Intelligent anesthesia drug delivery system and control method thereof

Abstract

The invention discloses an intelligent anesthesia drug delivery system and a control method thereof, and the method comprises the steps: obtaining operation information, predicting operation time, predicting operation state stability, determining a drug delivery strategy according to an operation time prediction value and an operation state prediction value, and controlling a vein micro pump based on a collaborative sliding mode control algorithm of an intelligent algorithm. And judging whether the anesthesia scheme needs to be adjusted. According to the method, prediction of the operation time and the operation state is achieved, a mode for objectively selecting anesthesia strategies is provided, anesthesia strategy selection errors caused by artificial omission or insufficient level are avoided, a collaborative sliding mode control algorithm based on an intelligent algorithm is used for controlling the multi-channel vein micro pump, and the operation efficiency is improved. Synergism of multiple drugs during injection is guaranteed, and optimal drug effect synergy is achieved.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to an intelligent anesthesia drug delivery system and a control method thereof. Background Art

[0003] During surgery, ensuring the dosage of anesthetics and accurately controlling the injection speed are key. Intravenous anesthesia is often combined to give full play to the characteristics of each drug to achieve stable anesthesia, mild physiological disturbance, few side effects, and rapid awakening, so it is also called intravenous compound anesthesia. Intravenous anesthesia is a method of producing general anesthesia by injecting drugs into the veins and acting on the central nervous system through the blood circulation. The issue of anesthesia safety during surgery is extremely important. Intelligent control of anesthesia is an advanced technology that reduces the occurrence of complications by enhancing the anesthetic effect and improving the safety of anesthesia. It is an indispensable part of modern medical technology.

[0004] In the prior art, intelligent control of the anesthesia process is limited to monitoring human vital signs and selecting the amount of anesthesia based on the interval values ​​of the vital signs. There is no predictive control based on the patient's physical condition and the type of surgery. In addition, abnormal vital signs often occur during surgery, which leads to prolonged surgery time. It is necessary to adjust the anesthesia time and control parameters in real time according to the actual situation. Summary of the invention

[0005] 1. Technical issues to be resolved

[0006] In order to solve the above technical problems, the present invention provides an intelligent anesthesia drug delivery system and a control method thereof.

[0007] (II) Technical solution

[0008] In order to solve the above-mentioned technical problems and achieve the purpose of the invention, the present invention is implemented by the following technical solutions:

[0009] The present invention also provides an intelligent anesthesia drug delivery system control method, comprising the following steps:

[0010] S1: Acquiring surgical information, wherein the surgical information includes surgical type, preoperative information, intraoperative dynamic monitoring information, and surgical status information;

[0011] The preoperative information is information obtained before the operation, including patient information, expected blood transfusion volume, and patient physiological data information;

[0012] The surgical status information includes the total blood transfusion volume minus the expected blood transfusion volume;

[0013] S2: Operation time prediction. The operation time prediction model is as follows:

[0014]

[0015] in, is the baseline surgery time, obtained from the surgery type and preoperative information, is the operation time deviation, which is obtained from the intraoperative dynamic monitoring information and the operation status information; wherein the benchmark operation time and the operation time deviation are obtained based on the optimized random forest prediction algorithm;

[0016] S3: Prediction of surgical status stability, predicting surgical status stability based on hemodynamic stability;

[0017] S4: Determine the medication strategy based on the predicted values ​​of surgical time and surgical status;

[0018] The drug administration strategy includes the duration of drug administration, the drug administration flow rate under the duration, and the selection of the type of anesthetic drugs; the drug administration strategy is determined based on the long short-term memory algorithm of the recurrent neural network;

[0019] S5: Control the intravenous micropump for drug delivery. The collaborative sliding mode control algorithm based on the intelligent algorithm controls the intravenous micropump.

[0020] S6: Determine whether the anesthesia regimen needs to be adjusted;

[0021] When the variables of the intraoperative dynamic monitoring information and the surgical condition exceed the set threshold, the anesthesia plan needs to be adjusted, and steps S1-S5 are repeated to update the anesthesia control parameters.

