Real-time hierarchical management method and system for delirium risk of ICU (Intensive Care Unit) multi-injury patient

Abstract

The invention provides an ICU multi-injury patient delirium risk real-time hierarchical management method and system, and the method comprises the steps: obtaining the multi-modal data of a patient, carrying out the preprocessing and standardization processing of the multi-modal data, and obtaining the standardized multi-modal data; performing feature extraction on the standardized multi-modal data to obtain a structured data set of the standardized multi-modal data; inputting the structured data set into a multi-algorithm parallel model, integrating output results of all algorithm models based on a soft voting strategy, and taking the model output result with the optimal comprehensive score as a prediction result; based on the prediction result, matching hierarchical levels meeting the prediction result; a personalized auxiliary intervention scheme is generated on the basis of the hierarchical level of the patient, and through risk hierarchical management, medical resource allocation is optimized, unnecessary excessive intervention is reduced, and it is ensured that the high-risk patient gets enough attention; according to the intelligent nursing and intervention system, personalized nursing and intervention schemes are provided, the nursing quality is effectively improved, the intervention schemes are adjusted in time through real-time feedback and dynamic updating, and the delirium occurrence rate is reduced.

Description

Technical Field

[0001] This application relates to the field of medical technology, and specifically to a real-time hierarchical management method and system for delirium risk in ICU multiple trauma patients. Background Art

[0002] ICU multiple trauma patients usually face multiple organ injuries, receive multiple treatments simultaneously, and experience frequent physiological fluctuations, which significantly increases their delirium risk compared to other inpatients. In the ICU environment, the patient's condition is usually in a rapid dynamic change, especially for multiple trauma patients, who often suffer from severe physical trauma, long-term use of sedative drugs, and treatment methods such as mechanical ventilation. The superposition of these factors significantly increases the risk of delirium. In addition, factors such as noise, light changes, and sleep deprivation in the ICU environment may also have an adverse impact on the patient's cognitive function and exacerbate the risk of delirium. Therefore, during the treatment of ICU multiple trauma patients, there is an urgent need for a more accurate and personalized auxiliary intervention plan to optimize management for the risks of different patients.

[0003] However, the application of current delirium risk prediction technologies in ICU multiple trauma patients faces the following challenges, thus limiting their actual effects: ① Single data source: Most traditional prediction models rely on a single data source (such as vital signs or laboratory test results), making it difficult to comprehensively reflect the patient's overall condition. The condition of ICU multiple trauma patients is complex and dynamically changing, and a single data source is insufficient to capture their multi-dimensional characteristics. ② Lack of personalization: Current prediction models usually do not fully consider the individual differences of patients (such as age, past medical history, and treatment interventions), making it difficult to provide accurate personalized prevention plans for each patient. This limitation may lead to underreporting or misreporting of delirium. ③ Insufficient real-time update: The condition of ICU patients changes rapidly, and traditional models mostly rely on static data and lack a real-time feedback mechanism, unable to dynamically adjust the prediction results according to the rapid changes in the condition. Therefore, how to overcome the above-mentioned technical problems and defects has become a problem that needs to be solved. Summary of the Invention

[0004] In order to overcome the above problems existing in the prior art, this application provides a real-time hierarchical management method and system for delirium risk in ICU multiple trauma patients, and adopts the following technical solutions:

[0005] In the first aspect, this application provides a real-time hierarchical management method for delirium risk in ICU multiple trauma patients, including:

[0006] Obtain the multi-modal data of the patient, perform preprocessing and standardization processing on the multi-modal data to obtain standardized multi-modal data.

[0007] Extract features from the standardized multi-modal data to obtain a structured data set of the standardized multi-modal data.

[0008] Input the structured data set into a multi-algorithm parallel model, integrate the output results of each algorithm model based on the soft voting strategy, and use the output result of the model with the best comprehensive score as the prediction result.

[0009] Based on the prediction result, match the hierarchical level that meets the prediction result.

[0010] Generate a personalized assisted intervention plan based on the patient's hierarchical level.

[0011] Furthermore, perform preprocessing and standardization on the multi-modal data, including:

[0012] Perform noise removal and missing value processing on the obtained multi-modal data to obtain preprocessed multi-modal data.

[0013] Perform standardization on the preprocessed multi-modal data to obtain standardized multi-modal data.

[0014] Furthermore, extract features from the standardized multi-modal data to obtain a structured data set of the standardized multi-modal data, including:

[0015] Construct the standardized multi-modal data into a comprehensive feature matrix, where each column of the comprehensive feature matrix represents a feature and each row corresponds to the observed data of a patient.

[0016] Use the Lasso regression algorithm to train the comprehensive feature matrix. Lasso regression automatically selects the features that contribute more to delirium prediction by adding an L1 regularization term to the loss function to reduce the coefficients of unimportant features to zero; adjust the regularization parameter of Lasso regression through cross-validation to determine the optimal balance point; during the training process, assign a coefficient to each feature, and this coefficient reflects the degree of contribution of the feature to the prediction target.

[0017] By analyzing the feature contribution values of Lasso regression, finally retain the feature set with contribution values greater than the set threshold as the key feature set.

[0018] Integrate the key feature set into a structured data set in a unified format.

