An individualized anesthesia regulation method and system based on multi-scale dynamics modeling

Abstract

The application discloses a kind of individualized anesthetic regulation methods and systems of multiscale dynamics modeling, it is related to anesthetic regulation technical field, including, acquisition patient intraoperative multimodal physiological signal, and carry out space-time alignment and feature extraction;Based on the preoperative prior information of patient and real-time data in operation, hierarchical causal dynamics model of molecular pharmacology layer, neuron network layer and organ system layer is constructed;The current infusion parameters of anesthetic drug are input into the causal dynamics model, and the virtual physiological state of patient is generated by real-time simulation, the deviation between the virtual physiological state and real physiological state is calculated, and the individualized parameters of the causal dynamics model are corrected online based on the deviation using reinforcement learning algorithm, which dynamically matches the actual physiological response of patient.This application analyzes strategy and issues instructions, controls anesthetic infusion equipment to adjust the infusion rate of various anesthetic drugs, and realizes individualized anesthetic precise regulation.

Description

Technical Field

[0001] This invention relates to the field of anesthesia control technology, and in particular to a multi-scale dynamic modeling-based individualized anesthesia control method and system. Background Technology

[0002] Clinical anesthesia control is a crucial aspect of surgical anesthesia management. With the continuous development of sensing technology, kinetic modeling, and intelligent algorithms, intelligent anesthesia control is gradually being promoted. Currently, clinical practice commonly employs control schemes based on single indicators and traditional kinetic models. This involves collecting patients' basic physiological data during surgery, combining population-based pharmacokinetic and pharmacodynamic models to calculate drug action patterns, and then using conventional control algorithms to output infusion parameters. This technology can meet the basic anesthesia management needs of routine surgeries, standardize anesthesia procedures, and is currently the mainstream application method in medical institutions. Furthermore, the development of multimodal signal acquisition technology and intelligent algorithms continues to drive anesthesia control towards automation.

[0003] Existing anesthetic control kinetic models focus only on a single organ system scale, failing to incorporate multi-scale modeling at the microscopic level, such as molecular pharmacology and neural networks. Furthermore, model parameters are set based on population data, lacking dynamic adaptive correction capabilities. This makes it difficult to match the individual physiological characteristics and intraoperative changes of different patients, resulting in discrepancies between model simulation results and the actual physiological state of patients. Consequently, the accuracy of multi-drug synergistic infusion strategies is affected, making it unsuitable for the clinical requirements of high-precision individualized anesthesia control. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a multi-scale dynamic modeling method for individualized anesthesia control, which solves the problem that traditional models have a single scale and cannot dynamically adapt to individual physiological states, resulting in insufficient accuracy of anesthetic drug infusion strategies.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for individualized anesthesia control based on multi-scale dynamic modeling, which includes collecting multimodal physiological signals of the patient during the operation and performing spatiotemporal alignment and feature extraction. Based on the patient’s preoperative prior information and intraoperative real-time data, a hierarchical causal dynamic model was constructed, consisting of molecular pharmacology layer, neural network layer and organ system layer. The current anesthetic drug infusion parameters are input into the causal dynamics model to generate the patient's virtual physiological state in real time. The deviation between the virtual physiological state and the real physiological state is calculated. The individualized parameters of the causal dynamics model are corrected online based on the deviation using a reinforcement learning algorithm to obtain a dynamic match with the patient's actual physiological response. The modified causal dynamics model is used as the environment model. The counterfactual multi-agent reinforcement learning algorithm is used to optimize decision-making and generate a drug co-infusion strategy that meets the preset anesthesia depth target and physiological safety constraints. Based on the drug synergistic infusion strategy, the drug infusion rate is adjusted by controlling the anesthesia infusion equipment.

[0007] As a preferred embodiment of the individualized anesthesia control method based on multi-scale dynamic modeling described in this invention, the step of collecting multimodal physiological signals from the patient during surgery and performing spatiotemporal alignment and feature extraction includes: Multimodal physiological signals of the patient during the operation were collected. The multimodal physiological signals included electroencephalogram (EEG) signals reflecting the state of the central nervous system and hemodynamic parameters reflecting the state of the circulatory system. The EEG signals and hemodynamic parameters were spatiotemporally aligned and feature extracted to obtain multimodal physiological signals after spatiotemporal alignment and feature extraction.

