Old people nutrition and health state evaluation system and method based on deep learning

Abstract

The invention relates to the technical field of nutrition and health status assessment, and particularly discloses a deep learning-based old people nutrition and health status assessment system and method, and the method comprises the steps: obtaining all nutrition and health status assessment indication information of a to-be-assessed old people object; a data processing and analysis algorithm based on artificial intelligence and deep learning is introduced to a back end to perform embedded semantic aggregation analysis on the various nutrition and health state assessment indication information, and then state assessment is performed based on the health state comprehensive feature representation of the elderly object to determine a state assessment result. Therefore, the quality of the nutrition and health state evaluation of the old people can be improved through a more intelligent data processing mode, and the score is not based on a fixed threshold but is automatically generated through a model, so that the difference between different regions or groups can be more flexibly reflected, meanwhile, the biological variation between individuals is also considered, and the evaluation accuracy is improved. Therefore, personalized risk prediction in a real sense is realized.

Description

Technical Field

[0001] This application relates to the technical field of nutritional and health status assessment, and more specifically, to a system and method for assessing the nutritional and health status of the elderly based on deep learning. Background Art

[0002] With the intensification of the global population aging, the assessment of the health and nutritional status of the elderly has become increasingly important. Traditional methods for assessing the health of the elderly mainly rely on manual assessments by doctors or professionals. This method is not only time-consuming and laborious, but also may lead to inaccurate assessment results due to subjective factors. In addition, the health status of the elderly is affected by multiple factors, including but not limited to age, gender, lifestyle, dietary structure, activity level, previous medical history, etc. These complex and variable factors increase the difficulty of assessment.

[0003] In response to the above problems, Chinese Patent CN114974570A proposes a system for assessing the nutritional and health status of the elderly and predicting risks based on machine learning. It collects and analyzes multi-dimensional health data of the elderly (such as basic information, body function status, dietary information, etc.), uses the CatBoost algorithm to build an assessment model, and provides personalized suggestions in combination with the SHAP algorithm. However, during the risk prediction process, the population division only relies on preset fixed thresholds. Such static threshold setting may not be able to adapt to the differences between different regions or groups, nor fully consider the biological variations between individuals, resulting in inaccurate and non-personalized assessment results and affecting the accuracy of risk prediction. Moreover, in the above-mentioned system for assessing the nutritional and health status of the elderly and predicting risks, traditional feature selection and preprocessing methods are adopted (such as K-nearest neighbor for filling missing values, sequential floating backward selection algorithm, etc.). These methods are not effective in dealing with complex data and are difficult to capture the deep features and complex relationships between data. In addition, this solution mainly relies on a single boosting tree model, CatBoost, and tunes parameters through grid search. Although CatBoost has excellent performance, the flexibility of this method is limited. Especially when facing different types of data, it does not have the ability of non-linear fitting and adaptive adjustment.

[0004] Therefore, an optimized system for assessing the nutritional and health status of the elderly based on machine learning is desired. Summary of the Invention

[0005] The present application provides a system and method for evaluating the nutritional and health status of the elderly based on deep learning, which can improve the quality of the evaluation of the nutritional and health status of the elderly through a more intelligent data processing method. Such a score is not based on a fixed threshold, but is automatically generated by a model, so it can more flexibly reflect the differences between different regions or groups, and at the same time takes into account the biological variations between individuals, so as to achieve a truly personalized risk prediction.

[0006] In a first aspect, there is provided a system for evaluating the nutritional and health status of the elderly based on deep learning, comprising:

[0007] An information acquisition module for the nutritional and health status of the elderly, configured to acquire various nutritional and health status evaluation indication information of an elderly object to be evaluated, where the nutritional and health status evaluation indication information includes basic information, matrix function status information, dietary information, quality of life information, biomarker - plasma biochemical examination information, biomarker - urine index information, biomarker - blood routine examination information;

[0008] An embedding and encoding module for the nutritional and health information, configured to perform embedding and encoding on the various nutritional and health status evaluation indication information to obtain a set of embedding and encoding features of the nutritional and health status evaluation indication information;

[0009] A dynamic compensation and aggregation analysis module for the nutritional and health features, configured to perform feature dynamic compensation and aggregation analysis on the set of embedding and encoding features of the nutritional and health status evaluation indication information to obtain significantly aggregated encoding features of the nutritional and health status evaluation indication information. Among them, the dynamic compensation and aggregation analysis module for the nutritional and health features includes: a node - hub complementary information calculation unit, configured to calculate the sequence hub features of the set of embedding and encoding features of the nutritional and health status evaluation indication information, and calculate the sequence of node - hub complementary information embedding and encoding features of the nutritional and health status evaluation indication information based on the sequence hub features; a compensation and aggregation analysis processing unit, configured to perform compensation and aggregation analysis based on an attention mechanism on the sequence of node - hub complementary information embedding and encoding features of the nutritional and health status evaluation indication information to obtain the significantly aggregated encoding features of the nutritional and health status evaluation indication information;

[0010] An elderly individual type label determination module, configured to determine a status evaluation result based on the significantly aggregated encoding features of the nutritional and health status evaluation indication information, where the status evaluation result is used to represent the individual type label of the elderly object to be evaluated, and the individual type label includes a healthy population, a low - risk population, a medium - risk population, or a high - risk population.

