Analgesia pump flow prediction method based on gradient lifting tree

Abstract

The invention relates to the technical field of anesthesia, and discloses an analgesia pump flow prediction method based on a gradient boosting tree, the pain tolerance value A of a patient is obtained according to the patient age Age, the body weight We, the metabolic rate MR of drugs and the operation type ST, the flow P'of the analgesia pump used in the first day after the operation of the patient is given according to the pain tolerance value A of the patient, and the flow P 'of the analgesia pump used in the second day after the operation of the patient is calculated. Collecting a heart rate sampling set E {HR1, HR2,..., HRj} when the patient uses the analgesia pump; according to the method for predicting the flow of the analgesia pump based on the gradient boosting tree, the flow of the analgesia pump used by the patient in the future is predicted through the physiological parameters of the patient and the actual rehabilitation condition, the method is more scientific, meanwhile, the analgesia requirement of the patient can be more accurately met through personalized setting, discomfort caused by insufficient or excessive analgesia is reduced, and the patient experience is improved. And meanwhile, the flow parameter of the patient when the patient uses the analgesia pump can be adjusted according to the real-time physiological parameter of the patient.

Description

Technical Field

[0001] The present invention relates to the field of anesthesia technology, and in particular to a method for predicting the flow rate of an analgesic pump based on a gradient boosting tree. Background Art

[0002] Analgesia pump is a device that can continuously infuse fluids and maintain a stable concentration of drugs in the blood, so that better analgesia can be achieved with less drugs. Patients are usually allowed to press on their own to add an additional infusion dose on the basis of the continuous infusion volume. Therefore, the treatment is more individualized, which conforms to the characteristics of large individual differences in pain perception. Since analgesia pumps are mainly used for acute postoperative analgesia and maternal labor analgesia: and strong opioid analgesics are often added to analgesia pumps, improper dosage of this drug can often cause life-threatening complications such as respiratory depression in patients. At present, most domestic hospitals are mainly responsible for the preparation and management of analgesia pumps by anesthesiologists throughout the hospital.

[0003] In reality, since there are many patients with pumps in the hospital and they are scattered, the analgesic needs of patients are different in different time periods after surgery. Due to insufficient staffing in the anesthesia department and cost factors, it is impossible to arrange a staff member in each surgical department to perform individualized regulation and management of the patient's analgesic pump. Continuous pumping at a single speed throughout the whole process can easily cause some patients to experience severe pain due to insufficient analgesia, and some patients to experience nausea, vomiting, drowsiness, urinary retention, lower limb numbness, skin itching and other adverse reactions due to excessive analgesia, which effectively affects the patient's analgesic effect and comfort feeling. At the same time, in the existing technology, the initial flow rate of an analgesic pump given to the patient is based on the experience of the anesthesiologist. However, each patient has different conditions, age, weight, etc., and it is not scientific to use the experience of the anesthesiologist. Therefore, a prediction method for the flow rate of an analgesic pump based on a gradient boosting tree is proposed. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a method for predicting the flow of analgesic pumps based on a gradient boosting tree, which has the advantages of scientific and accurate prediction and can adjust the flow of analgesic pumps in real time according to the patient's physiological condition, thereby solving the problems mentioned in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for predicting the flow rate of analgesic pump based on a gradient boosting tree, comprising the following steps: S1: According to the patient's age Age, weight We, drug metabolic rate MR and surgery type ST, the patient's pain tolerance value A is obtained, and the flow rate P' of the analgesic pump used by the patient after surgery is given according to the patient's pain tolerance value A; S2: Collect the heart rate sampling set E{HR1, HR2, ..., HR j}, respiratory rate sampling set F{RR1, RR2, ..., RR j}, blood pressure sampling set G{BP1, BP2, ..., BP j} and the bispectral index sampling set H{BIS1, BIS2, ..., BIS j}, where j is the number of acquisitions, j is a positive integer with a value of 1≤j≤n; S3: preprocessing the data collected in step S2, setting corresponding weight thresholds for each preprocessed data, and calculating the data into a comprehensive physiological index S; S4: Determine whether the comprehensive physiological index S belongs to the pain tolerance value range If yes, do not change the analgesic pump parameters, otherwise judge whether the comprehensive physiological index S is > , if yes, increase the analgesic pump flow rate P', otherwise reduce the analgesic pump flow rate P', where β is a reasonable threshold; S5: At the same time, the patient's recovery index Q is obtained according to the patient's daily recovery status, and multiple physiological parameter indexes S n The patient's recovery index Q is set as the total data set D, the total data set , where U i is the eigenvector, V i is the target variable, and the total data set D is divided into a training set D1 and a validation set D2; S6: Build a gradient boosting tree model and transform the feature vector U i Input to the gradient boosting tree model, initialize the gradient boosting tree model, use the training set D1 to iteratively train the gradient boosting tree model, use the validation set D2 to evaluate the performance of the gradient boosting tree model, and obtain the optimal gradient boosting tree model; S7: Predict the patient's analgesic pump flow rate P'' based on the gradient boosting tree model; S8: When using the analgesic pump flow rate P'', loop steps S2 to S4, and fine-tune the patient's analgesic pump flow rate P'' according to the patient's real-time physiological index S.

