Anesthesia drug dose optimization method based on artificial intelligence

Abstract

This invention relates to the field of medical informatics, and more particularly to an artificial intelligence-based method for optimizing anesthetic drug dosage. The method includes: collecting real-time and static physiological data from patients; converting the real-time and static data into standardized feature vectors through standardization processing; predicting the initial intraoperative infusion rate of anesthetic drugs using a long short-term memory network; collecting and standardizing the patient's intraoperative data to obtain standardized intraoperative data; generating personalized anesthetic drug infusion rates through a joint optimization algorithm of dynamic metabolic risk and neural sensitivity; determining the maximum permissible infusion rate and minimum effective infusion rate based on clinical guidelines; obtaining the remaining surgical time; performing safety verification; and generating the final infusion rate by combining the remaining surgical time. This method solves the technical problem of existing technologies relying heavily on anesthesiologist experience to determine the initial infusion rate, leading to inaccurate prediction of the initial preoperative anesthetic drug infusion rate.

Description

Technical Field

[0001] This invention relates to the field of medical informatics, and more particularly to an artificial intelligence-based method for optimizing anesthetic drug dosage. Background Technology

[0002] Optimizing anesthetic drug dosage is a crucial aspect of modern medicine, directly impacting surgical safety, patient recovery speed, and the effective utilization of medical resources. Traditional methods for determining anesthetic drug dosage rely primarily on the anesthesiologist's experience, patient physiological parameters (such as weight, age, and gender), and standardized dosing guidelines. However, due to individual differences, disease states, and the complexity of pharmacokinetics and pharmacodynamics, traditional methods often fail to achieve precise dosing, leading to underdosing or overdosing. In recent years, the rapid development of artificial intelligence (AI) technology has offered new possibilities for anesthetic drug dosage optimization. AI-based methods, by integrating big data, machine learning, deep learning, and real-time monitoring technologies, can achieve individualized and dynamic optimization of anesthetic drug dosage, thereby improving the accuracy and safety of anesthesia management.

[0003] However, the existing technologies generally adopt standardized anesthetic drug infusion protocols, which fail to fully consider individual patient differences, potentially leading to overdose or underdose in some patients, affecting surgical safety and efficacy. Traditional anesthesia management relies on the anesthesiologist's experience and lacks a real-time, automated, data-driven mechanism for optimizing anesthetic drug dosage. Existing technologies often do not fully consider the cumulative effect of previously infused drugs that have not yet fully taken effect, which may lead to excessively high or fluctuating dosages, increasing the risk of overdose anesthesia. Summary of the Invention

[0004] This invention provides an artificial intelligence-based method for optimizing anesthetic drug dosage, addressing the following issues: existing technologies rely heavily on anesthesiologist experience to determine the initial infusion rate, leading to inaccurate predictions of the initial preoperative anesthetic drug infusion rate; existing technologies struggle to dynamically adjust the infusion rate based on real-time intraoperative physiological data, lacking comprehensive analysis of metabolic risks and neurological sensitivity, resulting in delayed or inaccurate anesthetic drug dosage adjustments; existing technologies fail to effectively model the onset delay of anesthetic drugs, potentially leading to overdose risks from recent high-dose infusions, especially in surgical scenarios with rapid physiological changes; and the lack of dynamic safety constraints based on clinical guidelines may result in infusion rates exceeding safe limits.

[0005] The present invention provides an artificial intelligence-based method for optimizing anesthetic drug dosage, which specifically includes the following technical solutions:

[0006] An artificial intelligence-based method for optimizing anesthetic drug dosage includes the following steps:

[0007] S1. Collect the patient's real-time physiological data and static data, and convert the patient's real-time physiological data and static data into standardized feature vectors through standardization processing. Based on the standardized feature vectors, use long short-term memory networks to predict the initial infusion rate of anesthetic drugs during surgery.

[0008] S2. Collect and standardize the patient's intraoperative data to obtain standardized intraoperative data; based on the standardized intraoperative data, generate personalized anesthetic drug infusion rates through a dynamic metabolic risk and neurological sensitivity joint optimization algorithm;

[0009] S3. Determine the maximum permissible infusion rate and minimum effective infusion rate based on clinical guidelines, obtain the remaining surgical time, perform safety verification based on the personalized anesthetic drug infusion rate, and generate the final infusion rate by combining the remaining surgical time.