[0011] In a second aspect, there is provided a method for evaluating the nutritional and health status of the elderly based on deep learning, comprising:

[0012] Obtain the information of various nutritional and health status assessment indicators of the elderly object to be evaluated, where the nutritional and health status assessment indicator information includes basic information, matrix function status information, dietary information, quality of life information, biomarker - plasma biochemical examination information, biomarker - urine index information, biomarker - blood routine examination information;

[0013] Perform embedding coding on the above-mentioned various nutritional and health status assessment indicator information to obtain a set of embedded coding features of the nutritional and health status assessment indicator information;

[0014] Perform feature dynamic compensation aggregation analysis on the set of embedded coding features of the nutritional and health status assessment indicator information to obtain significantly aggregated coding features of the nutritional and health status assessment indicator information, including: calculating the sequence hub features of the set of embedded coding features of the nutritional and health status assessment indicator information, and calculating the sequence of node - hub complementary information embedded coding features of the nutritional and health status assessment indicator information based on the sequence hub features; performing compensation aggregation analysis based on the attention mechanism on the sequence of node - hub complementary information embedded coding features of the nutritional and health status assessment indicator information to obtain the significantly aggregated coding features of the nutritional and health status assessment indicator information;

[0015] Based on the significantly aggregated coding features of the nutritional and health status assessment indicator information, determine the status assessment result, where the status assessment result is used to represent the individual type label of the elderly object to be evaluated, and the individual type label includes healthy population, low - risk population, medium - risk population or high - risk population.

[0016] A system and method for evaluating the nutritional and health status of the elderly based on deep learning provided by this application can improve the quality of the evaluation of the nutritional and health status of the elderly through a more intelligent data - processing method. Moreover, such a score is not based on a fixed threshold but is automatically generated by the model. Therefore, it can more flexibly reflect the differences between different regions or groups, and at the same time takes into account the biological variations between individuals, thus realizing personalized risk prediction in the true sense. In addition, through this method, it can have stronger non - linear fitting ability and adaptive adjustment ability. This means that it can process more types of data, reduce the possibility of human error, and also maintain good performance when facing new situations. This not only improves the reliability of the evaluation results, provides a scientific basis for subsequent medical decisions, and ultimately is expected to improve the quality of life and health management level of the elderly. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of this application, the drawings of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below only relate to some embodiments of this application and do not limit this application.

[0018] Figure 1 It is a schematic block diagram of the elderly nutrition and health status evaluation system based on deep learning according to an embodiment of the present application.

[0019] Figure 2 It is a schematic diagram of data flow of the elderly nutrition and health status evaluation system based on deep learning according to an embodiment of the present application.

[0020] Figure 3 It is a schematic block diagram of the nutrition and health feature dynamic compensation aggregation analysis module in the elderly nutrition and health status evaluation system based on deep learning according to an embodiment of the present application.

[0021] Figure 4 It is a schematic block diagram of the compensation aggregation analysis processing unit in the elderly nutrition and health status evaluation system based on deep learning according to an embodiment of the present application.

[0022] Figure 5 It is a schematic flowchart of the elderly nutrition and health status evaluation method based on deep learning according to an embodiment of the present application. Detailed implementation manners

[0023] Next, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts also belong to the scope of protection of the present application.

[0024] In view of the above technical problems, in the technical solution of the present application, a system for evaluating the nutritional health status of the elderly based on deep learning is proposed, which aims to overcome the limitations of traditional evaluation methods and comprehensively and accurately evaluate the nutritional health status of the elderly by using advanced deep learning algorithms and artificial intelligence technologies. Specifically, the system will collect data in multiple aspects such as the basic information, physical function status, eating habits, quality of life, and biomarkers (such as plasma biochemical examinations, urine indicators, blood routine examinations, etc.) of the elderly to be evaluated. Then, by introducing data processing and analysis algorithms based on artificial intelligence and deep learning at the backend, embedded semantic aggregation analysis is performed on these various nutritional health status evaluation indication information to obtain a comprehensive feature representation that can reflect the individual's health status. Furthermore, based on the comprehensive feature representation of the health status of the elderly object, a status evaluation is carried out to determine which risk level the elderly belong to: healthy population, low-risk population, medium-risk population, or high-risk population. In this way, the quality of the evaluation of the nutritional health status of the elderly can be improved through a more intelligent data processing method, and such a score is not based on a fixed threshold but is automatically generated by the model, so it can more flexibly reflect the differences between different regions or groups, and at the same time take into account the biological variations between individuals, thus realizing personalized risk prediction in the true sense. In addition, through this method, it can have stronger non-linear fitting ability and adaptive adjustment ability. This means that it can process more types of data, reduce the possibility of human error, and also maintain good performance when facing new situations. This not only improves the reliability of the evaluation results, provides a scientific basis for subsequent medical decisions, and ultimately is expected to improve the quality of life and health management level of the elderly.

[0025] Specifically, in the technical solution of the present application, as Figure 1 and Figure 2As shown, the deep learning-based elderly nutrition and health status assessment system includes: an elderly nutrition and health information acquisition module 10, which is used to acquire various nutrition and health status assessment indication information of the elderly object to be evaluated, and the nutrition and health status assessment indication information includes basic information, matrix function status information, dietary information, quality of life information, biomarker - plasma biochemical examination information, biomarker - urine index information, biomarker - blood routine examination information; a nutrition and health information embedding and encoding module 20, which is used to perform embedding and encoding on the various nutrition and health status assessment indication information to obtain a set of nutrition and health status assessment indication information embedding and encoding features; a nutrition and health feature dynamic compensation and aggregation analysis module 30, which is used to perform feature dynamic compensation and aggregation analysis on the set of nutrition and health status assessment indication information embedding and encoding features to obtain significant aggregation encoding features of nutrition and health status assessment indication information; an elderly individual type label determination module 40, which is used to determine a status assessment result based on the significant aggregation encoding features of the nutrition and health status assessment indication information, and the status assessment result is used to represent the individual type label of the elderly object to be evaluated, and the individual type label includes healthy population, low-risk population, medium-risk population or high-risk population.

