ICU stroke patient nutrition balance real-time evaluation and optimization method based on artificial intelligence

Abstract

The invention relates to the field of nutrition balance evaluation, and discloses an ICU stroke patient nutrition balance real-time evaluation and optimization method based on artificial intelligence, and the method comprises the following steps: collecting the activity energy consumption of a patient through a wearable device, constructing an individualized nutrition metabolism model in combination with the diet intake and metabolism indexes, and carrying out the calculation of the individualized nutrition metabolism model; and calculating the nutrition balance degree in real time and generating a quantitative deviation index. And in combination with a quantitative deviation index and a rehabilitation stage target, generating a nutrition optimization scheme through a reinforcement learning algorithm, performing training on the scheme, and automatically iterating model parameters of an individualized nutrition metabolism model at the same time to achieve the purpose of continuous optimization. According to the invention, the pain points that traditional nutrition evaluation of the ICU stroke patient is lagged and depends on experience decision are overcome, accurate matching of nutrition supply and rehabilitation requirements is realized through an artificial intelligence form, and the residence time of the ICU stroke patient in the ICU is shortened.

Description

Technical Field

[0001] The present invention relates to the field of nutritional balance assessment, and in particular to a real-time assessment and optimization method of nutritional balance in ICU stroke patients based on artificial intelligence. Background Art

[0002] Stroke, also known as "cerebral infarction," is an acute brain injury caused by sudden rupture of a brain blood vessel (hemorrhagic stroke) or blockage, resulting in ischemic and hypoxic brain tissue (ischemic stroke). The core pathology of hemorrhagic stroke is rupture of a cerebral blood vessel, which leads to hematoma compression of brain tissue, while the core pathology of ischemic stroke is thrombosis blocking cerebral blood vessels, which leads to brain cell necrosis. Stroke can cause hemiplegia, impaired time domain and consciousness, and even death. Because stroke can cause metabolic disorders that worsen the condition, and can lead to functional impairments that hinder nutritional intake and nutritional imbalances that lead to a vicious cycle, nutritional balance is essential for ICU stroke patients. Compared with traditional clinical experience-based decision-making methods, artificial intelligence-based nutritional balance has advantages in response speed, individualized accuracy, and multi-indicator collaborative decision-making, which can achieve a reduction in mortality, a sharp decrease in complications, and resource optimization. For example, in a 65-year-old ischemic stroke patient, the AI ​​system, based on an early negative NDI trend (-0.15) and decreased BCAA utilization, adjusted the high-branched-chain amino acid diet 48 hours in advance, preventing 12% muscle loss (as measured by dual-energy X-ray) and ultimately shortening mechanical ventilation by 5 days. Therefore, an AI-based real-time nutritional balance assessment and optimization method for stroke patients in the ICU was proposed. Summary of the Invention

[0003] The present invention overcomes the deficiencies of the prior art and provides a real-time evaluation and optimization method for nutritional balance of ICU stroke patients based on artificial intelligence.

[0004] In order to achieve the above object, the technical solution adopted by the present invention is: The first aspect of the present invention provides a method for real-time evaluation and optimization of nutritional balance in ICU stroke patients based on artificial intelligence, comprising the following steps: Through the wearable activity monitoring device of ICU stroke patients, the personal status data of ICU stroke patients are collected in real time, and the personal status data of ICU stroke patients are combined to obtain the personal status time series data of the target patients; Through the time convolution network, the time series features of the target patient's personal status time series data are extracted, and a metabolic analysis model is constructed based on the time series features. Based on the metabolic analysis model, the real-time nutritional balance of the target patient is calculated, and a quantitative deviation index is generated at the same time; Based on the quantitative deviation index of the target patient, a nutrition optimization plan is generated in combination with the reinforcement learning algorithm.

[0005] Furthermore, in a preferred embodiment of the present invention, the wearable activity monitoring device of the ICU stroke patient is used to collect the personal status data of the ICU stroke patient in real time, and the personal status data of the ICU stroke patient is combined to obtain the personal status time series data of the target patient, specifically: Identify ICU stroke patients who require nutritional balance assessment and label them as target patients. Also obtain wearable devices used to monitor limb activity, heart rate, and exercise frequency in ICU stroke patients and label them as target monitoring devices. The target monitoring device is worn on the target patient and activated, and a standard device monitoring time is preset. During the standard device monitoring time, the target monitoring device collects the target patient's limb activity, heart rate, and exercise frequency in real time; An energy consumption algorithm is set in the target monitoring device. Based on the energy consumption algorithm, the target patient's limb activity, heart rate, and exercise frequency collected in real time are combined to obtain the target patient's real-time activity energy consumption; Obtaining a big data network and determining the target patient's daily meal type, volume, and cooking method, searching a nutrition database within the big data network, wherein the nutrition database includes the protein content, fat content, and carbohydrate content of different daily meal types, at different volumes, and with different cooking methods, collectively referred to as dietary states; Based on the nutrition database, determining the dietary state corresponding to the target patient, marking it as the target dietary state, and obtaining the metabolic indicators of the target patient through the hospital database; The metabolic indicators, dietary status and real-time activity energy consumption of the target patients are aligned in time and space to obtain the target patients' personal status time series data.

