Anesthetic gas output parameter determination method, medium and system

Abstract

The invention provides an anesthetic gas output parameter determination method, medium and system, and belongs to the technical field of anesthetic equipment.The anesthetic gas output parameter determination method comprises the steps that according to surgical anesthetic data of historical patients, the surgical anesthetic data comprise anesthetic demand parameters, patient physiological parameters and anesthetic gas output parameters; a simplified model of gas interaction and combination effects is established through molecular simulation and quantum chemistry calculation, and the accumulation effect of gas in a body is simulated by combining calculation fluid dynamics. And combining the analytical models with a neural network to form a mixed model, training the mixed model by using historical data, and finally obtaining a model capable of predicting the optimal anesthetic gas output parameters of the current patient. According to the method, the medium and the system, historical data, an advanced simulation technology and a machine learning method are comprehensively utilized, and the technical problem that according to an existing anesthetic gas output parameter determining technology, the accumulation effect of anesthetic medicine in the body of a patient is often neglected, and consequently judged output parameters are not accurate enough is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of anesthesia equipment, and in particular, relates to a method, a medium and a system for determining an anesthetic gas output parameter. Background Art

[0002] Surgical anesthesia is a key supporting technology in medical practice, and its precise control has an important impact on the success rate of surgery and patient safety. In the current clinical anesthesia practice, doctors mainly rely on personal experience to manually control the amount and type of anesthetic drugs according to the type of surgery, expected duration, and patient physiological indicators to maintain the patient in the best anesthesia state. However, this control method that relies on doctor experience has certain limitations.

[0003] First, the physiological characteristics of different patients vary greatly. Even for the same type of surgery, the required anesthesia depth and drug combination may vary significantly. It is difficult for doctors to accurately grasp the optimal anesthesia parameters for each individual patient, and they often need to make repeated adjustments to achieve the ideal anesthesia effect, which increases the risk of surgery.

[0004] Secondly, during the operation, the patient's physiological state will change dynamically over time, and the corresponding optimal anesthesia parameters also need to be adjusted in real time. However, it is difficult for doctors to capture these changes in time and make precise adjustments, which can easily lead to over-deep or insufficient anesthesia.

[0005] Furthermore, the interactions and accumulation effects of anesthetic drugs in the body are also relatively complex, and it is difficult to determine the best drug combination and dosage through empirical judgment. If used improperly, it may cause serious adverse reactions.

[0006] Existing technologies for determining anesthetic gas output parameters often ignore the accumulation effect of anesthetic drugs in the patient's body, resulting in inaccurate output parameters. Summary of the invention

[0007] In view of this, the present invention provides a method, medium and system for determining anesthetic gas output parameters, which can solve the technical problem that the existing anesthetic gas output parameter determination technology often ignores the accumulation effect of anesthetic drugs in the patient's body, resulting in inaccurate output parameters.

[0008] The present invention is achieved by comprising the following steps:

[0009] S10, obtaining multiple sets of historical anesthesia data of different stages of the patient's surgical process from the anesthesia information management system, including surgical anesthesia demand parameters, patient physiological parameters, and anesthetic gas output parameters;

[0010] S20, establishing a simplified gas interaction model using molecular simulation software Gaussian or GAMESS according to the anesthetic gas output parameters, taking into account the mutual influence of multiple anesthetic gases when mixed;

[0011] S30, establishing a simplified model of gas combination effects using quantum chemical calculation software Spartan or Q-Chem according to the anesthetic gas output parameters, for predicting the effects and interactions of different gas combinations;

[0012] S40, using computational fluid dynamics software ANSYS Fluent according to the anesthetic gas output parameters, establishing a simplified gas accumulation model that takes into account gas diffusion characteristics, so as to predict the cumulative effects of different anesthetic gases in the patient's body;

[0013] S50, establishing a hybrid model of a neural network + an analytical model, specifically: embedding the analytical model in the neural network, including the simplified gas interaction model, the simplified gas combination effect model and the simplified gas accumulation model, to form a hybrid model;

[0014] S60, using the anesthesia data to train the hybrid model to obtain an anesthetic gas output parameter model, wherein the training input is a surgical anesthesia requirement parameter and a patient physiological parameter, and the training output is an anesthetic gas output parameter;

[0015] S70, obtaining surgical anesthesia requirement parameters and real-time physiological parameters of the current patient, inputting the anesthetic gas output parameter model, and obtaining the optimal anesthetic gas output parameters of the current patient.

[0016] Among them, the surgical anesthesia requirement parameters include the type of surgery, expected duration of the surgery, anesthesia depth requirements, patient age, weight, height and medical history.

[0017] Among them, the patient's physiological parameters include heart rate, blood pressure, body temperature, respiratory rate, blood oxygen saturation, electroencephalogram index, degree of muscle relaxation and end-tidal carbon dioxide concentration.

[0018] The anesthetic gas output parameters include gas category, and flow rate and concentration of each component.

[0019] Wherein, the hybrid model also includes a multi-layer perceptron neural network as the main structure.

[0020] The input layer of the multi-layer perceptron neural network receives the surgical anesthesia demand parameters and the patient's physiological parameters, and is provided with multiple hidden layers using the ReLU activation function; a parallel branch is added after the first hidden layer, the simplified gas interaction model is embedded, and the gas interaction intensity vector is output; a parallel branch is added after the second hidden layer, the simplified gas combination effect model is embedded, and the gas combination effect vector is output; a parallel branch is added after the third hidden layer, the simplified gas accumulation model is embedded, and the estimated gas accumulation degree vector is output; the output of each analytical model is added to the output of the corresponding layer of the main neural network using residual connection; an attention layer is added after the last hidden layer, the weight of each sub-model output is dynamically calculated and multiplied with the sub-model output and then summed; finally, the anesthetic gas output parameter is output through the fully connected layer; the hybrid model simultaneously optimizes the neural network weights and the simplified parameters of the analytical model through the back propagation algorithm, and batch normalization and dropout technology are used to improve the generalization ability of the model. The analytical model is embedded in the neural network. This embedding method can greatly reduce the calculation amount of the hybrid model and reduce the calculation time, which is helpful to achieve the fine-grained time anesthetic gas output parameter adjustment during the operation.