[0026] Exemplarily, in the elderly nutrition and health information acquisition module 10, various nutrition and health status assessment indication information of the elderly object to be evaluated is acquired, and the nutrition and health status assessment indication information includes basic information, matrix function status information, dietary information, quality of life information, biomarker - plasma biochemical examination information, biomarker - urine index information, biomarker - blood routine examination information. It should be understood that the basic information, matrix function status information, dietary information, quality of life information, biomarkers such as plasma biochemical examination information, urine index information, and blood routine examination information together constitute a comprehensive perspective to understand the health status of the elderly. Each aspect provides data on different dimensions of an individual's health and is crucial for accurately assessing the nutrition and health status of the elderly. For example, the basic information may include factors such as age and gender that have a direct impact on health; the matrix function status information can reflect the basic operating ability of the body; the dietary information reveals eating habits and nutritional intake; the quality of life information can reflect the impact of an individual's living conditions and social support system; and the biomarker information provides specific data on the health status from a physiological level. By comprehensively analyzing this multi-source heterogeneous data, a solid foundation can be provided for subsequent risk assessment, thereby achieving personalized and precise health management.

[0027] Furthermore, in an embodiment of the present application, first, basic information such as age and gender can be directly extracted from the elderly's identity documents or medical records. Anthropometric information, such as height, weight, calf circumference, distance from the right knee to the ground, and waist circumference, needs to be accurately measured and recorded by professionals using standard equipment during medical institutions or community health examinations. Next, information on the body's functional status, such as the ability to perform daily activities like assisted bathing, dressing, and excretion, can be collected by designing detailed questionnaires. These questionnaires should be filled out by the elderly themselves or their family members, and can also be supplemented by professionals' actual observations and evaluations of the elderly's daily living activities. At the same time, for the ability to independently visit neighbors, go shopping, cook, do laundry, etc., detailed information can also be obtained through a combination of questionnaires and interviews. Dietary information, including the frequency of fresh fruit consumption, fresh vegetable consumption, meat consumption, etc., can be obtained by asking the elderly to recall and record their dietary situation over a past period through the 24-hour dietary recall method or a food frequency questionnaire (FFQ). In addition, regular dietary diaries can be combined to more accurately understand the elderly's long-term dietary patterns. This not only helps capture the elderly's eating habits but also provides an important basis for subsequent nutritional analysis. Information on quality of life, such as self-evaluation of personal quality of life and self-evaluation of personal health status, can be completed through standardized questionnaires. Tools such as the SF-36 Health Survey Short Form or other similar tools can be used to let the elderly self-evaluate their quality of life and health status, thereby obtaining subjective but valuable feedback. As for biomarker information, this part relies on laboratory tests. Plasma biochemical examination information, such as albumin, blood glucose, glycated serum protein, etc., must be obtained through professional medical testing means in a formal medical institution. Similarly, urine index information, such as urinary microalbumin, urinary creatinine, etc., also needs to be obtained through the collection and analysis of urine samples. Complete blood count examination information, including white blood cell count, red blood cell count, hemoglobin concentration, etc., requires the collection of blood samples and detailed blood analysis in the laboratory to ensure the scientificity and accuracy of the results.

[0028] Exemplarily, in the nutrition and health information embedding and encoding module 20, the various nutrition and health status assessment indication information is embedded and encoded to obtain a set of nutrition and health status assessment indication information embedding and encoding features. It should be understood that since the various nutrition and health status assessment indication information of the elderly object to be evaluated (such as basic information, dietary information, biomarkers, etc.) may have different data types and structures (for example, text, numerical value, category label). Therefore, in order to be able to convert these different types of data into a data form convenient for subsequent processing, so as to be able to more effectively capture the load patterns and correlation relationships between these data and provide a basis for the individual status assessment of the elderly, in the technical solution of this application, the various nutrition and health status assessment indication information is further embedded and encoded to obtain a set of nutrition and health status assessment indication information embedding and encoding vectors.

[0029] In one embodiment, the nutrition and health information embedding and encoding module is used to: embed and encode the various nutrition and health status assessment indication information to obtain a set of nutrition and health status assessment indication information embedding and encoding vectors as the set of nutrition and health status assessment indication information embedding and encoding features. The set of nutrition and health status assessment indication information embedding and encoding vectors includes a basic information embedding and encoding vector, a matrix function status information embedding and encoding vector, a dietary information embedding and encoding vector, a quality of life information embedding and encoding vector, a biomarker - plasma biochemical examination information embedding and encoding vector, a biomarker - urine index information embedding and encoding vector, and a biomarker - blood routine examination information embedding and encoding vector. That is, through the method of embedding and encoding, the various nutrition and health status assessment indication information of the elderly object to be evaluated can be embedded and mapped into a common space. Thus, these heterogeneous data are converted into a unified numerical vector form, so that they can be processed within the same model framework. In addition, embedding and encoding is not just a simple numerical conversion. It can also capture the deep semantic information or internal connections of the various nutrition and health status assessment indication information. Specifically, for the health assessment data of the elderly, embedding and encoding can help reveal the complex patterns and relationships between the various nutrition and health status assessment indication information hidden behind the data.