[0006] Furthermore, in a preferred embodiment of the present invention, the temporal convolutional network is used to extract the time series features of the target patient's personal status time series data, and a metabolic analysis model is constructed based on the time series features. The real-time nutritional balance of the target patient is calculated based on the metabolic analysis model, and a quantitative deviation index is generated at the same time, specifically: Obtaining a data processing terminal, introducing a temporal convolutional network into the data processing terminal, and importing the target patient's personal status time series data into the data processing terminal; Performing temporal convolution on the target patient's personal state time series data through a temporal convolution network, wherein the temporal convolution is to extract time series features corresponding to different time periods on the data, and generate a diurnal energy consumption fluctuation curve, a postprandial blood glucose response curve, and a nitrogen balance trend curve for the target patient's personal state time series data; The diurnal energy consumption fluctuation curve, postprandial blood glucose response curve, and nitrogen balance trend curve of the target patient's personal state time series data are calibrated as the target patient's personal state time series characteristics; The target patient's personal state time series features are subjected to wavelet denoising and normalization, and the missing values ​​in the target patient's personal state time series features are filled using the multiple interpolation method to obtain the target patient's personal state fusion features; A dual-channel metabolic analysis training model was introduced, and the model parameters of the dual-channel metabolic analysis training model were iteratively optimized through the Bayesian update method to obtain an optimized dual-channel metabolic analysis training model; The target patient's individual status fusion features are input into the optimized dual-channel metabolic analysis training model for training to generate the target patient's real-time nutritional balance; Based on the real-time nutritional balance of the target patient, the quantitative deviation index of the target patient is calculated.

[0007] Furthermore, in a preferred embodiment of the present invention, the dual-channel metabolic analysis training model is introduced, and the model parameters of the dual-channel metabolic analysis training model are iteratively optimized by a Bayesian update method to obtain an optimized dual-channel metabolic analysis training model, specifically: In the dual-channel metabolic training model, a Bayesian optimization framework is constructed, wherein the Bayesian optimization framework sets latent variable samples for different channels of the dual-channel metabolic training model; Energy consumption latent variable samples are set in channel 1 of the dual-channel metabolic training model, and trace element metabolism latent variable samples are set in channel 2 of the dual-channel metabolic training model; In the Bayesian optimization framework, the likelihood functions of the latent variable samples in channel 1 and channel 2 are calculated respectively, calibrated as the target likelihood function, and the dual-channel Bayesian update training is performed in combination with the target likelihood function; Among them, conjugate prior analytical training is performed in channel 1, and approximate Gaussian distribution training is performed in channel 2. After training, the distribution convergence of samples in channel 1 and channel 2 is calculated. When the distribution convergence reaches the preset range, the dual-channel Bayesian update training is stopped, and the optimized dual-channel metabolic analysis training model is output.

[0008] Furthermore, in a preferred embodiment of the present invention, the target patient's individual state fusion features are input into the optimized dual-channel metabolic analysis training model for training to generate the target patient's real-time nutritional balance, specifically: In the optimized dual-channel metabolic analysis training model, channel 1 is the model analysis channel for analyzing macronutrient balance. Channel 1 is connected to the BiLSTM layer in the optimized dual-channel metabolic analysis training model. The BiLSTM layer is used to calculate the real-time basal energy consumption value of the target patient under the current dietary state by combining the fusion features of the target patient's individual state. After obtaining the target patient's basal energy expenditure value from channel 1, the standard basal energy expenditure range of the target patient under the current dietary state is calculated by optimizing the dual-channel metabolic analysis training model; If the real-time basal energy consumption value is greater than the standard basal energy consumption range, the optimized dual-channel metabolic analysis training model outputs that the target patient is at risk of negative balance. If the real-time basal energy consumption value is less than the standard basal energy consumption range, the optimized dual-channel metabolic analysis training model outputs that the target patient is at risk of overfeeding. In the optimized dual-channel metabolic analysis training model, channel 2 is the model analysis channel for performing micro-metabolic analysis on the target patient's individual state fusion characteristics; In channel 2, by analyzing the metabolic index characteristics of the target patient corresponding to the target patient's personal status fusion characteristics, a nutritional metabolic graph is constructed, where the nodes in the nutritional metabolic graph represent the trace elements absorbed by the target patient, and the edges represent the biochemical reaction relationship between different trace elements; Performing iterative updates of the nutritional metabolism graph using a graph neural network. The graph neural network iterative updates the content of the trace elements represented by the nodes in real time based on the biochemical reaction relationships of the edges, and calculates the metabolic efficiency and metabolic pathway of the trace elements based on the content. According to the metabolic efficiency and metabolic pathway of trace elements, the real-time nutritional balance of the target patient is updated in the optimized dual-channel metabolic analysis training model. That is, the real-time nutritional balance of the target patient is calculated by combining the metabolic efficiency and metabolic pathway of trace elements and the real-time basal energy consumption value of the target patient in the current dietary state.