[0008] As a preferred embodiment of the individualized anesthesia regulation method based on multi-scale dynamic modeling described in this invention, the construction of a hierarchical causal dynamic model comprising a molecular pharmacology layer, a neuronal network layer, and an organ system layer includes, A molecular pharmacology layer is constructed based on the patient’s preoperative prior information and intraoperative real-time data. The molecular pharmacology layer is used to characterize the effects of the patient’s genotype and drug-metabolizing enzyme activity on pharmacokinetic parameters. The output of the molecular pharmacology layer is used as input to construct a neural network layer. The neural network layer uses a millisecond-level time sliding window strategy to extract EEG alpha band phase-locked connection features that reflect key pathways of consciousness state switching. The output of the neural network layer is used as input to construct the organ system layer, which integrates heart rate fractal dynamics features and end-tidal carbon dioxide parameters to establish a cross-scale physiological response mapping relationship. By hierarchically associating the molecular pharmacology layer, the neural network layer, and the organ system layer, a hierarchical causal dynamic model containing the molecular pharmacology layer, the neural network layer, and the organ system layer is obtained.

[0009] As a preferred embodiment of the individualized anesthesia control method based on multi-scale dynamic modeling described in this invention, the calculation of the deviation between the virtual physiological state and the real physiological state includes, The current anesthetic drug infusion parameters are input into the causal dynamic model to generate a virtual physiological state of the patient in real time. Based on the real physiological state and the patient's virtual physiological state, the deviation between the virtual physiological state and the real physiological state is calculated.

[0010] As a preferred embodiment of the individualized anesthesia control method based on multi-scale dynamic modeling described in this invention, the dynamic matching of the patient's actual physiological response includes: Based on the reinforcement learning algorithm, the deviation between the virtual physiological state and the real physiological state is received as a feedback signal to correct the action instructions of the causal dynamic model parameters. By using action commands to update the individualized parameters of the causal dynamics model online, the state transition mechanism within the causal dynamics model can be changed.

[0011] The causal dynamics model is run based on the updated individualized parameters, so that the generated simulation results are an individualized model that dynamically matches the patient's actual physiological response.

[0012] As a preferred embodiment of the individualized anesthesia control method based on multi-scale dynamic modeling described in this invention, the generation of a drug co-infusion strategy that satisfies the preset anesthesia depth target and physiological safety constraints includes, The modified causal dynamics model is used as the environment model. The counterfactual multi-agent reinforcement learning algorithm is used to obtain the current anesthesia state and extract the counterfactual baseline of each drug. Based on the deviation between the counterfactual baseline and the preset anesthesia depth target and physiological safety constraints, the counterfactual multi-agent reinforcement learning algorithm updates the strategy network parameters of the drug co-infusion action. A drug co-infusion strategy that meets preset anesthesia depth targets and physiological safety constraints is generated using strategy network parameters.

[0013] As a preferred embodiment of the individualized anesthesia control method based on multi-scale dynamic modeling described in this invention, wherein: controlling the anesthesia infusion device to adjust the drug infusion rate includes, The drug co-infusion strategy is analyzed, the target infusion rate command corresponding to each anesthetic drug is extracted, and the target infusion rate command is transmitted to the anesthesia infusion device to drive the actuator to adjust the operating state of the mechanical components; The speed at which the infusion tubing is advanced is changed based on the operating state of the mechanical components, thereby controlling the anesthesia infusion equipment to adjust the drug infusion rate.

[0014] Secondly, the present invention provides a multi-scale dynamic modeling-based individualized anesthesia control system, including an extraction module that collects multimodal physiological signals of the patient during surgery and performs spatiotemporal alignment and feature extraction. The module constructs a hierarchical causal dynamic model at the molecular pharmacology layer, neural network layer, and organ system layer, based on the patient's preoperative prior information and intraoperative real-time data. The matching module inputs the current anesthetic drug infusion parameters into the causal dynamics model, simulates and generates the patient's virtual physiological state in real time, calculates the deviation between the virtual physiological state and the real physiological state, and uses a reinforcement learning algorithm to correct the individualized parameters of the causal dynamics model online based on the deviation, thereby obtaining a dynamic match with the patient's actual physiological response. The strategy module uses the modified causal dynamics model as the environment model and employs a counterfactual multi-agent reinforcement learning algorithm to optimize decision-making and generate a drug co-infusion strategy that meets the preset anesthesia depth target and physiological safety constraints. The control module controls the anesthesia infusion equipment to adjust the drug infusion rate according to the drug co-infusion strategy.