[0030] Exemplarily, in the nutritional and health feature dynamic compensation aggregation analysis module 30, a feature dynamic compensation aggregation analysis is performed on the set of embedded coding features of the nutritional and health status evaluation indication information to obtain significantly aggregated coding features of the nutritional and health status evaluation indication information. It should be understood that since each of the embedded coding vectors of the nutritional and health status evaluation indication information in the set of embedded coding vectors of the nutritional and health status evaluation indication information represents the embedded semantic information of the nutritional and health status evaluation indication information of the elderly object to be evaluated, there are implicit correlation relationships and internal effects among these various nutritional and health status evaluation indication information, but not all of the nutritional and health status evaluation indication information is equally important for the subsequent evaluation of the elderly object status. Therefore, in order to be able to identify the key features that can best represent the core information of the entire embedded semantic data set of the nutritional and health status evaluation indication information, so as to reflect the important information and comprehensive feature representation in the embedded coding semantic feature distribution of these nutritional and health status evaluation indication information, and thus provide a basis for the subsequent evaluation of the elderly status and individual type judgment, in the technical solution of this application, a feature dynamic compensation aggregation analysis is further performed on the set of embedded coding features of the nutritional and health status evaluation indication information to obtain significantly aggregated coding features of the nutritional and health status evaluation indication information. In particular, through the feature dynamic compensation aggregation analysis process, the concept of a sequence hub can be introduced, that is, extracting representative core information from the input set of embedded coding vectors of the nutritional and health status evaluation indication information, to construct a feature compensation-aggregation mechanism based on the core information to enhance the model's understanding ability and aggregation analysis accuracy for the complex embedded coding semantic feature distribution of the nutritional and health status evaluation indication information.

[0031] In one embodiment, as Figure 3 shown, the nutritional and health feature dynamic compensation aggregation analysis module 30 includes: a node-hub complementary information calculation unit 31, configured to calculate the sequence hub features of the set of embedded coding features of the nutritional and health status evaluation indication information, and calculate the sequence of embedded coding features of the nutritional and health status evaluation indication information node-hub complementary information based on the sequence hub features; a compensation aggregation analysis processing unit 32, configured to perform a compensation aggregation analysis based on an attention mechanism on the sequence of embedded coding features of the nutritional and health status evaluation indication information node-hub complementary information to obtain the significantly aggregated coding features of the nutritional and health status evaluation indication information.

[0032] In one embodiment, the node-hub complementary information calculation unit 31 is configured to: input the set of embedded coding vectors of the nutritional and health status evaluation indication information into a sequence hub extraction network to obtain a sequence hub feature vector of the nutritional and health status evaluation indication information as the sequence hub feature. Specifically, this process can be represented by a formula as:

[0033]

[0034] Among them, X is the set of embedded coding vectors of the nutritional and health status assessment indication information, and x 1 , x 2 , x i , x n are the 1st, 2nd, ith, and nth nutritional and health status assessment indication information embedded coding vectors in the set of the nutritional and health status assessment indication information embedded coding vectors respectively. W 1i and b 1i represent the linear modulation weight matrix and the linear modulation bias vector corresponding to x i respectively. is matrix multiplication. is the transposed vector of the first modulation vector. e i is the feature extraction factor, softmax(·) is the normalization exponential function, a i is the feature extraction weight, n is the number of vectors in the set of the nutritional and health status assessment indication information embedded coding vectors, and v h is the hub feature vector of the nutritional and health status assessment indication information sequence.

[0035] Calculate the complementary information of each nutritional and health status assessment indication information embedded coding vector in the set of the nutritional and health status assessment indication information embedded coding vectors relative to the hub feature vector of the nutritional and health status assessment indication information sequence to obtain a sequence of nutritional and health status assessment indication information node-hub complementary information embedded coding vectors as the sequence of the nutritional and health status assessment indication information node-hub complementary information embedded coding features. Specifically, this process can be expressed by the formula:

[0036]

[0037] Among them, conv 1×1 (·) is the point convolution layer processing, W 11 is the first weight matrix, W 21 is the second weight matrix, Sigmoid(·) is the Sigmoid function, x i ' is the ith nutritional and health status assessment indication information embedded coding modulation vector, v h ' is the hub feature modulation vector of the nutritional and health status assessment indication information sequence. is subtraction by position, |·| is taking the absolute value, p i is the complementary weight vector, xb iThe \(i\)-th nutrition and health status assessment indication information node-hub complementary information embedded coding vector in the sequence of nutrition and health status assessment indication information node-hub complementary information embedded coding vectors, where \(g(\cdot,\cdot)\) is used to calculate the complementary information between two vectors.

[0038] It should be understood that the set of nutrition and health status assessment indication information embedded coding vectors is input into the sequence hub extraction network to extract the nutrition and health status assessment indication information sequence hub feature vector that can represent the core features of the entire data set. This nutrition and health status assessment indication information sequence hub feature vector is like a "data center", which condenses the most critical information in all embedded coding vectors and becomes the basis for subsequent analysis. In this way, the system can find the core elements that best reflect an individual's health status in a large amount of data, thereby enhancing the model's ability to understand complex data. Next, calculate the complementary information of each nutrition and health status assessment indication information embedded coding vector relative to the nutrition and health status assessment indication information sequence hub feature vector to obtain a sequence of node-hub complementary information embedded coding vectors. The significance of this step is that through the quantitative analysis of the relationship between each embedded coding vector and the sequence hub feature vector, the relative importance and unique contributions of different indication information in the overall health assessment can be revealed. Therefore, the calculation of this complementary information helps to identify key details that are easily overlooked by conventional statistical methods, enabling the model to more comprehensively understand the health status of the elderly. Finally, the sequence of nutrition and health status assessment indication information node-hub complementary information embedded coding vectors obtained by calculation not only supplements important information that may be missing or implicit in the original data set, but also provides a dynamic adjustment mechanism, enabling the model to flexibly adapt to different individual situations according to specific circumstances.

