Method and device for evaluating postoperative complications of old people based on artificial intelligence, and computer equipment

Abstract

The invention provides an artificial intelligence-based postoperative complication assessment method and device for old people, and computer equipment, and relates to the technical field of medical artificial intelligence. According to the method, within a specific postoperative monitoring period, laboratory inspection data including age, gender, basic weight information, heart rate, blood pressure, oxyhemoglobin saturation vital signs, blood sugar content, leukocyte count and C-reactive protein content of a patient are collected according to specified time intervals and recorded in a postoperative evaluation table, and table data are continuously updated within the period; the data is cleaned, missing values are interpolated by using a mode, abnormal values are corrected by using a box graph, and then standardization processing is performed by using Min-Max normalization. Secondly, calculating correlation between the data and complications by using the standardized data, and setting a Pearson's correlation coefficient threshold to screen high-correlation features to construct a data set; and finally, extracting features from the data set, constructing a prediction model based on regression analysis, and calculating the probability of occurrence of postoperative complications of the patient to realize evaluation of the postoperative complications of the old.

Description

Technical Field

[0001] The present invention relates to the technical field of medical artificial intelligence, and specifically to a method, device, and computer device for evaluating postoperative complications in the elderly based on artificial intelligence. Background Art

[0002] In the medical field, the evaluation of postoperative complications in the elderly has always been a crucial link. With the intensification of population aging, the number of elderly patients undergoing surgical treatment is increasing continuously, and the incidence of postoperative complications has also risen accordingly, making accurate and timely complication evaluation a key factor in improving the prognosis of patients.

[0003] Traditional postoperative complication evaluation mainly relies on the clinical experience of medical staff and regular routine examinations. Medical staff observe the symptoms of patients based on their own experience, such as the wound healing situation, whether there is fever, and whether there is difficulty breathing, and combine the measurement of basic vital signs at certain time intervals after surgery, such as heart rate, blood pressure, body temperature, etc., to roughly judge whether the patient has the risk of complications. However, this method has obvious limitations. On the one hand, human experience judgment is subjective and limited, and there are differences in the experience level and judgment criteria of different medical staff, which may lead to inconsistencies in evaluation results. On the other hand, routine examinations usually can only obtain limited and discrete data points, and it is difficult to comprehensively reflect the complex physiological state changes of patients after surgery.

[0004] Traditional evaluation methods usually only focus on whether these indicators exceed the normal range in isolation, while ignoring their mutual relationship and synergistic effect with other factors during the dynamic change process after surgery; in the face of a large amount of patient data, the traditional manual processing method is inefficient and cannot quickly screen out the data features that are crucial for complication evaluation; it is impossible to detect potential complication risks in time and delay the best treatment opportunity.

[0005] In summary, there is a need for a more scientific and efficient method, device, and computer device for evaluating postoperative complications in the elderly based on artificial intelligence to accurately evaluate postoperative complications in the elderly, reduce the errors and uncertainties brought by subjective judgment, and through analyzing multi-dimensional data of patients, dig out potential risk factors to achieve early warning of complications.

[0006] The above information disclosed in the background art section is only used to strengthen the understanding of the background of the present disclosure, and therefore it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0007] The purpose of the present invention is to provide a method, device, and computer device for evaluating postoperative complications in the elderly based on artificial intelligence to solve the problems raised in the above background art.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] An artificial intelligence-based method for evaluating postoperative complications in the elderly, the specific steps include:

[0010] Step 1: During the monitoring period, collect the patient's basic information, vital signs, and laboratory test data at regular time intervals and record them in the postoperative evaluation form. The basic information includes age, gender, and weight, the vital signs include heart rate, blood pressure, and blood oxygen saturation, and the laboratory test data includes blood glucose content, white blood cell count, and C-reactive protein content;

[0011] Step 2: Continuously update the postoperative evaluation form during the monitoring period, record whether postoperative complications occur and the types of complications for the patient, and after cleaning the data in the evaluation form, perform standardization processing;

[0012] Step 3: Through the standardized data, calculate the correlation between the basic information, vital signs, laboratory test data, and various types of complications, and screen out the highly correlated data as feature data to construct a feature dataset;

[0013] Step 4: Extract feature data from the constructed feature dataset, construct a prediction model for postoperative complications based on regression analysis, and calculate the probability of postoperative complications for each patient.

[0014] Further, start recording the patient's basic information, vital signs, and laboratory test data after surgery. During the 7-day monitoring period after surgery, record the data at intervals of every 4 hours. Among them, when recording gender information, use a digital coding method, assign 1 to "male" and 0 to "female".