[0015] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the individualized anesthesia control method of multi-scale dynamic modeling as described in the first aspect of the present invention.

[0016] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the individualized anesthesia control method for multi-scale dynamic modeling as described in the first aspect of the present invention.

[0017] The beneficial effects of this invention are as follows: It collects multimodal physiological signals such as intraoperative EEG and hemodynamics, completes spatiotemporal alignment and feature extraction, combines preoperative prior information with real-time intraoperative data, sequentially constructs a molecular pharmacology layer, a neuronal network layer, and an organ system layer, and associates them to form a hierarchical causal dynamic model. Then, it inputs drug infusion parameters to simulate and obtain a virtual physiological state. Combined with deviations, it uses reinforcement learning to online correct the model's individualized parameters to match the patient's actual physiological response. Using the corrected model as the environmental model, it employs a counterfactual multi-agent reinforcement learning algorithm combined with anesthesia depth and physiological safety constraints to optimize and generate a drug co-infusion strategy. The strategy is analyzed and instructions are issued to control the anesthesia infusion equipment to adjust the infusion rate of various anesthetic drugs, achieving precise individualized anesthesia control. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a hierarchical architecture diagram of a hierarchical causal dynamics model.

[0020] Figure 2 This is a flowchart of the parameter correction process for a causal dynamics model based on reinforcement learning. Detailed Implementation

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0023] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0024] Reference Figures 1-2 This is one embodiment of the present invention, which provides a method for individualized anesthesia control based on multi-scale dynamic modeling, comprising the following steps: Multimodal physiological signals of the patient during the operation were collected. The multimodal physiological signals included electroencephalogram (EEG) signals reflecting the state of the central nervous system and hemodynamic parameters reflecting the state of the circulatory system. The EEG signals and hemodynamic parameters were spatiotemporally aligned and feature extracted to obtain multimodal physiological signals after spatiotemporal alignment and feature extraction.

[0025] Furthermore, multimodal physiological signals were collected throughout the patient's surgical process. These signals specifically included electroencephalogram (EEG) signals that reflect the operation of the central nervous system and hemodynamic parameters that reflect the operation of the circulatory system. Spatiotemporal alignment processing was performed on the two types of signals to eliminate differences in the timing and spatial dimensions of signal acquisition. Effective features were then extracted from these signals to obtain standardized multimodal physiological signals with complete feature information, providing a reliable data foundation for subsequent model construction.

[0026] A molecular pharmacology layer is constructed based on the patient's preoperative information and intraoperative real-time data. This molecular pharmacology layer is used to characterize the effects of the patient's genotype and drug-metabolizing enzyme activity on pharmacokinetic parameters.

[0027] Furthermore, by integrating various prior information retained by patients before surgery and physiological data acquired in real time during surgery, a molecular pharmacology layer structure is constructed. This layer is specifically designed to model individual differences and can accurately reflect the effects of two major factors—the patient's own genotype and the activity of drug-metabolizing enzymes in the body—on the pharmacokinetic parameters of anesthetic drugs, thus completing the characterization of basic relationships at the microscopic pharmacological level.

[0028] The output of the molecular pharmacology layer is used as input to construct a neural network layer. The neural network layer uses a millisecond-level time sliding window strategy to extract EEG alpha band phase-locked connection features that reflect key pathways of consciousness state switching.

[0029] Furthermore, the results output from the molecular pharmacology layer are used as the raw input data to construct a neural network layer. This layer uses a millisecond-level time sliding window processing method to analyze the EEG data segment by segment, extract the phase-locked connection features of the EEG alpha band, and thus characterize the neural pathway changes corresponding to the switching of human consciousness states, achieving feature analysis at the neural activity level.

[0030] The output of the neural network layer is used as input to construct an organ system layer, which integrates heart rate fractal dynamics features and end-tidal carbon dioxide parameters to establish a cross-scale physiological response mapping relationship.

[0031] Furthermore, the organ system layer is built by taking over the output of the neural network layer. This layer summarizes and integrates the fractal dynamic characteristics of heart rate and the relevant parameters of end-tidal carbon dioxide. It builds a bridge between different scales based on multiple physiological parameters, establishes a complete cross-scale physiological response mapping relationship, and completes the modeling of the human macroscopic organ system.

[0032] By hierarchically associating the molecular pharmacology layer, the neural network layer, and the organ system layer, a hierarchical causal dynamic model containing the molecular pharmacology layer, the neural network layer, and the organ system layer is obtained.