[0021] The simplified gas interaction model takes into account parameters such as electrostatic interaction, van der Waals force, hydrogen bonding, and molecular orbital energy level differences between gas molecules, and is generated by geometry optimization and energy calculation using the B3LYP functional and 6-31G(d) basis set of Gaussian software according to the principle of density functional theory (DFT). The simplified gas interaction model (based on Gaussian or GAMESS software) is specifically expressed as follows:

[0022]

[0023] In the formula, I ij is the interaction strength between gas i and gas j; E int is the interaction energy calculated by Gaussian / GAMESS (kJ / mol); ΔG mix is the mixing Gibbs free energy (kJ / mol); HOMO i and LUMO j are the highest occupied molecular orbital energy of gas i and the lowest unoccupied molecular orbital energy of gas j (eV), respectively; k (r) is the basis set function; N basis is the number of basis set functions; k ij , α ij , β ij and γ ij is the unknown coefficient; c k is the molecular orbital coefficient.

[0024] Parameter acquisition method:

[0025] E int , ΔG mix HOMO i LUMO j : Obtained through quantum chemical calculations using Gaussian or GAMESS software.

[0026] χ k (r) and c k : Extracted from Gaussian or GAMESS output files.

[0027] k ij , α ij , β ij and γ ij :Obtained through regression analysis of historical data.

[0028] The simplified gas combination effect model takes into account the anesthetic efficacy, concentration, molecular polarity, ionization energy, electron affinity and other parameters of various gases, and is generated based on the principles of quantum chemistry and statistical thermodynamics using the complete active space self-consistent field (CASSCF) method and natural bond orbital (NBO) analysis function of Spartan software. The simplified gas combination effect model (based on Spartan or Q-Chem software) is specifically expressed as follows:

[0029]

[0030] In the formula, E comb is the overall effect of the gas combination; E i is the anesthetic effect of a single gas i; w i is the weight coefficient of gas i; k ij is the interaction coefficient between gas i and gas j; I ij is the interaction strength between gas i and gas j (calculated by the simplified gas interaction model); ΔE CASSCF is the self-consistent field energy difference in the complete active space (kJ / mol); Q i is the Mulliken charge of gas i; V i is the molecular electrostatic potential of gas i (V); S conf is the conformational entropy (J / mol·K); n is the number of anesthetic gas species; λ, μ and ν are unknown parameters.

[0031] Parameter acquisition method:

[0032] E i : The individual anesthetic effects of each gas were obtained by querying the pharmacology database.

[0033] w i: Determined according to the concentration ratio of each gas in the mixture.

[0034] k ij : Obtained through multiple regression analysis of historical data.

[0035] ΔE CASSCF , Q i 、V i and S conf : Obtained by quantum chemical calculations using Spartan or Q-Chem software.

[0036] λ, μ and ν: obtained by nonlinear fitting of historical data.

[0037] The simplified gas accumulation model takes into account the diffusion coefficient, solubility, metabolic rate, blood gas distribution coefficient, tissue perfusion rate and other parameters of the gas, and is generated based on the principles of computational fluid dynamics (CFD) and pharmacokinetics using the race transport model and user-defined function (UDF) of ANSYS Fluent software. The simplified gas accumulation model (based on ANSYS Fluent software) is specifically expressed as follows:

[0038]

[0039] Where ρ is the density of the mixture (kg / m 3 );Y i is the mass fraction of gas i; t is time (s); is the velocity vector (m / s); is the diffusion flux of gas i (kg / m 2 ·s); R i is the net generation rate of gas i (kg / m 3 ·s); S i For user-defined source terms (kg / m 3 ·s); D i,m is the mass diffusion coefficient of gas i in the mixture (m 2 / s); μ t is the turbulent viscosity (Pa·s); Sc t is the turbulent Schmidt number; D T,i is the thermal diffusion coefficient (kg / m·s); T is the temperature (K).

[0040] Parameter acquisition method:

[0041] ρ, T: Obtained by solving the Navier-Stokes equations and energy equation using ANSYS Fluent.

[0042] Y i:The initial value is obtained from the anesthetic gas output parameter, and then the transmission equation is solved by ANSYS Fluent.

[0043] D i,m : Obtained from the ANSYS Fluent material database or user-defined.

[0044] μ t : Calculated using the turbulence model selected in ANSYS Fluent.

[0045] Sc t : Usually 0.7-0.9, which can be set in ANSYS Fluent.

[0046] D T,i : The Soret effect can be enabled and set in ANSYS Fluent.

[0047] R i and S i : It is implemented through ANSYS Fluent's user-defined functions (UDFs) and can include chemical reaction kinetics and metabolic processes.

[0048] Specifically, step S10 includes: obtaining multiple groups of anesthesia data of different stages of historical patient operations from the anesthesia information management system, wherein the anesthesia data includes surgical anesthesia demand parameters, patient physiological parameters, and anesthetic gas output parameters. The surgical anesthesia demand parameters include the type of operation, expected operation duration, anesthesia depth requirements, patient age, weight, height, and medical history. The patient's physiological parameters include heart rate, blood pressure, body temperature, respiratory rate, blood oxygen saturation, electroencephalogram index, muscle relaxation, and end-tidal carbon dioxide concentration.

[0049] The specific implementation of step S20 is as follows: according to the anesthetic gas output parameters, a simplified model of gas interaction is established using molecular simulation software Gaussian or GAMESS, taking into account the mutual influence of multiple anesthetic gases when mixed. The simplified model of gas interaction can be expressed as: the interaction intensity I between gas i and gas j ij Equal to the interaction energy E int , Gibbs free energy of mixing ΔG mix 、Molecular orbital energy level difference; (HOMO i -LUMO j ) 2 And a linear combination of factors such as the molecular electron cloud distribution, where each coefficient needs to be obtained through regression analysis of historical data.

[0050] The specific implementation of step S30 is as follows: based on the anesthetic gas output parameters, a simplified model of the gas combination effect is established using quantum chemical calculation software Spartan or Q-Chem. The model takes into account the anesthetic efficacy, concentration, molecular polarity, ionization energy, electron affinity and other parameters of various gases, and combines the influence of gas interactions to predict the overall anesthetic effect of different gas combinations. The simplified model of the gas combination effect can be expressed as: The overall effect of the gas combination E comb Equal to the individual effect of each gas E and the interaction strength between gases I ij , quantum effect, ΔE CASSCF 、Molecular electrostatic properties Q i ·V i and conformational entropy S conf These parameters are obtained through quantum chemical calculations, while the weight coefficients need to be determined through regression analysis.