[0015] Further, continuously update the postoperative evaluation form during the monitoring period. Start recording the patient's basic information, vital signs, and laboratory test data after surgery. The data is updated every 4 hours, and the monitoring period is 7 days. During the monitoring period, each new data collection will replace the earliest data that exceeds the 7-day monitoring period, ensuring that the monitoring period always remains 7 days and retaining the new data within the period.

[0016] Further, for each patient, during the 7-day monitoring period after surgery, record the occurrence of complications for the i-th patient as Y in , where, if no complications are recorded, record as Y i0 , and assign a value of 0; if complications occur during the 7-day monitoring period after surgery, assign a value of 1 and record the serial number n of the complication type for the patient. At this time, n is a positive integer that increases sequentially starting from 1;

[0017] Clean the data in the postoperative evaluation form (except for basic information). For missing data, use the mode imputation method:

[0018]

[0019] Among them, represents the filled data, and mode(X) represents the data with the highest frequency of occurrence in the recorded data;

[0020] Use the box plot method to correct the recorded data:

[0021] For each column of data X in the postoperative evaluation form a , sort it in ascending order. Among them, a represents the serial number corresponding to each column of data, corresponding to age, gender, weight, heart rate, blood pressure, blood oxygen saturation, blood glucose content, white blood cell count, and C-reactive protein content respectively. Record the value Q1 at the 25% position of the data, the value Q2 at the 50% position of the data, and the value Q3 at the 75% position of the data, and calculate the interquartile range:

[0022] IQR = Q3 - Q1

[0023] In the formula, IQR represents the interquartile range, which reflects the dispersion degree of the middle 50% part of the data. Q1 represents the value at the 50% position of the data, and Q3 represents the value at the 75% position of the data;

[0024] Determine the lower limit of outliers:

[0025] IR1 = Q1 - 1.5 × IQR

[0026] In the formula, IR1 represents the lower limit of outliers;

[0027] Determine the upper limit of outliers:

[0028] IR2 = Q3 + 1.5 × IQR

[0029] In the formula, IR2 represents the upper limit of outliers;

[0030] Determine the data points less than the lower limit or greater than the upper limit in each column of data as outliers and replace them with the value Q2 at the 50% position of the data;

[0031] Then standardize the cleaned data and basic information (except for gender) using the Min-Max normalization method:

[0032]

[0033] In the formula, X anorm represents the normalized data, and X ajDenoted as the j-th value of the data in the a-th column, where j = 1, 2, … 42, X amin Denoted as the minimum value of the data in the a-th column, X amax Denoted as the maximum value of the data in the a-th column.

[0034] Furthermore, the Pearson correlation coefficient is used to calculate the correlation between each feature and postoperative complications:

[0035]

[0036] In the formula, r an Denoted as the Pearson correlation coefficient between the data in the a-th column and the n-th complication, Denoted as the mean value of the data in the a-th column of the i-th patient, Denoted as the average value of the data in the a-th column of all patients, Y' in Denoted as the occurrence of the n-th complication in the i-th patient, Denoted as the average incidence rate of the n-th complication in all patients;

[0037] The mean value of the data in the a-th column of the i-th patient The calculation formula is:

[0038]

[0039] In the formula, X aij Denoted as the j-th value of the data in the a-th column of the i-th patient;

[0040] The average incidence rate of the n-th complication in all patients The calculation formula is:

[0041]

[0042] In the formula, N denotes the total number of patients.

[0043] Furthermore, set the threshold r of the Pearson correlation coefficient thshold = 0.6. When r an ≥ r thshold it is determined that the a-th feature (the data in the a-th column) is highly correlated with the n-th complication, and it is recorded as X nsub = {X a | r an ≥ r thsold}.

[0044] Furthermore, for the different complications of each patient, a linear regression model is established:

[0045] Y in = β0 + β1X i1 + β2X i2 + … + βk X ik +∈ in

[0046] In the formula, Y in represents the occurrence of the nth complication in the ith patient, X i1 , X i2 , …, X ik represent the features related to complication n, where k represents the number of related features, β1, β2, …, β k represent the regression coefficients, ∈ in is the error term, representing the difference between the predicted and actual values of the nth complication in the ith patient;

[0047] Construct an N*1 matrix Y:

[0048] Y = [Y 1n , Y 2n , …, Y Nn T

[0049] In the formula, Y represents an N*1 matrix;

[0050] Construct a design matrix X:

[0051]

[0052] In the formula, X represents an N*K design matrix;

[0053] Construct a regression coefficient matrix β:

[0054] β = [β0, β1, β2, …, β k T

[0055] In the formula, β represents a regression coefficient matrix;

[0056] Construct an error matrix:

[0057] ε = [∈ 1n , ∈ 2n , ∈ 3n , …, ∈ Nn T

[0058] Minimize the residual sum of squares:

[0059]

[0060] In the formula, RSS represents the minimized residual sum of squares, represents the predicted value of the complication;

[0061] By taking the derivative and setting it to 0, the normal equations X​​​T Xβ = X T Y, and then calculate the estimated value of the regression coefficient:

[0062]

[0063] Finally, the prediction model for complications is:

[0064]

[0065] In the formula, denotes the prediction model for postoperative complications. The formula for calculating the probability of postoperative complications for each patient based on the prediction model for postoperative complications is:

[0066]

[0067] In the formula, P represents the probability of postoperative complications.