[0019] Furthermore, input the structured data set into a multi-algorithm parallel model, and integrate the output results of each model based on the soft voting strategy, including:

[0020] Obtain the first prediction result based on the random forest; obtain the second prediction result based on logistic regression; obtain the third prediction result based on the support vector machine; obtain the fourth prediction result based on the deep learning model.

[0021] Furthermore, obtain the first prediction result based on the random forest, including:

[0022] Predictions are made by constructing multiple decision trees. Each decision tree randomly selects a part of the data from the structured dataset for training through sampling with replacement, and at the same time randomly selects a subset of features for splitting each node.

[0023] When the model training is completed, the structured dataset will be independently processed by each decision tree to obtain a prediction result.

[0024] The prediction results of all trees are integrated through a majority voting mechanism to obtain the final delirium risk score as the first prediction result.

[0025] Furthermore, a second prediction result is obtained based on logistic regression, including:

[0026] The logistic regression model performs a linear combination on the structured dataset to obtain a weighted sum. The result of the linear combination is mapped between 0 and 1 through the Sigmoid function and converted into the probability of delirium occurrence, which is used as the second prediction result.

[0027] Furthermore, a third prediction result is obtained based on support vector machines, including:

[0028] Based on the historical patient structured dataset, the data points with and without delirium are separated. The structured dataset of the new patient is mapped into a high-dimensional space. The support vector machine obtains the probability estimation value of the category to which the new patient belongs based on the distance from the structured dataset of the new patient to the hyperplane as the third prediction result.

[0029] Furthermore, a fourth prediction result is obtained based on a deep learning model, including:

[0030] The structured dataset is processed through a multi-layer neural network. Each layer performs a non-linear transformation on the data in the way of weighted sum and gradually extracts high-level features.

[0031] A probability value is generated based on the high-level features, indicating the risk of the patient having delirium, as the fourth prediction result.

[0032] Furthermore, based on a soft voting strategy, the output results of each algorithm model are integrated, and the output result of the model with the best comprehensive score is used as the prediction result, including:

[0033] Through multiple evaluation indicators such as the area under the curve, accuracy, recall rate, precision rate, and F1 score, a comprehensive scoring and quantitative analysis of the performance of the multi-algorithm parallel model is carried out, and the output result of the model with the best comprehensive score is selected as the delirium risk prediction result.

[0034] In a second aspect, the present application also provides a real-time hierarchical management system for delirium risk of ICU multiple trauma patients, including:

[0035] A standardized multimodal data acquisition module for acquiring multimodal data of a patient, preprocessing and standardizing the multimodal data, and obtaining standardized multimodal data.

[0036] A structured data set acquisition module for extracting features from the standardized multimodal data and obtaining a structured data set of the standardized multimodal data.

[0037] A prediction result output module for inputting the structured data set into a multi-algorithm parallel model, integrating the output results of each algorithm model based on a soft voting strategy, and taking the output result of the model with the best comprehensive score as the prediction result.

[0038] A hierarchical level acquisition module for matching a hierarchical level that meets the prediction result based on the prediction result.

[0039] An auxiliary intervention plan generation module for generating a personalized auxiliary intervention plan based on the hierarchical level of the patient.

[0040] In a third aspect, the present application provides an electronic device, including:

[0041] One or more processors; a memory; and one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions that, when executed by the device, cause the device to execute the method described in the first aspect.

[0042] In a fourth aspect, the present application provides a computer-readable storage medium in which a computer program is stored, and when it runs on a computer, it causes the computer to execute the method described in the first aspect.

[0043] In a fifth aspect, the present application provides a computer program that, when executed by a computer, is used to execute the method described in the first aspect.

[0044] In a possible design, the program in the fifth aspect can be stored in whole or in part on a storage medium packaged together with the processor, or can be stored in whole or in part on a memory not packaged together with the processor.

[0045] The present application has the following beneficial effects:

[0046] 1. This application obtains multi-modal data of a patient, preprocesses and standardizes the multi-modal data to obtain standardized multi-modal data, extracts features from the standardized multi-modal data, and obtains a structured data set of the standardized multi-modal data. By preprocessing and standardizing the data, during the process of feature extraction from the patient's multi-modal data, the noise of the data can be reduced, and the speed and accuracy of data feature extraction can be improved. At the same time, by obtaining the multi-modal data of the patient, the overall condition of the patient can be comprehensively reflected, overcoming the problem of single data source.

[0047] 2. This application inputs the structured data set into a multi-algorithm parallel model, integrates the output results of each algorithm model based on a soft voting strategy, and takes the output result of the model with the best comprehensive score as the prediction result. Based on the prediction result, a hierarchical level that meets the prediction result is matched. A personalized assisted intervention plan is generated based on the patient's hierarchical level. By extracting features and predicting risks from the patient's multi-modal data, during the prediction process, the individual differences of the patient are fully considered, and thus a personalized assisted intervention plan that meets the individual differences of the patient is provided.

