Anesthetic dosage prediction method based on big data analysis

Abstract

The invention discloses an anesthetic dosage prediction method based on big data analysis, and relates to the field of anesthetic dosage prediction, and the method comprises the steps: obtaining patient information of anesthetic injection, screening useful information, including influence factors and prediction factors, according to the patient information by using a multi-round regression model, dividing the useful information into a training group and a control group according to a proportion, and establishing a linear regression model according to the predictive factor and the anesthesia site and area, visualizing the linear regression model in a column diagram form, and adjusting and evaluating the linear regression model to obtain a predictive model. Complex data of a patient is collected through big data, prediction to be performed is more accurate based on multiple regression model recursive analysis data, the interpretability of a prediction model is improved, the prediction data is visualized, the interactivity of the prediction data is increased, and the prediction efficiency is improved. Based on the linear regression model, the anesthetic injection dosage required by a patient individual can be quickly analyzed and predicted, so that the workload of an anesthetist is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of anesthetic dosage prediction, and in particular to an anesthetic dosage prediction method based on big data analysis. Background Art

[0002] In recent years, the dose of anesthetics has become a key medical parameter in surgery and treatment, directly affecting the surgical effect and safety of patients. Traditional methods of calculating the dose of anesthetics are based on the doctor's experience and the general characteristics of the patient, such as age and weight, and have problems of large individual differences and low accuracy. The development of big data technology enables the medical industry to collect, store and analyze large-scale medical data, including patients' physiological parameters, medical history, monitoring data during surgery, etc. These data are analyzed through appropriate algorithms and models to reveal the complex relationship between drug dose and patient characteristics, surgery type, physiological state, etc., and can predict the best anesthetic dose for each patient in a personalized manner, improving the safety and efficiency of surgery. This is an important development direction in the field of medical information technology.

[0003] At present, the Chinese invention patent with application number CN202311210971.4 discloses a drug dosage prediction system and method, which can enable users to independently add new prediction models, adapt to multiple development languages, support data introduction in various environments, and realize multiple forms of result presentation. It has good scalability for the docking of new development languages ​​and the introduction of models, and can realize rapid parameter adjustment and model training to complete the prediction of drug dosage of new drugs. However, the existing technology cannot improve the personalization, accuracy and interpretability of anesthesia management by making full use of large-scale medical data and advanced analysis technology based on big data analysis. The anesthetic dosage prediction method cannot quickly analyze the anesthetic injection dosage required by individual patients, thereby reducing the workload of anesthesiologists. Summary of the invention

[0004] The technical problem solved by the present invention is that the existing technology cannot use the anesthetic dosage prediction method based on big data analysis to improve the personalization, accuracy and interpretability of anesthesia management by making full use of large-scale medical data and advanced analysis technology, and cannot quickly analyze the anesthetic injection dosage required by individual patients.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for predicting anesthetic dosage based on big data analysis, comprising the following steps:

[0006] Step S1: Acquire patient information of anesthetic injection, wherein the patient information includes biological indicator data of the patient, and corresponding anesthetic injection method, injection dosage and anesthetic type;

[0007] Step S2: Based on the biological indicator data, a first characteristic factor is screened using a Lasso regression model, and the first characteristic factor is analyzed using a Logistic regression model to obtain a second characteristic factor. Based on the first characteristic factor and the second characteristic factor, other factor characteristics that affect the anesthetic dose are predicted, and the other factor characteristics are combined with the second characteristic factor to obtain a third characteristic factor;

[0008] Step S3: performing a multi-factor regression analysis on the third characteristic factor to obtain the influence of the third characteristic factor on the injection dose of the anesthetic, and dividing the third characteristic factor corresponding to the influence into a training group and a control group according to the proportion;

[0009] Step S4: putting the training group back into the Lasso regression model and combining the ten-fold cross validation method to screen the predictive factors, establishing a linear regression model based on the predictive factors and the anesthesia site and area, and visualizing the linear regression model in the form of a nomogram;

[0010] Step S5: The linear regression model is adjusted by adjusting the training group, and the linear regression model is evaluated by using the control group, and a prediction model is obtained when the evaluation error is less than a preset value.