[0068] Furthermore, (Y - Xβ) T (Y - Xβ) = Y T Y - Y T Xβ - β T X T Y + β T X T Xβ

[0069] Taking the derivative of β gives:

[0070]

[0071] Let That is, -2X T Y + 2X T Xβ = 0, and we get

[0072]

[0073] The present invention also provides an artificial intelligence-based postoperative complication assessment device for the elderly. The device is used to execute the above-mentioned artificial intelligence-based postoperative complication assessment method for the elderly, and includes:

[0074] A data collection module, which is used to collect the basic information, vital signs, and laboratory test data of patients at regular time intervals during the monitoring period and record them in the postoperative assessment form. The basic information includes age, gender, and weight, the vital signs include heart rate, blood pressure, and blood oxygen saturation, and the laboratory test data includes blood glucose content, white blood cell count, and C-reactive protein content;

[0075] A data update processing module, which is used to continuously update the postoperative evaluation form during the monitoring period, record whether postoperative complications occur in patients and the types of complications, and perform standardized processing after cleaning the data in the evaluation form;

[0076] A feature analysis module, which is used to calculate the correlation between basic information, vital signs, laboratory test data and various types of complications through the standardized data, screen out the highly relevant data as feature data, and construct a feature dataset;

[0077] A probability calculation module, which is used to extract feature data from the constructed feature dataset, construct a prediction model for postoperative complications based on regression analysis, and calculate the probability of postoperative complications for each patient.

[0078] To achieve the above object, the present invention also provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the computer program to implement the steps of the above-mentioned artificial intelligence-based postoperative complication assessment for the elderly.

[0079] Compared with the prior art, the beneficial effects of the present invention are:

[0080] In terms of data collection, it comprehensively covers various aspects of information. Compared with the traditional scattered and incomplete data collection methods, it can more completely reflect the postoperative state of patients, accurately find out the features highly related to complications, and overcome the problem that traditional empirical judgments cannot accurately identify key factors; the finally constructed prediction model can accurately calculate the probability of a certain complication, providing clear and quantitative evaluation results for medical staff, changing the previous fuzzy judgment mode, and helping to formulate targeted prevention and treatment measures in advance. Description of the Drawings

[0081] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0082] Figure 2 It is a schematic diagram of the overall system module of the present invention. Detailed Embodiments

[0083] To make the purpose, technical solution and advantages of the present invention clearer, the following further details the present invention with reference to specific embodiments.

[0084] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the ordinary meanings understood by those with ordinary skills in the field to which the present invention pertains. The "first", "second" and similar terms used in the present invention do not denote any order, quantity or importance, but are only used to distinguish different components. Words such as "comprising" or "including" mean that the elements or objects appearing before this word cover the elements or objects listed after this word and their equivalents, without excluding other elements or objects. Words such as "connected" or "linked" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Upper", "lower", "left", "right", etc. are only used to represent relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0085] Embodiment:

[0086] Please refer to Figure 1 , the present invention provides a technical solution:

[0087] An artificial intelligence-based method for evaluating postoperative complications in the elderly, the specific steps include:

[0088] Step 1: During the monitoring period, collect the patient's basic information, vital signs and laboratory test data at regular time intervals and record them in the postoperative evaluation form. The basic information includes age, gender and weight, the vital signs include heart rate, blood pressure and blood oxygen saturation, and the laboratory test data includes blood glucose content, white blood cell count and C-reactive protein content;

[0089] After the patient's operation, start a 7-day monitoring period. This is because most complications are in the high-incidence stage within 7 days after the operation. For example, common complications such as pulmonary infection and deep vein thrombosis usually show symptoms within one week after the operation. Monitoring the data during this period can timely capture the changes related to these complications and provide a basis for early intervention and treatment. During the monitoring period, collect the patient's basic information, vital signs and laboratory test data at 4-hour intervals in sequence. Start recording the patient's basic information, vital signs and laboratory test data after the operation. During the 7-day monitoring period after the operation, record data at 4-hour intervals. The 4-hour time interval can relatively timely capture the dynamic changes of these indicators and timely detect abnormal fluctuations;