[0006] The specific steps in step S2 are: S2.1: The heart rate sampling set E{HR1, HR2, ..., HR j}, respiratory rate sampling set F{RR1, RR2, ..., RR j}, blood pressure sampling set G{BP1, BP2, ..., BP j} and the bispectral index sampling set H{BIS1, BIS2, ..., BIS j} Denoising is performed through median filtering, the expression is: in, is the filtered heart rate sample set E1{HR1, HR2, ..., HR j}, respiratory rate sampling set F1{RR1, RR2, ..., RR j}, blood pressure sampling set G1{BP1, BP2, ..., BP j} and the bispectral index sampling set H1{BIS1, BIS2, ..., BIS j}, ζ is the heart rate sampling set E{HR1, HR2, ..., HR j}, respiratory rate sampling set F{RR1, RR2, ..., RR j}, blood pressure sampling set G{BP1, BP2, ..., BP j} and the bispectral index sampling set H{BIS1, BIS2, ..., BIS j}, M is the window size and is a positive integer; S2.2: The filtered heart rate sample set E1 {HR1, HR2, ..., HR j}, respiratory rate sampling set F{RR1, RR2, ..., RR j}, blood pressure sampling set G{BP1, BP2, ..., BP j} and the bispectral index sampling set H{BIS1, BIS2, ..., BIS j}Calculate the characteristic heart rate HR by arithmetic mean T , characteristic respiratory rate RR T , characteristic blood pressure BR T , characteristic heart rate BIS T , the expression is: S2.3: Characteristic heart rate HR T , characteristic respiratory rate RR T , characteristic blood pressure BR T , characteristic heart rate BIS T Normalize the data and map the result value to [0-1]. The expression is: in, Characteristic heart rate HR T , characteristic respiratory rate RR T , characteristic blood pressure BR T , characteristic heart rate BIS T , is the normalized characteristic heart rate HR α, characteristic respiratory rate RR α , characteristic blood pressure BR α , characteristic heart rate BIS α ; S2.4: According to the normalized characteristic heart rate HR α , characteristic respiratory rate RR α , characteristic blood pressure BR α , characteristic heart rate BIS α Calculate the comprehensive physiological index S, the expression is: S=HR α *W HR +RR α *W RR +BR α *W BP +BIS α *W BIS Among them, W HR Heart rate HR α The corresponding weight, W RR is the respiratory rate RR α The corresponding weight, W BP BR for blood pressure α The corresponding weight, W BIS Bispectral Index (BIS) α The corresponding weight.

[0007] The specific steps for calculating the pain tolerance value A are: S1.1: Label the surgical type ST to obtain ST encoded : S1.2: Define the pain tolerance value A, expressed as: Among them, λ1, λ2, λ3, and λ4 are weights corresponding to each parameter, and ε1 and ε2 are constant terms.

