Method and system for predicting delirium of postoperative patient through artificial intelligence

Abstract

The invention relates to the technical field of medical health information, in particular to a method and system for predicting delirium of a postoperative patient through artificial intelligence, and the method comprises the following steps: collecting electrocardiogram and electroencephalogram data of the postoperative patient, carrying out formatting processing on the data by adopting a time sequence analysis method, carrying out entropy calculation on the processed data, and obtaining the delirium of the postoperative patient. And obtaining a sample entropy data set by calculating sample entropy values in the differentiated time windows. According to the method, the specific frequency mode associated with the delirium state can be deeply analyzed and recognized through acquisition of the sample entropy and application of the frequency domain transformation technology, the data points deviating from the normal mode can be accurately marked by comparing the frequency change with the preset standard, accurate risk assessment and early warning are carried out, and the risk assessment accuracy is improved. Through application of real-time monitoring and frequency domain analysis results, prediction accuracy of postoperative delirium occurrence probability and time of hands is enhanced, and the method is used for optimizing patient nursing processes, reducing complications and improving patient recovery effects and life quality.

Description

Technical Field

[0001] The present invention relates to the field of medical health information technology, and in particular to a method and system for predicting postoperative patient delirium using artificial intelligence. Background Art

[0002] The field of healthcare information technology uses computer science and information technology to improve healthcare services, including data collection, management, analysis, and application in preventive patient care. Specific applications can be electronic health records, medical image processing, health information exchange, and real-time patient monitoring systems. The core purpose of this technology field is to improve the quality and accessibility of medical services through efficient information management while ensuring the security and privacy of patient information.

[0003] Among them, the method of using artificial intelligence to predict postoperative delirium in patients refers to the use of artificial intelligence technology to predict the delirium state that patients will experience after surgery, which is an acute brain dysfunction that is common in elderly patients after surgery. This method predicts the risk of postoperative delirium by analyzing the patient's medical condition, vital signs, drug use and other data. The purpose of this technology is to identify high-risk patients in advance, implement preventive measures, optimize the nursing process, reduce complications, and improve patients' recovery and quality of life.

[0004] Existing technologies mainly rely on static electronic health records and basic vital signs monitoring when dealing with the prediction of postoperative delirium, and rarely involve in-depth EEG analysis or the comprehensive application of real-time physiological data. This approach is limited to the type of data collected and the depth of analysis, resulting in the inability to effectively identify specific changes in brain function, affecting the accuracy and timeliness of the prediction. For example, when predicting the risk of delirium, traditional technology cannot accurately distinguish between patients who will develop into high-risk states, which makes it impossible to implement preventive measures in a targeted manner, increasing the patient's recovery cost and recovery time. Due to its reliance on basic data analysis, existing technologies have obvious deficiencies in identifying early signs of delirium, which clinically leads to delayed surgery and increases the risk of complications for patients. Summary of the invention

[0005] In order to solve the existing problems of EEG analysis or comprehensive application of real-time physiological data, the existing technology involves less and is limited to the type of data collected and the depth of analysis, which makes it impossible to effectively identify specific changes in brain function, affecting the accuracy and timeliness of prediction. For example, when predicting the risk of delirium, traditional technology cannot accurately distinguish patients who will develop into a high-risk state, which makes it impossible to implement preventive measures in a targeted manner, increasing the patient's recovery cost and recovery time. Due to its reliance on basic data analysis, existing technologies have obvious deficiencies in identifying early signs of delirium, which leads to delayed surgery in clinical practice and increases the risk of complications for patients.

[0006] Regarding the technical problem, an embodiment of the present invention provides a method and system for artificial intelligence to predict postoperative delirium in patients. The technical solution is as follows: On the one hand, a method for artificial intelligence to predict postoperative delirium in patients is provided, and the method includes: S1: Collect electrocardiogram and electroencephalogram data of postoperative patients, use time series analysis method to format the data, calculate the entropy value of the processed data, and obtain a sample entropy data set by calculating the sample entropy value within a differential time window; S2: Use the sample entropy data set, apply frequency domain transformation technology to convert time series data into frequency domain data, verify key frequency components through frequency domain analysis, and identify frequency patterns associated with the delirium state to obtain a frequency domain analysis result; S3: According to the frequency domain analysis result, compare with the normal frequency pattern, identify the frequency changes deviating from the normal pattern, and compare the sample entropy with a preset standard to mark the data points deviating from the threshold to obtain an abnormal entropy value mark; S4: Use the abnormal entropy value mark as the input of the model, utilize machine learning of artificial intelligence to verify the generalization ability of the model, predict the probability and time of postoperative delirium occurrence in patients, and construct a delirium prediction model; S5: Use the delirium prediction model to evaluate the state of postoperative patients, analyze the risk level of postoperative delirium in patients by real-time monitoring of physiological data, and obtain a delirium risk assessment result.