[0022] Furthermore, the random forest prediction algorithm optimized in step S2 includes constructing k decision tree models; each decision tree consists of a training subset Create, in the process of decision tree growth, select m feature variables from the T feature variables in the training subset as candidate feature variables, in the process of tree node splitting, calculate the entropy gain value of each candidate feature variable, and select the feature with the largest gain value as the split node.

[0023] Furthermore, the selection of m feature variables from the T feature variables of the training subset as the feature variables to be selected is performed in the following manner:

[0024] a. Divide the T features into m groups;

[0025] b. Randomly select a feature in each group to form the candidate feature variable.

[0026] Furthermore, step S3 predicts the stability of the surgical status based on the following method:

[0027]

[0028] in, , N is the total amount of data, It is the change of the monitored value of hemodynamic indexes compared with the initial state of surgery.

[0029] Furthermore, the long short-term memory algorithm of the recurrent neural network in step S4 includes the following steps:

[0030] S41: Establishing a network model, wherein the model adopts a multi-input multi-output model structure, is composed of a three-layer LSTM network, and uses a stacked network composed of multiple hidden layers to learn the time series characteristics of drug administration strategy data;

[0031] S42: Model training;

[0032] S43: Segment prediction, the segment prediction is performed in the following manner:

[0033] a. Assume that the current time is t, and obtain the historical medication strategy data from tnk to tn;

[0034] b. Divide the historical medication strategy data into p groups, which are: ;

[0035] c. Interval selection of data from group p to predict the dosing strategy at time t; for example, select

[0036] d. Output the prediction results.

[0037] Furthermore, the cooperative sliding mode control algorithm of the intelligent algorithm of step S5 includes multi-channel cooperative control based on deviation coupling control.

[0038] Furthermore, the mathematical model of the multi-channel collaborative control deviation coupling controller is:

[0039]

[0040] In the formula, is the output speed of the pump speed compensator of channel x; is the compensation gain of the speed compensator; are the speeds of the controlled channel pump and the comparison channel pump, respectively; n is the total number of channels, is the maximum pump speed, is the minimum value of the pump speed, is the average value fed back to the pump speed in the form of:

[0041] .

[0042] Furthermore, the deviation coupling control is dynamically compensated based on the fuzzy control strategy.

[0043] Also included is an intelligent anesthesia drug delivery system, comprising:

[0044] A surgery information acquisition module, which is used to acquire surgery information;

[0045] The surgical information includes surgical type, preoperative information, intraoperative dynamic monitoring information, and surgical status information;

[0046] The surgical status information includes the total blood transfusion volume minus the expected blood transfusion volume;

[0047] An operation time prediction module, which is used to predict the operation time based on the information obtained by the operation information acquisition module based on the optimized random forest;

[0048] A surgical status stability prediction module, which is used to predict surgical status changes based on hemodynamic stability;

[0049] A drug administration strategy determination module, which is used to determine the drug administration strategy based on the long short-term memory algorithm of the recurrent neural network, including the drug administration duration and the drug administration flow rate and the selection of the type of anesthetic drugs under the duration;

[0050] Intravenous micro-pump drug administration control module, which is used to control the multi-channel intravenous micro-pump based on the collaborative sliding mode control algorithm of the intelligent algorithm;

[0051] The judgment module is used to judge whether the anesthesia plan needs to be adjusted. When the variables of the intraoperative dynamic monitoring information and the surgical conditions exceed the set threshold, the anesthesia plan is adjusted and the anesthesia control parameters are updated.

[0052] In addition, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium, on which program instructions of an intelligent anesthesia drug delivery method based on artificial intelligence are stored, and the program instructions of the intelligent anesthesia drug delivery method based on artificial intelligence can be executed by one or more processors to implement the steps of the intelligent anesthesia drug delivery method based on artificial intelligence as described above.

[0053] (III) Beneficial effects

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] (1) The present invention predicts the operation time and operation status based on the operation information, and selects the anesthesia strategy according to the prediction results of the two, thereby realizing the selection of the appropriate anesthesia strategy according to the patient's own condition and the operation, providing an objective way to select the anesthesia strategy, and avoiding the selection errors of the anesthesia strategy due to human omissions or insufficient level.