[0008] The specific steps in step S6 are: S6.1: Initialize a model B0(U), expressed as: B0(U)=γ Among them, B0 (U) is the initialized model, and γ is a constant; S6.2: The constant γ is expressed as: Where n is the number of samples in D1 in the training set, y i is the target value of the i-th sample; S6.3: In each iteration C, first calculate the current model B C-1(U) The negative gradient at each sample point is used as the pseudo residual K for training the new decision tree ci , the expression is K ci =2{y i -B C-1 (U i )} Among them, K ci is the pseudo residual; S6.4: Using the pseudo residual K ci {(U1, K c1 ), (U2, K c2 ),……(U n , K cn ),} as the target variable, train a new decision tree Z c (u); S6.5: Calculate a learning rate ρ to control the new decision tree Z c (u) is the contribution value of the gradient boosting tree model, and the learning rate ρ is used to update the gradient boosting tree model, which is expressed as: B C (U) = B C-1 (U)+ρ c Z c (u) Among them, B C (U) is the gradient boosting tree model obtained after the Cth iteration, ρ c is the learning rate after the Cth iteration, Z c (u) is the decision tree after the Cth iteration; S6.6: Repeat steps S6.3 to S6.5 until the performance on the validation set D2 is no longer significantly improved; S6.7: Use k-fold cross validation to divide the training set D1 into θ subsets of similar size. θ-1 subsets are used as training set D1, and the remaining subset is used as validation set D2. In each training process, the validation set D2 is used to evaluate the performance of the gradient boosting tree model, and the performance indicators of the gradient boosting tree model in each iteration are recorded.

[0009] The specific steps in step S7 are: S7.1: Multiple physiological parameter indexes S and the patient's rehabilitation index Q are used as the new feature vector U new ; S7.2: Transform the eigenvector U new Input to the trained gradient boosting tree model B C (U), the analgesia pump flow rate P'' for the next day is obtained, and the expression is: Among them, P'' is the actual predicted value.

[0010] The specific steps to obtain the recovery status Q are: Based on the information of the rehabilitation patients’ muscle strength, pain level and daily living activities, the patients’ MMT score, VAS score and Barthel index can be obtained; The MMT score, VAS score and Barthel index were quantified, namely η MMT , η VAS With η Barthel , and calculate the recovery status Q, the expression is: Among them, ω1, ω2 and ω3 are adjustment factors, which are used to indirectly assign weights to different evaluation scores, and τ is a nonlinear factor.

[0011] Preferably, the specific steps in step S4 are: When the flow rate of the analgesic pump needs to be increased, the increased flow rate P1 is calculated as follows: In the equation, φ is the increase factor and is a positive number; When the flow rate of the analgesic pump needs to be reduced, the reduced flow rate P2 is calculated as follows: Among them, μ is the increase factor and is a positive number.

[0012] Compared with the prior art, the present invention provides a method for predicting the flow rate of analgesic pumps based on a gradient boosting tree, which has the following beneficial effects: This method of predicting the flow rate of analgesic pumps based on the gradient boosting tree predicts the flow rate of analgesic pumps used by patients in the future through the physiological parameters of patients and actual rehabilitation conditions. It is more scientific and can be personalized to more accurately meet the analgesic needs of patients, reduce the discomfort caused by insufficient or excessive analgesia, and is more suitable for each patient's own flow conditions. At the same time, the flow parameters of patients using analgesic pumps can also be adjusted according to the real-time physiological parameters of patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is a schematic flow chart of a method for predicting the flow of analgesic pumps based on a gradient boosting tree proposed in the present invention. DETAILED DESCRIPTION

[0014] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0015] See also Figure 1 , a method for predicting the flow rate of analgesic pump based on gradient boosting tree, comprising the following steps: S1: According to the patient's age Age, weight We, drug metabolic rate MR and surgery type ST, the patient's pain tolerance value A is obtained, and the flow rate P' of the analgesic pump used by the patient after surgery is given according to the patient's pain tolerance value A; By comprehensively considering the patient's age Age, weight We, drug metabolism rate MR and surgical type ST, a scientific pain tolerance value A is calculated, and the analgesic pump flow P' is initially set accordingly. This scientific setting can more accurately meet the patient's analgesic needs and reduce the discomfort caused by insufficient or excessive analgesia.