[0007] As a further solution of the present invention, the sample entropy data set includes entropy values of multiple time windows, time tags, and the change rate of entropy values. The frequency domain analysis result includes verification data of key frequency components, frequency patterns associated with the delirium state, the intensity and duration of frequency patterns. The abnormal entropy value mark includes the entropy value of the data point, the degree of deviation, and the corresponding timestamp. The result of the delirium prediction model includes the structure of the model, the training effect, and the performance data of the model on an independent test set. The delirium risk assessment result includes the delirium risk level of postoperative patients, the weights of risk factors, and the distribution of risk levels.

[0008] As a further solution of the present invention, the steps of collecting electrocardiogram and electroencephalogram data of postoperative patients, using time series analysis method to format the data, calculating the entropy value of the processed data, and obtaining a sample entropy data set by calculating the sample entropy value within a differential time window are specifically as follows: S101: Collect electrocardiogram and electroencephalogram data of postoperative patients, extract data records including timestamps and physiological parameters, organize the data format, and match the requirements of time series analysis to obtain a time series data set; S102: Use the time series data set to perform noise removal processing on the data, remove abnormal points, complete missing time points, and normalize the data to avoid the influence of dimensions, so as to obtain a formatted data set; S103: Based on the formatted data set, select a fixed-length sliding window and calculate the sample entropy of each window segment by segment, including selecting the window size and step length, and calculating the entropy value for each window one by one to obtain a sample entropy data set.

[0009] As a further solution of the present invention, the steps of using the sample entropy data set, applying frequency domain transformation technology to convert time series data into frequency domain data, verifying key frequency components through frequency domain analysis, and identifying frequency patterns associated with the delirium state to obtain the frequency domain analysis result are specifically as follows: S201: Use the sample entropy data set and adopt the fast Fourier transform algorithm to convert time series data into frequency domain data, verify that each data point is converted, and obtain a frequency domain data set; S202: According to the frequency domain data set, detect frequency components, identify key frequencies through the analysis of the amplitude and phase of the frequency components, and calculate the amplitude of the key frequencies to obtain the key frequency analysis result; S203: Through the key frequency analysis result, combined with the physiological parameters of the postoperative patient, identify the correlation between the target frequency pattern and the delirium state to obtain the frequency domain analysis result.

[0010] As a further solution of the present invention, the formula for calculating the amplitude of the key frequency is: Among them, is the amplitude value of the key frequency, represents the real part of the frequency of, represents the imaginary part of the frequency of, represents the total number of sample points, represents the compensation coefficient.

[0011] As a further solution of the present invention, according to the frequency domain analysis result, compare with the normal frequency pattern, identify the frequency changes deviating from the normal pattern, and compare the sample entropy with the preset standard to mark the data points deviating from the threshold to obtain the steps of marking abnormal entropy values are specifically as follows: S301: Use the frequency domain analysis result to compare the frequency pattern of the differential sample points with the frequency pattern in the normal state, identify the frequency changes different from the normal pattern, and obtain an abnormal frequency change data set; S302: Based on the abnormal frequency change data set, compare the sample entropy value of each sample point with the preset normal entropy value range, and use the threshold to determine and mark the sample points exceeding the normal range to obtain abnormal sample records; S303: Using the abnormal sample records, perform iterative statistical analysis and visual display to identify and analyze the changes in the physiological state of postoperative patients, and obtain abnormal entropy value markers.

[0012] As a further solution of the present invention, taking the abnormal entropy value marker as the input of the model, using machine learning of artificial intelligence to verify the generalization ability of the model, predict the probability and time of delirium occurrence in postoperative patients, and the steps of constructing a delirium prediction model are specifically as follows: S401: Based on the abnormal entropy value marker, collect the medical records of postoperative patients, including the type of surgery, anesthesia usage, and basic health indicators of the patients, and screen the records that meet the analysis criteria to obtain a structured data set; S402: Through the structured data set, screen key variables, including age, operation duration, and blood circulation parameters, perform outlier analysis and missing value processing on the key variables, check the quality of the data, and generate a processed data set; S403: Extract training data from the processed data set, use machine learning of artificial intelligence to perform random sampling to verify sample balance, and use the data to train the model, calculate the probability of delirium occurrence in postoperative patients, and construct a delirium prediction model.

[0013] As a further solution of the present invention, the formula for calculating the probability of delirium occurrence in the postoperative patient is as follows: Wherein, is the probability of delirium occurrence in the postoperative patient, represents the intercept of the model, represents the weight of the age factor, represents the weight of the operation duration, represents the weight of the anesthesia type, is the natural constant, is the age, is the operation duration, is the anesthesia type.