[0051] The specific implementation of step S40 is as follows: according to the anesthetic gas output parameters, a simplified gas accumulation model considering the gas diffusion characteristics is established using computational fluid dynamics software ANSYS Fluent. The model describes the accumulation process of anesthetic gas in the patient's body, including parameters such as gas diffusion coefficient, solubility, metabolic rate, blood gas distribution coefficient, tissue perfusion rate, etc. The simplified gas accumulation model can be expressed as: the mass fraction Y of gas i i The change with time t is equal to the gas diffusion flux Net generation rate R i and user-defined source term S i These parameters can be calculated and set through the solver, material database and user-defined functions of ANSYS Fluent software.

[0052] The specific implementation of step S50 is as follows: establishing a hybrid prediction model combining a neural network and an analytical model. In a multi-layer perceptron neural network, parallel branches are added after different hidden layers, and the simplified gas interaction model, the simplified gas combination effect model and the simplified gas accumulation model established in the above steps are respectively embedded. Through residual connection and attention mechanism, the output of the analytical model is integrated with the output of the neural network body. This hybrid model structure gives full play to the learning ability of the neural network and the interpretability of the analytical model, while also improving the computational efficiency.

[0053] The specific implementation of step S60 is as follows: the historical anesthesia data obtained in step S10 is used to train the hybrid model established in step S50 to obtain the final anesthetic gas output parameter prediction model. The surgical anesthesia requirement parameters and the patient's physiological parameters are used as training inputs, and the corresponding anesthetic gas output parameters are used as training target outputs. The back propagation algorithm is used to simultaneously optimize the neural network weights and analytical model parameters, and batch normalization and dropout techniques are used to improve the model generalization ability. After the training is completed, the model parameters are saved for subsequent predictions.

[0054] The specific implementation of step S70 is as follows: obtain the surgical anesthesia requirement parameters and real-time physiological parameters of the current patient, input the anesthetic gas output parameter prediction model trained in step S60, and obtain the optimal anesthetic gas output parameters required by the current patient. Monitor and obtain the current patient's surgical type, expected surgical duration, anesthesia depth requirements, age, weight, height, medical history and other demand parameters, as well as heart rate, blood pressure, body temperature, respiratory rate, blood oxygen saturation and other physiological parameters in real time. Input these parameters into the prediction model, and output the optimal anesthetic gas output parameters required by the current patient after forward calculation. Feedback the output parameters to the anesthesia equipment to adjust the output of the anesthetic gas in real time to ensure that the patient is in the best anesthesia state.

[0055] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the above-mentioned method for determining an anesthetic gas output parameter.

[0056] A third aspect of the present invention provides an anesthetic gas output parameter determination system, which includes the above-mentioned computer-readable storage medium.

[0057] Compared with the prior art, the method, medium and system for determining an anesthetic gas output parameter provided by the present invention have the following beneficial effects:

[0058] First, this method makes full use of historical anesthesia data and establishes a hybrid prediction model that can accurately predict the optimal anesthetic gas output parameters required by different patients. This model combines the powerful learning ability of neural networks and the interpretability of physical / chemical analytical models. It can not only effectively capture the complex mapping relationship between surgical anesthesia demand parameters, patient physiological parameters and optimal anesthesia parameters, but also fully consider the interaction, combination effect and accumulation characteristics of anesthetic gases in the body. Through this hybrid modeling method, the accuracy and reliability of prediction can be greatly improved, providing more scientific decision support for clinicians.

[0059] Secondly, this method realizes real-time monitoring of the patient's state changes during surgery and dynamic regulation of anesthesia parameters. By acquiring the patient's physiological parameters in real time and inputting them into the prediction model, the optimal anesthetic gas output parameters required at the moment can be output in a timely manner. This fine regulation based on real-time feedback ensures the optimal maintenance of anesthesia depth throughout the surgery, avoids excessive or insufficient anesthesia, and greatly improves surgical safety.

[0060] In addition, this method also fully considers the accumulation effect of anesthetic gases in the body. By establishing a simplified model of gas accumulation and combining gas interactions and combination effects, the accumulation of different gas combinations in the body can be more accurately predicted, providing doctors with a more comprehensive decision-making basis and reducing the risk of adverse reactions.

[0061] In summary, the present invention can use historical anesthesia data to accurately predict the optimal anesthetic gas combination and its output parameters for different patients, and can adapt to the dynamic changes of the patient's state during surgery to ensure the fine control of the depth of anesthesia. At the same time, the method can also consider the interaction of anesthetic gases and their cumulative effects in the body, providing more scientific and reliable decision support for clinicians.

[0062] Therefore, the present invention solves the technical problem that the existing anesthetic gas output parameter determination technology often ignores the accumulation effect of anesthetic drugs in the patient's body, resulting in inaccurate output parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 A flow chart of the method provided by the present invention. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0065] like Figure 1 FIG. 1 is a flow chart of a method for determining an anesthetic gas output parameter provided by the present invention. The method comprises the following steps:

[0066] S10, obtaining multiple sets of historical anesthesia data of different stages of the patient's surgical process from the anesthesia information management system, including surgical anesthesia demand parameters, patient physiological parameters, and anesthetic gas output parameters;

[0067] S20. Use molecular simulation software Gaussian or GAMESS to establish a simplified gas interaction model based on the anesthetic gas output parameters, taking into account the mutual influence of multiple anesthetic gases when mixed;

[0068] S30. Based on the anesthetic gas output parameters, a simplified model of gas combination effect is established using quantum chemical calculation software Spartan or Q-Chem to predict the effects and interactions of different gas combinations;

[0069] S40. Based on the anesthetic gas output parameters, a simplified gas accumulation model considering the gas diffusion characteristics is established using the computational fluid dynamics software ANSYS Fluent to predict the cumulative effects of different anesthetic gases in the patient's body;

[0070] S50, establishing a hybrid model of a neural network + an analytical model, specifically: embedding an analytical model in the neural network, including a simplified gas interaction model, a simplified gas combination effect model, and a simplified gas accumulation model, to form a hybrid model;

[0071] S60, using anesthesia data to train the hybrid model to obtain an anesthetic gas output parameter model, wherein the training input is the surgical anesthesia requirement parameter and the patient's physiological parameter, and the training output is the anesthetic gas output parameter;

[0072] S70, obtaining surgical anesthesia requirement parameters and real-time physiological parameters of the current patient, inputting into an anesthetic gas output parameter model, and obtaining optimal anesthetic gas output parameters of the current patient.