[0010] Preferably, S1 specifically includes:

[0011] During the standardization process, the mean and standard deviation of each feature are calculated based on the patient's historical data and healthy population data before surgery. Then, the mean is subtracted from the patient's real-time physiological data and static data, and the result is divided by the standard deviation to generate a standardized feature vector. The standardized feature vector contains the patient's preoperative physiological and static features.

[0012] Preferably, S1 specifically includes:

[0013] Based on standardized feature vectors, a long short-term memory network is used to predict the initial infusion rate of anesthetic drugs during surgery. The long short-term memory network is trained using the patient's preoperative physiological data and the corresponding initial infusion rate annotations to obtain the trained long short-term memory network.

[0014] Preferably, S1 specifically includes:

[0015] The standardized feature vectors are input into the trained Long Short-Term Memory (LSTM) network. The temporal relationship of the standardized features is nonlinearly mapped, and the output of the trained LSM network is mapped to the initial infusion rate value using a fully connected layer.

[0016] Preferably, S2 specifically includes:

[0017] The standardized intraoperative data includes: standardized heart rate variability, standardized bispectral index of EEG, and standardized mean arterial pressure; the dynamic metabolic risk and neurosensitivity joint optimization algorithm generates personalized anesthetic drug infusion rates by combining standardized intraoperative data and drug effect delay.

[0018] Preferably, S2 specifically includes:

[0019] In the process of implementing the joint optimization algorithm of dynamic metabolic risk and neural sensitivity, the standardized intraoperative data were analyzed by evaluating the interaction effect of standardized mean arterial pressure and standardized bispectral index of EEG; and standardized heart rate variability was introduced to adjust the personalized anesthetic drug infusion rate.

[0020] Preferably, S2 specifically includes:

[0021] In the implementation of the joint optimization algorithm for dynamic metabolic risk and neural sensitivity, the effect of delayed onset of anesthetic drugs is calculated. This is achieved by obtaining the final infusion rate within a time window from historical intraoperative anesthetic drug infusion rate records, using a weighted summation method combined with an exponential decay function to obtain a personalized anesthetic drug infusion rate. The specific calculation formula is as follows:

[0022] ,

[0023] in, Indicates time Personalized anesthetic drug infusion rates; This represents the initial infusion rate, i.e., the initial rate of continuous infusion during the procedure, which serves as a benchmark for dynamic adjustment. This represents the weighting adjustment coefficient; This represents the standardized mean arterial pressure and the standardized bispectral index of electroencephalogram (EEG). Interactive characterization of metabolic-neural synergistic effects; This represents the standardized bispectral index of the brain. This represents standardized heart rate variability. Indicates the weight of the delay effect; Indicates the delayed effect of drug efficacy; Indicates the current time; Indicates time window Historical time points within; Indicates the time constant for drug effect delay; Indicates the time step; Represents an exponentially decaying function; Indicates time The final infusion rate.

[0024] Preferably, S3 specifically includes:

[0025] In generating the final infusion rate, the remaining surgical time is proportionally adjusted by a time adjustment factor to influence the personalized anesthetic drug infusion rate, generating candidate values. The formula for calculating the final infusion rate is as follows:

[0026] ,

[0027] in, Indicates time The final infusion rate; Indicates the remaining time of the surgery; Indicates the time adjustment factor; Indicates the minimum effective infusion rate; Indicates the maximum permissible infusion rate; Represents the maximum value function; Describes the minimum value function; This represents the generated candidate value.

[0028] Preferably, S3 specifically includes:

[0029] The candidate value is compared with the maximum permissible infusion rate and the minimum effective infusion rate: if the candidate value exceeds the maximum permissible infusion rate, the maximum permissible infusion rate is used; if it is lower than the minimum effective infusion rate, the minimum effective infusion rate is used; otherwise, the candidate value is retained so that the final infusion rate is within a safe range.

[0030] The beneficial effects of the technical solution of the present invention are:

[0031] 1. By using specialized medical equipment to collect real-time physiological data from patients and static data from the electronic medical record system, standardized feature vectors are generated through standardized processing. The initial infusion rate of anesthetic drugs during surgery is predicted based on Long Short-Term Memory (LSTM) networks and individualized patient characteristics. This reduces the risk of over- or under-anesthetic treatment due to empirical selection of anesthetic drug dosages. Automated data collection and processing also reduces the workload of manual analysis by medical staff and improves the efficiency and accuracy of preoperative anesthesia planning.