[0056] (2) The present invention predicts the operation time based on the optimized random forest prediction algorithm, which not only takes into account the patient's own condition but also the operation status, thus achieving accurate operation time prediction and providing effective information for the selection of anesthesia strategy.

[0057] (3) The collaborative sliding mode control algorithm based on the intelligent algorithm of the present invention controls the multi-channel intravenous micropump, ensuring the synergy of multiple drug injections and achieving optimal drug efficacy synergy. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0059] Figure 1 It is a flow chart of a control method of an intelligent anesthesia drug delivery system based on artificial intelligence according to an embodiment of the present application;

[0060] Figure 2 It is an improved LSTM network model according to an embodiment of the present application;

[0061] Figure 3 It is a schematic diagram of the LSTM network structure according to an embodiment of the present application. DETAILED DESCRIPTION

[0062] The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0063] The following describes the embodiments of the present disclosure through specific examples, and those skilled in the art can easily understand other advantages and effects of the present disclosure from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0064] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present disclosure. The drawings only show components related to the present disclosure rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0065] See also Figure 1 , a control method of an intelligent anesthesia drug delivery system based on artificial intelligence comprises the following steps:

[0066] S1: Obtain surgical information;

[0067] The surgical information includes surgical type, preoperative information, intraoperative dynamic monitoring information, and surgical status information;

[0068] The preoperative information is the information obtained before the operation, including patient information, expected blood transfusion volume and patient physiological data information; the patient information includes the patient's age, gender, height, weight, health status, and historical medical records; the patient's physiological data information is the physiological data of the patient's physical examination before the operation, including blood pressure, blood routine, blood biochemistry, urine routine, and electrocardiogram.

[0069] The intraoperative dynamic monitoring information includes heart rate, blood pressure, blood oxygen saturation, respiratory rate, electrocardiogram, and body temperature.

[0070] The surgical status information includes the total blood transfusion volume minus the expected blood transfusion volume.

[0071] S2: surgery time prediction;

[0072] Predicting the operation time based on the operation information obtained in step S1 specifically includes the following steps:

[0073] S21: surgical information data preprocessing, including:

[0074] Normalize the surgical information data in the patient information;

[0075] Among them, the health status is divided into 10 levels from best to worst; the historical case is determined according to the severity of the patient's historical disease condition; the patient's physiological data information and intraoperative dynamic monitoring information are determined according to the degree of deviation from the normal range.

[0076] S22: establishing a prediction model;

[0077] The operation time prediction model is expressed as follows:

[0078]

[0079] in, is the baseline surgery time, obtained from the surgery type and preoperative information, is the surgical time deviation, which is obtained from the intraoperative dynamic monitoring information and surgical status information. The initial value is set to 0.

[0080] The benchmark operation time and the operation time deviation are obtained based on an optimized random forest prediction algorithm, including the following:

[0081] S221: sampling k training subsets;

[0082] For the dataset Perform repeatable random sampling to obtain a sampling subset of length N ; Repeat the sampling process k times to obtain k training subsets;

[0083] S222: construct k decision tree models;

[0084] Each decision tree consists of a training subset During the decision tree growth process, m feature variables are selected from the T feature variables in the training subset as candidate feature variables. During the splitting process of the tree node, the entropy gain value of each candidate feature variable is calculated, and the feature with the largest gain value is selected as the splitting node.

[0085] The selection of m feature variables from the T feature variables of the training subset as the feature variables to be selected is performed as follows:

[0086] a. Divide the T features into m groups;

[0087] b. Randomly select a feature in each group to form the candidate feature variable.

[0088] The random forest training model is optimized by grouping to avoid the destruction of the temporal characteristics of the time series during random selection.

[0089] S223: Using k decision trees to form a random forest model The set of k trained trees constitutes a random forest model:

[0090]

[0091] in, is a single decision tree classifier.

[0092] The prediction result of random forest is The predicted mean of the trees is:

[0093]

[0094] in, For the The accuracy of a decision tree.