[0048] 3. This application outputs a risk prediction result through a multi-algorithm parallel model to achieve accurate prediction of the occurrence of delirium and reduce the situation of missed reports and false reports. Through risk stratification management, the allocation of medical resources is optimized, unnecessary over-intervention is reduced, and it is ensured that high-risk patients receive sufficient attention. Provide personalized nursing and intervention plans to effectively improve the quality of nursing. Through real-time feedback and dynamic update, the intervention plan is adjusted in a timely manner to reduce the incidence of delirium. By establishing a real-time feedback mechanism with real-time data, the problem of insufficient real-time update is overcome. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is an exemplary system architecture diagram to which the embodiments of this application can be applied;

[0050] Figure 2 It is a flowchart of the real-time stratification management method for the delirium risk of multiple trauma patients in the ICU in the embodiments of this application;

[0051] Figure 3 It is a flowchart of obtaining a structured data set in the embodiments of this application;

[0052] Figure 4 It is the data processing and feature selection process in the embodiments of this application;

[0053] Figure 5 It is a flowchart of real-time feedback and dynamic adjustment in the embodiments of this application;

[0054] Figure 6 It is a system flowchart of the embodiments of this application;

[0055] Figure 7 It is a schematic diagram of a computer device according to an embodiment of the present application. Detailed implementation manners

[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field to which this application belongs; the terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above drawings are intended to cover non-exclusive inclusion. The terms "first", "second", etc. in the specification and claims of this application or the above drawings are used to distinguish different objects and not to describe a specific order.

[0057] Referring to "embodiments" herein means that a particular feature, structure, or characteristic described in connection with the embodiments can be included in at least one embodiment of this application. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0058] In order to enable those skilled in the art of this technology to better understand the solution of this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0059] As Figure 1 shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0060] Users can use the terminal devices 101, 102, 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications may be installed on the terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0061] The terminal devices 101, 102, and 103 can be various electronic devices with a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, and so on.

[0062] The server 105 can be a server that provides various services, such as a background server that supports the pages displayed on the terminal devices 101, 102, and 103.

[0063] It should be noted that the real-time hierarchical management method for the delirium risk of ICU multiple trauma patients provided by the embodiments of the present application is generally executed by the server / terminal device. Correspondingly, the real-time hierarchical management system for the delirium risk of ICU multiple trauma patients is generally set in the server / terminal device.

[0064] It should be understood that Figure 1 the numbers of the terminal devices, the network, and the server in

[0065] Continue to refer to Figure 2 , the figure shows a flowchart of a real-time hierarchical management method for the delirium risk of ICU multiple trauma patients in the present application. The method includes the following steps:

[0066] Step 201, obtain the multimodal data of the patient, perform preprocessing and standardization processing on the multimodal data, and obtain the standardized multimodal data.

[0067] In the embodiments of the present application, the multimodal data of the patient includes the physiological data of the patient collected in real time: such as vital signs such as heart rate, blood pressure, respiratory rate, body temperature, and oxygen saturation; laboratory test results: such as blood tests, urine analysis, liver and kidney function, and blood glucose levels; and past medical history: including the patient's age, gender, past medical history (such as hypertension, diabetes, heart disease), past surgeries and medication conditions, etc.; and treatment process: including drug use history, surgical records, and sedative drug use conditions, etc., to ensure the comprehensiveness and diversity of the data.

[0068] In a possible implementation manner, obtaining the multimodal data of the patient, performing preprocessing and standardization processing on the multimodal data, and obtaining the standardized multimodal data includes:

[0069] Perform noise removal and missing value processing on the obtained multimodal data to obtain the preprocessed multimodal data.

[0070] Normalize the preprocessed multimodal data to obtain normalized multimodal data.

[0071] In the embodiments of the present application, noise removal and missing value processing are performed on the multimodal data, including eliminating invalid or abnormal data through signal filtering or noise removal algorithms (such as low-pass filtering, outlier detection) to improve data quality. Strategies such as mean filling, KNN interpolation method, or regression interpolation are used to solve the problem of missing data and improve data integrity.

[0072] In the embodiments of the present application, normalizing the preprocessed multimodal data includes: normalizing the data using the Z-Score method or Min-Max scaling to eliminate the influence of different dimensions on the model calculation results and ensure the consistency of the data in the model.

[0073] Step 202: Extract features from the normalized multimodal data to obtain a structured data set of the normalized multimodal data.

[0074] In a possible implementation manner, to extract features from the normalized multimodal data to obtain a structured data set of the normalized multimodal data, please refer to Figure 3 , and the specific content includes:

[0075] Step 31: Construct the normalized multimodal data into a comprehensive feature matrix, where each column of the comprehensive feature matrix represents a feature, and each row corresponds to the observed data of a patient.

[0076] Step 32: Use the Lasso regression algorithm to train the comprehensive feature matrix. Lasso regression automatically selects the features that contribute more to delirium prediction by adding an L1 regularization term to the loss function to reduce the coefficients of unimportant features to zero.

[0077] Step 33: Adjust the regularization parameter (λ) of the Lasso regression through cross-validation to determine the optimal balance point. The optimal λ value ensures that the model can effectively remove redundant features while retaining enough key features to explain the risk of delirium occurrence.

[0078] Step 34: During the training process, Lasso regression assigns a coefficient to each feature, and this coefficient reflects the contribution degree of the feature to the prediction target. In the present application, a contribution threshold (for example, 0.05) is set, and only the features with coefficients greater than 0.05 are retained. This screening method ensures that only the features that contribute significantly to the model prediction are retained, avoiding the introduction of noise or redundant data.

[0079] Step 35: By analyzing the feature contribution values of Lasso regression, finally retain the feature set with contribution values greater than the set threshold as the key feature set; these key feature sets are used for subsequent machine learning model training to predict delirium risk.