[0039] In one embodiment, as Figure 4 shown, the compensation aggregation analysis processing unit 32 includes: an attention modulation processing subunit 321, which is used to perform complementary information significant marking on each nutrition and health status assessment indication information node-hub complementary information embedded coding vector in the sequence of nutrition and health status assessment indication information node-hub complementary information embedded coding vectors, and then perform attention modulation on the sequence of nutrition and health status assessment indication information node-hub complementary information embedded coding vectors with the sequence of marked attention weights to obtain a sequence of nutrition and health status assessment indication information significant node-hub complementary information embedded coding vectors; a significant aggregation coding processing subunit 322, which is used to fuse the nutrition and health status assessment indication information sequence hub feature vector and the sequence of nutrition and health status assessment indication information significant node-hub complementary information embedded coding vectors to obtain a nutrition and health status assessment indication information significant aggregation coding feature vector as the nutrition and health status assessment indication information significant aggregation coding feature.

[0040] In one embodiment, the attention modulation processing subunit 321 is configured to: input each piece of nutritional and health status assessment indication information node-hub complementary information embedded in the sequence of encoded vectors into the complementary information significant identification module based on the attention mechanism to obtain a sequence of nutritional and health status assessment indication information node complementary information attention weights. Specifically, this process can be represented by the formula:

[0041]

[0042] where, W 2i and b 2i are respectively the linear modulation weight matrix and the linear modulation bias vector corresponding to xb i , is the transposed vector of the second modulation vector, g i is the node complementary factor, exp(·) is the function to calculate the natural exponential function value with the natural constant e as the base, and w i is the i-th nutritional and health status assessment indication information node complementary information attention weight in the sequence of nutritional and health status assessment indication information node complementary information attention weights.

[0043] Perform attention modulation on the sequence of nutritional and health status assessment indication information node-hub complementary information embedded in the encoded vectors with the sequence of nutritional and health status assessment indication information node complementary information attention weights to obtain the sequence of nutritional and health status assessment indication information significant node-hub complementary information embedded in the encoded vectors. Specifically, this process can be represented by the formula:

[0044] Y = {xb 1 · w 1 , xb 2 · w 2 ,..., xb i · w i ,..., xb n · w n}

[0045] where, Y is the sequence of nutritional and health status assessment indication information significant node-hub complementary information embedded in the encoded vectors, xb 1 , xb 2 and xb n are respectively the 1st, 2nd, and nth nutritional and health status assessment indication information node-hub complementary information embedded in the encoded vectors in the sequence of nutritional and health status assessment indication information node-hub complementary information embedded in the encoded vectors, and w 1 , w 2 and w nThey are the first, second, and nth complementary information attention weights of the nutritional and health status assessment indication information nodes in the sequence, respectively.

[0046] It should be understood that after obtaining the complementary information embedding coding vectors of each nutritional and health status assessment indication information node - hub, these vectors are respectively input into the complementary information significant identification module based on the attention mechanism. The core of this module is that it can automatically calculate the importance of each complementary information embedding coding vector of the nutritional and health status assessment indication information node - hub relative to the entire sequence and assign an attention weight to each vector. This weight reflects the relative importance of specific indication information in the whole health assessment. For example, certain biomarkers may be more indicative of health risks than other factors in specific situations; while certain lifestyle habits may have a more significant impact on long - term health. In this way, the system can identify key details that are easily overlooked by conventional statistical methods, thus comprehensively understanding the health status of the elderly. Subsequently, the obtained sequence of complementary information attention weights of the nutritional and health status assessment indication information nodes is used to perform attention modulation on the original sequence of node - hub complementary information embedding coding vectors. This means that for each embedding coding vector, its contribution in the final aggregated representation will be adjusted according to the corresponding attention weight. Vectors with high weights will occupy a more important position in subsequent analyses, while vectors with low weights will be appropriately weakened. This dynamic adjustment mechanism ensures that the model can flexibly adapt to different individual situations according to specific circumstances, improving the accuracy and reliability of the prediction results. The finally formed sequence of complementary information embedding coding vectors of the significant nodes of the nutritional and health status assessment indication information - hub not only supplements important information that may be missing or implicit in the original dataset, but also provides an intelligent selective amplification mechanism, enabling the model to focus on the core elements that best reflect the individual's health status. For example, in practical applications, if an elderly person shows an abnormal value in a certain biomarker but has a relatively healthy lifestyle, then the model can, through the weighted processing of the attention mechanism, comprehensively consider these two aspects of factors and give a more reasonable risk assessment result. This helps medical staff formulate more appropriate intervention measures and improve the effectiveness of health management.

[0047] In one embodiment, the significant aggregation encoding processing subunit 322 is configured to: embed each of the significant node-hub complementary information of the nutritional and health status assessment indication information into the encoding vector sequence to obtain a comprehensive representation vector of the significant node-hub complementary information of the nutritional and health status assessment indication information by position points; concatenate the comprehensive representation vector of the significant node-hub complementary information of the nutritional and health status assessment indication information and the sequence hub feature vector of the nutritional and health status assessment indication information to obtain the significant aggregation encoding feature vector of the nutritional and health status assessment indication information. Specifically, this process can be represented by the formula:

[0048] v f = Concat{v h ; Y}

[0049] where Concat{·;·} represents the vector fusion operation, and v f is the significant aggregation encoding feature vector of the nutritional and health status assessment indication information.

[0050] In summary, the feature dynamic compensation aggregation analysis processing can skillfully combine these two types of features by extracting the sequence hub features of the nutritional and health status assessment indication information and calculating the node-hub complementary information of the nutritional and health status assessment indication information, providing a more comprehensive aggregated representation of the semantics of the nutritional and health status assessment indication information. Moreover, when calculating the complementary information, it also adopts the attention mechanism and significance modulation, enabling this aggregated representation of the semantics of the nutritional and health status assessment indication information to be dynamically adjusted according to the specific situation, improving the flexibility of the model and its ability to adapt to different individual situations of the elderly. By understanding the embedded encoding feature distributions of the nutritional and health status assessment indication information of these elderly objects to be evaluated and performing accurate aggregation analysis, the system can more accurately evaluate the nutritional and health status of the elderly and make more reliable predictions in the subsequent stage, providing strong support for subsequent medical decisions.