[0032] 2. Real-time acquisition of intraoperative patient data, including heart rate variability, bispectral index of electroencephalography, and mean arterial pressure. Standardized intraoperative data is generated through standardized processing. A dynamic metabolic risk and neurosensitivity joint optimization algorithm, combined with the drug effect delay effect, is used to generate personalized anesthetic drug infusion rates. The personalized anesthetic drug infusion rate adapts to the patient's physiological changes during surgery, avoiding excessively deep or shallow anesthesia, significantly reducing the risk of intraoperative patient awakening or over-anesthesia. The precise correction of the drug effect delay effect optimizes drug distribution, reduces unnecessary drug use, reduces postoperative drug residues, and shortens patient recovery time.

[0033] 3. Based on clinical guidelines, the maximum permissible infusion rate and minimum effective infusion rate are set. Safety verification is performed by combining the remaining surgical time with personalized anesthetic drug infusion rates to generate a clinically safe and reasonable final infusion rate. This rate is integrated with intelligent infusion pumps and electronic medical record systems. Rigorous safety verification effectively prevents overdose or underdose due to excessively high or low infusion rates, reducing intraoperative risks. Dynamic adjustments based on remaining surgical time reduce postoperative drug residue, shorten patient recovery time, and improve operating room turnover and resource utilization efficiency. Seamless integration with intelligent infusion pumps and electronic medical record systems automates, real-times, and digitizes anesthesia management, reducing the workload of medical staff and improving clinical workflow efficiency. Attached Figure Description

[0034] Figure 1 This is a flowchart of an artificial intelligence-based method for optimizing anesthetic drug dosage, as described in this invention. Detailed Implementation

[0035] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0037] The following description, in conjunction with the accompanying drawings, details a specific scheme for an artificial intelligence-based method for optimizing anesthetic drug dosage provided by the present invention.

[0038] See attached document Figure 1 The diagram illustrates a flowchart of an artificial intelligence-based anesthetic drug dosage optimization method according to an embodiment of the present invention, which includes the following steps:

[0039] S1. Collect the patient's real-time physiological data and static data, and convert the patient's real-time physiological data and static data into standardized feature vectors through standardized processing. Based on the standardized feature vectors, use long short-term memory networks to predict the initial infusion rate of anesthetic drugs during surgery.

[0040] One hour before the operation, real-time physiological data of the patient was collected using professional medical equipment, including heart rate (measured by an ECG monitor, in beats per minute), blood pressure (measured by a non-invasive blood pressure monitor, in millimeters of mercury), blood oxygen saturation (measured by a pulse oximeter, in percentage), and electroencephalogram (EEG) signals (measured by a portable EEG device, in microvolts). At the same time, the patient's static data was obtained from the electronic medical record system, including age (in years), weight (in kilograms), gender (categorical variable), liver and kidney function indicators (such as creatinine clearance rate, normalized to dimensionless values), and ASA classification (1-5).

[0041] The standardization process transforms the patient's real-time physiological and static data into standardized feature vectors. Specifically, the standardization process calculates the mean and standard deviation of each feature based on the patient's historical data or healthy population data before subtracting the mean from the patient's real-time physiological and static data and dividing by the standard deviation to generate the standardized feature vectors. ;

[0042] The standardized feature vector contains the patient's preoperative physiological and static characteristics, reflecting the patient's basal metabolism and health status, and is used to predict the initial infusion rate of anesthetic drugs during surgery.

[0043] Based on standardized feature vectors, the initial infusion rate of intraoperative anesthetic drugs is predicted using a long short-term memory network (LSTM), in milligrams per kilogram of body weight per hour, and applied via a smart infusion pump at the start of surgery (time point 0 minutes).

[0044] The Long Short-Term Memory (LSTM) network is trained using a historical surgical dataset (including preoperative physiological data of patients and corresponding initial infusion rate annotations). The trained LSTM network is then fed into a standardized feature vector, and the temporal relationships of the standardized features are non-linearly mapped. Finally, a fully connected layer (containing weights and biases) maps the output of the trained LSTM network to the initial infusion rate value, as shown in the following formula:

[0045] ,

[0046] in, This represents the initial infusion rate value; The weight matrix of the fully connected layer reflects the contribution weights of different features to the prediction of anesthetic drug dosage, and is optimized through training. This indicates the temporal dependence of capturing standardized feature vectors, such as small fluctuations in heart rate over time; This indicates a fully connected layer bias, used to adjust the baseline value for anesthetic drug dose prediction, ensuring that the predicted dose rate is within the clinical range;

[0047] The output initial infusion rate is used to reflect the patient's preoperative physiological state, but it cannot fully adapt to changes in the patient's physiological state during the operation, thus providing a benchmark for dynamic optimization of the patient's intraoperative state.