[0014] As a further solution of the present invention, use the delirium prediction model to evaluate the state of postoperative patients, and analyze the risk level of delirium occurrence in postoperative patients by real-time monitoring of physiological data, and the steps of obtaining a delirium risk assessment result are specifically as follows: S501: Use the delirium prediction model, combined with the real-time physiological data of postoperative patients, including heart rate, blood pressure, and brain waves, continuously monitor and real-time analyze the real-time physiological data, analyze the health status of postoperative patients and the potential risks of delirium, and obtain a real-time physiological monitoring data set; S502: Analyze the delirium occurrence risk of each patient by using the delirium prediction model based on the real-time physiological monitoring data set, and take preventive measures according to the risk level to obtain the delirium risk classification result; S503: Evaluate the delirium risk of the postoperative patients according to the delirium risk classification result, provide management strategies to manage and avoid the occurrence of delirium, and obtain the delirium risk assessment result.

[0015] On the other hand, an electric vehicle status monitoring system is provided. The electric vehicle status monitoring system is used to execute the above-mentioned electric vehicle status monitoring method, and the system includes: The data recording module collects the electrocardiogram data of postoperative patients, records the electroencephalogram information, performs time series analysis on the physiological data set, and performs formatting processing to obtain the formatted data set; The entropy value calculation module calculates the entropy value of the formatted data set, calculates the sample entropy within the differential time window, and obtains the sample entropy data set; The frequency domain conversion module applies frequency domain transformation technology to convert the sample entropy data set into frequency domain data, identifies the frequency patterns associated with the delirium state through frequency domain analysis, and obtains the frequency domain analysis result; The anomaly detection module compares the normal frequency pattern according to the frequency domain analysis result, identifies the frequency changes deviating from the normal pattern, and performs anomaly entropy value marking to obtain the anomaly entropy value marking; The model construction module uses the anomaly entropy value marking as the model input, and uses machine learning technology for model training to construct a delirium prediction model; The risk level identification module uses the delirium prediction model to evaluate the status of postoperative patients, analyzes the risk level of delirium occurrence, and obtains the delirium risk assessment result.

[0016] The beneficial effects brought by the technical solution provided by the embodiments of the present invention at least include: By collecting the electrocardiogram and electroencephalogram data of postoperative patients, and using time series analysis method to perform formatting processing and entropy value calculation on the data, the brain function state of the patients can be effectively revealed. Through the acquisition of sample entropy value and the application of frequency domain transformation technology, the specific frequency patterns associated with the delirium state can be deeply analyzed and identified. By comparing the frequency changes with the preset standards, the data points deviating from the normal pattern can be accurately marked, and more accurate risk assessment and early warning can be carried out for the patients. Through the application of real-time monitoring and frequency domain analysis results, not only the prediction accuracy of the occurrence probability and time of postoperative delirium is enhanced, but also it helps to optimize the patient care process, reduce complications, and improve the patient recovery effect and quality of life. Description of the Drawings

[0017] Figure 1 It is a schematic diagram of the working process of the present invention; Figure 2 It is the detailed flowchart of S1 of the present invention; Figure 3 It is the detailed flowchart of S2 of the present invention; Figure 4 It is the detailed flowchart of S3 of the present invention; Figure 5 It is the detailed flowchart of S4 of the present invention; Figure 6 It is the detailed flowchart of S5 of the present invention; Figure 7 It is the system flowchart of the present invention. Specific embodiments

[0018] Next, in conjunction with the accompanying drawings, the technical solutions in the present invention will be described.

[0019] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as an "example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of the word "example" is intended to present concepts in a specific manner. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either of the two can be selected.

[0020] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail in conjunction with the accompanying drawings and specific embodiments.

[0021] Please refer to Figure 1 , the embodiments of the present invention provide a method for artificial intelligence to predict postoperative delirium in patients. The processing flow of this method can include the following steps: S1: Collect the electrocardiogram and electroencephalogram data of postoperative patients, use time series analysis method to format the data, calculate the entropy value of the processed data, and obtain a sample entropy data set by calculating the sample entropy value within different time windows; S2: Use the sample entropy data set, apply frequency domain transformation technology to convert time series data into frequency domain data, verify key frequency components through frequency domain analysis, and identify frequency patterns associated with the delirium state to obtain the frequency domain analysis result; S3: According to the frequency domain analysis result, compare with the normal frequency pattern, identify the frequency changes deviating from the normal pattern, and compare the sample entropy with the preset standard to mark the data points deviating from the threshold to obtain abnormal entropy value marks; S4: Use the abnormal entropy value marks as the input of the model, utilize machine learning of artificial intelligence to verify the generalization ability of the model, predict the probability and time of postoperative delirium occurrence in patients, and construct a delirium prediction model; S5: Use the delirium prediction model to evaluate the status of postoperative patients. By real-time monitoring of physiological data, analyze the risk level of postoperative patients developing delirium to obtain the delirium risk assessment result.