[0051] Exemplarily, in the elderly individual type label determination module 40, based on the significantly aggregated encoded features of the nutritional and health status assessment indication information, a status assessment result is determined, and the status assessment result is used to represent the individual type label of the elderly object to be evaluated. The individual type label includes healthy population, low-risk population, medium-risk population, or high-risk population. In one embodiment, the elderly individual type label determination module is configured to: input the significantly aggregated encoded feature vector of the nutritional and health status assessment indication information into the status assessment module based on a classifier to obtain the status assessment result. That is to say, the status assessment is performed by using the semantic aggregated feature representation embedded in the nutritional and health status assessment indication information of the elderly object to be evaluated to determine which risk level the elderly belong to. In this way, a more intelligent data processing method can be used to better understand the nutritional and health status of the elderly, provide strong support for subsequent medical decisions, and ultimately help improve the quality of life and health management level of the elderly.

[0052] In one embodiment of the present application, the classifier-based status evaluation module uses a Multilayer Perceptron (MLP). A Multilayer Perceptron is a classic feedforward neural network structure, which consists of multiple layers of neurons, including an input layer, several hidden layers, and an output layer. By inputting the significantly aggregated encoded feature vectors of the nutritional and health status evaluation indicators into this MLP model, the status can be effectively evaluated and the final result can be obtained. Specifically, when the significantly aggregated encoded feature vectors of the nutritional and health status evaluation indicators are ready, they are passed to the input layer of the MLP. Each input node corresponds to a dimension in the significantly aggregated encoded feature vector of the nutritional and health status evaluation indicators, ensuring that all relevant information is completely introduced into the model. Next, this information is processed through a series of linear transformations and non-linear activation functions, and higher-level feature representations are gradually abstracted in the hidden layers. As the data is passed between layers, the hidden layers gradually extract the key features that can distinguish different health statuses and pass these features to the output layer. The output layer usually contains the same number of neurons as the number of target categories. In this application, they are the four categories of healthy people, low-risk people, medium-risk people, and high-risk people. Each output neuron is responsible for calculating the probability score of a specific category. In this application, it is achieved through the softmax function, which can convert the output into a set of probability values such that the sum of the probabilities of all categories is equal to 1. Finally, according to the probability distribution of the output layer, the category with the highest probability is selected as the final status evaluation result. For example, if the significantly aggregated encoded feature vectors of the nutritional and health status evaluation indicators of a certain elderly person are processed by the MLP and the probabilities of the four categories obtained in the output layer are: healthy people 0.15, low-risk people 0.25, medium-risk people 0.40, high-risk people 0.20, then this elderly person is classified as a medium-risk person.

[0053] In summary, the deep learning-based elderly nutritional and health status evaluation system according to the embodiments of the present application is clarified. It can improve the quality of the elderly nutritional and health status evaluation through a more intelligent data processing method. Moreover, such a score is not based on a fixed threshold but is automatically generated by the model. Therefore, it can more flexibly reflect the differences between different regions or groups, and at the same time takes into account the biological variations between individuals, thereby achieving personalized risk prediction in the true sense. In addition, through this method, it can have stronger non-linear fitting ability and adaptive adjustment ability. This means that it can process more types of data, reduce the possibility of human error, and also maintain good performance when facing new situations. This not only improves the reliability of the evaluation results, provides a scientific basis for subsequent medical decisions, but ultimately is expected to improve the quality of life and health management level of the elderly.

[0054] Figure 5 This is a schematic flowchart of the method for evaluating the nutritional and health status of the elderly based on deep learning according to an embodiment of the present application. As Figure 5 shown, the method for evaluating the nutritional and health status of the elderly based on deep learning includes: S1, obtaining various nutritional and health status evaluation indication information of the elderly object to be evaluated, where the nutritional and health status evaluation indication information includes basic information, matrix function status information, dietary information, quality of life information, biomarker - plasma biochemical examination information, biomarker - urine index information, and biomarker - blood routine examination information; S2, performing embedding encoding on the various nutritional and health status evaluation indication information to obtain a set of embedding encoding features of the nutritional and health status evaluation indication information; S3, performing feature dynamic compensation aggregation analysis on the set of embedding encoding features of the nutritional and health status evaluation indication information to obtain significantly aggregated encoding features of the nutritional and health status evaluation indication information; S4, based on the significantly aggregated encoding features of the nutritional and health status evaluation indication information, determining a status evaluation result, where the status evaluation result is used to represent the individual type label of the elderly object to be evaluated, and the individual type label includes healthy population, low - risk population, medium - risk population, or high - risk population.

[0055] In one embodiment, performing embedding encoding on the various nutritional and health status evaluation indication information to obtain a set of embedding encoding features of the nutritional and health status evaluation indication information includes: performing embedding encoding on the various nutritional and health status evaluation indication information to obtain a set of embedding encoding vectors of the nutritional and health status evaluation indication information as the set of embedding encoding features of the nutritional and health status evaluation indication information, and the set of embedding encoding vectors of the nutritional and health status evaluation indication information includes an embedding encoding vector of basic information, an embedding encoding vector of matrix function status information, an embedding encoding vector of dietary information, an embedding encoding vector of quality of life information, an embedding encoding vector of biomarker - plasma biochemical examination information, an embedding encoding vector of biomarker - urine index information, and an embedding encoding vector of biomarker - blood routine examination information.

