residual effect of the control applied at step j on the state at step i.

[0028] Formula (6) is used to recursively calculate all wound states for the next N steps from the known state at the current time t. Formula (7) is used to calculate the sensor's measurable output value at the corresponding time from the predicted future internal state.

[0029] The bleeding risk potential, obtained by fusing data from multiple sensors, can be represented as follows: (8) In the formula, Use the Sigmoid activation function; Let k be the rate of change of negative pressure at time k; The change in hemoglobin concentration in the drainage fluid at time k; Let k be the change in local temperature of the wound at time k; The patient's current coagulation index ratio; The patient's current platelet count; These are the pre-trained weight coefficients for each feature.

[0030] S3, based on the predicted future state and candidate control strategies, calculate the probability of adverse events occurring in the future period under the current strategy, and obtain the corresponding Value at Risk (VaR).

[0031] Based on the currently perceived patient physiological status (such as coagulation parameters INR, platelet count, local tissue oxygen saturation, etc.) and prediction results, the risk value of candidate control strategies is recalculated. This is the risk loss measure corresponding to the worst possible outcome at a given confidence level (e.g., 95%), ensuring its continued satisfaction. Constraints. If forecasts indicate that the current strategy may reach risk boundaries in the future, the system will adjust the control strategy in advance to mitigate the risk.

[0032] In this embodiment, adverse events include two categories: active bleeding and tissue ischemia and necrosis. Active bleeding is defined as a hemoglobin concentration in the drainage fluid exceeding a threshold for three consecutive cycles, or a bleeding risk potential. Tissue ischemia and necrosis are defined as local oxygen saturation at the wound site being below 60% for five consecutive cycles, or pressure exceeding -180 mmHg for more than 10 minutes.

[0033] The predicted state sequence for the next N control cycles is output through the wound physiological state reduction model of MPC. Each state vector contains six components: pressure, temperature, humidity, bleeding risk potential, oxygen saturation, and exudation rate, which serve as the core inputs for probability calculations. For a given candidate control strategy, the cumulative loss random variable over the next N steps can be defined as follows: (9) In the formula, Let be a random event that causes an adverse event at time t+i; For the event The severity of the injury is scored as follows: mild bleeding (no intervention required) is scored as 20 points, moderate bleeding (local hemostasis required) is scored as 50 points, severe bleeding (systemic medication / surgery required) is scored as 90 points, and local ischemia and necrosis is scored as 100 points. This is an indicator function; it takes a value of 1 if an event occurs, and 0 otherwise. This is a time-weighted coefficient; events closer to the current moment have higher weights to avoid excessive influence of long-term prediction errors on the results.

[0034] Based on the statistical distribution of a dataset of 1200 clinical adverse events, and combined with the current predicted state sequence, the cumulative loss is calculated. probability density function For each prediction time t+i, based on the prediction state By matching the frequency of adverse events under the same / similar conditions in the historical database, P( (i.e., the conditional probability of an event occurring in this state).

[0035] Based on the event independence assumption (adverse events occurring independently at different times), the cumulative loss over the next N steps is calculated using convolution. The complete probability distribution is used to derive the Value at Risk (VaR). In this embodiment, VaR is defined as the maximum cumulative loss that may occur within N steps in the future at a 95% confidence level, i.e., satisfying... .

[0036] S4, in Under the hard constraints, a multi-objective optimization problem in a finite time domain is solved to obtain the optimal control sequence for the current cycle. The control parameters include the target negative pressure value, the pressure change rate, the intermittent cycle parameter, and the auxiliary equipment command.

[0037] In this embodiment, in each control cycle, based on the future state prediction provided by the prediction module and the risk constraints provided by the VaR engine, a multi-objective optimization problem in a finite time domain is solved, and the optimal control sequence for the current cycle (including the target negative pressure value, pressure change rate, intermittent cycle parameters, auxiliary equipment commands, etc.) is output. The objective function of the multi-objective optimization problem can then be expressed as: (10) In the formula, , , As a weighting coefficient; for patients with severe wounds, increase Weighting; for patients with sensitive constitutions, increase Weighting; for portable device usage scenarios, increase The weight.

[0038] Meanwhile, the constraints are set as follows: (1) Hard constraint of value at risk: ; (2) Physical limit of pressure actuator: ; (3) Pressure change rate limit: Prevent tissue traction damage (4) Maximum duration of a single treatment session: .