[0022] The sample entropy data set includes entropy values, time tags, and change rates of entropy values for multiple time windows. The frequency domain analysis results include verification data of key frequency components, frequency patterns associated with the delirium state, intensity and duration of frequency patterns. The abnormal entropy value markers include entropy values of data points, deviation degrees, and corresponding timestamps. The results of the delirium prediction model include the structure of the model, training effect, and performance data of the model on the independent test set. The delirium risk assessment result includes the delirium risk level of postoperative patients, weights of risk factors, and distribution of risk levels.

[0023] Please refer to Figure 2 , collect electrocardiogram and electroencephalogram data of postoperative patients, use time series analysis method to format the data, and calculate the entropy value of the processed data. The steps to obtain the sample entropy data set by calculating the sample entropy value within different time windows are as follows: S101: Collect electrocardiogram and electroencephalogram data of postoperative patients, extract data records including timestamps and physiological parameters, organize the data format to match the requirements of time series analysis. The execution process to obtain the time series data set is as follows; Collect the original signal data of patients in various heart rate and brain wave activity states. The data includes timestamp information and corresponding physiological parameters, such as beats per minute of heart rate, brain wave frequency, etc. Organize the original data in a certain time order to ensure the continuity and integrity of the data during analysis. This process involves complex data verification and formatting to ensure that the data is not only complete in quantity but also meets the requirements of subsequent analysis in terms of quality. It is also necessary to preliminarily classify the data through software tools to exclude abnormal data that obviously does not conform to physiological norms, such as data points with extremely high or low heart rates, to obtain the time series data set.

[0024] S102: Use the time series data set to perform denoising on the data, remove abnormal points, fill in missing time points, and perform normalization to avoid the influence of dimensions. The execution process to obtain the formatted data set is as follows; Perform data normalization according to the formula: Calculate the normalized data value. In the formula, represents the original data value, represents the minimum value in the data set, represents the maximum value in the data set; Refer to a set of original electrocardiogram data , where the minimum value and the maximum value , using the above formula, the normalization process for the first data point is calculated as follows: Similarly, for the processing of , Continuing this calculation, the normalization results of the entire dataset can be obtained. The use of this formula ensures that data with different dimensions or units can be compared and analyzed under the same standard.

[0025] S103: Based on the formatted dataset, a fixed-length sliding window is selected, and the sample entropy of each window is calculated segment by segment. The execution process includes selecting the window size and step length, calculating the entropy value for each window one by one, and obtaining the sample entropy dataset as follows; Set the window size and step length. The parameters determine the amount of data covered by each window slide and the sliding distance. Move the window from the start to the end of the dataset. Each time it moves to a new position, calculate the sample entropy of the data within the window. Sample entropy is a method to quantify the complexity or unpredictability of data and is used to analyze the dynamic change characteristics of time series data. Each calculated entropy value will represent the complexity of the data within the corresponding window. The entropy values are summarized, providing a quantitative measure of complexity for subsequent data analysis. This process not only involves mathematical calculations but also requires appropriate algorithms and software support to obtain the sample entropy dataset.

[0026] Please refer to Figure 3 , using the sample entropy dataset, applying frequency domain transformation technology, converting the time series data into frequency domain data, verifying the key frequency components through frequency domain analysis, and identifying the frequency patterns associated with the delirium state. The specific steps to obtain the frequency domain analysis results are as follows: S201: Using the sample entropy dataset, adopt the fast Fourier transform algorithm to convert the time series data into frequency domain data. The execution process to verify that each data point is converted and obtain the frequency domain dataset is as follows; Convert each data point in the time series dataset into frequency domain data one by one. This transformation process is achieved through the fast Fourier transform algorithm, which can decompose the time series signal into various frequency components. Each frequency component represents the amplitude and phase of a specific frequency in the signal. The conversion of the entire data point ensures data integrity and improves the accuracy and efficiency of analysis, enabling further analysis and processing such as frequency screening and feature extraction, which is of great significance for understanding and interpreting the physiological changes of postoperative patients. Frequency domain analysis provides a new perspective for observing physiological signal changes and can help doctors better evaluate the patient's condition to obtain the frequency domain analysis results.