[0056] In one embodiment, performing feature dynamic compensation aggregation analysis on the set of embedding encoding features of the nutritional and health status evaluation indication information to obtain significantly aggregated encoding features of the nutritional and health status evaluation indication information includes: calculating the sequence hub feature of the set of embedding encoding features of the nutritional and health status evaluation indication information, and calculating the sequence of node - hub complementary information embedding encoding features of the nutritional and health status evaluation indication information based on the sequence hub feature; performing compensation aggregation analysis based on the attention mechanism on the sequence of node - hub complementary information embedding encoding features of the nutritional and health status evaluation indication information to obtain the significantly aggregated encoding features of the nutritional and health status evaluation indication information.

[0057] In one embodiment, the sequence hub feature of the set of nutrition and health status assessment indication information embedded coding features is calculated, and based on the sequence hub feature, the sequence of nutrition and health status assessment indication information node-hub complementary information embedded coding features is calculated, including: inputting the set of nutrition and health status assessment indication information embedded coding vectors into a sequence hub extraction network to obtain a nutrition and health status assessment indication information sequence hub feature vector as the sequence hub feature; calculating the complementary information of each nutrition and health status assessment indication information embedded coding vector in the set of nutrition and health status assessment indication information embedded coding vectors relative to the nutrition and health status assessment indication information sequence hub feature vector to obtain a sequence of nutrition and health status assessment indication information node-hub complementary information embedded coding vectors as the sequence of nutrition and health status assessment indication information node-hub complementary information embedded coding features.

[0058] Here, those skilled in the art can understand that the specific operations of each step in the above-described method for assessing the nutritional and health status of the elderly based on deep learning have been described in detail above with reference to Figures 1 to 4 the description of the system for assessing the nutritional and health status of the elderly based on deep learning, and thus, the repeated description thereof will be omitted.

[0059] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Skilled professionals can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0060] It should be understood that the specific examples herein are only for helping those skilled in the art better understand the embodiments of this application, rather than limiting the scope of the embodiments of this application.

[0061] It should also be understood that in various embodiments of this application, the magnitude of the sequence numbers of each process does not mean the order of execution, and the order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of this application.

[0062] It should also be understood that the various embodiments described in this specification can be implemented alone or in combination, and the embodiments of this application do not limit this.

[0063] Unless otherwise specified, all technical and scientific terms used in the embodiments of this application have the same meaning as commonly understood by those skilled in the technical field of this application. The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit the scope of this application. The term "and / or" used in this application includes any and all combinations of one or more of the related listed items. The singular forms "a", "above-mentioned", and "the" used in the embodiments of this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0064] In several embodiments provided by this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0065] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0066] In addition, in each embodiment of this application, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0067] When the above-mentioned functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art or a part of this technical solution can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0068] As described above, the above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, and all should be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. A system for evaluating the nutritional health status of the elderly based on deep learning, characterized in that: include: The module for acquiring nutrition and health information of the elderly is used to acquire various nutrition and health status assessment indicator information of the elderly subject to be assessed, wherein the nutrition and health status assessment indicator information includes basic information, matrix function status information, dietary information, quality of life information, biomarker-plasma biochemical examination information, biomarker-urine index information, and biomarker-routine blood examination information; A nutrition and health information embedding coding module, used for embedding and coding the various nutrition and health status assessment indication information to obtain a set of nutrition and health status assessment indication information embedding coding features; A nutrition and health feature dynamic compensation aggregation analysis module is used to perform feature dynamic compensation aggregation analysis on the set of nutrition and health status assessment indicator information embedded coding features to obtain significant aggregation coding features of the nutrition and health status assessment indicator information, wherein the nutrition and health feature dynamic compensation aggregation analysis module includes: a node-hub complementary information calculation unit, which is used to calculate the sequence hub feature of the set of nutrition and health status assessment indicator information embedded coding features, and calculate the sequence of nutrition and health status assessment indicator information node-hub complementary information embedded coding features based on the sequence hub feature; a compensation aggregation analysis processing unit, which is used to perform compensation aggregation analysis based on the attention mechanism on the sequence of nutrition and health status assessment indicator information node-hub complementary information embedded coding features to obtain the significant aggregation coding features of the nutrition and health status assessment indicator information; The module for determining individual type labels of the elderly is used to determine the status assessment results based on the significant aggregation coding features of the nutritional health status assessment indication information, and the status assessment results are used to represent the individual type labels of the elderly objects to be assessed, and the individual type labels include healthy people, low-risk people, medium-risk people or high-risk people.

2. The deep learning-based nutritional health status assessment system for the elderly according to claim 1 is characterized in that: The nutrition and health information embedding coding module is used to: embed the various nutrition and health status assessment indicator information to obtain a set of nutrition and health status assessment indicator information embedding coding vectors as a set of nutrition and health status assessment indicator information embedding coding features, and the set of nutrition and health status assessment indicator information embedding coding vectors includes basic information embedding coding vectors, matrix function status information embedding coding vectors, dietary information embedding coding vectors, quality of life information embedding coding vectors, biomarker-plasma biochemical examination information embedding coding vectors, biomarker-urine index information embedding coding vectors and biomarker-blood routine examination information embedding coding vectors.

3. The deep learning-based nutritional health status assessment system for the elderly according to claim 2 is characterized in that: The node-hub complementary information calculation unit is used to: The set of the nutritional health status assessment indication information embedded coding vectors is input into a sequence hub extraction network to obtain a nutritional health status assessment indication information sequence hub feature vector as the sequence hub feature; Calculate the complementary information of each nutrition and health status assessment indicator information embedded coding vector in the set of nutrition and health status assessment indicator information embedded coding vectors relative to the hub feature vector of the nutrition and health status assessment indicator information sequence to obtain a sequence of nutrition and health status assessment indicator information node-hub complementary information embedded coding vectors as a sequence of nutrition and health status assessment indicator information node-hub complementary information embedded coding features.