[0033] Furthermore, following the logic of data flow and action, the molecular pharmacology layer, the neural network layer, and the organ system layer are sequentially linked and structurally integrated to open up data interaction channels between each layer, forming a hierarchical causal dynamic model that integrates microscopic pharmacology, neural activity, and macroscopic organ systems, thus achieving multi-scale integrated modeling.

[0034] The current anesthetic drug infusion parameters are input into the causal dynamic model to generate a virtual physiological state of the patient in real time. Based on the real physiological state and the patient's virtual physiological state, the deviation between the virtual physiological state and the real physiological state is calculated.

[0035] The expression for the state deviation is: ; in, This is a state deviation. The total number of physiological indicators. For indicator serial number, For the first The virtual physiological state values ​​corresponding to each indicator For the first The actual physiological state value corresponding to each indicator.

[0036] The algorithm uses a reinforcement learning approach to receive the deviation between the virtual physiological state and the real physiological state as a feedback signal, and then corrects the action instructions for the parameters of the causal dynamics model.

[0037] Furthermore, the deviation between the virtual physiological state and the real physiological state calculated above is used as a feedback signal input into the reinforcement learning algorithm. The algorithm combines the magnitude and trend of the deviation to generate corresponding action instructions for adjusting the parameters of the causal dynamics model, thus completing the transformation of deviation information into control instructions.

[0038] By using action commands to update the individualized parameters of the causal dynamics model online, the state transition mechanism within the causal dynamics model can be changed.

[0039] Furthermore, the action instructions output by the reinforcement learning algorithm are executed to iteratively update the individualized parameters within the causal dynamics model online, synchronously adjust the data flow and state switching rules within the model, change the model's inherent state transition mechanism, and complete the dynamic optimization of the model parameters.

[0040] The causal dynamics model is run based on the updated individualized parameters, so that the generated simulation results are an individualized model that dynamically matches the patient's actual physiological response.

[0041] Furthermore, the simulation calculations are carried out again using the causal dynamic model after the parameters have been updated, so that the simulation results output by the model continuously match the patient's real-time physiological performance during the operation, creating a unique dynamic model that can follow the dynamic changes in the patient's state and adapt to individual characteristics.

[0042] The modified causal dynamics model is used as the environment model. The counterfactual multi-agent reinforcement learning algorithm is used to obtain the current anesthesia state and extract the counterfactual baseline of each drug. Furthermore, the causal dynamics model with completed parameter correction and high adaptability is used as the environment model for algorithm operation. The counterfactual multi-agent reinforcement learning algorithm is used to identify the patient's current actual anesthesia state. At the same time, the counterfactual baseline data corresponding to various anesthetic drugs are extracted one by one to provide a basis for strategy optimization.

[0043] Based on the deviation between the counterfactual baseline and the preset anesthesia depth target and physiological safety constraints, the counterfactual multi-agent reinforcement learning algorithm updates the strategy network parameters of the drug co-infusion action. Furthermore, by comparing and analyzing the counterfactual baselines of each drug, the preset target depth of anesthesia, and the clinical physiological safety constraints, the deviation information among the three is obtained. The counterfactual multi-agent reinforcement learning algorithm continuously iteratively updates the strategy network parameters corresponding to the drug synergistic infusion action based on this deviation, thereby optimizing the algorithm's decision-making logic.

[0044] A drug co-infusion strategy that meets preset anesthesia depth targets and physiological safety constraints is generated using strategy network parameters.

[0045] Furthermore, based on the updated strategy network parameters, computational decisions are made, and combined with the requirements for depth of anesthesia and the baseline of physiological safety, a reasonable and feasible multi-drug synergistic infusion strategy is generated to ensure that the anesthetic effect meets the target and that the patient's various physiological indicators are within a safe range.

[0046] The drug synergistic infusion strategy is analyzed, the target infusion rate command for each anesthetic drug is extracted, and the target infusion rate command is transmitted to the anesthesia infusion device to drive the actuator to adjust the operating state of the mechanical components.

[0047] Furthermore, the generated drug co-infusion strategy is analyzed and decomposed to separate the target infusion rate command corresponding to each anesthetic drug. The command is transmitted to the anesthesia infusion device in real time. After receiving the command, the device drives the internal actuator to operate and adjusts the working status of various mechanical components.

[0048] The speed at which the infusion tubing is advanced is changed based on the operating state of the mechanical components, thereby controlling the anesthesia infusion equipment to adjust the drug infusion rate.