[0080] Step 36: Integrate the key feature set into a structured data set in a unified format.

[0081] For example, integrate the key feature set into a structured data set in a unified format. For example, convert categorical data into one-hot encoding and standardize time series data to ensure that different data types can be processed by a unified model.

[0082] In the embodiment of the present application, setting the contribution threshold of Lasso regression to 0.05 helps to automatically screen out the most important features. These key features not only significantly improve the computational efficiency and prediction performance of the model, but also effectively reduce feature redundancy and the risk of model overfitting, thus ensuring that the present application can achieve high-precision and robust delirium risk prediction in a complex ICU environment.

[0083] Please refer to the data processing and feature selection process of steps 201-202 Figure 4 .

[0084] Step 203: Input the structured data set into a multi-algorithm parallel model, integrate the output results of each algorithm model based on the soft voting strategy, and use the output result of the model with the best comprehensive score as the prediction result.

[0085] In a possible implementation, input the structured data set into a multi-algorithm parallel model, and integrate the output results of each model based on the soft voting strategy, including:

[0086] Obtain the first prediction result based on the random forest; obtain the second prediction result based on logistic regression; obtain the third prediction result based on the support vector machine; obtain the fourth prediction result based on the deep learning model.

[0087] In a possible implementation, obtaining the first prediction result based on the random forest includes:

[0088] The random forest makes predictions by constructing multiple decision trees. Each decision tree randomly selects a part of the data from the structured data set for training by the method of sampling with replacement, and randomly selects a subset of features for splitting each node. In this way, the diversity of each tree is ensured, and overfitting is avoided.

[0089] After the model training is completed, the structured dataset is independently processed by each decision tree to obtain a prediction result. The result output by each tree is usually the predicted category of delirium or non-delirium, or the probability of belonging to the delirium category.

[0090] Finally, the prediction results of all trees are integrated through a majority voting mechanism to obtain the final delirium risk score as the first prediction result. Random forests are particularly good at capturing complex interactions between features and can handle missing values and imbalances in the data, so they can effectively predict whether a patient will develop delirium in a clinical scenario.

[0091] In one possible implementation, obtaining a second prediction result based on logistic regression includes:

[0092] The logistic regression model performs a linear combination of the structured dataset to obtain a weighted sum. The result of the linear combination is mapped between 0 and 1 through the Sigmoid function and converted into the probability of delirium occurrence. The weighted sum reflects the relationship between the structured dataset and the occurrence of delirium, and the probability of delirium occurrence is used as the second prediction result.

[0093] Specifically, if the calculated probability is greater than a preset threshold (such as 0.5), logistic regression will predict that the patient will develop delirium; if it is less than the preset threshold (such as 0.5), it will predict that the patient will not develop delirium.

[0094] To train the model, logistic regression adjusts the parameters by maximizing the likelihood function to ensure that the predicted probability on the training data is as close as possible to the true label. Logistic regression is a basic binary classification model. Although it is relatively simple, due to its fast calculation and easy interpretation, it is very suitable for probability prediction in clinical scenarios. It can give a clear probability of delirium occurrence, thus helping medical staff evaluate the risk of patients.

[0095] In one possible implementation, obtaining a third prediction result based on a support vector machine includes:

[0096] When predicting delirium risk, the support vector machine first inputs the patient's features into the model. The goal of the support vector machine is to find a hyperplane that separates the two types of data points (developing delirium and not developing delirium) as much as possible.

[0097] For linearly separable data, the support vector machine selects the optimal classification boundary by calculating the distance from each data point to the hyperplane, maximizing the interval between the two types of data points. When the data is not linearly separable, the support vector machine uses a kernel function to map the data into a higher-dimensional space where the data becomes linearly separable.

[0098] When training a support vector machine model, the optimal hyperplane is found through an optimization problem, and the distance from each data point to this hyperplane is calculated.

[0099] In the prediction stage, new patient data is mapped to this high-dimensional space, and the support vector machine determines the probability estimate value of the class it belongs to based on the distance from the data point to the hyperplane.

[0100] Support vector machines are particularly suitable for handling complex data distributions, can effectively handle high-dimensional data and non-linear classification tasks. Therefore, when facing multi-dimensional and non-linear patient data, support vector machines can provide accurate delirium risk predictions.

[0101] In one possible implementation, obtaining a fourth prediction result based on a deep learning model includes:

[0102] The deep learning model processes patient data through a multi-layer neural network. When predicting the delirium risk, it automatically extracts complex features from the data. The input data is processed through a multi-layer neural network, and each layer performs a non-linear transformation on the data in the form of a weighted sum, thereby gradually extracting higher-level features. These features can capture the dynamic changes in the patient's health status, especially in time series data. During the training process, the deep learning model uses the backpropagation algorithm to optimize the weights of each layer, so that the output value gradually approaches the true label. When the training is completed, the model generates a probability value based on the high-level features of the input data, indicating the risk of the patient developing delirium.

[0103] Deep learning models are particularly good at handling large-scale and high-dimensional data, and can automatically learn important patterns from the raw data without the need for manual feature engineering. Therefore, deep learning models can provide high-precision delirium predictions when dealing with complex and dynamically changing clinical data, and are particularly suitable for continuous monitoring and the analysis of time series data.