[0073] The specific implementation methods of the above steps are described in detail below:

[0074] The specific implementation method of step S10 is: first, multiple groups of historical anesthesia data of different stages of the patient's surgical process are obtained from the anesthesia information management system, including surgical anesthesia demand parameters, patient physiological parameters, and anesthetic gas output parameters. The surgical anesthesia demand parameters include the type of surgery, the expected duration of the operation, the depth of anesthesia required, the patient's age, weight, height, and medical history, etc.; the patient's physiological parameters include heart rate, blood pressure, body temperature, respiratory rate, blood oxygen saturation, electroencephalogram index, muscle relaxation, and end-tidal carbon dioxide concentration, etc. These historical anesthesia data are critical for the subsequent establishment of a gas output parameter prediction model, and can provide a basis for the training and verification of the model.

[0075] The specific implementation of step S20 is: according to the obtained anesthetic gas output parameters, a simplified model of gas interaction is established using molecular simulation software Gaussian or GAMESS. This step mainly considers the mutual influence of multiple anesthetic gases when mixed, and establishes a mathematical model to describe the intensity of interaction between gases. Specifically, the model can be expressed as:

[0076]

[0077]

[0078] Among them, I ij represents the interaction strength between gas i and gas j, E int The interaction energy calculated by Gaussian / GAMESS reflects the electrostatic interaction, van der Waals force, hydrogen bonding, etc. between gas molecules; ΔG mix is the mixing Gibbs free energy, which reflects the energy change during gas mixing; (HOMO i -LUMO j ) 2 The influence of the difference in gas molecular orbital energy levels on the interaction is taken into account; k (r) is the basis set function, c k is the molecular orbital coefficient, which describes the effect of the electron cloud distribution of gas molecules on the interaction. ij , α ij , β ij and γ ij The coefficient is unknown and needs to be obtained through regression analysis of historical data. The simplified gas interaction model established in this way can better describe the mutual influence of multiple anesthetic gases when mixed.

[0079] The specific implementation of step S30 is: based on the obtained anesthetic gas output parameters, a simplified model of the gas combination effect is established using quantum chemical calculation software Spartan or Q-Chem. The model mainly considers the anesthetic efficacy, concentration, molecular polarity, ionization energy, electron affinity and other parameters of various gases, and combines the influence of gas interactions to predict the overall anesthetic effect of different gas combinations. The model can be expressed as:

[0080]

[0081] Among them, E comb is the overall anesthetic effect of the gas combination, E i is the anesthetic effect of a single gas i, w i is the weight coefficient of gas i, k ij is the interaction coefficient between gas i and gas j, I ij is the interaction strength between the two, ΔE CASSCF is the energy difference of the self-consistent field in the completely active space, reflecting the quantum effect, Q i is the Mulliken charge of gas i, V i is the molecular electrostatic potential of gas i, S conf is the conformational entropy. These parameters are obtained by quantum chemical calculations using Spartan or Q-Chem, and k ij The coefficients need to be obtained through regression analysis. The simplified model of gas combination effect constructed in this way can better predict the impact of different gas combinations on anesthesia effect.

[0082] The specific implementation of step S40 is: based on the obtained anesthetic gas output parameters, using the computational fluid dynamics software ANSYS Fluent, a simplified gas accumulation model considering the gas diffusion characteristics is established. The model mainly describes the accumulation process of anesthetic gas in the patient's body, taking into account the gas diffusion coefficient, solubility, metabolic rate, blood gas distribution coefficient, tissue perfusion rate and other parameters. The mathematical expression of the model is as follows:

[0083]

[0084] Where ρ is the mixture density, Y i is the mass fraction of gas i, is the velocity vector, is the diffusion flux of gas i, R i is the net generation rate of gas i, S i is a user-defined source term, D i,m is the mass diffusion coefficient of gas i in the mixture, μ t is the turbulent viscosity, Sc t is the turbulent Schmidt number, D T,i is the thermal diffusion coefficient, and T is the temperature. These parameters can be calculated and set by the solver, material database, and user-defined function (UDF) of ANSYS Fluent. The simplified gas accumulation model can predict the cumulative effect of different anesthetic gases in the patient's body.

[0085] These models utilize specific functions and parameters of software such as Gaussian / GAMESS, Spartan / Q-Chem and ANSYS Fluent. These models can more accurately describe the interaction, combination effect and accumulation process of anesthetic gases, and also facilitate the use of these software for calculations and simulations in practical applications.

[0086] The specific implementation of step S50 is: establish a hybrid model of neural network + analytical model. Specifically, in the structure of a multi-layer perceptron (MLP) neural network, parallel branches are added after different hidden layers, and the three analytical models established in the above steps are embedded to form a hybrid model. The input layer of the neural network receives the surgical anesthesia demand parameters and the patient's physiological parameters, and is provided with multiple hidden layers using the ReLU activation function. Add parallel branches after the first hidden layer, embed the simplified gas interaction model, and output the gas interaction intensity vector; add parallel branches after the second hidden layer, embed the simplified gas combination effect model, and output the gas combination effect vector; add parallel branches after the third hidden layer, embed the simplified gas accumulation model, and output the estimated gas accumulation degree vector. Then, the output of each analytical model is added to the output of the corresponding layer of the main neural network using residual connection. Add an attention layer after the last hidden layer, dynamically calculate the weight of each sub-model output and multiply it with the sub-model output and sum it. Finally, the anesthetic gas output parameter is output through the fully connected layer.

[0087] This embedded hybrid model structure can give full play to the powerful learning ability of neural networks, while using the prior knowledge of physical / chemical analytical models, greatly improving the interpretability and generalization of the model. At the same time, through residual connections and attention mechanisms, it can effectively reduce the amount of calculation, improve real-time performance, and meet the time-granular adjustment requirements of anesthesia gas output parameters during surgery.

[0088] The specific implementation of step S60 is: using the historical anesthesia data obtained in step S10 to train the hybrid model established in step S50 to obtain the final anesthetic gas output parameter prediction model. The specific training process is as follows:

[0089] 1) The surgical anesthesia requirement parameters and the patient's physiological parameters are used as training inputs, and the corresponding anesthetic gas output parameters are used as training target outputs.

[0090] 2) The back-propagation algorithm is used to simultaneously optimize the weight parameters of the neural network and the simplified parameters in the analytical model.

[0091] 3) During the training process, techniques such as batch normalization and dropout are used to improve the generalization ability of the model.

[0092] 4) After the training is completed, the final trained model parameters are saved as the anesthetic gas output parameter model for subsequent predictions.

[0093] Through this training method, we can make full use of historical anesthesia data and learn the complex mapping relationship between surgical anesthesia demand parameters, patient physiological parameters and optimal anesthetic gas output parameters, providing a basis for subsequent real-time predictions.