[0011] Preferably, the step S1 comprises:

[0012] The biological indicator data include age, height, BMI, weight, age, ASA classification, anesthesia site and area, operation time, anesthesia time, combined medication, surgical procedure, heart rate, bispectral index value, blood oxygen saturation and mean arterial pressure value intervals at the time of patient admission, pre-induction time, intubation time, operation start time, operation end time and extubation time, and the patients are grouped according to the biological indicator data;

[0013] The anesthetic injection methods are grouped according to the part of the patient's body involved in the surgery.

[0014] Preferably, the step S2 comprises:

[0015] The first characteristic factor represents a factor related to the injection dose of anesthetic drugs, and the second characteristic factor represents a characteristic factor among the first characteristic factors that affects the injection dose of anesthetic drugs. Based on the first characteristic factor and the second characteristic factor, a stepwise regression analysis method is used to predict the characteristics of other factors that affect the anesthetic dose through big data retrieval, and the characteristics of the other factors are combined with the second characteristic factor to obtain a third characteristic factor, and the third characteristic factor represents a factor that affects the injection dose of anesthetic drugs.

[0016] Preferably, the first characteristic factor related to the injection dose of anesthetic drug in the biological indicator data and genotype data screened by using the Lasso regression model includes:

[0017] Performing Huffman coding of the same length on the biological indicator data, defining the Huffman coding as an independent variable, defining the anesthetic injection dose as a dependent variable, performing median filling on missing values ​​in the independent variable and the dependent variable, performing normalization operation on the independent variable, establishing a Lasso regression model based on the independent variable and the dependent variable, adding a regularization term to the Lasso regression model to achieve feature selection, wherein the coefficient of the regularization term is controlled by a user-defined parameter λ, and dividing the independent variable and the dependent variable into a first training set and a first test set based on individuals and group ratios;

[0018] Fitting a Lasso regression model on the training set, defining the objective of the Lasso regression model training as minimizing a loss function, wherein the loss function includes an original loss function and a regularization term;

[0019] A cross-validation method is used to select the optimal regularization parameter λ, λ is introduced into the Lasso regression model, and the Lasso regression model is repeatedly trained. According to the Lasso regression model completed by repeated training, coefficients of all characteristic factors to be obtained are obtained, all coefficients are sorted from large to small according to their absolute values, and characteristic factors whose coefficients are greater than a preset threshold are retained. The Lasso regression model is retrained using the retained characteristic factors, and the coefficients of several features of the retrained Lasso regression model are compressed to zero to obtain a first characteristic factor.

[0020] Preferably, the cross-validation method selects the optimal regularization parameter λ including:

[0021] The K-fold cross-validation method is used to divide the first training set into K parts, set the form of the Lasso regression model, select a set of candidate values ​​of λ, use K-1 parts of the first training set as the second training set, and use the remaining part of the first training set as the second validation set. The Lasso regression model is fitted on the second training set, and the performance of the model is evaluated according to the validation set using the mean square error. The validation set evaluation index of each λ value is recorded, the average performance index of K cross validations for each λ value is calculated, and the λ value that minimizes the MSE is selected as the optimal regularization parameter.

[0022] Preferably, step S3 comprises:

[0023] The third characteristic factor is subjected to multivariate regression analysis to obtain the influence of the third characteristic factor on the injection dose of the anesthetic, the influence being expressed in percentage, and the characteristic factors with influence less than 0 are removed to obtain the influencing factors, a pie-shaped trend chart is established according to the influence and the corresponding influencing factors, and the influencing factors are divided into a training group and a control group according to the proportion;

[0024] The influencing factors are divided into a training group and a control group according to proportion, including randomly selecting patients accounting for 20% of the total number of influencing factors of the same type in each influencing factor in turn as the control group, and the remaining patients accounting for 80% of the total number of influencing factors of the same type are defined as the training group.

[0025] Preferably, performing multivariate regression analysis includes:

[0026] A stepwise regression analysis method is used to gradually add or delete independent variables through the step function, and the dependent variable is preliminarily predicted using the Poisson regression method. Possible predictors are gradually introduced or excluded to obtain the optimal predictor. An independent variable combination is obtained according to the optimal predictor. A multifactor regression analysis is performed according to the independent variable combination. Several influencing factors are randomly selected from the results of the multifactor regression analysis. The influencing factors include the coefficients of the independent variables. The influencing factors are represented on the horizontal axis using a bar chart, and the anesthetic dose is represented on the vertical axis, and each influencing factor corresponds to a bar.