[0009] Furthermore, in a preferred embodiment of the present invention, the quantitative deviation index of the target patient is calculated based on the real-time nutritional balance of the target patient, specifically: In the big data network, obtain the standard nutritional balance corresponding to the target patients at different recovery stages, and calculate the standard nutritional balance weights at different recovery stages; Among them, the different rehabilitation stages of target patients include acute stage, muscle strength reconstruction stage and functional recovery stage; According to the standard nutritional balance weights of different rehabilitation stages and the real-time nutritional balance of the target patient, the difference between the real-time nutritional balance of the target patient at different rehabilitation stages and the standard value is calculated and calibrated as the real-time nutritional balance difference; According to the real-time difference in nutritional balance at different recovery stages, the index for correcting nutritional balance, namely the quantitative deviation index, is calculated.

[0010] Furthermore, in a preferred embodiment of the present invention, the nutritional optimization plan is generated based on the quantitative deviation index of the target patient in combination with a reinforcement learning algorithm, specifically: Within the big data network, the initial nutritional supplement plan required for target patients at different recovery stages is obtained and calibrated as a nutritional optimization plan with a content to be determined; Constructing a reinforcement learning framework, within which, based on the quantitative deviation index of the target patient, content weights are calculated for the nutritional optimization plan with a content to be determined, and a positive and negative analysis of the quantitative deviation index is performed; If the quantitative deviation index is positive, a positive reward mechanism is triggered, wherein the positive reward mechanism is to set priority for the nutritional supplement scheme in the nutritional optimization scheme with the content to be determined; If the quantitative deviation index is negative, a negative penalty mechanism is triggered, wherein the negative penalty mechanism is to set a priority for the nutritional conditioning scheme in the nutritional optimization scheme with a content to be determined; Based on the quantitative deviation index and combined with the content weight calculation of the nutritional optimization plan to be determined, the nutritional optimization plan to be analyzed is output, and the contraindicated nutritional supplements for the target patients are determined through the hospital database. Based on the contraindicated nutritional supplements for the target patients, the corresponding nutritional supplements are eliminated from the nutritional optimization plan to be analyzed to obtain the target nutritional optimization plan.

[0011] A second aspect of the present invention further provides an artificial intelligence-based real-time nutritional balance assessment and optimization system for ICU stroke patients, the assessment and optimization system comprising a memory and a processor, wherein the memory stores an assessment and optimization method, and when the assessment and optimization method is executed by the processor, the following steps are implemented: Through the wearable activity monitoring device of ICU stroke patients, the personal status data of ICU stroke patients are collected in real time, and the personal status data of ICU stroke patients are combined to obtain the personal status time series data of the target patients; Through the time convolution network, the time series features of the target patient's personal status time series data are extracted, and a metabolic analysis model is constructed based on the time series features. Based on the metabolic analysis model, the real-time nutritional balance of the target patient is calculated, and a quantitative deviation index is generated at the same time; Based on the quantitative deviation index of the target patient, a nutrition optimization plan is generated in combination with the reinforcement learning algorithm.

[0012] The present invention solves the technical defects existing in the background technology, and the present invention has the following beneficial effects: the patient's activity energy consumption is collected through wearable devices, and a personalized nutritional metabolism model is constructed in combination with dietary intake and metabolic indicators. The nutritional balance is calculated in real time and a quantitative deviation index is generated. In combination with the quantitative deviation index and the goals of the rehabilitation stage, a nutritional optimization plan is generated through a reinforcement learning algorithm, the plan is trained for execution, and the model parameters of the personalized nutritional metabolism model are automatically iterated to achieve the purpose of continuous optimization. The present invention overcomes the pain points of the lag in traditional nutritional assessment of ICU stroke patients and reliance on experience-based decision-making. Through the form of artificial intelligence, it achieves a precise match between nutritional supply and rehabilitation needs, shortening the stay time of ICU stroke patients in the ICU. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, without paying any creative work, they can also obtain drawings of other embodiments based on these drawings.

[0014] Figure 1 A flowchart showing a method for real-time assessment and optimization of nutritional balance in ICU stroke patients based on artificial intelligence is shown; Figure 2 A flow chart of a method for calculating the real-time nutritional balance and quantitative deviation index of a target patient is shown; Figure 3 A program view of an artificial intelligence-based real-time evaluation and optimization system for nutritional balance of ICU stroke patients is shown. DETAILED DESCRIPTION

[0015] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features therein can be combined with each other.

[0016] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0017] Figure 1 A flowchart of an artificial intelligence-based method for real-time assessment and optimization of nutritional balance in ICU stroke patients is shown, comprising the following steps: S102: Using a wearable activity monitoring device of an ICU stroke patient, personal status data of the ICU stroke patient is collected in real time, and the personal status data of the ICU stroke patient is combined to obtain personal status time series data of the target patient; S104: extracting the time series features of the target patient's personal status time series data through a time convolutional network, building a metabolic analysis model based on the time series features, calculating the target patient's real-time nutritional balance based on the metabolic analysis model, and generating a quantitative deviation index; S106: Generate a nutrition optimization plan based on the quantitative deviation index of the target patient and the reinforcement learning algorithm.