[0094] The specific implementation method of step S70 is: obtain the surgical anesthesia requirement parameters and real-time physiological parameters of the current patient, input the anesthetic gas output parameter prediction model trained in step S60, and obtain the optimal anesthetic gas output parameters of the current patient.

[0095] This process can be described as:

[0096] 1) Real-time monitoring and acquisition of the patient's surgical anesthesia requirement parameters (type of surgery, estimated duration of surgery, anesthesia depth requirements, patient age, weight, height, medical history, etc.) and physiological parameters (heart rate, blood pressure, body temperature, respiratory rate, blood oxygen saturation, electroencephalogram index, degree of muscle relaxation, end-tidal carbon dioxide concentration, etc.).

[0097] 2) Inputting these parameters acquired in real time into the anesthetic gas output parameter prediction model trained in step S60.

[0098] 3) After forward calculation, the model outputs the optimal anesthetic gas output parameters required by the current patient, including gas category and the flow and concentration of each component.

[0099] 4) Feedback the optimal output parameters obtained to the anesthesia equipment, adjust the output of anesthetic gas in real time, and ensure that the patient is in the best anesthesia state.

[0100] Through such real-time prediction and feedback control, the depth of anesthesia during surgery can be accurately controlled to avoid excessive or insufficient anesthesia, thereby improving surgical safety and effectiveness.

[0101] In summary, the specific implementation of this invention involves multiple steps, including obtaining historical anesthesia data, establishing a simplified model of gas interaction, gas combination effect and gas accumulation, constructing a hybrid prediction model of neural network + analytical model, and using this model to perform real-time anesthetic gas output parameter prediction and feedback control. These steps fully combine physical / chemical knowledge and machine learning technology to form a novel method for determining anesthetic gas output parameters, which can significantly improve the accuracy and safety of surgical anesthesia.

[0102] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the above-mentioned method for determining an anesthetic gas output parameter.

[0103] A third aspect of the present invention provides an anesthetic gas output parameter determination system, which includes the above-mentioned computer-readable storage medium.

[0104] Specifically, the principle of the present invention is:

[0105] 1. Obtain historical anesthesia data: Obtain multiple sets of historical anesthesia data of different stages of the patient's surgical process from the anesthesia information management system, including surgical anesthesia demand parameters, patient physiological parameters, and actual anesthetic gas output parameters. These historical data provide the basis for subsequent modeling and training.

[0106] 2. Establish a simplified model of gas interaction: Based on the obtained anesthetic gas output parameters, use the molecular simulation software Gaussian or GAMESS to establish a mathematical model that describes the intensity of gas interaction. This model takes into account factors such as electrostatic interaction, van der Waals force, and hydrogen bonding between molecules, providing a basis for evaluating the compatibility of different gas combinations.

[0107] 3. Establish a simplified model of gas combination effect: Based on the output parameters of anesthetic gases, a model is established to predict the overall anesthetic effect of different gas combinations using quantum chemical calculation software Spartan or Q-Chem. This model comprehensively considers factors such as the anesthetic efficacy, molecular polarity, and ionization energy of various gases, and combines gas interactions to provide guidance for determining the optimal gas combination.

[0108] 4. Establish a simplified gas accumulation model: Using the computational fluid dynamics software ANSYS Fluent, an accumulation model that takes into account the gas diffusion characteristics is established. This model describes the dynamic accumulation process of anesthetic gases in the patient's body, including factors such as diffusion and metabolism, and provides support for predicting the long-term effects of different gas combinations in the body.

[0109] 5. Construct a hybrid prediction model: The above three analytical models are embedded into a neural network structure to form a hybrid prediction model. This model fully utilizes the learning ability of the neural network and the interpretability of the analytical model, and can accurately predict the optimal anesthetic gas output parameters required by different patients. Through residual connections and attention mechanisms, the computational efficiency of the model is also greatly improved.

[0110] 6. Model training and application: Use historical anesthesia data to train the hybrid prediction model and obtain the final anesthetic gas output parameter prediction model. In practical applications, the current patient's surgical anesthesia demand parameters and real-time physiological parameters are obtained, and the optimal anesthetic gas output parameters are input into the prediction model to obtain the required optimal anesthetic gas output parameters, which are then fed back to the anesthesia equipment for real-time regulation.

[0111] This hybrid modeling method based on machine learning and physical / chemical analysis makes full use of the patient characteristics contained in historical data, while incorporating prior knowledge on the interaction, combination effect, and accumulation of anesthetic gases in the body. Through this collaborative modeling approach, the accuracy and interpretability of predictions can be greatly improved, meeting the need for accurate and real-time decision support during surgical anesthesia.

[0112] In order to better understand and implement the present invention, a specific embodiment 1 of the present invention is provided below: The specific steps of this embodiment 1 are described in detail as follows: The specific implementation method of step S10 is: obtaining multiple groups of historical anesthesia data of different stages of the patient's surgical process from the anesthesia information management system. The acquired data includes the surgical anesthesia requirement parameter X m 、Patient physiological parameters X p And the anesthetic gas output parameter Y.

[0113] Surgical anesthesia requirement parameter X m Includes: Surgery Type x m1 , Estimated duration of surgery x m2 , Anesthesia depth requirement x m3 , patient age x m4 , weight x m5 , height x m6 And past medical history x m7 wait.

[0114] Patient physiological parameters X p Includes: Heart Rate x p1 , blood pressure x p2 , body temperature x p3 , breathing rate x p4 , blood oxygen saturation x p5 , EEG index x p6 , muscle relaxation degree x p7 and end-tidal carbon dioxide concentration x p8 wait.

[0115] Anesthetic gas output parameters Y include: Gas category y 1 , and the flow rate y of each component 2 and concentration y 3 .

[0116] These historical anesthesia data will provide a basis for the subsequent establishment of a gas output parameter prediction model.

[0117] The specific implementation of step S20 is: according to the obtained anesthetic gas output parameter Y, a simplified model of gas interaction is established using molecular simulation software Gaussian or GAMESS. This model mainly considers the mutual influence of multiple anesthetic gases when mixed, which can be expressed as:

[0118]

[0119] Among them, I ij represents the interaction strength between gas i and gas j. E int ΔG is the interaction energy calculated by Gaussian / GAMESS, which reflects the electrostatic interaction, van der Waals force, hydrogen bonding, etc. between gas molecules.mix is the mixing Gibbs free energy, which reflects the energy change during gas mixing. i and LUMO j are the highest occupied molecular orbital energy of gas i and the lowest unoccupied molecular orbital energy of gas j, respectively, reflecting the influence of the difference in gas molecular orbital energy levels on the interaction. k (r) is the basis set function, c k is the molecular orbital coefficient, which describes the effect of the electron cloud distribution of gas molecules on the interaction. ij , α ij , β ij and γ ij It is an undetermined coefficient and needs to be obtained through regression analysis of historical data.

