A whole-course nursing management system for tumor patients in combination of traditional Chinese and western medicine

Abstract

The application provides a traditional Chinese and western medicine combined whole-process nursing management system for tumor patients, relates to the field of intelligent health management, and first collects physiological, exercise, diet and psychological condition data of the tumor patients, then identifies potential health risks of the tumor patients by analyzing the physiological, exercise and diet data, simultaneously evaluates the positivity of the patients according to the psychological condition data, and subsequently inputs all the collected data and the identification and evaluation results into a traditional Chinese and western medicine combined nursing health management suggestion engine based on a large language model to generate traditional Chinese and western medicine combined nursing health management suggestions conforming to the actual conditions of the tumor patients. In this way, comprehensive and personalized health management services can be provided for the tumor patients.

Description

Technical Field

[0001] This application relates to the field of intelligent health management, and more specifically, to a comprehensive nursing management system for cancer patients that integrates traditional Chinese and Western medicine. Background Technology

[0002] With advancements in medical technology and increased health awareness, cancer treatment has transcended the limitations of traditional surgery, radiotherapy, and chemotherapy. Research indicates that patient self-management and lifestyle are crucial for disease treatment. Chemotherapy patients, in particular, often face side effects such as nausea, vomiting, fatigue, and weakened immunity, which not only reduce their quality of life but may also affect treatment outcomes. Therefore, providing comprehensive and effective nursing support to cancer patients during treatment has become a critical issue that urgently needs to be addressed.

[0003] However, traditional nursing management that relies solely on one medical system (such as using only Western medicine or traditional Chinese medicine) can easily overlook other needs. For example, Western medicine may focus more on the biological mechanisms of disease while neglecting the patient's psychological state; traditional Chinese medicine, while emphasizing holistic treatment, may not be timely or effective enough in treating certain acute conditions. This singular approach can lead to incomplete treatment and affect patient recovery. Furthermore, knowledge barriers and communication difficulties exist between the two medical systems; differences in terminology and diagnostic standards further complicate collaborative efforts.

[0004] Therefore, there is a need for a comprehensive nursing management program that integrates traditional Chinese and Western medicine for cancer patients. Summary of the Invention

[0005] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a comprehensive nursing management system integrating traditional Chinese and Western medicine for cancer patients.

[0006] According to one aspect of this application, a comprehensive nursing management system integrating traditional Chinese and Western medicine for cancer patients is provided, comprising:

[0007] The system includes a data acquisition module for acquiring physiological, exercise, dietary, and psychological data of cancer patients; a risk identification module for identifying potential health risks of cancer patients based on their physiological, exercise, and dietary data; an assessment module for assessing the level of positivity of cancer patients based on their psychological data; and a suggestion generation module for inputting the physiological, exercise, dietary, and psychological data of cancer patients, along with the identification and assessment results, into a large language model-based integrated traditional Chinese and Western medicine nursing health management suggestion engine to obtain integrated traditional Chinese and Western medicine nursing health management suggestions. The risk identification module includes:

[0008] A physiological feature extraction unit is used to extract physiological features from the physiological data of the tumor patient to obtain temporal implicit features of physiological features;

[0009] A motion feature extraction unit is used to extract motion features from the motion data of the tumor patient to obtain temporal implicit features of motion features;

[0010] A dietary feature extraction unit is used to extract dietary features from the dietary data of the tumor patients to obtain temporal implicit features of dietary features.

[0011] The fusion unit is used to fuse the temporal implicit features of the exercise features and the temporal implicit features of the diet features to obtain the temporal implicit features of the exercise-diet physiological regulation features;

[0012] The interaction unit is used to perform principal component feature adaptive compensation interaction on the temporal latent features of the exercise-diet physiological regulation features and the temporal latent features of the physiological features to obtain physiological-physiological regulation variable compensation interaction features, and to obtain the recognition result based on the physiological-physiological regulation variable compensation interaction features.

[0013] Compared with existing technologies, the integrated traditional Chinese and Western medicine nursing management system for cancer patients provided in this application first collects data on the physiological, exercise, dietary, and psychological conditions of cancer patients. Then, it analyzes the physiological, exercise, and dietary data to identify potential health risks in cancer patients, and simultaneously assesses patient motivation based on psychological data. Subsequently, all collected data and the identification and assessment results are input into an integrated traditional Chinese and Western medicine nursing health management suggestion engine based on a large language model to generate integrated traditional Chinese and Western medicine nursing health management suggestions tailored to the specific circumstances of the cancer patient. This provides comprehensive and personalized health management services for cancer patients. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0015] Figure 1 This is a system block diagram of the integrated traditional Chinese and Western medicine nursing management system for cancer patients according to an embodiment of this application.