[0039] Where N is the prediction time domain length, pk is the control input for step k, where pk is the negative pressure value. The objective function J makes a weighted trade-off between maximizing the healing rate, maximizing patient comfort, and minimizing system energy consumption. The weights themselves can be adaptively adjusted according to the patient type. For example, the weight of the healing rate is increased for patients with severe wounds, and the weight of comfort is increased for patients with sensitive constitutions.

[0040] By employing MPC (Moving Conversion) in the rolling time domain, the system does not solve the parameter sequence for the entire treatment cycle all at once. Instead, it re-solves the optimization problem in each sampling cycle, executing only the optimal control for the current step and re-optimizing based on the latest perceived information in the next cycle. This allows the system to continuously track the dynamic changes in the wound state and maintain a close approximation of the optimal treatment trajectory under risk constraints.

[0041] S5 executes the optimal control command for the current cycle, and repeats S2~S4 in the next cycle to form a rolling closed-loop optimization control until the end of the operation.

[0042] Extract the first step (i.e., the optimal control command for the current step) from the control sequence and send it to each actuator for execution. Discard subsequent steps in the sequence (which will be recalculated in the next cycle). Repeat steps S2 to S4 in the next cycle to form a rolling closed-loop optimization control, enabling the system to continuously track the dynamic changes of the wound until the end of the operation.

[0043] It also includes anomaly detection and supervised intervention. In this embodiment, the anomaly handling subroutine is entered when any of the following situations are detected: (a) The deviation between the actual perceived state and the predicted trajectory exceeds a preset threshold; (b) The sensor readings have reached the absolute safety boundary; (c) Received personnel alarm instruction.

[0044] Upon entering the abnormal subroutine, the system will: (1) suspend the automatic optimization and execution of MPC; (2) display the current abnormal diagnosis information, risk assessment results, and recommended emergency response measures in the form of a standardized procedure (SOP) on the human-machine interface; and (3) record the perception data and system response log of the entire event process. The system can only reactivate the MPC optimizer and resume automatic treatment under the adjusted parameter constraints after receiving a clear "confirmation" or "resume" instruction from medical staff. Through this supervision mechanism, the computational advantages of the MPC+VaR framework in dynamic optimization and risk management are fully utilized, while the final decision-making power is retained in the hands of medical staff, ensuring the clinical safety of the system in marginal cases and unexpected events.

[0045] In this embodiment, the biological wound is treated as a time-varying controlled object with strong noise, clinical guidelines are transformed into VaR hard constraints, and the treatment process is transformed into a constrained MPC infinite time-domain approximation problem. This is not only a technological upgrade of negative pressure equipment, but also a technological leap from empirical medical tools to systems engineering medicine, with extremely strong technical barriers and non-replicability.

[0046] Option 2 In this embodiment, a patient-type adaptive intelligent negative pressure therapy system is also provided, applied to the above-mentioned patient-type adaptive intelligent negative pressure therapy system control method, as shown in the attached figure. Figure 2 As shown, it includes: The physiological state sensing module uses a multimodal flexible sensor array integrated within the dressing, including pressure sensors, temperature sensors, humidity sensors, and drainage fluid optical sensors, to collect pressure, temperature, humidity, and drainage fluid optical signals of the wound microenvironment in real time and output an estimate of the current wound state.

[0047] Based on the collected historical signal sequences, a built-in wound physiological state reduction model is used to make rolling predictions of key physiological indicators (such as local tissue perfusion pressure, exudation trend, and bleeding risk probability) over several future time steps. This reduction model can be pre-identified based on clinical big data and can also be adaptively corrected through online learning during treatment.

[0048] The patient type classification engine receives input patient clinical characteristic data, including age, wound type, coagulation indicators, and comorbidity information, and classifies patients into different risk level categories.

[0049] In this embodiment, the patient type classification engine has a built-in clinical rule base, which can update the classification rules according to new clinical data and supports the configuration of custom patient risk types and corresponding risk thresholds.

[0050] The Risk Value Constraint Engine is used to adaptively set the maximum acceptable risk threshold VaR_max according to the patient's risk level, and dynamically calculate the risk value VaR corresponding to the strategy based on the current wound status and candidate control strategies.

[0051] For each control cycle, the engine calculates the probability distribution of "unacceptable adverse events" (such as tissue ischemia and necrosis, active bleeding) in the future under the current control strategy, based on the currently perceived patient physiological status (such as coagulation index INR, platelet count, local tissue oxygen saturation, etc.) and historical data distribution, and determines the value at risk (VaR) of the control strategy accordingly, which is the risk loss measure corresponding to the worst possible outcome at a given confidence level (such as 95%).