[0120] The specific implementation of step S30 is: based on the obtained anesthetic gas output parameter Y, using quantum chemical calculation software Spartan or Q-Chem, a simplified model of gas combination effect is established. This model mainly considers the anesthetic efficacy E of various gases. i , concentration w i , molecular polarity, ionization energy, electron affinity and other parameters, combined with the influence of gas interactions, to predict the overall anesthetic effect E of different gas combinations comb The model can be expressed as:

[0121]

[0122] Among them, E i is the anesthetic effect of a single gas i, w i is the weight coefficient of gas i, k ij is the interaction coefficient between gas i and gas j, I ij is the interaction strength between the two (calculated by the simplified gas interaction model established in step S20). CASSCF Q is the energy difference of the self-consistent field in the complete active space, reflecting the quantum effect. i is the Mulliken charge of gas i, V i is the molecular electrostatic potential of gas i, S conf is the conformational entropy. λ, μ and ν are unknown parameters that need to be obtained through nonlinear fitting of historical data.

[0123] The specific implementation of step S40 is: based on the obtained anesthetic gas output parameter Y, using the computational fluid dynamics software ANSYS Fluent, a simplified gas accumulation model considering the gas diffusion characteristics is established. The model mainly describes the accumulation process of anesthetic gas in the patient's body, which can be expressed as:

[0124]

[0125] Where ρ is the mixture density, Y i is the mass fraction of gas i, is the velocity vector, is the diffusion flux of gas i, R i is the net generation rate of gas i, S i A user-defined source item. i,m is the mass diffusion coefficient of gas i in the mixture, μ t is the turbulent viscosity, Sc t is the turbulent Schmidt number, D T,i is the thermal diffusion coefficient, and T is the temperature. These parameters can be calculated and set through the ANSYS Fluent solver, material database, and user-defined functions.

[0126] The following is a description of the specific method of obtaining the parameters of each simplified model:

[0127] 1. Method for obtaining parameters of simplified gas interaction model:

[0128] E int ,ΔG mix ,HOMO i ,LUMO j : Use GaussView, the graphical interface of Gaussian software, to build the three-dimensional structure of gas molecules. In the input file of Gaussian, select DFT method B3LYP and basis set 6-31G(d). Set the keyword "opt freq" for geometry optimization and frequency analysis. Select "Energy" in "Job Type" to calculate single-point energy. After running the calculation, find ΔG in the "Thermochemistry" section of the output file. mix , find HOMO in the "Population analysis" section i and LUMO j Energy, find E in the "Counterpoise correction" section int .

[0129] χ k (r) and c k : In the Gaussian output file, find the "Basis Set" section to obtain the basis set function χ k (r). Get the molecular orbital coefficient c in the "Molecular Orbital Coefficients" section k The numerical value of .

[0130] k ij ,αij ,β ij and γ ij : Use statistical software such as SPSS or R. Import the historical anesthesia data into the software, select "Regression" in the "Analyze" menu, and then select "Linear" or "Nonlinear" regression. Use the quantum chemical calculation results as the independent variable and the actual observed interaction strength as the dependent variable. Run the regression analysis and obtain the optimal values ​​of these coefficients in the output results.

[0131] 2. Method for obtaining parameters of simplified model of gas combination effect:

[0132] E i : Use professional drug database software such as DrugBank. Enter the gas name in the search bar and find the anesthetic effect related data in the "Pharmacology" section of the result page.

[0133] w i :Use Excel or similar spreadsheet software. Enter the concentration of each gas using the formula w i =C i / ∑C j Calculate the weight coefficient.

[0134] k ij : The same statistical software and methods as the gas interaction simplified model were used, but the independent and dependent variables needed to be adjusted according to the gas combination effect model.

[0135] ΔE CASSCF ,Q i ,V i and S conf :Use Spartan software. Create a 3D structure of a gas molecule or gas combination in Spartan's building blocks. Select "Calculations" in the "Setup" menu, select the CASSCF method and set the active space. Select "Geometry Optimization" and "Energy" in the "Submit" menu. After the calculation is completed, select "Properties" in the "Display" menu to view the results. Select the "Surfaces" menu to create an electrostatic potential map to obtain V i . Use the "Conformer Distribution" function in the "Analysis" menu to perform conformation analysis and obtain S conf .

[0136] λ, μ and ν: Use MATLAB's Curve Fitting Toolbox. Import the historical anesthesia data into MATLAB. Open the Curve Fitting Toolbox, select "Custom Equation" and enter the model equation. Use the quantum chemical calculation results as the independent variable and the actual observed anesthesia effect as the dependent variable. Use the "Nonlinear Least Squares" method to fit the curve and obtain the optimal values ​​of λ, μ and ν.

[0137] 3. Method for obtaining parameters of simplified gas accumulation model:

[0138] ρ, T: Set up the computational domain and mesh in ANSYS Fluent, set the boundary conditions and initial conditions. Select the appropriate turbulence model (such as the k-ε model) and solver settings. Enable the energy equation in the "Models" panel. Run the calculation until convergence, then extract the density ρ, velocity vector from the result file and temperature T.

[0139] Y i : Enable "Species Transport" in the "Species" model in ANSYS Fluent. Set the initial mass fractions of the gas species at each inlet in the "BoundaryConditions" panel. Select the appropriate discretization format and solution algorithm in the "Solution Methods" panel. Solve the transport equation to obtain the Y i distributed.

[0140] D i,m : In ANSYS Fluent's "Materials" panel, select the mixture and the pure species. For each gas, set its diffusion coefficient in "Mixture Species". If a temperature-dependent diffusion coefficient is required, it can be defined using the "piecewise-linear" or "polynomial" function.

[0141] μ t : Select an appropriate turbulence model (such as the k-ε model) in the "Viscous Model" option of the "Models" panel. Set the parameters of the turbulence model, such as turbulence intensity and turbulence length scale. ANSYS Fluent will automatically calculate the turbulent viscosity μ during the solution process. t .