[0090] Among the basic information, age is an important risk factor for various postoperative complications. Due to the decline of physical functions in the elderly, they are more likely to have pulmonary infections and cardiovascular complications such as arrhythmia and heart failure after the operation;

[0091] There are differences in the risk of occurrence of certain complications among genders. The risk of deep vein thrombosis after surgery is relatively higher in women than in men, which is related to the physiological characteristics of women, such as hormone levels, vascular structure and other factors;

[0092] Both being overweight or underweight may affect the occurrence of postoperative complications. Obese patients have a higher risk of postoperative incision infection and pulmonary complications because obesity leads to abdominal fat accumulation, which affects respiratory function and increases the risk of pulmonary infection. While patients with too low body weight often have malnutrition problems, their body resistance decreases, and they are prone to complications such as infection and poor incision healing after surgery;

[0093] Among vital signs, the change in heart rate can reflect the cardiovascular function of the patient and the stress state of the body. A too fast heart rate after surgery may indicate problems such as bleeding, infection, pain, etc., or it may also be an early manifestation of arrhythmia;

[0094] Abnormal blood pressure is a common problem after surgery. Too high blood pressure may increase the risks of cerebrovascular accident, cardiovascular rupture, etc.; too low blood pressure may indicate conditions such as insufficient blood volume and shock;

[0095] Blood oxygen saturation reflects the oxygenation state of the body. A decrease in blood oxygen saturation after surgery may be an early manifestation of pulmonary complications such as atelectasis, pulmonary infection, acute respiratory distress syndrome, etc., or it may also be related to oxygen delivery disorders caused by cardiovascular insufficiency;

[0096] Among laboratory test data, abnormal blood glucose levels are closely related to complications such as postoperative infection and poor incision healing. A high blood glucose state is conducive to the growth and reproduction of bacteria, which will increase the risk of postoperative infection, especially urinary tract infection and pulmonary infection. At the same time, it will also lead to complications such as delayed incision healing and dehiscence;

[0097] White blood cell count is an important indicator reflecting the body's inflammatory response. An increase in white blood cell count after surgery often indicates the presence of infection, such as incision infection, pulmonary infection, urinary tract infection, etc.; if the white blood cell count is too low, it may indicate that the body's immune function is low and various infectious complications are likely to occur;

[0098] The content of C-reactive protein is an acute-phase reaction protein, which will increase rapidly in cases of inflammation, tissue injury, etc., and is related to complications such as infection and thrombosis;

[0099] Among them, when recording gender information, a digital coding method is adopted, and "male" is assigned a value of 1, and "female" is assigned a value of 0. This is because when constructing an artificial intelligence or statistical-based postoperative complication assessment model, the model usually requires the input data to be in numerical form. Coding gender as 0 and 1 can be used as a feature variable of the model to directly participate in the training and operation of the model.

[0100] Step 2: Continuously update the postoperative evaluation form during the monitoring period, record whether postoperative complications occur in the patient and the types of complications. After cleaning the data in the evaluation form, perform standardization processing;

[0101] Throughout the monitoring process, ensure that only the data of the most recent 7 days is always retained. Whenever new data is collected, the system will automatically check the time. For example, monitoring starts at 0:00 on January 1st, and the latest data is collected at 0:00 on January 8th. At this time, the system will identify that the data at 0:00 on January 1st has exceeded the 7-day monitoring period, so it will delete it from the postoperative evaluation form and supplement the new data collected at 0:00 on January 8th. In this way, no matter when the postoperative evaluation form is viewed, the data in it is from the most recent 7 days, which can timely reflect the patient's current physical condition, provide the latest and most effective data support for doctors to subsequently analyze the condition, evaluate the patient's recovery situation, and predict postoperative complications. At the same time, for each patient, during the 7-day monitoring period after surgery, record the occurrence of complications of the i-th patient as Y in , where if any complication occurs, n is 0 and it is recorded as Y i0 , and it is assigned a value of 0; if a complication occurs during the 7-day monitoring period after surgery, it is assigned a value of 1, and record the serial number n of the type of complication of this patient. At this time, n is a positive integer that increases sequentially starting from 1; by representing the occurrence of complications with 0 and 1, it provides a unified and clear quantification standard for the data, and can also quickly identify and classify the occurrence of complications for each patient.