[0048] S2. Collect and standardize the patient's intraoperative data to obtain standardized intraoperative data; based on the standardized intraoperative data, generate personalized anesthetic drug infusion rates through a dynamic metabolic risk and neurosensitivity joint optimization algorithm.

[0049] During the surgery, intraoperative data of the patient is collected. Specifically, heart rate variability is calculated in real time using an ECG monitor, and time-domain analysis (such as the standard deviation of heart rate sequences) or frequency-domain analysis (such as the low-frequency / high-frequency ratio) is used to reflect metabolic activity. The bispectral index of the electroencephalogram (EEG) is measured using a BIS monitor to reflect the depth of anesthesia, ranging from 0 to 100, with a range of 40-60 considered the appropriate depth of anesthesia. Mean arterial pressure (unit: mmHg) is measured using an arterial pressure sensor to reflect the pressure of the circulatory system.

[0050] The patient's intraoperative data was standardized by calculating the mean and standard deviation of the patient's intraoperative data using a time window of the past 10 minutes, and then subtracting the mean of the time window from the patient's intraoperative data and dividing by the standard deviation to obtain the standardized intraoperative data.

[0051] Based on standardized intraoperative data, a dynamic metabolic risk and neurosensitivity joint optimization algorithm is used to analyze the standardized intraoperative data in real time. The standardized intraoperative data includes: standardized heart rate variability, standardized bispectral index (BPI), and standardized mean arterial pressure. The dynamic metabolic risk and neurosensitivity joint optimization algorithm combines standardized intraoperative data with the drug effect delay to generate personalized anesthetic drug infusion rates, suitable for general anesthesia surgeries, such as continuous propofol infusion. This ensures appropriate and safe intraoperative anesthesia depth, avoids overdose due to delayed drug onset, and adapts to dynamic changes in the patient's physiological state. The specific implementation process is as follows:

[0052] The standardized intraoperative data were analyzed by evaluating the interaction effect between standardized mean arterial pressure and standardized bispectral index of EEG. This reflected the synergistic state of circulatory system pressure and depth of anesthesia. A higher interaction effect indicated metabolic activity or insufficient anesthesia. Furthermore, standardized heart rate variability was introduced to adjust the personalized infusion rate of anesthetic drugs. During the patient's operation, a higher standardized heart rate variability suggested that the dosage of anesthetic drugs needed to be increased.

[0053] The calculation of the delayed effect of anesthetic drugs takes into account the amount of drug that has been injected but has not yet fully taken effect within the past time window. Specifically, the final infusion rate within the time window is obtained from the intraoperative historical anesthetic drug infusion rate record. Using a weighted summation method combined with an exponential decay function, the recent infusion volume is weighted according to the time distance. The recent infusion volume has a greater impact on the current anesthetic drug dose, while the impact of the long-term infusion volume on the current anesthetic drug dose gradually decreases.

[0054] The formula for calculating the personalized anesthetic drug infusion rate is as follows:

[0055] ,

[0056] in, Indicates time Personalized anesthetic drug infusion rates; This represents the initial infusion rate, i.e. the initial rate of continuous infusion during the operation, which serves as a benchmark for dynamic adjustment to ensure that the personalized infusion rate of anesthetic drugs is based on preoperative prediction. This represents the weighting adjustment coefficient, used to balance the contribution of the patient's intraoperative physiological data to the personalized adjustment of the anesthetic drug infusion rate, and to regulate the combined effect of metabolic-neural synergy and standardized heart rate variability on the anesthetic drug dosage. The value ranges from 0.1 to 0.5 and can be set according to the specific implementation scenario, without being limited here. This represents the standardized mean arterial pressure, i.e., over time. Circulatory system stress, and standardized bispectral index of EEG Interactive characterization of metabolic-neural synergistic effects; This represents the standardized bispectral index of brainwaves, i.e., the time-frequency response. The depth of anesthesia reflects the brain's sensitivity to anesthetic drugs. When the standardized bispectral index of EEG is low (too deep), the infusion rate of anesthetic drugs needs to be reduced. When the standardized bispectral index of EEG is high (too shallow), the infusion rate of anesthetic drugs needs to be increased. This represents standardized heart rate variability, i.e., over time. The metabolic activity level reflects the dynamic balance of the sympathetic-parasympathetic nervous system; a high value indicates that the infusion rate of anesthetic drugs needs to be increased. This represents the weight of the delayed effect, used to adjust the impact of delayed drug effect on the dosage adjustment of anesthetic drugs. The value ranges from 0.1 to 0.3 and can be set according to the specific implementation scenario. No limit is imposed here. This represents the delayed effect term of drug efficacy, estimating past effects. The amount of drug injected within the window but not yet fully effective is used to reduce the adjustment range of anesthetic drug dosage when the recent infusion volume is high, and to avoid over-injection due to delayed drug effect; Indicates the current time; Indicates time window Historical time points within; This represents the time constant for drug effect delay, in minutes, with a typical value of 1.5 min, based on clinical pharmacokinetic studies; Indicates the time step, in minutes, and is fixed at 1 (sampled once per minute); This represents the exponential decay function, used to simulate the delayed onset characteristics of a drug. Indicates time The final infusion rate;

[0057] By comprehensively considering the patient's intraoperative metabolic status, depth of anesthesia, and circulatory system pressure, a personalized anesthetic drug infusion rate is generated to adapt to individual patient differences, improve anesthetic effect, and significantly reduce the risk of overdose through the delayed effect of drug action.

[0058] S3. Determine the maximum permissible infusion rate and minimum effective infusion rate based on clinical guidelines, obtain the remaining surgical time, perform safety verification based on personalized anesthetic drug infusion rates, and generate the final infusion rate by combining the remaining surgical time.

[0059] The maximum permissible infusion rate and minimum effective infusion rate are determined based on clinical guidelines;

[0060] The remaining time of surgery is obtained from the surgical planning system or a real-time timer, reflecting the remaining time before the expected end of the surgery. Since the anesthetic drugs are continuously infused, the infusion rate is updated every minute, so the remaining time before the expected end of the surgery is checked every minute.

[0061] Based on the personalized anesthetic drug infusion rate, safety verification is performed, and combined with the remaining surgical time, a clinically safe and reasonable final infusion rate is generated.

[0062] Specifically, the remaining surgical time is proportionally adjusted by a time adjustment factor to influence the personalized anesthetic drug infusion rate, generating a candidate value. This candidate value is then compared with the maximum permissible infusion rate and the minimum effective infusion rate: if the candidate value exceeds the maximum permissible infusion rate, the maximum permissible infusion rate is used; if it is lower than the minimum effective infusion rate, the minimum effective infusion rate is used; otherwise, the candidate value is retained to ensure that the final infusion rate remains within a safe range, while also taking into account the dynamic changes in the surgical process, which aligns with clinical practice.

[0063] The formula for calculating the final infusion rate is:

[0064] ,

[0065] in, Indicates time The final infusion rate is used for clinical anesthetic drug administration, ensuring that the infusion rate of anesthetic drugs is within a safe range, optimizing drug distribution, transferring to the infusion pump and recording it; Indicates the remaining time of the surgery, indicating the time. The estimated duration until the end of the surgery is used to fine-tune the individualized anesthetic drug infusion rate; This represents the time adjustment factor, used to adjust the effect of the remaining surgical time on the dosage of anesthetic drugs. The value is 0.01, and it can be set according to the specific implementation scenario. There is no limitation here. This indicates the minimum effective infusion rate, expressed in mg / kg / h. It sets the lower limit for the infusion rate of anesthetic drugs to ensure anesthetic efficacy and is determined based on clinical guidelines. This indicates the maximum permissible infusion rate, expressed in mg / kg / h. It sets the upper limit for the infusion rate of anesthetic drugs to avoid overdose and is determined based on clinical guidelines. This represents the maximum value function, ensuring that the infusion rate of anesthetic drugs is not lower than the minimum effective infusion rate; This represents a minimum value function to ensure that the infusion rate of anesthetic drugs does not exceed the maximum permissible infusion rate; Indicates the generated candidate values;

[0066] Through rigorous safety constraint verification, it effectively prevents overdose or underdose of anesthetic drugs due to excessively high infusion rates, thus reducing intraoperative risks. Combined with dynamic adjustment of the remaining surgical time, it optimizes drug distribution, reduces postoperative drug residue, shortens patient recovery time, and improves surgical efficiency. Seamless integration with intelligent infusion pumps and electronic medical record systems achieves real-time and automation, updating the anesthetic drug infusion rate every minute, consistent with the sampling frequency of clinical monitoring equipment, making the operation highly efficient and closely aligned with general anesthesia practice.