[0052] The model predictive control optimizer is used to solve a finite-time multi-objective optimization problem in each control cycle based on the predicted future wound state and VaR hard constraints, and outputs the optimal control sequence including negative pressure parameters and auxiliary treatment parameters.

[0053] The multimodal collaborative execution module receives unified control commands output by the MPC optimizer and converts them into precise drive signals for each actuator, including: speed and start / stop control of the negative pressure pump, opening adjustment of the electric valve, temperature control and power adjustment of the external cold compress device, and release triggering of the hemostatic microcapsule. It drives the negative pressure pump, electric valve, cold compress device, and hemostatic microcapsule release mechanism to perform corresponding operations, realizing collaborative control of multiple treatment modalities. The control commands of all actuators come from the unified solution of the same MPC optimizer, rather than independent decisions of each modality, thus ensuring collaborative optimality under coupling constraints.

[0054] This embodiment also includes a human-computer interaction module, used to display real-time monitoring data, predicted trajectories, risk assessment results, and control scheme details, and to receive parameter adjustment and intervention instructions from medical staff. It employs a touchscreen design to visually display to medical staff: real-time perceived data of the patient's current physiological state, the future state trajectory predicted by MPC, the VaR assessment results of the current control strategy, parameter details of the recommended control scheme, and its risk-benefit trade-off information.

[0055] Medical staff can confirm, adjust, or reject the system's recommended plans. The system will only execute the control plan after receiving a confirmation instruction, thus forming a supervised intelligent architecture where a person is within the loop.

[0056] In this embodiment, by combining VaR hard constraints with a multi-objective optimization function, the system can maximize treatment efficiency while strictly ensuring patient safety: low-risk patients can receive higher treatment intensity to accelerate healing, while high-risk patients are given priority in risk control to avoid adverse events. Clinical test data show that this approach can shorten the average healing time by 27% and reduce the incidence of adverse events by 63%, achieving a win-win situation for both safety and efficiency.

[0057] Simultaneously, the control commands of all actuators, such as the negative pressure pump, cold compress device, and hemostatic microcapsule release mechanism, are incorporated into a single MPC optimizer for unified solution, completely resolving the coupling conflict problem in multimodal treatment. For example, the system can automatically coordinate the combination of negative pressure value and cold compress power, ensuring both the effectiveness of low-temperature treatment and avoiding the risk of frostbite caused by negative pressure leakage. The output is a globally coordinated optimal solution, rather than a locally optimal solution based on independent decisions of each modality.

[0058] Based on accurate predictions of future conditions, the system can identify potential risks in advance and proactively adjust control strategies, rather than passively responding after risk events occur or indicators exceed thresholds. This upgrades safety protection from post-event remediation to pre-event prevention, significantly improving the safety and controllability of the treatment process, and is especially suitable for high-risk patient groups such as those with abnormal coagulation function or frail elderly individuals.

[0059] The above descriptions are merely embodiments of the present invention, and common knowledge such as specific technical solutions and / or characteristics are not described in detail here. It should be noted that those skilled in the art can make various modifications and improvements without departing from the technical solutions of the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A control method for an intelligent negative pressure therapy system based on patient type adaptation, characterized in that, Includes the following steps: S1, acquire the patient's clinical characteristic data, and classify the patient into different risk levels based on preset classification rules; The maximum acceptable risk value threshold is automatically matched based on the risk level. ; S2 collects multidimensional physiological signals of the wound microenvironment in real time, updates the current wound state estimate, and constructs a wound physiological state reduction model based on the current wound state estimate and historical sequence to predict the evolution trend of the wound physiological state in the next N control cycles. S3, based on the predicted future state and candidate control strategies, calculate the probability of adverse events occurring in the future period under the current strategy, and obtain the corresponding Value at Risk (VaR); S4, in Under the hard constraints, a multi-objective optimization problem in a finite time domain is solved to obtain the optimal control sequence for the current cycle. The control parameters include the target negative pressure value, the pressure change rate, the intermittent cycle parameter, and the auxiliary equipment command. S5 executes the optimal control command for the current cycle, and repeats S2~S4 in the next cycle to form a rolling closed-loop optimization control until the end of the operation.

2. The control method for the patient-type adaptive intelligent negative pressure therapy system according to claim 1, characterized in that, In S1, the clinical characteristic data includes age, wound type, coagulation function indicators, and comorbidity information. The maximum acceptable risk value threshold is calculated as follows: ; In the formula, Basic contribution threshold; , , These are risk adjustment coefficients for age, coagulation function, and comorbidities, respectively, with each adjustment coefficient dynamically assigned based on the patient's clinical characteristics.