[0102] Clean the data in the postoperative evaluation form (except for basic information). For missing data, use the method of mode imputation:

[0103]

[0104] Among them, represents the filled data, mode(X) represents the data with the highest frequency of occurrence in the recorded data. Mode imputation is a simple method for processing missing data values, applicable to categorical data, which can maintain the dominant position of this category in the data and avoid introducing unreasonable data;

[0105] Use the box plot method to correct the recorded data:

[0106] For each column of data X in the postoperative evaluation form a , arrange them in ascending order. Among them, a represents the serial number corresponding to each column of data, corresponding to age, gender, weight, heart rate, blood pressure, blood oxygen saturation, blood glucose content, white blood cell count, and C-reactive protein content respectively. Record the value Q1 at the 25% position of the data, the value Q2 at the 50% position of the data, and the value Q3 at the 75% position of the data, and calculate the interquartile range:

[0107] IQR = Q3 - Q1

[0108] Wherein, IQR represents the interquartile range, which reflects the dispersion degree of the middle 50% part of the data. Q1 represents the value at the 50% position of the data, and Q3 represents the value at the 75% position of the data;

[0109] Determine the lower limit of outliers:

[0110] IR1 = Q1 - 1.5 × IQR

[0111] Wherein, IR1 represents the lower limit of outliers;

[0112] Determine the upper limit of outliers:

[0113] IR2 = Q3 + 1.5 × IQR

[0114] Wherein, IR2 represents the upper limit of outliers;

[0115] Determine the data points in each column of data that are less than the lower limit or greater than the upper limit as outliers, and replace them with the value Q2 at the 50% position of the data. The box plot method is an efficient outlier detection method that can clearly find the outliers in the data. These outliers may be caused by measurement errors, data entry errors or special circumstances. Identifying them helps to improve the quality of the data;

[0116] If the data is not standardized, the differences between the data may cause some data to have too much influence on model training, resulting in bias in the model. Therefore, standardize the cleaned data and basic information (except gender) using the Min - Max normalization method:

[0117]

[0118] Wherein, X anorm represents the normalized data, X aj represents the j - th value of the a - th column of data, where j = 1, 2,... 42, X amin represents the minimum value of the a - th column of data, X amax represents the maximum value of the a - th column of data. Min - Max normalization scales the data to the interval [0, 1], unifies it to the same range, retains the relative size relationship of the original data, and eliminates the scale differences between different data.

[0119] Step 3: Through the standardized data, calculate the correlations between the basic information, vital signs, laboratory examination data and various types of complications, screen out the highly correlated data as feature data, and construct a feature dataset;

[0120] Medical data were collected, and a linear relationship was found between many clinical features and the occurrence of complications, such as pulmonary infection, cardiovascular complications, deep vein thrombosis, incision infection, vascular infection, etc. Pearson's correlation coefficient can be used to intuitively reflect this relationship. Pearson's correlation coefficient is used to measure the linear correlation between two variables, so as to calculate the correlation between each feature (age, blood pressure, body temperature, etc.) and postoperative complications:

[0121]

[0122] In the formula, r an represents the Pearson correlation coefficient between the data in the a-th column and the n-th complication. represents the mean value of the data in the a-th column of the i-th patient, representing the general level of the patient in this feature, eliminating the influence of factors such as short-term fluctuations and measurement errors of the patient's data, and more accurately reflecting the relationship between the overall state of this feature of the patient and postoperative complications. represents the average value of the data in the a-th column of all patients, reflecting the average level of the data in the entire patient group, and serving as an overall reference benchmark Y'. in represents the occurrence of the n-th complication of the i-th patient, with a value of 1 or 0, clearly defining the state of each patient in a specific complication. represents the average incidence rate of the n-th complication of all patients, measuring the prevalence of this complication in the entire study population;

[0123] Among them, the mean value of the data in the a-th column of the i-th patient The calculation formula is:

[0124]

[0125] In the formula, X aij represents the j-th value of the data in the a-th column of the i-th patient. Among them, because the monitoring period is 7 days and the monitoring is carried out every 4 hours, the total number of monitoring times is 42;

[0126] The average incidence rate of the n-th complication of all patients The calculation formula is:

[0127]

[0128] In the formula, N represents the total number of patients

[0129] Pearson's correlation coefficient reflects the strength of the linear relationship. Set the threshold r of Pearson's correlation coefficient thshold= 0.6, which is also a relatively common choice. According to general experience in statistics, it is regarded as moderately to highly correlated. Even if a small number of patient individuals account for a relatively small proportion in the overall sample, the correlation between their characteristics and complications can be reflected by the Pearson correlation coefficient. As long as the correlation coefficient reaches or exceeds the threshold, it will be identified and recorded, so that it will not be missed. When r an ≥ r thshold , it is determined that the a-th characteristic (the data in the a-th column) is highly correlated with the n-th complication, avoiding the inclusion of some characteristics with weak correlation with the complication, and recording it as X nsub ={X a |r an ≥ r thshold}, which is convenient for subsequent construction of the model to call the feature data;