[0027] S202: Detect frequency components based on the frequency-domain data set. Through the analysis of the amplitude and phase of the frequency components, identify the key frequencies, calculate the amplitudes of the key frequencies, and the execution process for obtaining the key frequency analysis results is as follows; The formula for calculating the amplitude of the key frequency is: Where, is the amplitude value of the key frequency, represents the real part of the frequency and represents the imaginary part of the frequency ; represents the total number of sample points, represents the compensation coefficient; Parameter meanings and set values: is the real part of the frequency and is provided by the fast Fourier transform (FFT) algorithm. For example, at a specific frequency , the measured real part value is 30; is the imaginary part of the frequency and is also obtained by the FFT algorithm. When set at , the measured imaginary part value is 40; is the total number of samples, which is related to the accuracy and range of the frequency-domain conversion. For example, a 1024-point FFT is used; is the compensation coefficient, which is adjusted according to the processing gain of the signal. If the signal is weak, it is set to 1.5 to ensure that the amplitude is not distorted by changes in the signal strength; Substitute the parameters into the formula for calculation: Calculate the sum of the squares of the real and imaginary parts of the amplitude: ; ; ; Multiply the square root of this sum by the square root of the number of samples and the compensation coefficient: ; ; Calculate the amplitude value of the key frequency: The result 2.34 indicates the normalized amplitude value at the 60 Hz frequency. This value reflects the signal strength at a specific frequency, helps analyze and compare the significance of different frequency components, can accurately obtain the key frequency components in the signal, and perform feature recognition in frequency-domain analysis.

[0028] S203: Based on the results of key frequency analysis and combined with the physiological parameters of patients after surgery, the process of identifying the association between the target frequency pattern and the delirium state and obtaining the frequency domain analysis results is as follows; Determine the association between the delirium state and some specific frequency patterns, including comparing the frequency domain data of different patients in the delirium state and the normal state. By identifying the significantly changed frequency patterns, the relationship between the changes and the delirium state of the patients can be inferred. This not only helps to understand the physiological basis of the delirium state but also provides insights for clinical practice on how to predict and prevent delirium. The entire analysis process relies on advanced data processing techniques and in-depth physiological knowledge to obtain the frequency domain analysis results.

[0029] Please refer to Figure 4 , according to the frequency domain analysis results, compare with the normal frequency pattern, identify the frequency changes deviating from the normal pattern, and compare the sample entropy with the preset standard to mark the data points deviating from the threshold. The specific steps for obtaining the marked abnormal entropy values are as follows: S301: Adopt the frequency domain analysis results, compare the frequency pattern of the differential sample points with the frequency pattern in the normal state, and identify the frequency changes different from the normal pattern. The execution process of obtaining the abnormal frequency change data set is as follows; Compare the frequency pattern of the differential sample points with the frequency pattern in the normal state. The comparison process needs to be converted from time series data to frequency domain data through Fourier transform. Then, match and analyze the converted frequency pattern with the standard frequency pattern in the normal state. By calculating the frequency deviation, identify the frequency changes significantly different from the normal pattern, including all sample points deviating from the normal frequency pattern. The whole process involves complex data processing techniques such as frequency extraction and difference analysis to ensure that abnormal frequency points can be accurately found for monitoring and analyzing the abnormal state in physiological data and obtaining the abnormal frequency change data set.

[0030] S302: Based on the abnormal frequency change data set, compare the sample entropy value of each sample point with the preset normal entropy value range, and use threshold judgment to mark the sample points exceeding the normal range. The execution process of obtaining the abnormal sample record is as follows; Compare the sample entropy value of each sample point with the preset normal entropy value range, which involves the calculation of entropy value and threshold judgment. By setting a threshold based on the entropy value range of the normal physiological state obtained from long-term observation, if the entropy value of the sample point exceeds this range, it is marked as an abnormal sample point, providing important data basis for subsequent analysis. In this way, the abnormal physiological state can be quickly and effectively identified for more accurate monitoring and prediction of the condition, and the abnormal sample record can be obtained.

[0031] S303: Using the abnormal sample records, perform iterative statistical analysis and visual display to identify and analyze the physiological state changes of postoperative patients. The execution process for obtaining the abnormal entropy value markers is as follows; Perform iterative statistical analysis according to the formula: Update the data value. In the formula, represents the data value of the previous iteration, represents the target data value, represents the learning rate; Set the abnormal entropy value of a certain patient in the initial iteration to and the normal entropy value target to with the learning rate set to 0.1; Substitute into the formula for calculation: This iterative process reflects how to gradually adjust the data to make it tend to the normal range. Continuing to apply this formula can gradually reduce the abnormal entropy value and more accurately identify and analyze the physiological state changes of postoperative patients.

[0032] Please refer to Figure 5 Using the abnormal entropy value marker as the input of the model and leveraging machine learning in artificial intelligence to verify the generalization ability of the model, predict the probability and time of postoperative delirium in patients. The specific steps for constructing the delirium prediction model are as follows: S401: Based on the abnormal entropy value marker, collect the medical records of postoperative patients, including the type of surgery, anesthesia usage, and the patient's basic health indicators, and screen the records that meet the analysis criteria to obtain the structured data set. The execution process is as follows; Determine the types of medical records to be collected, which include the type of surgery, anesthesia usage, and the patient's basic health indicators such as heart rate, blood pressure, etc. According to the set analysis criteria, screen out the eligible medical records. After collection, the records need to be structured for subsequent data analysis. This is achieved through a standardized input format and unified data fields to ensure that each piece of data can be correctly used in the analysis. The entire process requires not only accurate data entry but also involves a preliminary check of data integrity and accuracy to obtain the structured data set.