[0018] Furthermore, in a preferred embodiment of the present invention, the wearable activity monitoring device of the ICU stroke patient is used to collect the personal status data of the ICU stroke patient in real time, and the personal status data of the ICU stroke patient is combined to obtain the personal status time series data of the target patient, specifically: Identify ICU stroke patients who require nutritional balance assessment and label them as target patients. Also obtain wearable devices used to monitor limb activity, heart rate, and exercise frequency in ICU stroke patients and label them as target monitoring devices. The target monitoring device is worn on the target patient and activated, and a standard device monitoring time is preset. During the standard device monitoring time, the target monitoring device collects the target patient's limb activity, heart rate, and exercise frequency in real time; An energy consumption algorithm is set in the target monitoring device. Based on the energy consumption algorithm, the target patient's limb activity, heart rate, and exercise frequency collected in real time are combined to obtain the target patient's real-time activity energy consumption; Obtaining a big data network and determining the target patient's daily meal type, volume, and cooking method, searching a nutrition database within the big data network, wherein the nutrition database includes the protein content, fat content, and carbohydrate content of different daily meal types, at different volumes, and with different cooking methods, collectively referred to as dietary states; Based on the nutrition database, determining the dietary state corresponding to the target patient, marking it as the target dietary state, and obtaining the metabolic indicators of the target patient through the hospital database; The metabolic indicators, dietary status and real-time activity energy consumption of the target patients are aligned in time and space to obtain the target patients' personal status time series data.

[0019] It should be noted that stroke patients need to calculate their energy expenditure based on their daily activity level, and combine dietary status and metabolic indicators to determine the nutritional supplements they need and whether their current nutritional intake is excessive or insufficient. Therefore, the target monitoring device collects the target patient's limb activity, heart rate, and exercise frequency in real time to calculate energy consumption. The target patient's daily meal type, volume, and cooking method are determined to calculate dietary status. Finally, daily real-time metabolic indicators are recorded in the hospital database, including 15 metabolic indicators such as blood glucose, prealbumin, and urea nitrogen, to build a patient-specific biochemical time series database.

[0020] Furthermore, in a preferred embodiment of the present invention, the nutritional optimization plan is generated based on the quantitative deviation index of the target patient in combination with a reinforcement learning algorithm, specifically: Within the big data network, the initial nutritional supplement plan required for target patients at different recovery stages is obtained and calibrated as a nutritional optimization plan with a content to be determined; Constructing a reinforcement learning framework, within which, based on the quantitative deviation index of the target patient, content weights are calculated for the nutritional optimization plan with a content to be determined, and a positive and negative analysis of the quantitative deviation index is performed; If the quantitative deviation index is positive, a positive reward mechanism is triggered, wherein the positive reward mechanism is to set priority for the nutritional supplement scheme in the nutritional optimization scheme with the content to be determined; If the quantitative deviation index is negative, a negative penalty mechanism is triggered, wherein the negative penalty mechanism is to set a priority for the nutritional conditioning scheme in the nutritional optimization scheme with a content to be determined; Based on the quantitative deviation index and combined with the content weight calculation of the nutritional optimization plan to be determined, the nutritional optimization plan to be analyzed is output, and the contraindicated nutritional supplements for the target patients are determined through the hospital database. Based on the contraindicated nutritional supplements for the target patients, the corresponding nutritional supplements are eliminated from the nutritional optimization plan to be analyzed to obtain the target nutritional optimization plan.

[0021] It should be noted that the quantitative deviation index is currently known. Since the quantitative deviation index is used to provide conditions for accurately adjusting the nutritional content within the plan during the output of the nutritional balance plan for the patient, the nutritional optimization plan with a to-be-determined content is first obtained. The nutritional optimization plan with a to-be-determined content only contains nutrients that the patient needs to supplement, and the content needs to be determined based on the quantitative deviation index. Among them, the quantitative deviation index includes positive and negative values. In the positive state, the patient needs to be nutritionally supplemented, that is, the current nutritional supplementation of the patient is insufficient, and nutritional supplementation needs to be given priority; on the contrary, the negative state requires nutritional conditioning, that is, the current nutrition is excessive, and it is necessary to supplement the known nutrients in batches or even excessively. In the process of outputting the nutritional optimization plan, it is also necessary to determine the patient's contraindications to nutrition to prevent the patient from being in danger, and finally output the target nutritional optimization plan.