[0130] Step 4: Extract the feature data from the constructed feature dataset, construct a prediction model for postoperative complications based on regression analysis, and calculate the probability of postoperative complications for each patient;

[0131] Select the features highly correlated with the complications, and extract the columns where these features are located from the feature dataset. In medical research, many postoperative complications have a linear relationship with certain features. Therefore, for different complications of each patient, a linear regression model is established:

[0132] Y in = β0 + β1X i1 + β2X i2 +…+ β k X ik + ∈ in

[0133] In the formula, Y in represents the occurrence of the n-th complication of the i-th patient, X i1 , X i2 ,…, X ik represent the features related to complication n. Among them, k represents the number of related features, that is, the number of features highly correlated with the n-th complication, β1, β2,…, β k represent the regression coefficients, ∈ in is the error term, representing the difference between the prediction and the actual value of the n-th complication of the i-th patient. This linear regression model makes personalized predictions for the specific complications of each patient. The occurrence of postoperative complications for each patient is affected not only by general factors (such as age, gender, etc.) but also by the specific health status of the patient. By establishing a regression model for each patient separately, the risk of their postoperative complications can be evaluated more accurately;

[0134] Construct an N*1 matrix Y:

[0135] Y = [Y1n , Y 2n , …, Y Nn T

[0136] In the formula, Y represents an N*1 matrix. The occurrence situations of this complication for all N patients are arranged in sequence to form a matrix, providing actual values for the calculation of residuals;

[0137] Construct the design matrix X:

[0138]

[0139] In the formula, X represents an N*K design matrix. Integrating multiple feature information in one matrix and multiplying it by the subsequent regression coefficient matrix β can reflect the influence of each feature on the occurrence situation of the complication, that is, obtaining the predicted value;

[0140] Construct the regression coefficient matrix β:

[0141] β = [β0, β1, β2, …, β k T

[0142] In the formula, β represents the regression coefficient matrix, corresponding to the intercept term and the coefficients of each feature respectively. Each of its elements represents the influence degree of the corresponding feature on the occurrence situation of the complication, so as to determine the estimated value of the specific regression coefficient through β and determine the linear regression model of a certain complication;

[0143] Construct the error matrix:

[0144] ε = [∈ 1n , ∈ 2n , ∈ 3n , …, ∈ Nn T

[0145] In the formula, ε represents the prediction error of the nth complication of the ith patient, that is It is used to measure the difference between the model predicted value and the actual observed value, so as to calculate the residual sum of squares RSS to obtain the predicted value of the complication

[0146] Minimize the residual sum of squares:

[0147]

[0148] In the formula, RSS represents the minimized residual sum of squares, represents the predicted value of the complication;

[0149] By taking the derivative and setting the derivative to 0, we get

[0150] ​​​(Y - Xβ) T (Y - Xβ) = Y T Y - Y T Xβ - β T X T Y + β T X T Xβ

[0151] Derivative with respect to β gives:

[0152]

[0153] Let That is, -2X T Y + 2X T Xβ = 0, we get

[0154]

[0155] Finally, the prediction model for complications is:

[0156]

[0157] In the formula, Denoted as the prediction model for postoperative complications. Since the reasonable value range for the probability of postoperative complications is [0, 1], a logistic function is used based on the prediction results of postoperative complications, which can better capture the probability of complications. The calculation formula for constructing the probability of postoperative complications for each patient is:

[0158]

[0159] In the formula, P represents the probability of postoperative complications. When tends to positive infinity, P approaches 1. When tends to negative infinity, P approaches 0.

[0160] For doctors and medical teams, it provides an intuitive and easy - to - understand indicator for the probability and type of postoperative complications. For example, if it is found that certain features are closely related to the probability of complications, such as the values of blood glucose content, white blood cell count, and C - reactive protein content, and the calculated probability of infectious complications is relatively large, doctors can adjust the treatment plan accordingly, control the blood glucose content more strictly, adjust the insulin dosage or the diet plan, and thus take corresponding preventive and treatment measures.