[0142] Sc t: In the "Materials" panel of ANSYS Fluent, find "Mixture Properties" and set the turbulent Schmidt number. The value is usually in the range of 0.7-0.9, which can be adjusted according to the specific application.

[0143] D T,i : In ANSYS Fluent's "Models" panel, find the "Species" option and enable "ThermalDiffusion" (Soret effect). Then set the thermal diffusion coefficient for each gas in "Mixture Species" in the "Materials" panel. You can use a constant value or a temperature-dependent function.

[0144] R i and S i : In the "Cell Zone Conditions" panel of ANSYS Fluent, select the fluid zone and enable "Source Terms". For each species, you can define mass source terms. Use built-in mathematical functions and variables such as temperature, pressure, mass fraction, etc. to express reaction rates and source terms. For example, you can use the Arrhenius equation to describe chemical reaction rates, or use a simplified metabolic model to represent the consumption rate of gases in the body.

[0145] The specific implementation method of step S50 is to establish a hybrid prediction model combining a neural network and an analytical model. Specifically, in a multi-layer perceptron (MLP) neural network structure, parallel branches are added after different hidden layers, and the three analytical models established in the above steps are embedded to form a hybrid model.

[0146] The input layer of the neural network receives the surgical anesthesia requirement parameter X m and patient physiological parameters X p , with multiple hidden layers using ReLU activation function. A parallel branch is added after the first hidden layer, a simplified gas interaction model is embedded, and the gas interaction intensity vector I is output. ij ; Add parallel branches after the second hidden layer, embed the simplified model of gas combination effect, and output the gas combination effect E comb ; Add a parallel branch after the third hidden layer, embed the simplified gas accumulation model, and output the estimated gas accumulation degree Y i Then, the residual connection is used to add the output of each analytical model to the output of the corresponding layer of the main neural network. An attention layer is added after the last hidden layer to dynamically calculate the weight of each sub-model output and multiply it with the sub-model output and then sum it. Finally, the anesthetic gas output parameter Y is output through the fully connected layer.

[0147] This embedded hybrid model structure can give full play to the powerful learning ability of neural networks, while also utilizing the prior knowledge of physical / chemical analytical models, greatly improving the interpretability and generalization of the model. At the same time, through residual connections and attention mechanisms, the amount of computation can be effectively reduced and real-time performance can be improved.

[0148] The specific implementation of step S60 is: using the historical anesthesia data obtained in step S10 to train the hybrid model established in step S50 to obtain the final anesthetic gas output parameter prediction model. The specific training process is as follows:

[0149] 1) The surgical anesthesia requirement parameter X m and patient physiological parameters X p As training input, the corresponding anesthetic gas output parameter Y is output as the training target.

[0150] 2) Using the back propagation algorithm, the weight parameters of the neural network and the simplified parameters in the analytical model are optimized simultaneously. The optimization objectives are:

[0151]

[0152] Among them, θ is the neural network weight parameter, φ is the analytical model simplification parameter, is the loss function (such as mean square error loss), is the regularization term.

[0153] 3) During the training process, techniques such as batch normalization and dropout are used to improve the generalization ability of the model.

[0154] 4) After the training is completed, the final trained model parameters are saved as the anesthetic gas output parameter model for subsequent predictions.

[0155] Through this training method, we can make full use of historical anesthesia data and learn the surgical anesthesia requirement parameters X. m 、Patient physiological parameters X p The complex mapping relationship between the optimal anesthetic gas output parameter Y provides a basis for subsequent real-time predictions.

[0156] The specific implementation of step S70 is: obtaining the surgical anesthesia requirement parameter X of the current patient m and real-time physiological parameters X p , input the anesthetic gas output parameter prediction model trained in step S60 to obtain the optimal anesthetic gas output parameter Y required by the current patient.

[0157] This process can be described as:

[0158] 1) Real-time monitoring and acquisition of the current patient's surgery type x m1, Estimated duration of surgery x m2 , Anesthesia depth requirement x m3 , age x m4 , weight x m5 , height x m6 , Past medical history x m7 And other demand parameters, as well as heart rate x p1 , blood pressure x p2 , body temperature x p3 , breathing rate x p4 , blood oxygen saturation x p5 And other physiological parameters.

[0159] 2) These real-time acquired parameters X m and X p The anesthetic gas output parameter prediction model trained in step S60 is input.

[0160] 3) After forward calculation, the model outputs the optimal anesthetic gas output parameters Y required by the current patient, including gas category y 1 and the flow rate y of each component 2 and concentration y 3 .

[0161] 4) The optimal output parameter Y obtained is fed back to the anesthesia equipment to adjust the output of anesthetic gas in real time to ensure that the patient is in the best anesthesia state.

[0162] Through such real-time prediction and feedback control, the depth of anesthesia during surgery can be accurately controlled to avoid excessive or insufficient anesthesia, thereby improving surgical safety and effectiveness.

[0163] In summary, the specific implementation of this invention involves multiple steps, which fully combines physical / chemical knowledge and machine learning technology. Among them, the simplified gas interaction model established in step S20, I ij It reflects factors such as electrostatic interaction and van der Waals force between molecules; the simplified model of gas combination effect established in step S30, E comb Factors such as anesthetic efficacy and molecular polarity are taken into consideration; the simplified gas accumulation model established in step S40, Y i The diffusion and metabolism of gases in the body are described. These analytical models are established through quantum chemical calculations, computational fluid dynamics and other principles. In step S50, these analytical models are embedded in a neural network structure to form a hybrid prediction model that can give full play to the advantages of machine learning and physical knowledge. Finally, in step S60, the final anesthetic gas output parameter prediction model is obtained through training, and real-time prediction and feedback control of the current patient is realized in step S70.

[0164] In order to further better understand and implement the present invention, the following is a specific application scenario of the present invention, Example 2: Patient Li, 67 years old, 170 cm tall, 80 kg in weight, with a history of hypertension. Today, he needs to undergo a lower limb joint replacement surgery, the estimated operation time is about 3 hours, and the depth of anesthesia is required to reach level 4. Li's surgical anesthesia requirement parameters are as follows: The type of surgery is joint replacement surgery (x m1 ), the estimated duration of the operation is 3 hours (x m2 ), the depth of anesthesia is required to be level 4 (x m3 ), the patient's age is 67 years old (x m4 ), weight 80kg(x m5 )、Height 170cm(x m6 ), history of hypertension (x m7 ).