[0022] Figure 2 A flow chart of a method for calculating the real-time nutritional balance and quantitative deviation index of a target patient is shown, comprising the following steps: S202: extracting time series features of the target patient's personal status time series data through a time convolutional network, building a metabolic analysis model based on the time series features, calculating the target patient's real-time nutritional balance based on the metabolic analysis model, and generating a quantitative deviation index; S204: introducing a dual-channel metabolic analysis training model, and iteratively optimizing model parameters of the dual-channel metabolic analysis training model through a Bayesian update method to obtain an optimized dual-channel metabolic analysis training model; S206: Inputting the target patient's individual state fusion features into the optimized dual-channel metabolic analysis training model for training to generate the target patient's real-time nutritional balance; S208: Based on the real-time nutritional balance of the target patient, a quantitative deviation index of the target patient is calculated.

[0023] Furthermore, in a preferred embodiment of the present invention, the temporal convolutional network is used to extract the time series features of the target patient's personal status time series data, and a metabolic analysis model is constructed based on the time series features. The real-time nutritional balance of the target patient is calculated based on the metabolic analysis model, and a quantitative deviation index is generated at the same time, specifically: Obtaining a data processing terminal, introducing a temporal convolutional network into the data processing terminal, and importing the target patient's personal status time series data into the data processing terminal; Performing temporal convolution on the target patient's personal state time series data through a temporal convolution network, wherein the temporal convolution is to extract time series features corresponding to different time periods on the data, and generate a diurnal energy consumption fluctuation curve, a postprandial blood glucose response curve, and a nitrogen balance trend curve for the target patient's personal state time series data; The diurnal energy consumption fluctuation curve, postprandial blood glucose response curve, and nitrogen balance trend curve of the target patient's personal state time series data are calibrated as the target patient's personal state time series characteristics; The target patient's personal state time series characteristics are subjected to wavelet denoising and normalization, and the missing values ​​in the target patient's personal state time series characteristics are filled by the multiple interpolation method to obtain the target patient's personal state fusion characteristics.

[0024] It should be noted that the data processing terminal is a modeling terminal device used to build a model for analyzing the patient's personal status data. First, the time series features of the target patient's personal status time series data are extracted through the time convolution network to build a metabolic analysis model. Modeling requires feature data training, and the extracted time series features are the diurnal energy consumption fluctuation curve, the postprandial blood glucose response curve, and the nitrogen balance trend curve. The extracted features may have fluctuation noise points, etc., and need to be preprocessed, including denoising and missing value interpolation filling. Therefore, the target patient's personal status time series features are subjected to wavelet denoising and normalization, and the multiple interpolation method is used to fill the missing values ​​in the target patient's personal status time series features to obtain the target patient's personal status fusion features.

[0025] Furthermore, in a preferred embodiment of the present invention, the dual-channel metabolic analysis training model is introduced, and the model parameters of the dual-channel metabolic analysis training model are iteratively optimized by a Bayesian update method to obtain an optimized dual-channel metabolic analysis training model, specifically: In the dual-channel metabolic training model, a Bayesian optimization framework is constructed, wherein the Bayesian optimization framework sets latent variable samples for different channels of the dual-channel metabolic training model; Energy consumption latent variable samples are set in channel 1 of the dual-channel metabolic training model, and trace element metabolism latent variable samples are set in channel 2 of the dual-channel metabolic training model; In the Bayesian optimization framework, the likelihood functions of the latent variable samples in channel 1 and channel 2 are calculated respectively, calibrated as the target likelihood function, and the dual-channel Bayesian update training is performed in combination with the target likelihood function; Among them, conjugate prior analytical training is performed in channel 1, and approximate Gaussian distribution training is performed in channel 2. After training, the distribution convergence of samples in channel 1 and channel 2 is calculated. When the distribution convergence reaches the preset range, the dual-channel Bayesian update training is stopped, and the optimized dual-channel metabolic analysis training model is output.

[0026] It should be noted that the Bayesian optimization framework is a framework that uses the Bayesian algorithm to update model parameters. The purpose of the Bayesian algorithm is to enhance the calibration ability of key parameters, such as trace element metabolism, nutrient consumption and other latent variables. By setting sample data, the model is Bayesian optimized to ensure that the analysis results of the model in the subsequent dual-channel metabolic analysis do not show excessive deviations, thereby enhancing accuracy. Different channels represent different variables, and the likelihood function of the latent variable samples in channel 1 and channel 2 is calculated. The likelihood function is the conditional function for Bayesian update, and the dual-channel independence training is divided into conjugate prior analytical training and Gaussian distribution training. The purpose is to improve the accuracy of patient nutritional balance assessment in clinical practice.