4. The deep learning-based nutritional health status assessment system for the elderly according to claim 3 is characterized in that: The compensation aggregation analysis processing unit comprises: An attention modulation processing subunit is used to respectively mark the complementary information of each nutrition and health status assessment indicator information node-hub complementary information embedded coding vector in the sequence of the nutrition and health status assessment indicator information node-hub complementary information embedded coding vector, and then perform attention modulation on the sequence of the nutrition and health status assessment indicator information node-hub complementary information embedded coding vector with the sequence of marked attention weights to obtain a sequence of nutrition and health status assessment indicator information significant node-hub complementary information embedded coding vectors; The significant aggregation coding processing sub-unit is used to fuse the hub feature vector of the nutrition and health status assessment indication information sequence and the sequence of significant node-hub complementary information embedded coding vectors of the nutrition and health status assessment indication information to obtain a significant aggregation coding feature vector of the nutrition and health status assessment indication information as the significant aggregation coding feature of the nutrition and health status assessment indication information.

5. The deep learning-based nutritional health status assessment system for the elderly according to claim 4 is characterized in that: The attention modulation processing subunit is used to: Inputting each nutrition and health status assessment indication information node-hub complementary information embedding coding vector in the sequence of nutrition and health status assessment indication information node-hub complementary information embedding coding vector into a complementary information significant identification module based on an attention mechanism to obtain a sequence of nutrition and health status assessment indication information node complementary information attention weights; Using the sequence of attention weights of the nutritional health status assessment indicator information node complementary information, the sequence of nutritional health status assessment indicator information node-hub complementary information embedded coding vectors is attention modulated to obtain the sequence of nutritional health status assessment indicator information salient node-hub complementary information embedded coding vectors.

6. The deep learning-based nutritional health status assessment system for the elderly according to claim 5 is characterized in that: The significant aggregation coding processing subunit is used to: Each of the nutrition and health status assessment indicator information significant node-hub complementary information embedded coding vectors in the sequence of the nutrition and health status assessment indicator information significant node-hub complementary information embedded coding vectors is embedded according to the position point to obtain a nutrition and health status assessment indicator information significant node-hub complementary information comprehensive representation vector; The nutritional health status assessment indicator information significant node-hub complementary information comprehensive representation vector and the nutritional health status assessment indicator information sequence hub feature vector are cascaded to obtain the nutritional health status assessment indicator information significant aggregation coding feature vector.

7. The deep learning-based nutritional health status assessment system for the elderly according to claim 6 is characterized in that: The elderly individual type label determination module is used to: input the nutritional health status assessment indication information significant aggregation encoding feature vector into a classifier-based status assessment module to obtain the status assessment result.

8. A method for evaluating the nutritional health status of the elderly based on deep learning, characterized in that: include: Obtaining various nutritional health status assessment indicator information of the elderly subject to be assessed, wherein the nutritional health status assessment indicator information includes basic information, matrix function status information, dietary information, quality of life information, biomarker-plasma biochemical examination information, biomarker-urine index information, and biomarker-routine blood examination information; Embedding and encoding the various nutritional and health status assessment indicator information to obtain a set of nutritional and health status assessment indicator information embedding and encoding features; The set of nutrition and health status assessment indication information embedded coding features is subjected to feature dynamic compensation aggregation analysis to obtain significant aggregation coding features of nutrition and health status assessment indication information, including: calculating the sequence hub feature of the set of nutrition and health status assessment indication information embedded coding features, and calculating the sequence of nutrition and health status assessment indication information node-hub complementary information embedded coding features based on the sequence hub feature; the sequence of nutrition and health status assessment indication information node-hub complementary information embedded coding features is subjected to compensation aggregation analysis based on the attention mechanism to obtain significant aggregation coding features of nutrition and health status assessment indication information; Based on the significant aggregate coding features of the nutritional health status assessment indication information, a status assessment result is determined, and the status assessment result is used to represent the individual type label of the elderly object to be assessed, and the individual type label includes healthy people, low-risk people, medium-risk people or high-risk people.

9. The method for evaluating the nutritional health status of the elderly based on deep learning according to claim 8, characterized in that: Embedding and encoding the various nutrition and health status assessment indicator information to obtain a set of nutrition and health status assessment indicator information embedded coding features, including: embedding and encoding the various nutrition and health status assessment indicator information to obtain a set of nutrition and health status assessment indicator information embedded coding vectors as the set of nutrition and health status assessment indicator information embedded coding features, the set of nutrition and health status assessment indicator information embedded coding vectors includes basic information embedded coding vectors, matrix function status information embedded coding vectors, dietary information embedded coding vectors, quality of life information embedded coding vectors, biomarker-plasma biochemical examination information embedded coding vectors, biomarker-urine index information embedded coding vectors and biomarker-routine blood examination information embedded coding vectors.

10. The method for evaluating the nutritional health status of the elderly based on deep learning according to claim 9, characterized in that: Calculating the sequence hub feature of the set of nutritional health status assessment indication information embedded coding features, and calculating the sequence of nutritional health status assessment indication information node-hub complementary information embedded coding features based on the sequence hub feature, including: The set of the nutritional health status assessment indication information embedded coding vectors is input into a sequence hub extraction network to obtain a nutritional health status assessment indication information sequence hub feature vector as the sequence hub feature; Calculate the complementary information of each nutrition and health status assessment indicator information embedded coding vector in the set of nutrition and health status assessment indicator information embedded coding vectors relative to the hub feature vector of the nutrition and health status assessment indicator information sequence to obtain a sequence of nutrition and health status assessment indicator information node-hub complementary information embedded coding vectors as a sequence of nutrition and health status assessment indicator information node-hub complementary information embedded coding features.

Images