[0016] S2: Collect the heart rate sampling set E{HR1, HR2, ..., HR j}, respiratory rate sampling set F{RR1, RR2, ..., RR j}, blood pressure sampling set G{BP1, BP2, ..., BP j} and the bispectral index sampling set H{BIS1, BIS2, ..., BIS j}, where j is the number of acquisitions, j is a positive integer with a value of 1≤j≤n; S3: preprocessing the data collected in step S2, setting corresponding weight thresholds for each preprocessed data, and calculating the data into a comprehensive physiological index S; S4: Determine whether the comprehensive physiological index S belongs to the pain tolerance value range If yes, do not change the analgesic pump parameters, otherwise judge whether the comprehensive physiological index S is > , if yes, increase the analgesic pump flow rate P', otherwise reduce the analgesic pump flow rate P', where β is a reasonable threshold value, and the threshold value β can be 10%; The analgesic pump flow rate for general patients is set at 2 ml / h, and the reference threshold β, i.e. 1.8 ml / h-2.2 ml / h, is a reasonable range.

[0017] Through real-time physiological monitoring and intelligent analgesic pump flow adjustment, the patient's pain can be effectively relieved and their comfort after surgery can be improved.

[0018] S5: At the same time, the patient's recovery index Q is obtained according to the patient's daily recovery status, and multiple physiological parameter indexes S n The patient's recovery index Q is set as the total data set D, the total data set , where U i is the eigenvector, V i is the target variable, and the total data set D is divided into a training set D1 and a validation set D2; Reasonable analgesia management can help alleviate patients' stress response and reduce the risk of postoperative complications, thereby accelerating the patient's recovery process. At the same time, by monitoring the patient's rehabilitation index Q and combining it with the physiological parameter index Sn for analysis, the patient's recovery status can be more comprehensively evaluated, providing a basis for formulating personalized rehabilitation plans.

[0019] S6: Build a gradient boosting tree model and transform the feature vector U i Input to the gradient boosting tree model, initialize the gradient boosting tree model, use the training set D1 to iteratively train the gradient boosting tree model, use the validation set D2 to evaluate the performance of the gradient boosting tree model, and obtain the optimal gradient boosting tree model; S7: Predict the patient's analgesic pump flow rate P'' based on the gradient boosting tree model; S8: When using the analgesic pump flow rate P'', loop steps S2 to S4, and fine-tune the patient's analgesic pump flow rate P'' according to the patient's real-time physiological index S.

[0020] Constructing and applying a gradient boosting tree model to predict and optimize the analgesic pump flow rate P'' based on the patient's physiological parameters and rehabilitation status can reduce the subjectivity and uncertainty of human judgment and improve the scientificity and accuracy of analgesia management.

[0021] This detection or adjustment can be performed twice or three times a day, and can be determined according to the patient's own physiological condition.

[0022] The specific steps in step S2 are: S2.1: The heart rate sampling set E{HR1, HR2, ..., HR j}, respiratory rate sampling set F{RR1, RR2, ..., RR j}, blood pressure sampling set G{BP1, BP2, ..., BP j} and the bispectral index sampling set H{BIS1, BIS2, ..., BIS j} Denoising is performed through median filtering, the expression is: in, is the filtered heart rate sample set E1{HR1, HR2, ..., HR j}, respiratory rate sampling set F1{RR1, RR2, ..., RR j}, blood pressure sampling set G1{BP1, BP2, ..., BP j} and the bispectral index sampling set H1{BIS1, BIS2, ..., BIS j}, ζ is the heart rate sampling set E{HR1, HR2, ..., HR j}, respiratory rate sampling set F{RR1, RR2, ..., RR j}, blood pressure sampling set G{BP1, BP2, ..., BP j} and the bispectral index sampling set H{BIS1, BIS2, ..., BIS j}, M is the window size and is a positive integer; S2.2: The filtered heart rate sample set E1 {HR1, HR2, ..., HR j}, respiratory rate sampling set F{RR1, RR2, ..., RR j}, blood pressure sampling set G{BP1, BP2, ..., BP j} and the bispectral index sampling set H{BIS1, BIS2, ..., BIS j}Calculate the characteristic heart rate HR by arithmetic average T , characteristic respiratory rate RR T , characteristic blood pressure BR T , characteristic heart rate BIS T , the expression is: S2.3: Characteristic heart rate HR T , characteristic respiratory rate RR T , characteristic blood pressure BR T , characteristic heart rate BIS T Normalize the data and map the result value to [0-1]. The expression is: in, Characteristic heart rate HR T , characteristic respiratory rate RR T , characteristic blood pressure BR T , characteristic heart rate BIS T , is the normalized characteristic heart rate HR α , characteristic respiratory rate RR α , characteristic blood pressure BR α, characteristic heart rate BIS α ; S2.4: According to the normalized characteristic heart rate HR α , characteristic respiratory rate RR α , characteristic blood pressure BR α , characteristic heart rate BIS α Calculate the comprehensive physiological index S, the expression is: S=HR α *W HR +RR α *W RR +BR α *W BP +BIS α *W BIS Among them, W HR Heart rate HR α The corresponding weight, W RR is the respiratory rate RR α The corresponding weight, W BP BR for blood pressure α The corresponding weight, W BIS Bispectral Index (BIS) α The corresponding weight.