[0027] Furthermore, in a preferred embodiment of the present invention, the target patient's individual state fusion features are input into the optimized dual-channel metabolic analysis training model for training to generate the target patient's real-time nutritional balance, specifically: In the optimized dual-channel metabolic analysis training model, channel 1 is the model analysis channel for analyzing macronutrient balance. Channel 1 is connected to the BiLSTM layer in the optimized dual-channel metabolic analysis training model. The BiLSTM layer is used to calculate the real-time basal energy consumption value of the target patient under the current dietary state by combining the fusion features of the target patient's individual state. After obtaining the target patient's basal energy expenditure value from channel 1, the standard basal energy expenditure range of the target patient under the current dietary state is calculated by optimizing the dual-channel metabolic analysis training model; If the real-time basal energy consumption value is greater than the standard basal energy consumption range, the optimized dual-channel metabolic analysis training model outputs that the target patient is at risk of negative balance. If the real-time basal energy consumption value is less than the standard basal energy consumption range, the optimized dual-channel metabolic analysis training model outputs that the target patient is at risk of overfeeding. In the optimized dual-channel metabolic analysis training model, channel 2 is the model analysis channel for performing micro-metabolic analysis on the target patient's individual state fusion characteristics; In channel 2, by analyzing the metabolic index characteristics of the target patient corresponding to the target patient's personal status fusion characteristics, a nutritional metabolic graph is constructed, where the nodes in the nutritional metabolic graph represent the trace elements absorbed by the target patient, and the edges represent the biochemical reaction relationship between different trace elements; Performing iterative updates of the nutritional metabolism graph using a graph neural network. The graph neural network iterative updates the content of the trace elements represented by the nodes in real time based on the biochemical reaction relationships of the edges, and calculates the metabolic efficiency and metabolic pathway of the trace elements based on the content. According to the metabolic efficiency and metabolic pathway of trace elements, the real-time nutritional balance of the target patient is updated in the optimized dual-channel metabolic analysis training model. That is, the real-time nutritional balance of the target patient is calculated by combining the metabolic efficiency and metabolic pathway of trace elements and the real-time basal energy consumption value of the target patient in the current dietary state.

[0028] It should be noted that channel 1 is a model analysis channel for macronutrient balance analysis. This channel analyzes the patient's energy intake from dietary status and the energy expended through activity. Channel 1 connects to the BiLSTM layer within the optimized dual-channel metabolic analysis training model. This BiLSTM layer combines the fused features of the target patient's individual state to calculate the target patient's real-time basal energy expenditure under the current dietary status. Energy consumption determines the patient's current state, which can be categorized as either overfeeding or negative balance. Different states require different patient management approaches. For example, an overfeeding state requires discontinuing high-nutrient intake, while a negative balance indicates insufficient energy intake and requires increased energy intake. Channel 2 is a model analysis channel for micronutrient metabolism analysis based on the fused features of the target patient's individual state. In the human body, the levels of trace elements consumed can affect health. Trace elements include, but are not limited to, leucine, glutamine, and vitamin D. Using graph neural network analysis, we can generate metabolic efficiency and metabolic pathway analysis within the patient, thereby providing a secondary update of the patient's current nutritional balance. This combines energy intake and consumption with trace element metabolism to accurately determine the patient's nutritional balance.

[0029] Furthermore, in a preferred embodiment of the present invention, the quantitative deviation index of the target patient is calculated based on the real-time nutritional balance of the target patient, specifically: In the big data network, obtain the standard nutritional balance corresponding to the target patients at different recovery stages, and calculate the standard nutritional balance weights at different recovery stages; Among them, the different rehabilitation stages of target patients include acute stage, muscle strength reconstruction stage and functional recovery stage; According to the standard nutritional balance weights of different rehabilitation stages and the real-time nutritional balance of the target patient, the difference between the real-time nutritional balance of the target patient at different rehabilitation stages and the standard value is calculated and calibrated as the real-time nutritional balance difference; According to the real-time difference in nutritional balance at different recovery stages, the index for correcting nutritional balance, namely the quantitative deviation index, is calculated.

[0030] It should be noted that patients have different stages of recovery, and the corresponding nutritional balance varies at different stages, so standard nutritional balance weights are calculated for different stages. By mapping the target patient's real-time nutritional balance onto the standard nutritional balance weights at different stages, the difference between the patient's current nutritional balance and the standard value at that stage can be immediately determined. A quantitative deviation index can also be obtained, which is used to accurately adjust the nutritional content within the nutritional balance plan when outputting the plan for the patient.

[0031] like Figure 3As shown, the second aspect of the present invention further provides an artificial intelligence-based real-time evaluation and optimization system for nutritional balance of ICU stroke patients, wherein the evaluation and optimization system includes a memory 31 and a processor 32. The memory 31 stores an evaluation and optimization method. When the evaluation and optimization method is executed by the processor 32, the following steps are implemented: Through the wearable activity monitoring device of ICU stroke patients, the personal status data of ICU stroke patients are collected in real time, and the personal status data of ICU stroke patients are combined to obtain the personal status time series data of the target patients; Through the time convolution network, the time series features of the target patient's personal status time series data are extracted, and a metabolic analysis model is constructed based on the time series features. Based on the metabolic analysis model, the real-time nutritional balance of the target patient is calculated, and a quantitative deviation index is generated at the same time; Based on the quantitative deviation index of the target patient, a nutrition optimization plan is generated in combination with the reinforcement learning algorithm.