[0049] Furthermore, by changing the operating status of the internal mechanical components of the equipment, the advancing speed of the infusion pipeline is adjusted, ultimately precisely controlling the infusion rate of various anesthetic drugs by the anesthesia infusion equipment, and completing the final execution action of the entire anesthesia control.

[0050] This embodiment also provides a multi-scale dynamic modeling-based individualized anesthesia control system, including: The extraction module collects multimodal physiological signals from patients during surgery and performs spatiotemporal alignment and feature extraction. The module constructs a hierarchical causal dynamic model at the molecular pharmacology layer, neural network layer, and organ system layer, based on the patient's preoperative prior information and intraoperative real-time data. The matching module inputs the current anesthetic drug infusion parameters into the causal dynamics model, simulates and generates the patient's virtual physiological state in real time, calculates the deviation between the virtual physiological state and the real physiological state, and uses a reinforcement learning algorithm to correct the individualized parameters of the causal dynamics model online based on the deviation, thereby obtaining a dynamic match with the patient's actual physiological response. The strategy module uses the modified causal dynamics model as the environment model and employs a counterfactual multi-agent reinforcement learning algorithm to optimize decision-making and generate a drug co-infusion strategy that meets the preset anesthesia depth target and physiological safety constraints. The control module controls the anesthesia infusion equipment to adjust the drug infusion rate according to the drug co-infusion strategy.

[0051] This embodiment also provides a computer device applicable to the individualized anesthesia control method based on multi-scale dynamic modeling, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the individualized anesthesia control method based on multi-scale dynamic modeling as proposed in the above embodiment.

[0052] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0053] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the individualized anesthesia control method for multi-scale dynamic modeling as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0054] In summary, this invention collects multimodal physiological signals such as intraoperative EEG and hemodynamics, completes spatiotemporal alignment and feature extraction, and combines preoperative prior information with real-time intraoperative data to sequentially build a molecular pharmacology layer, a neuronal network layer, and an organ system layer, and associates them to form a hierarchical causal dynamic model. Then, drug infusion parameters are input to simulate and obtain a virtual physiological state. Based on the deviation, reinforcement learning is used to correct the individualized parameters of the model online to match the patient's actual physiological response. Using the corrected model as the environmental model, a counterfactual multi-agent reinforcement learning algorithm is used to optimize and generate a drug co-infusion strategy in combination with anesthesia depth and physiological safety constraints. The strategy is analyzed and instructions are issued to control the anesthesia infusion equipment to adjust the infusion rate of various anesthetic drugs, thereby achieving precise individualized anesthesia control.

[0055] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for individualized anesthesia control through multi-scale dynamic modeling, characterized in that: This includes collecting multimodal physiological signals from patients during surgery and performing spatiotemporal alignment and feature extraction; Based on the patient’s preoperative prior information and intraoperative real-time data, a hierarchical causal dynamic model was constructed, consisting of molecular pharmacology layer, neural network layer and organ system layer. The current anesthetic drug infusion parameters are input into the causal dynamics model to generate the patient's virtual physiological state in real time. The deviation between the virtual physiological state and the real physiological state is calculated. The individualized parameters of the causal dynamics model are corrected online based on the deviation using a reinforcement learning algorithm to obtain a dynamic match with the patient's actual physiological response. The modified causal dynamics model is used as the environment model. The counterfactual multi-agent reinforcement learning algorithm is used to optimize decision-making and generate a drug co-infusion strategy that meets the preset anesthesia depth target and physiological safety constraints. Based on the drug synergistic infusion strategy, the drug infusion rate is adjusted by controlling the anesthesia infusion equipment.

2. The individualized anesthesia control method based on multi-scale dynamic modeling as described in claim 1, characterized in that: The process of collecting multimodal physiological signals from patients during surgery and performing spatiotemporal alignment and feature extraction includes: Multimodal physiological signals of the patient during the operation were collected. The multimodal physiological signals included electroencephalogram (EEG) signals reflecting the state of the central nervous system and hemodynamic parameters reflecting the state of the circulatory system. The EEG signals and hemodynamic parameters were spatiotemporally aligned and feature extracted to obtain multimodal physiological signals after spatiotemporal alignment and feature extraction.