[0033] S402: Through the structured data set, screen the key variables, including age, operation duration, and blood circulation parameters, perform outlier analysis and missing value processing on the key variables, check the data quality, and generate the processed data set. The execution process is as follows; During the process of screening key variables for outlier analysis and missing value handling, the analysis team defines key variables, including the patient's age, surgical duration, and blood circulation parameters, etc. These variables are crucial for evaluating surgical outcomes and patient recovery. Statistical methods are used to identify and handle outliers in the key variables, and at the same time, missing data values are filled to ensure the integrity and usability of the dataset for further statistical analysis and model building. In this process, maintaining data quality is of utmost importance because the quality of the data directly affects the accuracy and reliability of the analysis results, and a processed dataset is generated.

[0034] S403: Extract training data from the processed dataset, use machine learning in artificial intelligence for random sampling to verify sample balance, and apply the data to train the model. The execution process of constructing a delirium prediction model is as follows; The formula for calculating the probability of postoperative delirium in patients is as follows: Where, is the probability of postoperative delirium in patients, represents the intercept of the model, represents the weight of the age factor, represents the weight of the surgical duration, represents the weight of the anesthesia type, is the natural constant, is the age, is the surgical duration, is the anesthesia type; Parameter meanings and set values: is the intercept, set to 0.85, based on the average baseline risk in previous studies; is the weight of the age factor, set to 0.05, representing that for each additional year of age, the risk of delirium increases by 5%; is the weight of the surgical duration, set to 0.04, representing that for each additional hour of surgical time, the risk of delirium increases by 4%; is the weight of the anesthesia type, set to 0.5, representing that using some types of anesthesia (such as general anesthesia) compared with the remaining types, the delirium risk increases by 50%; Substitute the parameters into the formula for calculation: Set a 70-year-old patient with a surgical time of 3 hours and general anesthesia, then , , (Non-numerical data is represented by 0 or 1 for presence or absence); The result of 0.657 indicates that the probability of the patient developing delirium after surgery is 65.7%. This value is higher than the conventional risk level, indicating that advanced age, longer surgical time, and the use of general anesthesia all significantly increase the risk of delirium, guiding the medical team to take corresponding preventive measures before surgery.

[0035] Please refer to Figure 6 , use the delirium prediction model to evaluate the status of patients after surgery, and analyze the risk level of delirium in patients after surgery by real-time monitoring of physiological data. The steps to obtain the delirium risk assessment result are as follows: S501: Use the delirium prediction model, combined with the real-time physiological data of the patient after surgery, including heart rate, blood pressure, and brain waves, to continuously monitor and analyze the real-time physiological data, analyze the health status of the patient after surgery and the potential risk of delirium. The execution process of obtaining the real-time physiological monitoring data set is as follows; Continuously monitor key physiological parameters such as the patient's heart rate, blood pressure, and brain waves. The data is collected in real time by high-precision medical devices and transmitted to the analysis system through a data interface. The system analyzes the data in real time to detect any changes indicating abnormal health status, such as an abnormal increase in heart rate or a sudden drop in blood pressure. These indicators help to identify the potential risk of delirium in patients after surgery in a timely manner. This process not only involves the real-time collection of data but also includes complex data processing and analysis techniques such as signal denoising, data smoothing, and anomaly detection, etc., to obtain the real-time physiological monitoring data set.

[0036] S502: Through the real-time physiological monitoring data set, use the delirium prediction model to analyze the risk of delirium occurrence in each patient, and take preventive measures according to the risk level. The execution process of obtaining the delirium risk classification result is as follows; Use the delirium prediction model to evaluate the delirium risk of the patient. According to the formula: Calculate the delirium risk score. In the formula, represents the heart rate, represents the blood pressure, represents the brain wave activity, , and represent the weights of each factor; Set the patient's heart rate , blood pressure , brain wave activity , and set the weights of the heart rate, blood pressure, and brain waves to be , , ; Calculate the delirium risk score through the formula: This calculation shows that based on the given physiological parameters and weights, the delirium risk of patients can be quantitatively evaluated, which helps to identify high-risk patients and take corresponding preventive measures.

[0037] S503: According to the delirium risk classification results, evaluate the delirium risk of postoperative patients, provide management strategies, manage and avoid the occurrence of delirium, and the implementation process for obtaining the delirium risk assessment results is as follows; The process of evaluating the delirium risk of postoperative patients involves the formulation and implementation of management strategies. This process requires a detailed analysis of the risk classification results, and personalized management measures are formulated according to the specific situation of each patient. The measures include environmental control, psychological support, etc., to avoid the occurrence of delirium. The entire evaluation process is the result of the collaboration of a multidisciplinary team, involving professionals in various fields such as doctors, nurses, and psychologists, working together to ensure that each measure can specifically address the actual needs of patients and obtain the delirium risk assessment results.