[0032] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. An artificial intelligence-based real-time nutritional balance assessment and optimization method for ICU stroke patients, characterized by: The following steps are involved: Through the wearable activity monitoring device of ICU stroke patients, the personal status data of ICU stroke patients are collected in real time, and the personal status data of ICU stroke patients are combined to obtain the personal status time series data of the target patients; Through the time convolution network, the time series features of the target patient's personal status time series data are extracted, and a metabolic analysis model is constructed based on the time series features. Based on the metabolic analysis model, the real-time nutritional balance of the target patient is calculated, and a quantitative deviation index is generated at the same time; Based on the quantitative deviation index of the target patient, a nutrition optimization plan is generated in combination with the reinforcement learning algorithm.

2. The method for real-time evaluation and optimization of nutritional balance in ICU stroke patients based on artificial intelligence according to claim 1, characterized in that: The wearable activity monitoring device of the ICU stroke patient is used to collect the personal status data of the ICU stroke patient in real time, and the personal status data of the ICU stroke patient is combined to obtain the personal status time series data of the target patient, specifically: Identify ICU stroke patients who require nutritional balance assessment and label them as target patients. Also obtain wearable devices used to monitor limb activity, heart rate, and exercise frequency in ICU stroke patients and label them as target monitoring devices. The target monitoring device is worn on the target patient and activated, and a standard device monitoring time is preset. During the standard device monitoring time, the target monitoring device collects the target patient's limb activity, heart rate, and exercise frequency in real time; An energy consumption algorithm is set in the target monitoring device. Based on the energy consumption algorithm, the target patient's limb activity, heart rate, and exercise frequency collected in real time are combined to obtain the target patient's real-time activity energy consumption; Obtaining a big data network and determining the target patient's daily meal type, volume, and cooking method, searching a nutrition database within the big data network, wherein the nutrition database includes the protein content, fat content, and carbohydrate content of different daily meal types, at different volumes, and with different cooking methods, collectively referred to as dietary states; Based on the nutrition database, determining the dietary state corresponding to the target patient, marking it as the target dietary state, and obtaining the metabolic indicators of the target patient through the hospital database; The metabolic indicators, dietary status and real-time activity energy consumption of the target patients are aligned in time and space to obtain the target patients' personal status time series data.

3. The method for real-time evaluation and optimization of nutritional balance in ICU stroke patients based on artificial intelligence according to claim 1, characterized in that: The temporal convolutional network is used to extract the temporal features of the target patient's personal status time series data, and a metabolic analysis model is constructed based on the temporal features. The real-time nutritional balance of the target patient is calculated based on the metabolic analysis model, and a quantitative deviation index is generated at the same time, specifically: Obtaining a data processing terminal, introducing a temporal convolutional network into the data processing terminal, and importing the target patient's personal status time series data into the data processing terminal; Performing temporal convolution on the target patient's personal state time series data through a temporal convolution network, wherein the temporal convolution is to extract time series features corresponding to different time periods on the data, and generate a diurnal energy consumption fluctuation curve, a postprandial blood glucose response curve, and a nitrogen balance trend curve for the target patient's personal state time series data; The diurnal energy consumption fluctuation curve, postprandial blood glucose response curve, and nitrogen balance trend curve of the target patient's personal state time series data are calibrated as the target patient's personal state time series characteristics; The target patient's personal state time series features are subjected to wavelet denoising and normalization, and the missing values ​​in the target patient's personal state time series features are filled using the multiple interpolation method to obtain the target patient's personal state fusion features; A dual-channel metabolic analysis training model was introduced, and the model parameters of the dual-channel metabolic analysis training model were iteratively optimized through the Bayesian update method to obtain an optimized dual-channel metabolic analysis training model; The target patient's individual status fusion features are input into the optimized dual-channel metabolic analysis training model for training to generate the target patient's real-time nutritional balance; Based on the real-time nutritional balance of the target patient, the quantitative deviation index of the target patient is calculated.

4. The method for real-time evaluation and optimization of nutritional balance in ICU stroke patients based on artificial intelligence according to claim 3, characterized in that: The dual-channel metabolic analysis training model is introduced, and the model parameters of the dual-channel metabolic analysis training model are iteratively optimized by the Bayesian update method to obtain the optimized dual-channel metabolic analysis training model, specifically: In the dual-channel metabolic training model, a Bayesian optimization framework is constructed, wherein the Bayesian optimization framework sets latent variable samples for different channels of the dual-channel metabolic training model; Energy consumption latent variable samples are set in channel 1 of the dual-channel metabolic training model, and trace element metabolism latent variable samples are set in channel 2 of the dual-channel metabolic training model; In the Bayesian optimization framework, the likelihood functions of the latent variable samples in channel 1 and channel 2 are calculated respectively, calibrated as the target likelihood function, and the dual-channel Bayesian update training is performed in combination with the target likelihood function; Among them, conjugate prior analytical training is performed in channel 1, and approximate Gaussian distribution training is performed in channel 2. After training, the distribution convergence of samples in channel 1 and channel 2 is calculated. When the distribution convergence reaches the preset range, the dual-channel Bayesian update training is stopped, and the optimized dual-channel metabolic analysis training model is output.