[0161] Please refer to Figure 2 , the present invention also provides an artificial - intelligence - based device for evaluating postoperative complications in the elderly. The system is used to execute the above - mentioned artificial - intelligence - based method for evaluating postoperative complications in the elderly, including:

[0162] A data collection module, which is used to collect the basic information, vital signs and laboratory test data of patients at regular time intervals during the monitoring period and record them in the postoperative evaluation form. The basic information includes age, gender and weight, the vital signs include heart rate, blood pressure and blood oxygen saturation, and the laboratory test data includes blood glucose content, white blood cell count and C-reactive protein content;

[0163] A data update and processing module, which is used to continuously update the postoperative evaluation form during the monitoring period, record whether postoperative complications occur and the types of complications of the patient, and perform standardization processing after cleaning the data in the evaluation form;

[0164] A feature analysis module, which is used to calculate the correlation between the basic information, vital signs, laboratory test data and various types of complications through the standardized data, screen out the highly correlated data as feature data, and construct a feature dataset;

[0165] A probability calculation module, which is used to extract feature data from the constructed feature dataset, construct a prediction model for postoperative complications based on regression analysis, and calculate the probability of postoperative complications for each patient.

[0166] In an embodiment, a computer device is further provided, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned method for evaluating postoperative complications of the elderly based on artificial intelligence are implemented.

[0167] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the real situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0168] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0169] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units. They can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0170] As described above, it is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application.

Claims

1. An artificial intelligence-based method for assessing postoperative complications in the elderly, characterized in that: The specific steps include: Step 1: During the monitoring period, the patient's basic information, vital signs and laboratory test data are collected at certain time intervals and recorded in the postoperative evaluation form. The basic information includes age, gender and weight, the vital signs include heart rate, blood pressure and blood oxygen saturation, and the laboratory test data include blood sugar content, white blood cell count and C-reactive protein content; Step 2: Continuously update the postoperative evaluation form during the monitoring period, record whether the patient has postoperative complications and the type of complications, clean the data in the evaluation form, and perform standardization; Step 3: Calculate the correlation between basic information, vital signs, laboratory test data and various types of complications through standardized data, select highly relevant data as feature data, and construct a feature data set; Step 4: Extract feature data from the constructed feature data set, build a prediction model for postoperative complications based on regression analysis, and calculate the probability of postoperative complications for each patient.

2. The artificial intelligence-based method for evaluating postoperative complications in the elderly according to claim 1, characterized in that: The method for collecting basic information, vital signs and laboratory test data of patients at certain time intervals is as follows: The patient's basic information, vital signs, and laboratory test data were recorded after surgery. During the 7-day monitoring period after surgery, data were recorded at intervals of every 4 hours. When recording gender information, digital coding was used, with "male" assigned a value of 1 and "female" assigned a value of 0.

3. The artificial intelligence-based method for evaluating postoperative complications in the elderly according to claim 1, characterized in that: Continuous updating of the postoperative assessment form during the monitoring period includes the following steps: The postoperative evaluation form is continuously updated during the monitoring period. The patient's basic information, vital signs and laboratory test data are recorded after the operation. The data is updated every 4 hours. The monitoring period is 7 days. During the monitoring period, each new data collection will replace the earliest data that exceeds the 7-day monitoring period, ensuring that the monitoring period is always maintained at 7 days and the new data is retained within the period.

4. The artificial intelligence-based method for evaluating postoperative complications in the elderly according to claim 1, characterized in that: After cleaning the data in the evaluation form, standardization includes the following steps: For each patient, within the 7-day postoperative monitoring period, the complication of the i-th patient is recorded as Y in If no complications were recorded, it was recorded as Y i0 , and assign a value of 0; if a complication occurs within the 7-day monitoring period after surgery, assign a value of 1 and record the patient's complication type number n, where n is a positive integer that increases from 1; Clean the data in the postoperative evaluation form and use the mode interpolation method for missing data: in, It represents the data after filling, and modeX represents the data with the highest frequency in the recorded data; The recorded data were corrected using the box plot method: For each column of data X in the postoperative evaluation table a , arranged in ascending order, where a represents the serial number corresponding to each column of data, corresponding to age, gender, weight, heart rate, blood pressure, blood oxygen saturation, blood sugar content, white blood cell count and C-reactive protein content, record the value Q1 at the 25% position of the data, the value Q2 at the 50% position of the data, and the value Q3 at the 75% position of the data, and calculate the interquartile range: IQR=Q3-Q1 In the formula, IQR represents the interquartile range, which reflects the degree of dispersion of the middle 50% of the data, Q1 represents the value at the 50% position of the data, and Q3 represents the value at the 75% position of the data; Determine the lower outlier limit: IR1=Q1-1.5×IQR In the formula, IR1 represents the lower limit of outliers; Determine the upper outlier limit: IR2=Q3+1.5×IQR In the formula, IR2 represents the upper limit of outliers; The data points in each column of data that are less than the lower limit or greater than the upper limit are determined as outliers and replaced with the value Q2 at the 50% position of the data; Then the cleaned data and basic information are standardized (except gender) using the Min-Max normalization method: Where, X anorm Represented as normalized data, X aj It is represented as the jth value of the data in the ath column, where j = 1, 2, ... 42, X amin Represents the minimum value of the data in column a, X amax Represents the maximum value of the data in column a.