[0038] Please refer to Figure 7 , on the other hand, an electric vehicle status monitoring system is provided. The electric vehicle status monitoring system is used to execute the above-mentioned electric vehicle status monitoring method, and the system includes: The data recording module collects the electrocardiogram data of postoperative patients, records the electroencephalogram information, performs time series analysis on the physiological data set, and conducts formatting processing to obtain a formatted data set; The entropy value calculation module calculates the entropy value of the formatted data set, calculates the sample entropy within the differential time window, and obtains a sample entropy data set; The frequency domain conversion module applies frequency domain transformation technology to convert the sample entropy data set into frequency domain data, identifies the frequency patterns associated with the delirium state through frequency domain analysis, and obtains the frequency domain analysis results; The anomaly detection module, based on the frequency domain analysis results, compares with the normal frequency pattern, identifies the frequency changes deviating from the normal pattern, and performs abnormal entropy value marking to obtain an abnormal entropy value marking; The model construction module uses the abnormal entropy value marking as the model input, and uses machine learning technology for model training to construct a delirium prediction model; The risk level identification module uses the delirium prediction model to evaluate the status of postoperative patients, analyzes the risk level of delirium occurrence, and obtains the delirium risk assessment results.

[0039] As described above, only the specific embodiments of the present invention are provided, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for predicting postoperative delirium in patients using artificial intelligence, characterized in that: The following steps are involved: Collect the electrocardiogram and electroencephalogram data of patients after surgery, use time series analysis method to format the data, calculate the entropy value of the processed data, and obtain the sample entropy data set by calculating the sample entropy value within the differentiated time window; The sample entropy data set is used, and the frequency domain transformation technology is applied to convert the time series data into frequency domain data, and the key frequency components are verified through frequency domain analysis, and the frequency pattern associated with the delirium state is identified to obtain the frequency domain analysis results; According to the frequency domain analysis result, the normal frequency pattern is compared to identify the frequency change that deviates from the normal pattern, and the sample entropy is compared with the preset standard, and the data points that deviate from the threshold are marked to obtain the abnormal entropy value mark; The abnormal entropy value mark is used as the input of the model, and the generalization ability of the model is verified by using the machine learning of artificial intelligence, so as to predict the probability and time of delirium in postoperative patients and construct a delirium prediction model; The delirium prediction model is used to evaluate the postoperative patient status, and the risk level of postoperative delirium in patients is analyzed by real-time monitoring of physiological data to obtain a delirium risk assessment result.

2. The method for predicting postoperative delirium by artificial intelligence according to claim 1, characterized in that: The sample entropy data set includes entropy values, time labels, and rates of change of entropy values ​​for multiple time windows; the frequency domain analysis results include verification data of key frequency components, frequency patterns associated with delirium states, and the intensity and duration of frequency patterns; the abnormal entropy value markers include entropy values ​​of data points, degrees of deviation, and corresponding timestamps; the results of the delirium prediction model include the structure of the model, training effects, and performance data of the model on an independent test set; the delirium risk assessment results include the delirium risk level of postoperative patients, the weights of risk factors, and the distribution of risk levels.

3. The method for predicting postoperative delirium by artificial intelligence according to claim 1, characterized in that: Collect the electrocardiogram and electroencephalogram data of patients after surgery, use time series analysis method to format the data, calculate the entropy value of the processed data, and calculate the sample entropy value within the differentiated time window to obtain the sample entropy data set. The specific steps are as follows: Collect the ECG and EEG data of patients after surgery, extract data records including timestamps and physiological parameters, organize the data format, match the needs of time series analysis, and obtain a time series data set; Using the time series data set, denoising the data, removing outliers, filling in missing time points, normalizing the data to avoid dimensional influence, and obtaining a formatted data set; Based on the formatted data set, a fixed-length sliding window is selected, and the sample entropy of each window is calculated segment by segment, including selecting a window size and a step size, and calculating entropy values ​​for each window to obtain a sample entropy data set.

4. The method for predicting postoperative delirium by artificial intelligence according to claim 1, characterized in that: The sample entropy data set is used, and the frequency domain transformation technology is applied to convert the time series data into frequency domain data. The key frequency components are verified through frequency domain analysis, and the frequency patterns associated with the delirium state are identified. The specific steps of obtaining the frequency domain analysis results are as follows: Using the sample entropy data set, a fast Fourier transform algorithm is used to convert the time series data into frequency domain data, verify that each data point is converted, and obtain a frequency domain data set; According to the frequency domain data set, the frequency components are detected, the key frequencies are identified by analyzing the amplitude and phase of the frequency components, the amplitude of the key frequencies is calculated, and the key frequency analysis results are obtained; By combining the key frequency analysis results with the physiological parameters of the postoperative patient, the correlation between the target frequency pattern and the delirium state is identified to obtain the frequency domain analysis results.