[0104] In a possible implementation, the output results of each model are integrated based on a soft voting strategy, including: comprehensively scoring and quantifying the performance of the multi-algorithm parallel model through multiple evaluation metrics such as the area under the curve (AUC), accuracy, recall, precision, and F1-score, and selecting the output result of the model with the best comprehensive score as the delirium risk prediction result. The above metrics evaluate from the perspectives of the model's discrimination ability, prediction accuracy, false negative rate, false positive rate, and comprehensive balance respectively; finally, the model with the best comprehensive score is selected as the main tool for delirium risk prediction, and the prediction results of other models are integrated to optimize the overall performance, so as to achieve accurate prediction of the patient's delirium risk. Through the collaborative work of multiple algorithms, this application not only improves the accuracy and robustness of the prediction, but also can dynamically adapt to the changes in patient data, providing reliable support for subsequent stratified management and personalized intervention.

[0105] In a possible implementation, the output result of the model with the best comprehensive score is used as the prediction result, where the calculation method of the comprehensive score is:

[0106] Comprehensive score = w1 * AUC + w2 * Accuracy + w3 * Recall + w4 * Precision + w5 * F1Score

[0107] where w1, w2, w3, w4, and w5 are the weights of each metric respectively, and the sum of the weights of each metric is 1; AUC, Accuracy, Recall, Precision, and F1Score are the specific values of the corresponding metrics of each model. The model evaluation metrics are shown in Table 1, and the weight assignment and comprehensive score calculation are as

[0108] shown in Table 2:

[0109]

[0110]

[0111] Table 1: Model evaluation metrics

[0112]

[0113] Table 2: Weight assignment and comprehensive score calculation

[0114] Evaluation metrics: The method of weighted comprehensive scoring is used to quantitatively evaluate the model performance to ensure the selection of the optimal model for deployment. The specific metrics and weight assignments are as follows:

[0115] Area under the curve (AUC) (weight 0.2): Measures the overall discrimination ability of the model, reflects the classification performance of the model for positive and negative samples, and is suitable for stratification requirements.

[0116] Accuracy (weight 0.1): As a secondary metric, it reflects the overall correctness of the model's prediction results.

[0117] Recall (weight 0.4): As the most important metric, it is used to reduce false negatives and ensure the discovery ability of the model in practical applications.

[0118] Precision (weight 0.1): An auxiliary metric used to reduce false positives and avoid over-intervention.

[0119] F1 Score (weight 0.2): A balanced performance metric that comprehensively considers recall and precision, suitable for overall evaluation.

[0120] Through the weighted scoring method, this application can scientifically balance multiple performance metrics, select the optimal model as the final tool for delirium risk prediction, and ensure the best effect between the accuracy of the model and the actual application requirements.

[0121] Step 204: Based on the prediction result, match the hierarchical level that meets the prediction result.

[0122] In a possible implementation, matching the hierarchical level that meets the prediction result based on the prediction result includes:

[0123] When the prediction result meets the first preset threshold, the patient's hierarchical level is high risk; when the prediction result meets the second preset threshold, the patient's hierarchical level is medium risk; when the prediction result meets the third preset threshold, the patient's hierarchical level is low risk.

[0124] In the embodiments of this application, the range of the first preset threshold is higher than 0.7; the range of the second preset threshold is between 0.4 and 0.7 (including the two end values of 0.4 and 0.7); the range of the third preset threshold is lower than 0.4.

[0125] In the embodiments of this application, according to the generated prediction result, the delirium risk of the patient is divided into three levels: high risk, medium risk, and low risk. When the prediction result is higher than 0.7, it is determined as high risk. The vital signs of high-risk patients usually fluctuate significantly, and the laboratory test results often show severe electrolyte disorders or persistent heart rate abnormalities. Medium-risk patients refer to those with a prediction result between 0.4 and 0.7, indicating that there is a certain degree of risk of delirium occurrence. Medium-risk patients may have minor physiological abnormalities or a history of mental health problems, so they need moderate monitoring and intervention. Low-risk patients refer to those with a prediction result lower than 0.4, relatively normal physiological indicators, no obvious abnormalities in laboratory test results, and a lower likelihood of delirium occurrence.

[0126] Explanation: The condition of ICU multiple trauma patients changes rapidly and there are significant individual differences. Therefore, there is an urgent need to implement effective stratified management strategies for patients. Stratified management can not only formulate intervention measures according to different risk levels, but also optimize the allocation of medical resources and improve nursing efficiency. However, existing delirium risk prediction models often fail to fully consider the needs of stratified management and do not provide accurate personalized interventions for patients with different risk levels. Therefore, this application provides a stratified management plan based on risk assessment to ensure the rational allocation of medical resources and optimize the treatment effect of patients.

[0127] Step 205, generate a personalized assisted intervention plan based on the stratified level of the patient.

[0128] Explanation: This application effectively responds to the real-time changes in the patient's condition by dynamically adjusting the treatment plan. This application can timely adjust the treatment plan or intervention measures according to the patient's real-time monitoring data (such as vital signs, laboratory test results, etc.) and prediction results. To ensure the accuracy of the treatment plan adjustment, it is necessary to accurately assess and classify the patient's delirium risk. This application divides patients into three levels: high risk, medium risk, and low risk according to the delirium prediction results, so as to provide appropriate intervention measures for different patients.