[0095] Establish two groups of random forests and obtain , where the baseline surgery time is calculated The model input parameters are surgery type and preoperative information, and the surgery time deviation is calculated. The model input parameters are intraoperative dynamic monitoring information and surgical status information.

[0096] S23: Calculate and predict the operation time;

[0097] The predicted value of the operation time is calculated based on the operation time prediction model in combination with the reference operation time and the operation time deviation obtained in step S22.

[0098] S3: prediction of surgical status stability;

[0099] Changes in intraoperative hemodynamic stability reflect changes in surgical status. Fluctuating hemodynamic stability requires further adjustment of the anesthetic dose to keep hemodynamic indicators in a stable state. Therefore, control of intraoperative anesthesia requires effective prediction of the surgical status. The present invention predicts surgical status stability based on the following method:

[0100]

[0101] in, , N is the total amount of data, It is the change of the monitored value of hemodynamic indexes compared with the initial state of surgery.

[0102] S4: Determine the medication strategy based on the predicted values ​​of surgical time and surgical status;

[0103] The drug administration strategy includes the duration of drug administration, the drug administration flow rate within the duration, and the selection of the type of anesthetic drugs.

[0104] Drugs that enter the body through intravenous injection and act on the central nervous system through the blood circulation to produce general anesthetic effects. Commonly used drugs include sodium thiopental, ketamine, etomidate, propofol, etc. Different drugs have different hypnotic and analgesic effects, effective time, usage and dosage. Therefore, the drug flow rate and type of anesthetic drugs need to be determined according to the operation time and status.

[0105] It is difficult to establish an accurate model for determining the drug administration strategy, which is a complex problem with multiple variables, nonlinearity, and strong coupling. Therefore, the present invention proposes a long short-term memory algorithm based on a recurrent neural network to model the drug administration strategy problem. The algorithm not only takes into account the influencing factors of time, but also solves the problem of long sequence information loss.

[0106] The steps include:

[0107] S41: Network model establishment;

[0108] The model input is the predicted value of operation time and operation status, and the output is the type of anesthetic drug, total administration time, initial value of administration flow rate, flow change time point and corresponding administration flow rate.

[0109] Network models such as Figure 2 As shown, the model adopts a multi-input multi-output model structure, which is composed of a three-layer LSTM network. A stacked network composed of multiple hidden layers is used to learn the time series characteristics of drug administration strategy data. The Dropput algorithm is used to reduce the interdependent learning between neurons, thereby improving the robustness and generalization of the network model.

[0110] like Figure 3 As shown, the present invention improves the structure of the LSTM network. The network of the present invention is composed of multiple LSTM networks with different numbers of cell units stacked together. Figure 3 middle, represents a time series data sample, express The information output at each moment, the feature learning process of the LSTM network is affected by the output value at the previous moment and the current input value Therefore, in At this moment, the input of the memory unit is and .

[0111] Input Gate The role of is to control the addition of new information, through Functions are implemented with control. The function generates a candidate state , input gate Produces a value in The vector within is controlled The amount to be added to the next calculation. pass Functions are mapped to Within the range, pass The activation function maps to , the expression is as follows:

[0112]

[0113]

[0114] in, and Represent the weight matrices of the input gate and candidate state respectively, and denote the bias of the input gate and candidate state respectively, express Activation function, Represents the vector and Put together into a long vector.

[0115] Forget Gate Control the output of the previous moment With the current input The degree of being forgotten, pass Functions are mapped to Range, the calculation formula of the forget gate is:

[0116]

[0117] in, represents the weight matrix of the forget gate, Represents the bias of the forget gate.

[0118] Through the input gate and forget gate You can determine the current internal state , and its calculation formula is as follows:

[0119]

[0120] in, Represents element-wise multiplication.

[0121] Output Gate The function is to control how much of the current internal state is passed Information for the next moment . Output Gate Also through Functions are mapped to The calculation formula is as follows:

[0122]

[0123] in, represents the weight matrix of the output gate, Represents the bias of the output gate.