5. The artificial intelligence-based method for assessing postoperative complications in the elderly according to claim 4, characterized in that: Calculation of the correlation between basic information, vital signs, laboratory test data and various types of complications includes the following steps: The Pearson correlation coefficient was used to calculate the correlation between each characteristic and postoperative complications: In the formula, r an It is expressed as the Pearson correlation coefficient between the data in column a and the nth complication, It is represented as the mean of the data in column a of the ith patient. It is expressed as the average value of the data in column a of all patients, Y' in It represents the occurrence of the nth complication of the i-th patient. It is expressed as the average incidence of the nth complication among all patients; The mean of the data in column a of the ith patient The calculation formula is: Where, X aij It is represented as the jth value of the ath column data of the i-th patient; The average incidence of the nth complication among all patients The calculation formula is: In the formula, N represents the total number of patients.

6. The artificial intelligence-based method for evaluating postoperative complications in the elderly according to claim 1, characterized in that: Screening out highly relevant features includes the following steps: Set the threshold r of the Pearson correlation coefficient thshold =0.6, when r an ≥r thshold When the ath characteristic (the ath column data) is determined to be highly correlated with the nth complication, it is recorded as X nsub ={X a |r an ≥r thshold }.

7. The artificial intelligence-based method for evaluating postoperative complications in the elderly according to claim 5, characterized in that: The prediction model of postoperative complications was constructed based on regression analysis. The probability of postoperative complications for each patient was calculated, which included the following steps: According to the different complications of each patient, a linear regression model is established: Y in =β0+β1X i1 +β2X i2 +…+b k X ik +∈ in Where Y in It represents the occurrence of the nth complication of the ith patient, X i1 , X i2 , …, X ik is represented by the features related to complication n, where k represents the number of related features, β1, β2, …, β k Expressed as regression coefficient, ∈ in is the error term, expressed as the difference between the predicted and actual value of the nth complication of the ith patient; Construct an N*1 matrix Y: And=[And 1n ,AND 2n ,…,AND Nn ] T In the formula, Y is represented by an N*1 matrix; Construct the design matrix X: Where X represents the design matrix of N*K; Construct the regression coefficient matrix β: β＝[β0，β1，β2，…，β k ] T In the formula, β represents the regression coefficient matrix; Construct the error matrix: ε=[∈ 1n ,∈ 2n ,∈ 3n ,…,∈ Nn ] T Minimize the residual sum of squares: In the formula, RSS is expressed as minimizing the residual sum of squares, Expressed as the predicted value of complications; By taking the derivative and setting it to 0, we get the normal equation system X T Xβ=X T Y, and then calculate the estimated value of the regression coefficient: Finally, the prediction model for complications was: In the formula, It is expressed as a prediction model for postoperative complications. The calculation formula for calculating the probability of postoperative complications for each patient based on the prediction model for postoperative complications is: Where P represents the probability of postoperative complications.

8. The artificial intelligence-based method for evaluating postoperative complications in the elderly according to claim 7, characterized in that: By taking the derivative and setting it to 0, we get the normal equation system X T Xβ=X T The calculation process of Y is to expand RSS to get: Y-Xβ T Y-Xβ=Y T YY T Xβ-β T X T Y+β T X T Xβ Taking the derivative of β, we get: make That is -2X T Y+2X T Xβ=0, we get 9. An artificial intelligence-based device for assessing postoperative complications in the elderly, characterized by: The device is used to implement the artificial intelligence-based method for evaluating postoperative complications in the elderly as described in any one of claims 1 to 8: A data collection module is used to collect the patient's basic information, vital signs and laboratory test data at certain time intervals during the monitoring period and record them in a postoperative evaluation form. The basic information includes age, gender and weight, the vital signs include heart rate, blood pressure and blood oxygen saturation, and the laboratory test data include blood sugar content, white blood cell count and C-reactive protein content; The data update processing module is used to continuously update the postoperative evaluation form within the monitoring period, record whether the patient has postoperative complications and the type of complications, and perform standardization after cleaning the data in the evaluation form; The feature analysis module is used to calculate the correlation between basic information, vital signs, laboratory test data and various types of complications through standardized data, screen out highly relevant data as feature data, and construct a feature data set; The probability calculation module is used to extract feature data from the constructed feature data set, build a prediction model for postoperative complications based on regression analysis, and calculate the probability of postoperative complications for each patient.

10. A computer device, characterized in that: The computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the artificial intelligence-based postoperative complication assessment method for the elderly as described in any one of claims 1-8.

Images