[0129] Intervention plans for patients with different stratified levels:

[0130] High-risk patients:

[0131] Close monitoring: Strengthen the continuous dynamic monitoring of vital signs (such as heart rate, blood pressure, body temperature, etc.), and use dynamic monitoring equipment to quickly respond to changes in the patient's condition;

[0132] Personalized intervention measures:

[0133] Drug monitoring: Increase the frequency of drug monitoring, especially for drugs that may induce delirium (such as sedatives, antipsychotics), to ensure medication safety;

[0134] Environmental optimization: Provide a quiet and comfortable environment for the patient, and reduce external stimuli by controlling noise and light intensity;

[0135] Multidisciplinary collaboration: A team composed of doctors, nurses, pharmacists, and psychology experts regularly evaluates the patient's condition and formulates personalized intervention plans.

[0136] Medium-risk patients:

[0137] Routine monitoring: Regularly check the patient's vital signs (such as blood pressure, heart rate, oxygen saturation, etc.) to ensure that potential risks can be detected in a timely manner;

[0138] Adjustment of nursing measures: Flexibly adjust the nursing strategy according to the risk changes, including monitoring drug side effects, optimizing the nursing process, and providing psychological support;

[0139] Preventive intervention: Introduce drug or non-drug intervention measures (such as cognitive behavioral therapy, improving sleep, etc.) according to the patient's risk factors in a timely manner.

[0140] Low-risk patients:

[0141] Routine nursing: Continue with routine nursing measures to ensure the comfort and safety of the patient during hospitalization;

[0142] Continuous monitoring of vital sign changes: Although the patient is evaluated as low-risk, it is still necessary to continuously monitor the vital sign changes to maintain vigilance against sudden situations;

[0143] Early detection and intervention: Regularly evaluate the patient's health status to ensure that appropriate intervention measures can be taken immediately once changes occur to prevent delirium.

[0144] In the embodiment of the present application, a personalized assisted intervention plan is generated based on the hierarchical level of the patient, and it also includes real-time feedback and dynamic adjustment. When the patient data monitored in real time changes beyond the preset threshold, steps 201-205 are re-executed to further adjust the personalized assisted intervention plan for the patient. Please refer to Figure 5 , specifically manifested as: dynamically adjusting the risk assessment and intervention plan according to the patient's latest physiological state and data changes during the treatment process. The condition of ICU multiple trauma patients changes rapidly, so it is necessary to have the ability to respond quickly. The specific functions include:

[0145] 1. Real-time data monitoring

[0146] Monitoring content: Through integrated biological monitoring devices, continuously and dynamically monitor the patient's key vital sign data (such as heart rate, blood pressure, blood oxygen saturation, body temperature, etc.). This process has a high degree of automation and accuracy to ensure the real-time and continuity of the data.

[0147] Data transmission: The real-time data collected is transmitted to the centralized database or cloud storage system through a secure data transmission module to ensure that the data can be used by the prediction model in a timely and accurate manner.

[0148] 2. Dynamic adjustment of the prediction model

[0149] Model input update: As the patient's vital sign data changes dynamically, the real-time feedback module transmits the new data to the prediction model, and the model is updated and predicted in real time based on the latest data to ensure a quick response to the patient's status.

[0150] Model adjustment mechanism: According to the real-time data changes, the model dynamically re-evaluates the delirium risk level of the patient. For example, if there is a significant fluctuation in heart rate, the model will automatically adjust the risk prediction level and update the intervention plan.

[0151] Risk assessment feedback: When the model outputs a new risk level (such as high risk, medium risk, low risk), the system will trigger corresponding intervention strategies to ensure that the nursing plan always matches the patient's risk level.

[0152] 3. Dynamic intervention measures (high, medium, low risk)

[0153] High-risk patients:

[0154] Rapid implementation of intensive intervention strategies: including measures such as increasing the frequency of drug monitoring, optimizing the environment (such as avoiding noise and light stimulation), and strengthening psychological support.

[0155] Multi-disciplinary team participation: A nursing team consisting of nurses, doctors, pharmacists, and psychiatrists regularly assesses the patient's condition to ensure timely and effective management of the patient's delirium risk.

[0156] Medium-risk patients:

[0157] Routine intervention: Take routine nursing measures, but pay attention to potential risk changes. For example, regularly check vital signs and adjust the drug use plan to prevent risk escalation.

[0158] Low-risk patients:

[0159] Regular monitoring: Continuously conduct routine nursing and vital sign monitoring to ensure early identification of delirium and prevent potential accidents.

[0160] 4. Closed-loop feedback to real-time data monitoring

[0161] Real-time monitoring: By connecting with the real-time monitoring equipment in the ICU, real-time collection of the patient's vital sign data (such as heart rate, blood oxygen saturation, blood pressure, etc.) is carried out, and the prediction results and intervention measures are dynamically adjusted according to the data changes.

[0162] Dynamic update: As the patient's condition changes, the multi-algorithm parallel model is updated in real time to ensure the accuracy and timeliness of the delirium risk assessment results.

[0163] Feedback mechanism: Through interaction and feedback with medical staff, provide suggestions for optimizing treatment and nursing, such as adjusting the drug dosage or frequency, or adding additional soothing intervention plans for the patient.