[0023] The specific steps for calculating the pain tolerance value A are: S1.1: Label the surgical type ST to obtain ST encoded : S1.2: Define the pain tolerance value A, expressed as: Among them, λ1, λ2, λ3, and λ4 are weights corresponding to each parameter, and ε1 and ε2 are constant terms.

[0024] The specific steps in step S6 are: S6.1: Initialize a model B0(U), expressed as: B0(U)=γ Among them, B0(U) is the initialized model, γ is a constant, usually equal to the mean of the target variable in the training set; S6.2: The constant γ is expressed as: Where n is the number of samples in D1 in the training set, y i is the target value of the i-th sample, and the target value is the flow value P''; S6.3: In each iteration C, first calculate the current model B C-1 (U) The negative gradient at each sample point is used as the pseudo residual K for training the new decision treeci , the expression is K ci =2{y i -B C-1 (U i )} Among them, K ci is the pseudo residual; S6.4: Using the pseudo residual K ci {(U1, K c1 ), (U2, K c2 ),……(U n , K cn ),} as the target variable, train a new decision tree Z c (u); S6.5: Calculate a learning rate ρ to control the new decision tree Z c (u) is the contribution value of the gradient boosting tree model, and the learning rate ρ is used to update the gradient boosting tree model, which is expressed as: B C (U) = B C-1 (U)+ρ c Z c (u) Among them, B C (U) is the gradient boosting tree model obtained after the Cth iteration, ρ c is the learning rate after the Cth iteration, Z c (u) is the decision tree after the Cth iteration; S6.6: Repeat steps S6.3 to S6.5 until the performance on the validation set D2 is no longer significantly improved; S6.7: Use k-fold cross validation to divide the training set D1 into θ subsets of similar size. θ-1 subsets are used as training set D1, and the remaining subset is used as validation set D2. In each training process, the validation set D2 is used to evaluate the performance of the gradient boosting tree model, and the performance indicators of the gradient boosting tree model in each iteration are recorded.

[0025] Through intelligent analgesia management methods, doctors can manage patients' analgesia needs more efficiently and reduce unnecessary waste of medical resources. At the same time, the application of this method will also help improve the overall quality and efficiency of medical services.

[0026] The specific steps in step S7 are: S7.1: Multiple physiological parameter indexes S and the patient's rehabilitation index Q are used as the new feature vector U new ; S7.2: Transform the eigenvector U new Input to the trained gradient boosting tree model B C(U), the analgesia pump flow rate P'' for the next day is obtained, and the expression is: Among them, P'' is the actual predicted value.

[0027] The specific steps to obtain the recovery status Q are: Based on the information of the rehabilitation patients’ muscle strength, pain level and daily living activities, the patients’ MMT score, VAS score and Barthel index can be obtained; The MMT score, VAS score and Barthel index were quantified, namely η MMT , η VAS With η Barthel , and calculate the recovery status Q, the expression is: Among them, ω1, ω2 and ω3 are adjustment factors, which are used to indirectly assign weights to different evaluation scores, and τ is a nonlinear factor.

[0028] The specific steps in step S4 are: When the flow rate of the analgesic pump needs to be increased, the increased flow rate P1 is calculated as follows: Among them, φ is the increase factor and is a positive number; When the flow rate of the analgesic pump needs to be reduced, the reduced flow rate P2 is calculated as follows: Among them, μ is the increase factor and is a positive number.

[0029] In summary, the analgesic pump flow prediction method based on the gradient boosting tree predicts the patient's future analgesic pump flow through the patient's physiological parameters and actual rehabilitation status. It is more scientific and can be personalized to more accurately meet the patient's analgesic needs, reduce the discomfort caused by insufficient or excessive analgesia, and is more suitable for each patient's own flow situation. At the same time, it can also adjust the patient's flow parameters when using the analgesic pump according to the patient's real-time physiological parameters.