3. The control method for the patient-type adaptive intelligent negative pressure therapy system according to claim 2, characterized in that: In S2, the reduced-order model of the wound's physiological state is a discrete-time linear time-invariant state-space model, expressed as follows: Status Update: ; Observation output: ; In the formula, The physiological state of the wound at time k; This is the control input vector at time k; Let k be the uncontrollable perturbation input vector at time k; This is the process noise vector; Let be the sensor output vector at time k; To measure the noise vector; This is the state transition matrix; To control the input matrix; The perturbation input matrix; This is the output mapping matrix.

4. The control method for the patient-type adaptive intelligent negative pressure therapy system according to claim 3, characterized in that: In S2, the predicted trend of wound physiological state evolution is represented as follows: ; ; In the formula, This is the current moment in the control cycle; To predict the step size; To predict the length of the time domain; The residual effect of the control applied at step j on the state at step i.

5. The control method for the patient-type adaptive intelligent negative pressure therapy system according to claim 1, characterized in that: In S2, the multidimensional physiological signals include physiological signals related to pressure, temperature, humidity, and drainage fluid properties; the wound condition estimation includes bleeding risk potential, obtained by fusing the collected data, and is represented as follows. ; In the formula, Use the Sigmoid activation function; Let k be the rate of change of negative pressure at time k; The change in hemoglobin concentration in the drainage fluid at time k; Let k be the change in local temperature of the wound at time k; The patient's current coagulation index ratio; The patient's current platelet count; These are the pre-trained weight coefficients for each feature.

6. The control method for the patient-type adaptive intelligent negative pressure therapy system according to claim 5, characterized in that: In S3, adverse events include two categories: active bleeding and tissue ischemia and necrosis. The predicted state sequence for the next N control cycles is output through a wound physiological state reduction model. Each state vector contains 6 components: pressure, temperature, humidity, bleeding risk potential, oxygen saturation, and exudation rate, which serve as the core input for probability calculation. For a given candidate control strategy, a cumulative loss random variable is defined for the next N steps, the probability density of the cumulative loss is calculated, and the Value at Risk (VaR) is obtained based on the probability distribution.

7. The control method for the patient-type adaptive intelligent negative pressure therapy system according to claim 3, characterized in that: In S4, the objective function of the multi-objective optimization problem is: ; In the formula, , , As a weighting coefficient; for patients with severe wounds, increase Weighting; for patients with sensitive constitutions, increase Weighting; for portable device usage scenarios, increase The weight.

8. The control method for the intelligent negative pressure therapy system based on patient type adaptation according to claim 1, characterized in that, This also includes entering the exception handling subroutine when any of the following conditions are detected: (a) The deviation between the actual perceived state and the predicted trajectory exceeds a preset threshold; (b) The sensor readings have reached the absolute safety boundary; (c) Received personnel alarm instruction.

9. A patient-type adaptive intelligent negative pressure therapy system, characterized in that, The control method for the patient-type adaptive intelligent negative pressure therapy system according to any one of claims 1-8 includes: The physiological state sensing module, through a multimodal flexible sensor array integrated within the dressing, collects pressure, temperature, humidity, and drainage fluid optical signals of the wound microenvironment in real time, and outputs an estimate of the current wound state. The patient type classification engine receives input patient clinical characteristic data, including age, wound type, coagulation indicators, and comorbidity information, and classifies patients into different risk level categories. The Risk Value Constraint Engine is used to adaptively set the maximum acceptable risk threshold VaR_max according to the patient's risk level, and dynamically calculate the risk value VaR corresponding to the strategy based on the current wound status and candidate control strategies. The model predictive control optimizer is used to solve a finite-time multi-objective optimization problem in each control cycle based on the predicted future wound state and VaR hard constraints, and outputs the optimal control sequence including negative pressure parameters and auxiliary treatment parameters. The multimodal collaborative execution module is used to receive unified control commands output by the MPC optimizer and drive the negative pressure pump, electric valve, cold compress device, and hemostatic microcapsule release mechanism to perform corresponding operations, thereby realizing the collaborative control of multiple treatment modalities.

10. The patient-type adaptive intelligent negative pressure therapy system according to claim 9, characterized in that: It also includes a human-computer interaction module, which displays real-time monitoring data, predicted trajectories, risk assessment results and control plan details, and receives parameter adjustment and intervention instructions from medical staff.