3. The individualized anesthesia control method based on multi-scale dynamic modeling as described in claim 2, characterized in that: The hierarchical causal dynamic model constructed, comprising molecular pharmacology layer, neuronal network layer, and organ system layer, includes... A molecular pharmacology layer is constructed based on the patient’s preoperative prior information and intraoperative real-time data. The molecular pharmacology layer is used to characterize the effects of the patient’s genotype and drug-metabolizing enzyme activity on pharmacokinetic parameters. The output of the molecular pharmacology layer is used as input to construct a neural network layer. The neural network layer uses a millisecond-level time sliding window strategy to extract EEG alpha band phase-locked connection features that reflect key pathways of consciousness state switching. The output of the neural network layer is used as input to construct the organ system layer, which integrates heart rate fractal dynamics features and end-tidal carbon dioxide parameters to establish a cross-scale physiological response mapping relationship. By hierarchically associating the molecular pharmacology layer, the neural network layer, and the organ system layer, a hierarchical causal dynamic model containing the molecular pharmacology layer, the neural network layer, and the organ system layer is obtained.

4. The individualized anesthesia control method based on multi-scale dynamic modeling as described in claim 3, characterized in that: The calculation of the deviation between the virtual physiological state and the real physiological state includes... The current anesthetic drug infusion parameters are input into the causal dynamic model to generate a virtual physiological state of the patient in real time. Based on the real physiological state and the patient's virtual physiological state, the deviation between the virtual physiological state and the real physiological state is calculated.

5. The individualized anesthesia control method based on multi-scale dynamic modeling as described in claim 4, characterized in that: The dynamic matching of the patient's actual physiological response includes... Based on the reinforcement learning algorithm, the deviation between the virtual physiological state and the real physiological state is received as a feedback signal to correct the action instructions of the causal dynamic model parameters. By using action commands to update the individualized parameters of the causal dynamics model online, the state transition mechanism within the causal dynamics model can be changed. The causal dynamics model is run based on the updated individualized parameters, so that the generated simulation results are an individualized model that dynamically matches the patient's actual physiological response.

6. The individualized anesthesia control method based on multi-scale dynamic modeling as described in claim 5, characterized in that: The generation of a drug co-infusion strategy that satisfies the preset anesthesia depth target and physiological safety constraints includes, The modified causal dynamics model is used as the environment model. The counterfactual multi-agent reinforcement learning algorithm is used to obtain the current anesthesia state and extract the counterfactual baseline of each drug. Based on the deviation between the counterfactual baseline and the preset anesthesia depth target and physiological safety constraints, the counterfactual multi-agent reinforcement learning algorithm updates the strategy network parameters of the drug co-infusion action. A drug co-infusion strategy that meets preset anesthesia depth targets and physiological safety constraints is generated using strategy network parameters.

7. The individualized anesthesia control method based on multi-scale dynamic modeling as described in claim 6, characterized in that: The control of the anesthesia infusion device includes adjusting the drug infusion rate. The drug co-infusion strategy is analyzed, the target infusion rate command corresponding to each anesthetic drug is extracted, and the target infusion rate command is transmitted to the anesthesia infusion device to drive the actuator to adjust the operating state of the mechanical components; The speed at which the infusion tubing is advanced is changed based on the operating state of the mechanical components, thereby controlling the anesthesia infusion equipment to adjust the drug infusion rate.

8. A multi-scale dynamic modeling-based individualized anesthesia control system, based on the multi-scale dynamic modeling-based individualized anesthesia control method according to any one of claims 1 to 7, characterized in that: This includes an extraction module that collects multimodal physiological signals from patients during surgery and performs spatiotemporal alignment and feature extraction. The module constructs a hierarchical causal dynamic model at the molecular pharmacology layer, neural network layer, and organ system layer, based on the patient's preoperative prior information and intraoperative real-time data. The matching module inputs the current anesthetic drug infusion parameters into the causal dynamics model, simulates and generates the patient's virtual physiological state in real time, calculates the deviation between the virtual physiological state and the real physiological state, and uses a reinforcement learning algorithm to correct the individualized parameters of the causal dynamics model online based on the deviation, thereby obtaining a dynamic match with the patient's actual physiological response. The strategy module uses the modified causal dynamics model as the environment model and employs a counterfactual multi-agent reinforcement learning algorithm to optimize decision-making and generate a drug co-infusion strategy that meets the preset anesthesia depth target and physiological safety constraints. The control module controls the anesthesia infusion equipment to adjust the drug infusion rate according to the drug co-infusion strategy.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the individualized anesthesia control method based on multi-scale dynamic modeling as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the individualized anesthesia control method based on multi-scale dynamic modeling as described in any one of claims 1 to 7.

Images