[0165] Before the operation, the nurse monitored and recorded Li's physiological parameters in real time: heart rate 72 beats / min (x p1 ), systolic pressure 135mmHg, diastolic pressure 85mmHg (x p2 )、body temperature 36.5℃(x p3 ), respiratory rate 18 times / min (x p4 ), blood oxygen saturation 98% (x p5 ), EEG index 92(x p6 ), good muscle relaxation (x p7 ), end-tidal carbon dioxide concentration 38mmHg (x p8 ).

[0166] These surgical anesthesia requirement parameters X m =[x m1 ,x m2 ,x m3 ,x m4 ,x m5 ,x m6 ,x m7 ] and real-time monitoring of patient physiological parameters X p =[x p1 ,x p2 ,x p3 ,x p4 ,x p5 ,x p6 ,x p7 ,x p8 ] and is input into the anesthetic gas output parameter determination system of the present invention pre-installed in the hospital.

[0167] The system first calls the hybrid prediction model trained in step S60 and converts X m and X pAs input, after forward calculation, the optimal anesthetic gas output parameter Y = [y 1 ,y 2 ,y 3 ].

[0168] Specifically include:

[0169] Gas Category 1 It is a mixture of sevoflurane and oxygen;

[0170] Sevoflurane flow rate ml / min,concentration

[0171] Oxygen flow rate L / min, concentration

[0172] The determination of the output parameters in this embodiment 2 fully considers the following aspects:

[0173] 1. Gas Interaction:

[0174] According to the simplified gas interaction model established in step S20, I ij , the interaction strength between sevoflurane and oxygen is 0.85. This indicates that the two have strong mutual compatibility and can form a stable anesthetic gas mixture.

[0175] 2. Gas combination effect:

[0176] According to the simplified model of gas combination effect established in step S30, E comb ,The mixed ratio of sevoflurane and oxygen can achieve the best overall anesthetic effect and meet the requirements of level 4 anesthesia depth.

[0177] 3. Gas accumulation:

[0178] According to the simplified gas accumulation model established in step S40, Y i , the cumulative effect of this gas combination in Li's body is controllable and will not cause serious adverse reactions. Considering Li's age of 67 and previous history of hypertension, the model predicts that his metabolic capacity is relatively weak, so the dosage of sevoflurane needs to be appropriately reduced to reduce the burden on liver and kidney function.

[0179] Based on the analysis of the above three aspects, the hybrid prediction model provides the optimal anesthetic gas output parameter solution that meets Li's surgical needs.

[0180] Subsequently, the parameter plan was fed back to the anesthesiologist, who, after professional judgment, output it to the anesthesia equipment to adjust the output of sevoflurane and oxygen in real time. During the entire operation, medical staff needed to continuously monitor Li's physiological indicators and input them into the system in real time so that the system could adjust the output of anesthetic gas in a timely manner according to the changes.

[0181] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for determining an anesthetic gas output parameter, characterized in that: The following steps are involved: S10, obtaining multiple sets of historical anesthesia data of different stages of the patient's surgical process from the anesthesia information management system, including surgical anesthesia demand parameters, patient physiological parameters, and anesthetic gas output parameters; S20, establishing a simplified gas interaction model using molecular simulation software Gaussian or GAMESS according to the anesthetic gas output parameters, taking into account the mutual influence of multiple anesthetic gases when mixed; S30, establishing a simplified model of gas combination effects using quantum chemical calculation software Spartan or Q-Chem according to the anesthetic gas output parameters, for predicting the effects and interactions of different gas combinations; S40, using computational fluid dynamics software ANSYS Fluent according to the anesthetic gas output parameters, establishing a simplified gas accumulation model that takes into account gas diffusion characteristics, so as to predict the cumulative effects of different anesthetic gases in the patient's body; S50, establishing a hybrid model of a neural network + an analytical model, specifically: embedding the analytical model in the neural network, including the simplified gas interaction model, the simplified gas combination effect model and the simplified gas accumulation model, to form a hybrid model; S60, using the anesthesia data to train the hybrid model to obtain an anesthetic gas output parameter model, wherein the training input is a surgical anesthesia requirement parameter and a patient physiological parameter, and the training output is an anesthetic gas output parameter; S70, obtaining surgical anesthesia requirement parameters and real-time physiological parameters of the current patient, inputting the anesthetic gas output parameter model, and obtaining the optimal anesthetic gas output parameters of the current patient.

2. The method for determining an anesthetic gas output parameter according to claim 1, characterized in that: The surgical anesthesia requirement parameters include the type of surgery, expected duration of the surgery, required depth of anesthesia, patient age, weight, height, and medical history.

3. The method for determining an anesthetic gas output parameter according to claim 1, characterized in that: The patient's physiological parameters include heart rate, blood pressure, body temperature, respiratory rate, blood oxygen saturation, electroencephalogram index, degree of muscle relaxation and end-tidal carbon dioxide concentration.

4. The method for determining an anesthetic gas output parameter according to claim 1, characterized in that: The anesthetic gas output parameters include gas type, and flow rate and concentration of each component.

5. The method for determining an anesthetic gas output parameter according to claim 1, characterized in that: The hybrid model also includes a multi-layer perceptron neural network as the main structure.

6. A method for determining an anesthetic gas output parameter according to claim 5, characterized in that: The simplified gas interaction model takes into account parameters such as electrostatic interaction, van der Waals force, hydrogen bonding, and molecular orbital energy level differences between gas molecules, and is generated by geometry optimization and energy calculation using the B3LYP functional and 6-31G(d) basis set of Gaussian software based on the principles of density functional theory.

7. A method for determining an anesthetic gas output parameter according to claim 6, characterized in that: The simplified model of gas combination effect takes into account parameters such as anesthetic efficacy, concentration, molecular polarity, ionization energy, electron affinity, etc. of various gases, and is generated based on the principles of quantum chemistry and statistical thermodynamics using the full active space self-consistent field method and natural bond orbital analysis function of Spartan software.

8. A method for determining an anesthetic gas output parameter according to claim 7, characterized in that: The simplified gas accumulation model takes into account parameters such as gas diffusion coefficient, solubility, metabolic rate, blood gas distribution coefficient, tissue perfusion rate, etc., and is generated based on the principles of computational fluid dynamics and pharmacokinetics using the race transport model and user-defined function of ANSYS Fluent software.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the method for determining an anesthetic gas output parameter according to any one of claims 1 to 8.

10. A system for determining an anesthetic gas output parameter, characterized in that: A computer-readable storage medium comprising the computer-readable storage medium of claim 9.

Images