[0067] In summary, an artificial intelligence-based method for optimizing anesthetic drug dosage has been developed.

[0068] The order of the embodiments is for illustrative purposes only and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0069] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0070] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for optimizing anesthetic drug dosage based on artificial intelligence, characterized in that, Includes the following steps: S1. Collect the patient's real-time physiological data and static data, and convert the patient's real-time physiological data and static data into standardized feature vectors through standardization processing. Based on the standardized feature vectors, use long short-term memory networks to predict the initial infusion rate of anesthetic drugs during surgery. S2. Collect and standardize the patient's intraoperative data to obtain standardized intraoperative data, including standardized heart rate variability, standardized bispectral index (BPI), and standardized mean arterial pressure (MAP). Introduce a dynamic metabolic risk and neurosensitivity joint optimization algorithm to analyze the standardized intraoperative data by evaluating the interaction effect between standardized MAP and standardized BPI. Furthermore, incorporate standardized heart rate variability and the drug effect delay effect to generate personalized anesthetic drug infusion rates. The specific calculation formula is as follows: ， in, Indicates time Personalized anesthetic drug infusion rates; This represents the initial infusion rate, i.e., the initial rate of continuous infusion during the procedure, which serves as a benchmark for dynamic adjustment. This represents the weighting adjustment coefficient; This represents the standardized mean arterial pressure; This represents the standardized bispectral index of the brain. This represents standardized heart rate variability. Indicates the weight of the delay effect; Indicates the delayed effect of drug efficacy; Indicates the current time; Indicates time window Historical time points within; Indicates the time constant for drug effect delay; Indicates the time step; Represents an exponentially decaying function; Indicates time The final infusion rate; S3. Based on clinical guidelines, determine the maximum permissible infusion rate and the minimum effective infusion rate, and obtain the remaining surgical time; the remaining surgical time is used to proportionally influence the personalized anesthetic drug infusion rate through a time adjustment factor, generating candidate values, and finally obtaining the final infusion rate; the formula for calculating the final infusion rate is: ， in, Indicates time The final infusion rate; Indicates the remaining time of the surgery; Indicates the time adjustment factor; Indicates the minimum effective infusion rate; Indicates the maximum permissible infusion rate; Represents the maximum value function; Describes the minimum value function; This represents the generated candidate value.

2. The method for optimizing anesthetic drug dosage based on artificial intelligence according to claim 1, characterized in that, S1 specifically includes: During the standardization process, the mean and standard deviation of each feature are calculated based on the patient's historical data and healthy population data before surgery. Then, the mean is subtracted from the patient's real-time physiological data and static data, and the result is divided by the standard deviation to generate a standardized feature vector. The standardized feature vector contains the patient's preoperative physiological and static features.

3. The method for optimizing anesthetic drug dosage based on artificial intelligence according to claim 2, characterized in that, S1 specifically includes: Based on standardized feature vectors, a long short-term memory network is used to predict the initial infusion rate of anesthetic drugs during surgery. The long short-term memory network is trained using the patient's preoperative physiological data and the corresponding initial infusion rate annotations to obtain the trained long short-term memory network.

4. The method for optimizing anesthetic drug dosage based on artificial intelligence according to claim 3, characterized in that, S1 specifically includes: The standardized feature vectors are input into the trained Long Short-Term Memory (LSTM) network. The temporal relationship of the standardized feature vectors is processed by nonlinear mapping. The output of the trained LSM network is then mapped to the initial infusion rate value using a fully connected layer.

5. The method for optimizing anesthetic drug dosage based on artificial intelligence according to claim 1, characterized in that, S3 specifically includes: The candidate value is compared with the maximum permissible infusion rate and the minimum effective infusion rate: if the candidate value exceeds the maximum permissible infusion rate, the maximum permissible infusion rate is used; if it is lower than the minimum effective infusion rate, the minimum effective infusion rate is used; otherwise, the candidate value is retained so that the final infusion rate is within a safe range.

Images