[0027] 80% of the influencing factors are divided into a training group, and 20% of the influencing factors are divided into a control group.

[0028] Preferably, step S4 comprises:

[0029] The ten-fold cross validation method is to change K in the K-fold cross validation method to 10;

[0030] The prediction factor is defined as a second dependent variable, the training group is defined as a second independent variable, and in the Lasso regression model regression process, the prediction factor is selected as a feature whose corresponding coefficient is not zero when a certain value is reached;

[0031] The mathematical regression model established based on the predictive factors and the anesthesia site and area includes:

[0032] Define the predictor as the independent variable code x, define the anesthesia site and area as the independent variable code y, establish a linear regression model based on the independent variable code x and the independent variable code y, and use a nomogram to draw a visual two-dimensional graph of the linear regression model established based on the independent variable code x and the independent variable code y.

[0033] Preferably, step S5 comprises:

[0034] The regression coefficient of the linear regression model is calculated by the reduction method to obtain a new fitting line, which is plotted on the two-dimensional graph for comparison. The fitting degree of the predicted value and the data is used to analyze the effect, and a fitting threshold is set. If the fitting degree is higher than the fitting threshold, the linear regression model has a good fitting effect and high prediction accuracy. If the fitting degree is lower than the fitting threshold, the encoding method and parameters are adjusted, the linear regression model is re-established and re-evaluated until the fitting degree reaches the fitting threshold to obtain a prediction model.

[0035] Preferably, the linear regression model evaluation index and prediction efficiency include:

[0036] The area under the ROC curve of the linear regression model was calculated, and the consistency index, accuracy, sensitivity, specificity, positive predictive value, negative predictive value and harmonic mean score index of precision and recall in machine learning of the linear regression model were analyzed. The accuracy of the linear regression model was evaluated by calibration curve and the clinical decision curve analysis was used to evaluate the clinical benefit of the model. Python software was used for data analysis. The chi-square test was used among all groups to perform Lasso regression on the characteristic factors with an error of <0.2 for re-variable selection. When the error was <0.05, the prediction model was obtained.

[0037] The beneficial effects of the present invention are as follows: it can recursively analyze data based on big data analysis and multiple regression models, improve the personalization, accuracy and interpretability of anesthesia management by making full use of large-scale medical data and advanced analysis techniques, and can quickly analyze and predict the anesthetic injection dose required by individual patients, thereby reducing the workload of anesthesiologists. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A basic flow chart of a method for predicting anesthetic dosage based on big data analysis is provided for one embodiment of the present invention. DETAILED DESCRIPTION

[0039] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0040] Example 1, reference Figure 1 , as an embodiment of the present invention, provides an anesthetic dosage prediction method based on big data analysis, comprising the following steps:

[0041] Step S1: Acquire patient information of anesthetic injection, wherein the patient information includes biological indicator data of the patient, and corresponding anesthetic injection method, injection dosage and anesthetic type;

[0042] Step S2: Based on the biological indicator data, a first characteristic factor is screened using a Lasso regression model, and the first characteristic factor is analyzed using a Logistic regression model to obtain a second characteristic factor. Based on the first characteristic factor and the second characteristic factor, other factor characteristics that affect the anesthetic dose are predicted, and the other factor characteristics are combined with the second characteristic factor to obtain a third characteristic factor;

[0043] Step S3: performing a multi-factor regression analysis on the third characteristic factor to obtain the influence of the third characteristic factor on the injection dose of the anesthetic, and dividing the third characteristic factor corresponding to the influence into a training group and a control group according to the proportion;

[0044] Step S4: putting the training group back into the Lasso regression model and combining the ten-fold cross validation method to screen the predictive factors, establishing a linear regression model based on the predictive factors and the anesthesia site and area, and visualizing the linear regression model in the form of a nomogram;

[0045] Step S5: The linear regression model is adjusted by adjusting the training group, and the linear regression model is evaluated by using the control group, and a prediction model is obtained when the evaluation error is less than a preset value.

[0046] Through big data, complex data of patients are collected, and recursive data analysis is performed based on multiple regression models to screen out useful data. Based on the linear regression model, the anesthetic injection dose required by individual patients can be quickly analyzed and predicted.