5. The method for predicting postoperative delirium by artificial intelligence according to claim 4, characterized in that: The formula for calculating the amplitude of the critical frequency is: in, is the amplitude value of the key frequency, Representative frequency The real part of Representative frequency The imaginary part of Represents the total number of sample points, Represents the compensation factor.

6. The method for predicting postoperative delirium by artificial intelligence according to claim 1, characterized in that: According to the frequency domain analysis result, the normal frequency pattern is compared, the frequency change that deviates from the normal pattern is identified, and the sample entropy is compared with the preset standard, and the data points that deviate from the threshold are marked, and the steps of obtaining the abnormal entropy value mark are specifically as follows: Using the frequency domain analysis results, the frequency pattern of the differentiated sample points is compared with the frequency pattern in the normal state, the frequency changes that are different from the normal pattern are identified, and the abnormal frequency change data set is obtained; Based on the abnormal frequency change data set, the sample entropy value of each sample point is compared with a preset normal entropy value range, and the sample points that exceed the normal range are marked by using a threshold value to obtain an abnormal sample record; The abnormal sample records are used to perform iterative statistical analysis and visual display to identify and analyze changes in the patient's physiological state after surgery and obtain abnormal entropy value markers.

7. The method for predicting postoperative delirium by artificial intelligence according to claim 1, characterized in that: The abnormal entropy value mark is used as the input of the model, and the generalization ability of the model is verified by using artificial intelligence machine learning to predict the probability and time of delirium in postoperative patients. The specific steps of constructing the delirium prediction model are as follows: Based on the abnormal entropy value markers, medical records of postoperative patients are collected, including the type of surgery, anesthesia usage, and basic health indicators of patients, and records that meet the parsing criteria are screened to obtain a structured data set; Through the structured data set, key variables are screened, including age, operation duration, and blood circulation parameters, outlier analysis and missing value processing are performed on the key variables, the quality of the data is checked, and a processed data set is generated; Training data is extracted from the processed data set, and random sampling is performed to verify sample balance using artificial intelligence machine learning. The data is used to train the model, calculate the probability of delirium occurring in postoperative patients, and construct a delirium prediction model.

8. The method for predicting postoperative delirium by artificial intelligence according to claim 7, characterized in that: The formula for calculating the probability of delirium occurring in patients after the surgery is as follows: in, is the probability of delirium in patients after surgery, represents the intercept of the model, represents the weight of the age factor, represents the weight of the operation duration, represents the weight of the anesthesia type, is a natural constant, For age, The duration of the operation, Type of anesthesia.

9. The method for predicting postoperative delirium by artificial intelligence according to claim 1, characterized in that: The steps of using the delirium prediction model to evaluate the postoperative patient status, analyzing the risk level of postoperative delirium in patients by real-time monitoring of physiological data, and obtaining the delirium risk assessment result are as follows: Using the delirium prediction model, combined with the real-time physiological data of the postoperative patient, including heart rate, blood pressure and brain waves, the real-time physiological data is continuously monitored and analyzed in real time, the health status of the postoperative patient and the potential risk of delirium are analyzed, and a real-time physiological monitoring data set is obtained; By using the real-time physiological monitoring data set and the delirium prediction model, the risk of delirium occurrence of each patient is analyzed, and preventive measures are taken according to the risk level to obtain a delirium risk classification result; According to the delirium risk classification results, the delirium risk of postoperative patients is assessed, and management strategies are provided to manage and avoid the occurrence of delirium, thereby obtaining delirium risk assessment results.

10. An artificial intelligence system for predicting postoperative delirium in patients, characterized in that: According to any one of claims 1 to 9, the method for predicting postoperative delirium in patients using artificial intelligence, the system comprising: The data recording module collects the electrocardiogram data of the patient after surgery, records the electroencephalogram information, performs time series analysis on the physiological data set, and formats it to obtain a formatted data set; The entropy value calculation module performs entropy value calculation on the formatted data set, calculates the sample entropy within the differentiated time window, and obtains a sample entropy data set; The frequency domain conversion module applies frequency domain transformation technology to convert the sample entropy data set into frequency domain data, identifies the frequency pattern associated with the delirium state through frequency domain analysis, and obtains frequency domain analysis results; The anomaly detection module compares the normal frequency pattern according to the frequency domain analysis result, identifies the frequency change that deviates from the normal pattern, performs an abnormal entropy value mark, and obtains an abnormal entropy value mark; The model building module uses the abnormal entropy value mark as a model input, uses machine learning technology to train the model, and builds a delirium prediction model; The risk level identification module uses the delirium prediction model to evaluate the patient's status after surgery, analyze the risk level of delirium, and obtain a delirium risk assessment result.

Images