[0124] The final output of the LSTM network is calculated as follows:

[0125]

[0126] in, Represents the output at the current moment, Represents element-wise multiplication.

[0127] S42: Model training

[0128] S421: Initialize model parameters. Set the maximum number of iterations, number of hidden layers, number of neurons, and batch size.

[0129] S422: Determine the size relationship between two hyperparameters based on the characteristics of the time series data.

[0130] S423: Update model parameters:

[0131] (a) Check if an episode has been completed. If an episode has not been completed, adjust the number of neurons in the hidden layer according to the regularization probability p.

[0132] (b) Calculate the loss value according to the loss function. Then, use the Adam optimizer to update the weight W and bias b, and repeat step 3 with the next batch. If an iteration has been completed, proceed to the next step.

[0133] S424: Determine whether the termination condition is met, if so, end the calculation, otherwise return to step S423. Finally, save the optimized model parameters, including the Dropout algorithm probability p, weight matrix W and bias b.

[0134] Optionally, the network model is that the first LSTM layer contains 64 neurons, the second LSTM layer contains 128 neurons, and the third layer contains 32 neurons.

[0135] S43: Segment Prediction

[0136] Since the prediction time is relatively long, direct prediction will result in a large cumulative error. Therefore, the present invention adopts a segmented prediction method to determine the drug administration strategy. The segmented prediction is performed in the following manner:

[0137] a. Assume that the current time is t, and obtain the historical medication strategy data from tnk to tn;

[0138] b. Divide the historical medication strategy data into p groups, which are: ;

[0139] c. Interval selection of data from group p to predict the dosing strategy at time t; for example, select

[0140] d. Output the prediction results.

[0141] S5: Control intravenous micropump drug delivery;

[0142] During surgery, in order to achieve mixed delivery of multiple drugs including anesthesia, a multi-channel intravenous micropump and an extension tube are usually used to mix multiple drugs. In order to achieve the best drug efficacy synergy, the collaborative sliding mode control algorithm based on the intelligent algorithm of the present invention controls the intravenous micropump. The specific steps are as follows:

[0143] S51: Intelligent sliding mode control of single-channel micro pumps

[0144] S511: Build controller model

[0145] The intermediate variable of the model position change is shown as follows:

[0146]

[0147] In the formula, is the position error value; is the expected displacement; is the actual displacement. The sliding mode switching surface is defined as follows:

[0148]

[0149] In the formula, is the sliding surface; is a proportional variable; is the integral variable. The following formula can be obtained by taking the derivative:

[0150]

[0151] choose The function replaces the traditional symbolic function. The function curve is a continuously transformed and smooth image. When the independent variable tends to infinity, the function value tends to 1 infinitely. The two are combined as the approach law of the stepping turntable, as shown in the following formula:

[0152]

[0153] In the formula, is the switching gain coefficient; is the design parameter; X is the state variable; is the exponential value of the state variable, .

[0154]

[0155] In the formula, is any number greater than zero. s is the sliding surface. When s tends to positive infinity, the dependent variable is 1; when s tends to negative infinity, the dependent variable is -1. Combining the above two equations and rearranging them, we can get:

[0156]

[0157] The angular velocity error value can be obtained by differentiating the position error, as shown in the following formula:

[0158]

[0159] in, is the desired angular velocity, is the actual angular velocity.

[0160] The reaching law of position control can be obtained, , we can get:

[0161]

[0162] In the formula, is the control rate of the exponential approach law; the larger the k, the faster the speed of approaching the sliding surface; the switching gain coefficient The larger the value, the faster the convergence speed and the faster the rate of approaching the sliding surface, but there is obvious chattering phenomenon; switch gain coefficient The smaller the value, the chattering phenomenon will be weakened to a certain extent, but the convergence speed will be slower. By adjusting the appropriate sliding mode parameter value, increasing the design parameter as much as possible, and reducing the switching gain coefficient, not only the effective time is shortened, but also the chattering phenomenon is weakened.

[0163] Sliding mode variable structure control requires the selection of control law, and the design needs to meet the reachability condition, that is, the control system can move quickly at any initial position and reach the sliding surface. The stability of the control law function is determined by Lyapunov's theorem. Judgment function selection , it can be proved that the system is stable.