[0047] The step S1 comprises:

[0048] The biological indicator data include age, height, BMI, weight, age, ASA classification, anesthesia site and area, operation time, anesthesia time, combined medication, surgical procedure, heart rate, bispectral index value, blood oxygen saturation and mean arterial pressure value intervals at the time of patient admission, pre-induction time, intubation time, operation start time, operation end time and extubation time, and the patients are grouped according to the biological indicator data;

[0049] The anesthetic injection method is grouped according to the body parts involved in the patient's surgery. The body parts are divided into the head, neck, upper limbs, chest, abdomen, lower limbs, reproductive organs, back and other parts based on medical basis. The other parts include all parts of the human body except the head, neck, upper limbs, chest, abdomen, lower limbs, reproductive organs and back, including but not limited to skin, bones, muscles, blood vessels and nerves.

[0050] All patients were prohibited from eating or drinking before surgery, and their non-invasive blood pressure, electrocardiogram, and blood oxygen saturation were routinely monitored. Invasive blood pressure monitoring was also performed, and a bispectral index monitor was connected to monitor the depth of anesthesia. Additional muscle relaxants were expected to be stopped 1 hour before the end of the surgery, and the infusion of anesthetic drugs was stopped immediately after the surgery. After the patient woke up and resumed spontaneous breathing, the endotracheal tube was removed and the patient was sent to the anesthesia recovery room.

[0051] The step S2 comprises:

[0052] The first characteristic factor represents a factor related to the injection dose of anesthetic drugs, and the second characteristic factor represents a characteristic factor among the first characteristic factors that affects the injection dose of anesthetic drugs. Based on the first characteristic factor and the second characteristic factor, a stepwise regression analysis method is used to predict the characteristics of other factors that affect the anesthetic dose through big data retrieval, and the characteristics of the other factors are combined with the second characteristic factor to obtain a third characteristic factor, and the third characteristic factor represents a factor that affects the injection dose of anesthetic drugs.

[0053] The first characteristic factors related to the anesthetic injection dose in the biological indicator data and genotype data screened by using the Lasso regression model include:

[0054] Perform Huffman coding of the same length on the biological indicator data, define the Huffman coding as an independent variable, define the anesthetic injection dose as a dependent variable, perform median filling on missing values ​​in the independent and dependent variables, and perform normalization on the independent variables to ensure that the numerical ranges of different features are similar to avoid the impact of different numerical ranges on the model. Establish a Lasso regression model based on the independent and dependent variables, add a regularization term to the Lasso regression model to achieve feature selection, the coefficient of the regularization term is controlled by a user-defined parameter λ, and divide the independent and dependent variables into a first training set and a first test set based on individuals and group ratios;

[0055] The groups included in the stated proportions were:

[0056] For example, if the age groups are grouped into age group 1, age group 2, ..., age group k, then 80% of the age group 1, age group 2, ..., age group k are taken as training sets, and 20% of the age group 1, age group 2, ..., age group k are taken as test sets;

[0057] Fitting a Lasso regression model on the training set, defining the objective of the Lasso regression model training as minimizing a loss function, wherein the loss function includes an original loss function and a regularization term;

[0058] A cross-validation method is used to select the optimal regularization parameter λ, λ is introduced into the Lasso regression model, and the Lasso regression model is repeatedly trained. According to the Lasso regression model completed by repeated training, coefficients of all characteristic factors to be obtained are obtained, all coefficients are sorted from large to small according to their absolute values, and characteristic factors whose coefficients are greater than a preset threshold are retained. The Lasso regression model is retrained using the retained characteristic factors, and the coefficients of several features of the retrained Lasso regression model are compressed to zero to obtain a first characteristic factor.

[0059] Features with non-zero coefficients are considered to have a significant impact on the dependent variable and are retained as the final relevant factors.

[0060] The cross-validation method selects the optimal regularization parameter λ including:

[0061] The K-fold cross-validation method is used to divide the first training set into K parts, set the form of the Lasso regression model, select a set of candidate values ​​of λ, start from a smaller range, such as exponential growth or decrease, and the selection of λ is usually on a logarithmic scale, such as 0.001, 0.01, 0.1, 1. K-1 parts of the first training set are used as the second training set, and the remaining part of the first training set is used as the second validation set. The Lasso regression model is fitted on the second training set, and the performance of the model is evaluated according to the validation set using the mean square error. The validation set evaluation index of each λ value is recorded, and the average performance index of K cross validations of each λ value is calculated. The λ value that minimizes the MSE is selected as the optimal regularization parameter.