[0164] Closed-loop management: Throughout the inpatient treatment cycle, the closed-loop management system can continuously and dynamically adjust the patient's delirium intervention strategy to ensure that the risk is always effectively controlled and reduce the incidence of delirium.

[0165] The feedback mechanism of this application not only supports automatic adjustment but also helps medical staff to respond quickly and make accurate decisions through efficient interaction with them:

[0166] Automated alerts and notifications: When a high-risk event is detected, it will automatically send early warnings to medical staff and push accurate intervention suggestions to ensure that medical staff can take effective measures in a timely manner.

[0167] Interactive feedback: Medical staff interact with the model, adjust the intervention plan according to actual clinical data, and optimize the treatment and care measures for patients in real time.

[0168] Continuous monitoring and reporting: This application regularly generates health reports for patients, including real-time risk assessment, evaluation of the effectiveness of intervention measures, etc., providing a scientific basis for subsequent treatment decisions.

[0169] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. This computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, an optical disk, a Read-Only Memory (ROM), or a Random Access Memory (RAM), etc.

[0170] It should be understood that although the steps in the flowchart of the accompanying drawings are shown in sequence according to the indication of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit and can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time but can be executed at different times. Their execution order is not necessarily sequential but can be executed alternately or in turn with at least a part of other steps or sub-steps or stages of other steps.

[0171] Continue to refer to Figure 6 , the real-time hierarchical management system for delirium risk of multiple trauma patients in the ICU described in this embodiment includes:

[0172] A standardized multimodal data acquisition module 601, which is used to acquire multimodal data of a patient, preprocess and standardize the multimodal data, and obtain standardized multimodal data.

[0173] The structured data set acquisition module 602 is used to extract features from the standardized multimodal data and obtain the structured data set of the standardized multimodal data.

[0174] The prediction result output module 603 is used to input the structured data set into the multi-algorithm parallel model, integrate the output results of each algorithm model based on the soft voting strategy, and use the output result of the model with the best comprehensive score as the prediction result.

[0175] The hierarchical level acquisition module 604 is used to match the hierarchical level that meets the prediction result based on the prediction result.

[0176] The auxiliary intervention plan generation module 605 is used to generate a personalized auxiliary intervention plan based on the hierarchical level of the patient.

[0177] To solve the above technical problems, the embodiments of the present application also provide a computer device. For details, please refer to Figure 7 , Figure 7 which is the basic structural block diagram of the computer device in this embodiment.

[0178] The computer device 7 includes a memory 7a, a processor 7b, and a network interface 7c that are communicatively connected to each other through a system bus. It should be noted that only the computer device 7 with components 7a - 7c is shown in the figure, but it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Among them, those skilled in the art of the present technology can understand that a computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to a microprocessor, an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a digital signal processor (DSP), an embedded device, etc.

[0179] The computer device can be a desktop computer, a notebook, a palm computer, a cloud server, or other computing devices. The computer device can perform human-computer interaction with the user through a keyboard, a mouse, a remote control, a touchpad, a voice control device, or other means.

[0180] The memory 7a at least includes one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (such as SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disc, etc. In some embodiments, the memory 7a may be an internal storage unit of the computer device 7, such as the hard disk or memory of the computer device 7. In other embodiments, the memory 7a may also be an external storage device of the computer device 7, such as a plug-in hard disk, SmartMediaCard (SMC), Secure Digital (SD) card, FlashCard, etc. equipped on the computer device 7. Of course, the memory 7a may also include both the internal storage unit and the external storage device of the computer device 7. In this embodiment, the memory 7a is generally used to store the operating system and various application software installed on the computer device 7, such as the program code of the real-time hierarchical management method for the delirium risk of ICU multiple trauma patients. In addition, the memory 7a can also be used to temporarily store various data that have been output or will be output.

[0181] In some embodiments, the processor 7b may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips. The processor 7b is generally used to control the overall operation of the computer device 7. In this embodiment, the processor 7b is used to run the program code stored in the memory 7a or process data, such as running the program code of the real-time hierarchical management method for the delirium risk of ICU multiple trauma patients.

[0182] The network interface 7c may include a wireless network interface or a wired network interface, and the network interface 7c is generally used to establish a communication connection between the computer device 7 and other electronic devices.

[0183] This application also provides another implementation manner, that is, to provide a non-volatile computer-readable storage medium storing a program of the real-time hierarchical management method for the delirium risk of ICU multiple trauma patients, and the real-time hierarchical management of the delirium risk of ICU multiple trauma patients can be executed by at least one processor, so that the at least one processor executes the steps of the real-time hierarchical management method for the delirium risk of ICU multiple trauma patients as described above.

[0184] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-described example methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases, the former is a better implementation method. Based on such an understanding, the technical solution of the present application, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disc) and includes several instructions for causing a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in various embodiments of the present application.

[0185] Obviously, the above-described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The accompanying drawings show preferred embodiments of the present application, but do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosed content of the present application more thorough and comprehensive. Although the present application has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions described in the foregoing specific embodiments, or perform equivalent replacements for some of the technical features. Any equivalent structure made by using the specification and drawings of the present application, directly or indirectly applied in other related technical fields, is similarly within the scope of the patent protection of the present application.