[0164]

[0165] In the formula, , , 0, that is It is always less than 0, ensuring that the control system has a faster convergence speed and can reach the sliding mode switching surface.

[0166] S512: Improved Sliding Mode Controller

[0167] In order to compensate for the change in resistance of microbubbles in the intravenous infusion tube due to temperature changes, the present invention improves the sliding mode controller based on the dynamic determination of the controller parameters , , , , , including the following steps:

[0168] a. Get real-time temperature value;

[0169] b. Obtain controller parameters corresponding to the real-time temperature from the database.

[0170] S52: Multi-channel collaborative control

[0171] During surgery, in addition to the need for intravenous injection of anesthetic drugs, multiple drugs are often mixed and injected simultaneously. In this case, a multi-channel intravenous micropump is required. In order to avoid uneven mixing of two or more drugs in the extension tube, the pump of each channel needs to be coordinated and controlled. The present invention controls the multi-channel pump based on the improved deviation coupling synchronous control, which specifically includes the following steps:

[0172] Deviation coupling control is a deviation coupling expanded on the basis of cross-coupling control, which can realize the global coupling of the control system. Its control strategy is to use the feedback compensation principle of cross-coupling control to multiply the speed signal deviation between each channel pump by the corresponding gain and then add it to the speed signal of each channel pump, thereby reducing the influence of the speed of each channel pump after interference and improving the control accuracy.

[0173] The coupling control in the deviation coupling control is adjusted through the speed compensator to maintain the stable operation of the synchronous system. The mathematical model of the deviation coupling controller is:

[0174]

[0175] In the formula, is the output speed of the pump speed compensator of channel x; is the compensation gain of the speed compensator; are the speeds of the controlled channel pump and the comparison channel pump, respectively; n is the total number of channels, is the maximum pump speed, is the minimum value of the pump speed, is the average value fed back to the pump speed in the form of:

[0176]

[0177] Compensation Gain for:

[0178]

[0179] In the formula, , are the moments of inertia of x and y respectively.

[0180] Taking the first channel pump as an example, the output speed compensation of the speed compensator of the channel pump should be:

[0181]

[0182] In the formula, , , They are the compensation gains of the speed differences between the first channel pump and the second, third and fourth channel pumps, that is, the corresponding rotational inertia ratios.

[0183] Furthermore, the deviation coupling control is dynamically compensated. The idea of ​​dynamic compensation is that when the synchronization error is large, the compensation strength increases accordingly, and when the synchronization error is small, the compensation strength decreases accordingly, so that it has good tracking performance and synchronization performance at the same time. The present invention performs dynamic compensation based on fuzzy control strategy:

[0184] a. Divide according to the error size and select the scaling factor based on the division result;

[0185] b. Select the domain and subset;

[0186] c. Select the membership function and fuzzy reasoning and defuzzification method. Optionally, the membership functions used are triangular, Z-type and S-type, the Mamdani type reasoning method is used as the fuzzy reasoning method, and the centroid method is used as the defuzzification method.

[0187] d. Design relevant fuzzy rules of synchronization error compensator according to the control law.

[0188] S6: Determine whether the anesthesia regimen needs to be adjusted;

[0189] When the variables of the intraoperative dynamic monitoring information and the surgical condition exceed the set threshold, the anesthesia plan needs to be adjusted, and steps S1-S5 are repeated to update the anesthesia control parameters.

[0190] In this embodiment, the operation time and operation status are predicted based on the surgical information, and the anesthesia strategy is selected according to the prediction results of the two, so that the appropriate anesthesia strategy can be selected according to the patient's own condition and the operation, and an objective way of selecting anesthesia strategy is provided, which avoids the selection errors of anesthesia strategy due to human omissions or insufficient level. The operation time is predicted based on the optimized random forest prediction algorithm, which not only takes into account the patient's own condition, but also the surgical status, and realizes accurate prediction of operation time, providing effective information for the selection of anesthesia strategy. The collaborative sliding mode control algorithm based on the intelligent algorithm controls the multi-channel intravenous micropump, ensuring the synergy of multiple drug injections and achieving the best drug efficacy synergy.