[0062] The step S3 comprises:

[0063] The third characteristic factor is subjected to multivariate regression analysis to obtain the influence of the third characteristic factor on the injection dose of the anesthetic, the influence being expressed in percentage, characteristic factors with influence less than 0 are removed to obtain influencing factors, a pie-shaped trend chart is established according to the influence and the corresponding influencing factors to visualize the influencing factors, and the influencing factors are divided into a training group and a control group according to proportion;

[0064] The influencing factors are divided into a training group and a control group according to proportion, including randomly selecting patients accounting for 20% of the total number of influencing factors of the same type in each influencing factor in turn as the control group, and the remaining patients accounting for 80% of the total number of influencing factors of the same type are defined as the training group.

[0065] Multivariate regression analysis included:

[0066] A stepwise regression analysis method is adopted, and independent variables are gradually added or deleted through the step function. The dependent variable is preliminarily predicted using the Poisson regression method, and possible predictors are gradually introduced or excluded to obtain the optimal predictor. The independent variable combination is obtained according to the optimal predictor, which helps to intuitively understand the influence trend of each factor on the anesthetic dose. A multifactor regression analysis is performed according to the independent variable combination, and several influencing factors are randomly selected from the results of the multifactor regression analysis. The influencing factors include the coefficients of the independent variables. The influencing factors are represented on the horizontal axis using a bar chart, and the anesthetic dose is represented on the vertical axis. Each influencing factor corresponds to a bar, and the bar shows the influence trend of the independent variable on the anesthetic dose.

[0067] 80% of the influencing factors are divided into a training group, and 20% of the influencing factors are divided into a control group.

[0068] The step S4 comprises:

[0069] The ten-fold cross validation method is to change K in the K-fold cross validation method to 10;

[0070] The prediction factor is defined as the second dependent variable, and the training group is defined as the second independent variable. In the regression process of the Lasso regression model, as the penalty coefficient increases, the coefficients of several features are compressed to zero, and the selected prediction factor is a feature whose corresponding coefficient is not zero when a certain value is reached;

[0071] The mathematical regression model established based on the predictive factors and the anesthesia site and area includes:

[0072] Define the predictor as the independent variable code x, define the anesthesia site and area as the independent variable code y, establish a linear regression model based on the independent variable code x and the independent variable code y, and use a nomogram to draw a visual two-dimensional graph of the linear regression model established based on the independent variable code x and the independent variable code y, which is helpful for understanding and analyzing the data.

[0073] The step S5 comprises:

[0074] The regression coefficient of the linear regression model is calculated by the reduction method to obtain a new fitting line, which is plotted on the two-dimensional graph for comparison. The fitting degree of the predicted value and the data is used to analyze the effect, and a fitting threshold is set. If the fitting degree is higher than the fitting threshold, the linear regression model has a good fitting effect and high prediction accuracy. If the fitting degree is lower than the fitting threshold, the encoding method and parameters are adjusted, the linear regression model is re-established and re-evaluated until the fitting degree reaches the fitting threshold to obtain a prediction model.

[0075] Use regression to predict a value given an input, predict the category label of continuous data. Assuming that the input data is stored in the matrix z and the regression coefficients are stored in the vector w, the prediction result for the given data will be given by the product of z and w.

[0076] Finding the vector w involves:

[0077] Find the w that minimizes the error, where the error refers to the difference between the predicted independent variable code y and the true independent variable code y, and use simple accumulation of the errors and find the average value to offset the positive and negative differences.

[0078] The linear regression model evaluation indicators and prediction performance include:

[0079] The area under the ROC curve of the linear regression model was calculated, and the consistency index, accuracy, sensitivity, specificity, positive predictive value, negative predictive value and harmonic mean score index of precision and recall in machine learning of the linear regression model were analyzed. The accuracy of the linear regression model was evaluated by calibration curve and the clinical decision curve analysis was used to evaluate the clinical benefit of the model. Python software was used for data analysis. The chi-square test was used among all groups to perform Lasso regression on the characteristic factors with an error of <0.2 for re-variable selection. When the error was <0.05, the prediction model was obtained.

[0080] When the error is less than 0.05, it is statistically significant, that is, the linear regression model is successful in predicting the anesthetic dose.