Claims

1. A real-time hierarchical management method for the delirium risk of ICU multiple trauma patients, characterized in that, Including: Obtain the multimodal data of the patient, preprocess and standardize the multimodal data to obtain the standardized multimodal data; Extract features from the standardized multimodal data to obtain a structured data set of the standardized multimodal data; Input the structured data set into the multi-algorithm parallel model, integrate the output results of each algorithm model based on the soft voting strategy, and take the output result of the model with the best comprehensive score as the prediction result; Based on the prediction result, match the hierarchical level that meets the prediction result; Generate a personalized assisted intervention plan based on the patient's hierarchical level.

2. The real-time hierarchical management method for delirium risk of ICU multiple trauma patients according to claim 1, wherein Preprocess and standardize the multimodal data, including: Remove noise and handle missing values from the obtained multimodal data to obtain preprocessed multimodal data; Standardize the preprocessed multimodal data to obtain the standardized multimodal data.

3. The real-time hierarchical management method for delirium risk of ICU multiple trauma patients according to claim 1, characterized in that, Extract features from the standardized multimodal data to obtain a structured data set of the standardized multimodal data, including: Construct the standardized multimodal data into a comprehensive feature matrix, where each column of the comprehensive feature matrix represents a feature and each row corresponds to the observed data of a patient; Use the Lasso regression algorithm to train the comprehensive feature matrix. Lasso regression reduces the coefficients of unimportant features to zero by adding an L1 regularization term to the loss function, and automatically selects the features that contribute more to delirium prediction; adjust the regularization parameter of Lasso regression through cross-validation to determine the optimal balance point; during the training process, assign a coefficient to each feature, and this coefficient reflects the contribution degree of the feature to the prediction target; By analyzing the feature contribution values of Lasso regression, finally retain the feature set with contribution values greater than the set threshold as the key feature set; Integrate the key feature set into a structured data set in a unified format.

4. The real-time hierarchical management method for delirium risk of ICU multiple trauma patients according to claim 1, characterized in that, Input the structured data set into the multi-algorithm parallel model, and integrate the output results of each model based on the soft voting strategy, including: Obtain the first prediction result based on the random forest; obtain the second prediction result based on logistic regression; obtain the third prediction result based on the support vector machine; obtain the fourth prediction result based on the deep learning model.

5. The real-time hierarchical management method for the delirium risk of ICU multiple trauma patients according to claim 4, characterized in that Obtain the first prediction result based on the random forest, including: Perform prediction by constructing multiple decision trees. Each decision tree randomly selects a part of the data from the structured data set for training through the sampling method with replacement, and randomly selects a subset of features for splitting at each node; When the model training is completed, the structured data set will be independently processed by each decision tree to obtain a prediction result; Integrate the prediction results of all trees through the majority voting mechanism to obtain the final delirium risk score as the first prediction result.

6. The real-time stratified management method for delirium risk of ICU multiple trauma patients according to claim 4, characterized in that, Obtain the second prediction result based on logistic regression, including: The logistic regression model performs a linear combination of the structured data set to obtain a weighted sum. The result of the linear combination is mapped between 0 and 1 through the Sigmoid function and converted into the probability of delirium occurrence, and the probability of delirium occurrence is used as the second prediction result.

7. The real-time stratified management method for delirium risk of ICU multiple trauma patients according to claim 4, wherein Obtain the third prediction result based on the support vector machine, including: Separate the data points of delirium occurrence and non-occurrence based on the structured dataset of historical patients, map the structured dataset of new patients to a high-dimensional space, and the support vector machine obtains the probability estimate of the category to which the new patient belongs based on the distance from the structured dataset of the new patient to the hyperplane, as the third prediction result.

8. The real-time hierarchical management method for delirium risk of ICU multiple trauma patients according to claim 4, characterized in that, Obtain the fourth prediction result based on the deep learning model, including: The structured dataset is processed by a multi-layer neural network. Each layer performs a non-linear transformation on the data through weighted summation, gradually extracting high-level features; Generate a probability value based on the high-level features, indicating the risk of the patient having delirium, as the fourth prediction result.

9. The real-time hierarchical management method for delirium risk of ICU multiple trauma patients according to claim 1, characterized in that, Integrate the output results of each algorithm model based on the soft voting strategy, and use the output result of the model with the best comprehensive score as the prediction result, including: Conduct a comprehensive scoring and quantitative analysis of the performance of the multi-algorithm parallel model through multiple evaluation indicators such as the area under the curve, accuracy, recall rate, precision rate, and F1 score, and select the output result of the model with the best comprehensive score as the delirium risk prediction result.

10. An ICU multiple trauma patient delirium risk real-time stratification management system for implementing the ICU multiple trauma patient delirium risk real-time stratification management method of claims 1-9, characterized in that, Including: A standardized multi-modal data acquisition module for acquiring the multi-modal data of patients, preprocessing and standardizing the multi-modal data to obtain standardized multi-modal data; A structured dataset acquisition module for extracting features from the standardized multi-modal data to obtain the structured dataset of the standardized multi-modal data; A prediction result output module for inputting the structured dataset into the multi-algorithm parallel model, integrating the output results of each algorithm model based on the soft voting strategy, and using the output result of the model with the best comprehensive score as the prediction result; A hierarchical level acquisition module for matching the hierarchical level that meets the prediction result based on the prediction result; An auxiliary intervention plan generation module for generating a personalized auxiliary intervention plan based on the hierarchical level of the patient.

Images