[0191] According to one aspect of the present invention, there is provided an intelligent anesthesia drug delivery system based on artificial intelligence, comprising:

[0192] A surgery information acquisition module, which is used to acquire surgery information;

[0193] The surgical information includes surgical type, preoperative information, intraoperative dynamic monitoring information, and surgical status information;

[0194] The preoperative information is the information obtained before the operation, including patient information, expected blood transfusion volume and patient physiological data information; the patient information includes the patient's age, gender, height, weight, health status, and historical medical records; the patient's physiological data information is the physiological data of the patient's physical examination before the operation, including blood pressure, blood routine, blood biochemistry, urine routine, and electrocardiogram.

[0195] The intraoperative dynamic monitoring information includes heart rate, blood pressure, blood oxygen saturation, respiratory rate, electrocardiogram, and body temperature.

[0196] The surgical status information includes the total blood transfusion volume minus the expected blood transfusion volume.

[0197] An operation time prediction module, which is used to predict the operation time based on the information obtained by the operation information acquisition module based on the optimized random forest;

[0198] The operation time prediction model is expressed as follows:

[0199]

[0200] in, is the baseline surgery time, obtained from the surgery type and preoperative information, is the surgical time deviation, which is obtained from the intraoperative dynamic monitoring information and surgical status information. The initial value is set to 0.

[0201] The benchmark operation time and the operation time deviation are obtained based on a random forest prediction algorithm.

[0202] A surgical status stability prediction module, which is used to predict surgical status changes based on hemodynamic stability;

[0203] The drug administration strategy determination module is used to determine the drug administration strategy based on the LSTM network, including the drug administration duration, the drug administration flow rate within the duration, and the selection of the type of anesthetic drugs.

[0204] The intravenous micro-pump drug administration control module is used to control the multi-channel intravenous micro-pump based on the collaborative sliding mode control algorithm of the intelligent algorithm.

[0205] The judgment module is used to judge whether the anesthesia plan needs to be adjusted. When the variables of the intraoperative dynamic monitoring information and the surgical conditions exceed the set threshold, the anesthesia plan is adjusted and the anesthesia control parameters are updated.

[0206] In addition, an embodiment of the present invention also proposes a computer-readable storage medium, on which program instructions of a control method of an intelligent anesthesia drug delivery system based on artificial intelligence are stored. The control program instructions of the intelligent anesthesia drug delivery system based on artificial intelligence can be executed by one or more processors to implement the steps of the control method of the intelligent anesthesia drug delivery system based on artificial intelligence as described above.

[0207] The embodiments described above are only descriptions of the preferred implementation modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for controlling an intelligent anesthesia drug delivery system, characterized in that: The steps include: S1: Acquiring surgical information, wherein the surgical information includes surgical type, preoperative information, intraoperative dynamic monitoring information, and surgical status information; The preoperative information is information obtained before the operation, including patient information, expected blood transfusion volume, and patient physiological data information; The surgical status information includes the total blood transfusion volume minus the expected blood transfusion volume; S2: Operation time prediction. The operation time prediction model is as follows: ； in, is the baseline surgery time, obtained from the surgery type and preoperative information, is the operation time deviation, which is obtained from the intraoperative dynamic monitoring information and the operation status information; wherein the benchmark operation time and the operation time deviation are obtained based on the optimized random forest prediction algorithm; S3: Prediction of surgical status stability, predicting surgical status stability based on hemodynamic stability; S4: Determine the medication strategy based on the predicted values ​​of surgical time and surgical status; The drug administration strategy includes the duration of drug administration, the drug administration flow rate under the duration, and the selection of the type of anesthetic drugs; the drug administration strategy is determined based on the long short-term memory algorithm of the recurrent neural network; S5: Control the intravenous micropump for drug delivery. The collaborative sliding mode control algorithm based on the intelligent algorithm controls the intravenous micropump. S6: Determine whether the anesthesia regimen needs to be adjusted; When the variables of the intraoperative dynamic monitoring information and the surgical condition exceed the set threshold, the anesthesia plan needs to be adjusted, and steps S1-S5 are repeated to update the anesthesia control parameters.