[0081] By collecting complex data of patients through big data, recursively analyzing the data based on multiple regression models, screening out useful data, recursively narrowing the range of characteristic factors required for predicting anesthetic doses, making future predictions more accurate, and improving the interpretability of the prediction model, visualizing the prediction data, and increasing the interactivity of the prediction data, the linear regression model can quickly analyze and predict the anesthetic injection dose required by individual patients, thereby reducing the workload of anesthesiologists.

[0082] It should be understood by those skilled in the art that the embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program codes. Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, disk or optical disk. These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0083] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for predicting anesthetic dosage based on big data analysis. It is characterized in that The following steps are involved: Step S1: Acquire patient information of anesthetic injection, wherein the patient information includes biological indicator data of the patient, and corresponding anesthetic injection method, injection dosage and anesthetic type; Step S2: Based on the biological indicator data, a first characteristic factor is screened using a Lasso regression model, and the first characteristic factor is analyzed using a Logistic regression model to obtain a second characteristic factor. Based on the first characteristic factor and the second characteristic factor, other factor characteristics that affect the anesthetic dose are predicted, and the other factor characteristics are combined with the second characteristic factor to obtain a third characteristic factor; Step S3: performing a multi-factor regression analysis on the third characteristic factor to obtain the influence of the third characteristic factor on the injection dose of the anesthetic, and dividing the third characteristic factor corresponding to the influence into a training group and a control group according to the proportion; Step S4: putting the training group back into the Lasso regression model and combining the ten-fold cross validation method to screen the predictive factors, establishing a linear regression model based on the predictive factors and the anesthesia site and area, and visualizing the linear regression model in the form of a nomogram; Step S5: The linear regression model is adjusted by adjusting the training group, and the linear regression model is evaluated by using the control group, and a prediction model is obtained when the evaluation error is less than a preset value.

2. The method for predicting anesthetic dosage based on big data analysis according to claim 1, It is characterized in that The step S1 comprises: The biological indicator data include age, height, BMI, weight, age, ASA classification, anesthesia site and area, operation time, anesthesia time, combined medication, surgical procedure, heart rate, bispectral index value, blood oxygen saturation and mean arterial pressure value intervals at the time of patient admission, pre-induction time, intubation time, operation start time, operation end time and extubation time, and the patients are grouped according to the biological indicator data; The anesthetic injection methods are grouped according to the part of the patient's body involved in the surgery.

3. The method for predicting anesthetic dosage based on big data analysis as claimed in claim 2, It is characterized in that The step S2 comprises: The first characteristic factor represents a factor related to the injection dose of anesthetic drugs, and the second characteristic factor represents a characteristic factor among the first characteristic factors that affects the injection dose of anesthetic drugs. Based on the first characteristic factor and the second characteristic factor, a stepwise regression analysis method is used to predict the characteristics of other factors that affect the anesthetic dose through big data retrieval, and the characteristics of the other factors are combined with the second characteristic factor to obtain a third characteristic factor, and the third characteristic factor represents a factor that affects the injection dose of anesthetic drugs.

4. The method for predicting anesthetic dosage based on big data analysis according to claim 3, It is characterized in that The first characteristic factors related to the anesthetic injection dose in the biological indicator data and genotype data screened by using the Lasso regression model include: Performing Huffman coding of the same length on the biological indicator data, defining the Huffman coding as an independent variable, defining the anesthetic injection dose as a dependent variable, performing median filling on missing values ​​in the independent variable and the dependent variable, performing normalization operation on the independent variable, establishing a Lasso regression model based on the independent variable and the dependent variable, adding a regularization term to the Lasso regression model to achieve feature selection, wherein the coefficient of the regularization term is controlled by a user-defined parameter λ, and dividing the independent variable and the dependent variable into a first training set and a first test set based on individuals and group ratios; Fitting a Lasso regression model on the training set, defining the objective of the Lasso regression model training as minimizing a loss function, wherein the loss function includes an original loss function and a regularization term; A cross-validation method is used to select the optimal regularization parameter λ, λ is introduced into the Lasso regression model, and the Lasso regression model is repeatedly trained. According to the Lasso regression model completed by repeated training, coefficients of all characteristic factors to be obtained are obtained, all coefficients are sorted from large to small according to their absolute values, and characteristic factors whose coefficients are greater than a preset threshold are retained. The Lasso regression model is retrained using the retained characteristic factors, and the coefficients of several features of the retrained Lasso regression model are compressed to zero to obtain a first characteristic factor.