[0030] It should be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0031] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for predicting the flow rate of analgesic pump based on gradient boosting tree, characterized in that: The following steps are involved: S1: According to the patient's age Age, weight We, drug metabolic rate MR and surgery type ST, the patient's pain tolerance value A is obtained, and the flow rate P' of the analgesic pump used by the patient after surgery is given according to the patient's pain tolerance value A; S2: Collect the heart rate sampling set E{HR1, HR2, ..., HR j }, respiratory rate sampling set F{RR1, RR2, ..., RR j }, blood pressure sampling set G{BP1, BP2, ..., BP j } and the bispectral index sampling set H{BIS1, BIS2, ..., BIS j }, where j is the number of acquisitions, j is a positive integer with a value of 1≤j≤n; S3: preprocessing the data collected in step S2, setting corresponding weight thresholds for each preprocessed data, and calculating the data into a comprehensive physiological index S; S4: Determine whether the comprehensive physiological index S belongs to the pain tolerance value range If yes, do not change the analgesic pump parameters, otherwise judge whether the comprehensive physiological index S is > , if yes, increase the analgesic pump flow rate P', otherwise reduce the analgesic pump flow rate P', where β is a reasonable threshold; S5: At the same time, the patient's recovery index Q is obtained according to the patient's daily recovery status, and multiple physiological parameter indexes S n The patient's recovery index Q is set as the total data set D, the total data set , where U i is the eigenvector, V i is the target variable, and the total data set D is divided into a training set D1 and a validation set D2; S6: Build a gradient boosting tree model and transform the feature vector U i Input to the gradient boosting tree model, initialize the gradient boosting tree model, use the training set D1 to iteratively train the gradient boosting tree model, use the validation set D2 to evaluate the performance of the gradient boosting tree model, and obtain the optimal gradient boosting tree model; S7: Predict the patient's analgesic pump flow rate P'' based on the gradient boosting tree model; S8: When using the analgesic pump flow rate P'', loop steps S2 to S4, and fine-tune the patient's analgesic pump flow rate P'' according to the patient's real-time physiological index S.

2. The method for predicting the flow rate of analgesic pump based on gradient boosting tree according to claim 1, characterized in that: The specific steps in step S2 are: S2.1: The heart rate sampling set E{HR1, HR2, ..., HR j }, respiratory rate sampling set F{RR1, RR2, ..., RR j }, blood pressure sampling set G{BP1, BP2, ..., BP j } and the bispectral index sampling set H{BIS1, BIS2, ..., BIS j } Denoising is performed through median filtering, the expression is: in, is the filtered heart rate sample set E1{HR1, HR2, ..., HR j }, respiratory rate sampling set F1{RR1, RR2, ..., RR j }, blood pressure sampling set G1{BP1, BP2, ..., BP j } and the bispectral index sampling set H1{BIS1, BIS2, ..., BIS j }, ζ is the heart rate sampling set E{HR1, HR2, ..., HR j }, respiratory rate sampling set F{RR1, RR2, ..., RR j }, blood pressure sampling set G{BP1, BP2, ..., BP j } and the bispectral index sampling set H{BIS1, BIS2, ..., BIS j }, M is the window size and is a positive integer; S2.2: The filtered heart rate sample set E1 {HR1, HR2, ..., HR j }, respiratory rate sampling set F{RR1, RR2, ..., RR j }, blood pressure sampling set G{BP1, BP2, ..., BP j } and the bispectral index sampling set H{BIS1, BIS2, ..., BIS j }Calculate the characteristic heart rate HR by arithmetic average T , characteristic respiratory rate RR T , characteristic blood pressure BR T , characteristic heart rate BIS T , the expression is: S2.3: Characteristic heart rate HR T , characteristic respiratory rate RR T , characteristic blood pressure BR T , characteristic heart rate BIS T Normalize the data and map the result value to [0-1]. The expression is: in, Characteristic heart rate HR T , characteristic respiratory rate RR T , characteristic blood pressure BR T , characteristic heart rate BIS T , is the normalized characteristic heart rate HR α , characteristic respiratory rate RR α , characteristic blood pressure BR α , characteristic heart rate BIS α ; S2.4: According to the normalized characteristic heart rate HR α , characteristic respiratory rate RR α , characteristic blood pressure BR α , characteristic heart rate BIS α Calculate the comprehensive physiological index S, the expression is: S=HR α *W HR +RR α *W RR +BR α *W BP +BIS α *W BIS Among them, W HR Heart rate HR α The corresponding weight, W RR is the respiratory rate RR α The corresponding weight, W BP BR for blood pressure α The corresponding weight, W BIS Bispectral Index (BIS) α The corresponding weight.