2. The intelligent anesthesia drug delivery system control method according to claim 1, characterized in that: The random forest prediction algorithm optimized in step S2 includes constructing k decision tree models; each decision tree consists of a training subset Create, in the process of decision tree growth, select m feature variables from the T feature variables in the training subset as candidate feature variables, in the process of tree node splitting, calculate the entropy gain value of each candidate feature variable, and select the feature with the largest gain value as the split node.

3. The intelligent anesthesia drug delivery system control method according to claim 2, characterized in that: The selection of m feature variables from the T feature variables of the training subset as the feature variables to be selected is performed as follows: a. Divide the T features into m groups; b. Randomly select a feature in each group to form the candidate feature variable.

4. The intelligent anesthesia drug delivery system control method according to claim 1, characterized in that: The step S3 predicts the stability of the surgical status based on the following method: ； in, , N is the total amount of data, It is the change of the monitored value of hemodynamic indexes compared with the initial state of surgery.

5. The intelligent anesthesia drug delivery system control method according to claim 1, characterized in that: The long short-term memory algorithm of the recurrent neural network in step S4 comprises the following steps: S41: Establishing a network model, wherein the model adopts a multi-input multi-output model structure, is composed of a three-layer LSTM network, and uses a stacked network composed of multiple hidden layers to learn the time series characteristics of drug administration strategy data; S42: Model training; S43: Segment prediction, the segment prediction is performed in the following manner: a. Assume that the current time is t, and obtain the historical medication strategy data from tnk to tn; b. Divide the historical medication strategy data into p groups, which are: ; c. Interval selection of data from group p to predict the dosing strategy at time t; for example, select ; d. Output the prediction results.

6. The intelligent anesthesia drug delivery system control method according to claim 1, characterized in that: The cooperative sliding mode control algorithm of the intelligent algorithm of step S5 includes multi-channel cooperative control based on deviation coupling control.

7. The intelligent anesthesia drug delivery system control method according to claim 6, characterized in that: The mathematical model of the multi-channel cooperative control based on deviation coupling control is: ； In the formula, is the output speed of the pump speed compensator of channel x; is the compensation gain of the speed compensator; are the speeds of the controlled channel pump and the comparison channel pump, respectively; n is the total number of channels, is the maximum pump speed, is the minimum value of the pump speed, is the average value fed back to the pump speed in the form of: 。 8. The intelligent anesthesia drug delivery system control method according to claim 7, characterized in that: The deviation coupling control is dynamically compensated based on fuzzy control strategy.

9. A system based on the intelligent anesthesia drug delivery system control method according to any one of claims 1 to 8, comprising: A surgery information acquisition module, which is used to acquire surgery information; The surgical information includes surgical type, preoperative information, intraoperative dynamic monitoring information, and surgical status information; The surgical status information includes the total blood transfusion volume minus the expected blood transfusion volume; An operation time prediction module, which is used to predict the operation time based on the information obtained by the operation information acquisition module based on the optimized random forest; A surgical status stability prediction module, which is used to predict surgical status changes based on hemodynamic stability; A drug administration strategy determination module, which is used to determine the drug administration strategy based on the long short-term memory algorithm of the recurrent neural network, including the drug administration duration and the drug administration flow rate and the type of anesthetic drugs under the duration; Intravenous micro-pump drug administration control module, which is used to control the multi-channel intravenous micro-pump based on the collaborative sliding mode control algorithm of the intelligent algorithm; The judgment module is used to judge whether the anesthesia plan needs to be adjusted. When the variables of the intraoperative dynamic monitoring information and the surgical conditions exceed the set threshold, the anesthesia plan is adjusted and the anesthesia control parameters are updated.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions of the intelligent anesthesia drug delivery system control method, and the program instructions of the intelligent anesthesia drug delivery system control method can be executed by one or more processors to implement the steps of the intelligent anesthesia drug delivery system control method as described in any one of claims 1-8.

Images