5. The method for real-time evaluation and optimization of nutritional balance in ICU stroke patients based on artificial intelligence according to claim 3, characterized in that: The target patient's individual state fusion features are input into the optimized dual-channel metabolic analysis training model for training to generate the target patient's real-time nutritional balance, specifically: In the optimized dual-channel metabolic analysis training model, channel 1 is the model analysis channel for analyzing macronutrient balance. Channel 1 is connected to the BiLSTM layer in the optimized dual-channel metabolic analysis training model. The BiLSTM layer is used to calculate the real-time basal energy consumption value of the target patient under the current dietary state by combining the fusion features of the target patient's individual state. After obtaining the target patient's basal energy expenditure value from channel 1, the standard basal energy expenditure range of the target patient under the current dietary state is calculated by optimizing the dual-channel metabolic analysis training model; If the real-time basal energy consumption value is greater than the standard basal energy consumption range, the optimized dual-channel metabolic analysis training model outputs that the target patient is at risk of negative balance. If the real-time basal energy consumption value is less than the standard basal energy consumption range, the optimized dual-channel metabolic analysis training model outputs that the target patient is at risk of overfeeding. In the optimized dual-channel metabolic analysis training model, channel 2 is the model analysis channel for performing micro-metabolic analysis on the target patient's individual state fusion characteristics; In channel 2, by analyzing the metabolic index characteristics of the target patient corresponding to the target patient's personal status fusion characteristics, a nutritional metabolic graph is constructed, where the nodes in the nutritional metabolic graph represent the trace elements absorbed by the target patient, and the edges represent the biochemical reaction relationship between different trace elements; Performing iterative updates of the nutritional metabolism graph using a graph neural network. The graph neural network iterative updates the content of the trace elements represented by the nodes in real time based on the biochemical reaction relationships of the edges, and calculates the metabolic efficiency and metabolic pathway of the trace elements based on the content. According to the metabolic efficiency and metabolic pathway of trace elements, the real-time nutritional balance of the target patient is updated in the optimized dual-channel metabolic analysis training model. That is, the real-time nutritional balance of the target patient is calculated by combining the metabolic efficiency and metabolic pathway of trace elements and the real-time basal energy consumption value of the target patient in the current dietary state.

6. The method for real-time evaluation and optimization of nutritional balance in ICU stroke patients based on artificial intelligence according to claim 3, characterized in that: The quantitative deviation index of the target patient is calculated based on the real-time nutritional balance of the target patient, specifically: In the big data network, obtain the standard nutritional balance corresponding to the target patients at different recovery stages, and calculate the standard nutritional balance weights at different recovery stages; Among them, the different rehabilitation stages of target patients include acute stage, muscle strength reconstruction stage and functional recovery stage; According to the standard nutritional balance weights of different rehabilitation stages and the real-time nutritional balance of the target patient, the difference between the real-time nutritional balance of the target patient at different rehabilitation stages and the standard value is calculated and calibrated as the real-time nutritional balance difference; According to the real-time difference in nutritional balance at different recovery stages, the index for correcting nutritional balance, namely the quantitative deviation index, is calculated.

7. The method for real-time evaluation and optimization of nutritional balance in ICU stroke patients based on artificial intelligence according to claim 1, characterized in that: The nutritional optimization plan is generated based on the quantitative deviation index of the target patient and combined with the reinforcement learning algorithm, specifically: Within the big data network, the initial nutritional supplement plan required for target patients at different recovery stages is obtained and calibrated as a nutritional optimization plan with a content to be determined; Constructing a reinforcement learning framework, within which, based on the quantitative deviation index of the target patient, content weights are calculated for the nutritional optimization plan with a content to be determined, and a positive and negative analysis of the quantitative deviation index is performed; If the quantitative deviation index is positive, a positive reward mechanism is triggered, wherein the positive reward mechanism is to set priority for the nutritional supplement scheme in the nutritional optimization scheme with the content to be determined; If the quantitative deviation index is negative, a negative penalty mechanism is triggered, wherein the negative penalty mechanism is to set a priority for the nutritional conditioning scheme in the nutritional optimization scheme with a content to be determined; Based on the quantitative deviation index and combined with the content weight calculation of the nutritional optimization plan to be determined, the nutritional optimization plan to be analyzed is output, and the contraindicated nutritional supplements for the target patients are determined through the hospital database. Based on the contraindicated nutritional supplements for the target patients, the corresponding nutritional supplements are eliminated from the nutritional optimization plan to be analyzed to obtain the target nutritional optimization plan.

8. An artificial intelligence-based real-time nutritional balance assessment and optimization system for ICU stroke patients, characterized by: The evaluation and optimization system includes a memory and a processor, wherein the memory stores an evaluation and optimization method program. When the evaluation and optimization method program is executed by the processor, the evaluation and optimization method steps as described in any one of claims 1 to 7 are implemented.

Images