3. The method for predicting the flow rate of analgesic pump based on gradient boosting tree according to claim 1, characterized in that: The specific steps of calculating the pain tolerance value A are: S1.1: Label the surgical type ST to obtain ST encoded : S1.2: Define the pain tolerance value A, expressed as: Among them, λ1, λ2, λ3, and λ4 are weights corresponding to each parameter, and ε1 and ε2 are constant terms.

4. The method for predicting the flow rate of analgesic pump based on gradient boosting tree according to claim 1, characterized in that: The specific steps in step S6 are: S6.1: Initialize a model B0(U), expressed as: B0(U)=γ Among them, B0 (U) is the initialized model, and γ is a constant; S6.2: The constant γ is expressed as: Where n is the number of samples in D1 in the training set, y i is the target value of the i-th sample; S6.3: In each iteration C, first calculate the current model B C-1 (U) The negative gradient at each sample point is used as the pseudo residual K for training the new decision tree ci , the expression is K ci =2{y i -B C-1 (U i )} Among them, K ci is the pseudo residual; S6.4: Using the pseudo residual K ci {(U1, K c1 ), (U2, K c2 ),……(U n , K cn ),} as the target variable, train a new decision tree Z c (u); S6.5: Calculate a learning rate ρ to control the new decision tree Z c (u) is the contribution value of the gradient boosting tree model, and the learning rate ρ is used to update the gradient boosting tree model, which is expressed as: B C (U)=B C-1 (U)+ρ c Z c (u) Among them, B C (U) is the gradient boosting tree model obtained after the Cth iteration, ρ c is the learning rate after the Cth iteration, Z c (u) is the decision tree after the Cth iteration; S6.6: Repeat steps S6.3 to S6.5 until the performance on the validation set D2 is no longer significantly improved; S6.7: Use k-fold cross validation to divide the training set D1 into θ subsets of similar size. θ-1 subsets are used as training set D1, and the remaining subset is used as validation set D2. In each training process, the validation set D2 is used to evaluate the performance of the gradient boosting tree model, and the performance indicators of the gradient boosting tree model in each iteration are recorded.

5. The method for predicting the flow rate of analgesic pump based on gradient boosting tree according to claim 1, characterized in that: The specific steps in step S7 are: S7.1: Multiple physiological parameter indexes S and the patient's rehabilitation index Q are used as the new feature vector U new ; S7.2: Transform the eigenvector U new Input to the trained gradient boosting tree model B C (U), the analgesia pump flow rate P'' for the next day is obtained, and the expression is: P'' = B C (U new ) Where P'' is the actual predicted value.

6. The method for predicting the flow rate of analgesic pump based on gradient boosting tree according to claim 1, characterized in that: The specific steps for obtaining the recovery status Q are: Based on the information of the rehabilitation patients’ muscle strength, pain level and daily living activities, the patients’ MMT score, VAS score and Barthel index can be obtained; The MMT score, VAS score and Barthel index were quantified, namely η MMT , η VAS With η Barthel , and calculate the recovery status Q, the expression is: Among them, ω1, ω2 and ω3 are adjustment factors, which are used to indirectly assign weights to different evaluation scores, and τ is a nonlinear factor.

7. The method for predicting the flow rate of analgesic pump based on gradient boosting tree according to claim 1, characterized in that: The specific steps in step S4 are: When the flow rate of the analgesic pump needs to be increased, the increased flow rate P1 is calculated as follows: Among them, φ is the increase factor and is a positive number; When the flow rate of the analgesic pump needs to be reduced, the reduced flow rate P2 is calculated as follows: Among them, μ is the increase factor and is a positive number.

Images