5. The method for predicting anesthetic dosage based on big data analysis according to claim 4, It is characterized in that The cross-validation method selects the optimal regularization parameter λ including: The K-fold cross-validation method is used to divide the first training set into K parts, set the form of the Lasso regression model, select a set of candidate values ​​of λ, use K-1 parts of the first training set as the second training set, and use the remaining part of the first training set as the second validation set. The Lasso regression model is fitted on the second training set, and the performance of the model is evaluated according to the validation set using the mean square error. The validation set evaluation index of each λ value is recorded, the average performance index of K cross validations for each λ value is calculated, and the λ value that minimizes the MSE is selected as the optimal regularization parameter.

6. The method for predicting anesthetic dosage based on big data analysis according to claim 5, It is characterized in that The step S3 comprises: The third characteristic factor is subjected to multivariate regression analysis to obtain the influence of the third characteristic factor on the injection dose of the anesthetic, the influence being expressed in percentage, and the characteristic factors with influence less than 0 are removed to obtain the influencing factors, a pie-shaped trend chart is established according to the influence and the corresponding influencing factors, and the influencing factors are divided into a training group and a control group according to the proportion; The influencing factors are divided into a training group and a control group according to proportion, including randomly selecting patients accounting for 20% of the total number of influencing factors of the same type in each influencing factor in turn as the control group, and the remaining patients accounting for 80% of the total number of influencing factors of the same type are defined as the training group.

7. The method for predicting anesthetic dosage based on big data analysis according to claim 6, It is characterized in that Multivariate regression analysis included: A stepwise regression analysis method is used to gradually add or delete independent variables through the step function, and the dependent variable is preliminarily predicted using the Poisson regression method. Possible predictors are gradually introduced or excluded to obtain the optimal predictor. An independent variable combination is obtained according to the optimal predictor. A multifactor regression analysis is performed according to the independent variable combination. Several influencing factors are randomly selected from the results of the multifactor regression analysis. The influencing factors include the coefficients of the independent variables. The influencing factors are represented on the horizontal axis using a bar chart, and the anesthetic dose is represented on the vertical axis, and each influencing factor corresponds to a bar. 80% of the influencing factors are divided into a training group, and 20% of the influencing factors are divided into a control group.

8. The method for predicting anesthetic dosage based on big data analysis according to claim 7, It is characterized in that The step S4 comprises: The ten-fold cross validation method is to change K in the K-fold cross validation method to 10; The prediction factor is defined as a second dependent variable, the training group is defined as a second independent variable, and in the Lasso regression model regression process, the prediction factor is selected as a feature whose corresponding coefficient is not zero when a certain value is reached; The mathematical regression model established based on the predictive factors and the anesthesia site and area includes: Define the predictor as the independent variable code x, define the anesthesia site and area as the independent variable code y, establish a linear regression model based on the independent variable code x and the independent variable code y, and use a nomogram to draw a visual two-dimensional graph of the linear regression model established based on the independent variable code x and the independent variable code y.

9. The method for predicting anesthetic dosage based on big data analysis according to claim 8, It is characterized in that The step S5 comprises: The regression coefficient of the linear regression model is calculated by the reduction method to obtain a new fitting line, which is plotted on the two-dimensional graph for comparison. The fitting degree of the predicted value and the data is used to analyze the effect, and a fitting threshold is set. If the fitting degree is higher than the fitting threshold, the linear regression model has a good fitting effect and high prediction accuracy. If the fitting degree is lower than the fitting threshold, the encoding method and parameters are adjusted, the linear regression model is re-established and re-evaluated until the fitting degree reaches the fitting threshold to obtain a prediction model.

10. The method for predicting anesthetic dosage based on big data analysis according to claim 9, It is characterized in that The linear regression model evaluation indicators and prediction performance include: The area under the ROC curve of the linear regression model was calculated, and the consistency index, accuracy, sensitivity, specificity, positive predictive value, negative predictive value and harmonic mean score index of precision and recall in machine learning of the linear regression model were analyzed. The accuracy of the linear regression model was evaluated by calibration curve and the clinical decision curve analysis was used to evaluate the clinical benefit of the model. Python software was used for data analysis. The chi-square test was used among all groups to perform Lasso regression on the characteristic factors with an error of <0.2 for re-variable selection. When the error was <0.05, the prediction model was obtained.

Images