[0016] Figure 2 This is a block diagram of the risk identification module in the integrated traditional Chinese and Western medicine nursing management system for cancer patients according to an embodiment of this application.

[0017] Figure 3This is a block diagram of the physiological feature extraction unit in the integrated traditional Chinese and Western medicine nursing management system for cancer patients according to an embodiment of this application.

[0018] Figure 4 This is a block diagram of the interactive unit in the integrated traditional Chinese and Western medicine nursing management system for cancer patients according to an embodiment of this application. Detailed Implementation

[0019] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0020] With the continuous advancement of medical technology and the increasing health awareness of the public, cancer treatment has gradually broken through the limitations of traditional single-method treatments such as surgery, radiotherapy, and chemotherapy. Studies show that patients' self-management abilities and lifestyles have a significant impact on the treatment outcome. Patients undergoing chemotherapy, in particular, often experience side effects such as nausea, vomiting, fatigue, and weakened immunity, which not only significantly reduce their quality of life but may also adversely affect the overall treatment effect. Therefore, providing comprehensive and effective nursing support for cancer patients during treatment has become an important issue that urgently needs to be addressed.

[0021] However, relying solely on one medical system (such as Western medicine or traditional Chinese medicine) for nursing management often fails to fully meet patients' needs. For example, while Western medicine excels at addressing diseases through biological mechanisms, it may sometimes overlook the patient's psychological state; and traditional Chinese medicine, which emphasizes holistic conditioning, may be insufficient in dealing with certain acute illnesses. Furthermore, knowledge barriers, terminology differences, and inconsistent diagnostic standards between Western and traditional Chinese medicine further complicate collaboration, posing challenges to comprehensive nursing management.

[0022] Based on this, this application proposes a comprehensive nursing management system for cancer patients that integrates traditional Chinese and Western medicine. Through intelligent and precise technologies, it breaks down knowledge barriers between disciplines and promotes the deep integration of the two, aiming to provide cancer patients with comprehensive and personalized health management services. Figure 1 This is a system block diagram of the integrated traditional Chinese and Western medicine nursing management system for cancer patients according to an embodiment of this application. Figure 1As shown, the integrated traditional Chinese and Western medicine nursing management system 100 for cancer patients includes: a data acquisition module 110, used to acquire physiological data, exercise data, dietary data, and psychological status data of cancer patients; a risk identification module 120, used to identify potential health risks of cancer patients based on their physiological, exercise, and dietary data to obtain identification results; an assessment module 130, used to assess the positive level of cancer patients based on their psychological status data to obtain assessment results; and a suggestion generation module 140, used to input the physiological, exercise, dietary, and psychological status data of cancer patients, as well as the identification results and assessment results, into an integrated traditional Chinese and Western medicine nursing health management suggestion engine based on a large language model to obtain integrated traditional Chinese and Western medicine nursing health management suggestions.

[0023] In this embodiment, the data acquisition module 110 is used to acquire physiological data, exercise data, dietary data, and psychological status data of cancer patients. Specifically, in this embodiment, the physiological data includes time series data of heart rate, blood pressure, and weight. It should be understood that physiological data, including time series data such as heart rate, blood pressure, and weight, can reflect the patient's basic health status and trends in bodily functions. For example, fluctuations in heart rate and blood pressure can indicate the state of the cardiovascular system, while changes in weight may be related to nutritional status or the effects of treatment side effects. Exercise data specifically includes information such as exercise intensity, duration, and frequency for cancer patients. Dietary data involves information on the types and amounts of food consumed daily by cancer patients, as well as their dietary habits. Psychological status data reflects information such as the emotional state and psychological stress level of cancer patients.

[0024] The following is a detailed explanation of the specific implementation process for "obtaining physiological data, exercise data, dietary data, and psychological data of cancer patients":

[0025] Firstly, for physiological data (time series of heart rate, blood pressure, and weight), modern technology offers multiple avenues for accurate monitoring. Smartwatches or fitness trackers, worn daily as personal health assistants, incorporate photoelectric sensors to continuously measure a user's heart rate. These devices not only capture real-time heart rate changes in daily life but also seamlessly transmit data to terminal devices via Bluetooth, ensuring data immediacy and accuracy. Furthermore, dedicated medical-grade devices such as continuous blood pressure monitors play a crucial role in obtaining more detailed physiological parameters. These instruments can stably record blood pressure values ​​over extended periods without disrupting the patient's daily life, creating time series data for subsequent analysis. Simultaneously, electronic scales, a common household device, can also become part of physiological data collection when connected to smart systems. They can record weight trends, providing models with a more comprehensive assessment of physical condition.

[0026] When it comes to exercise data, smart wearable devices once again demonstrate their irreplaceable value. Smartwatches and fitness trackers not only track steps, distance, and calorie consumption, but more importantly, they can recognize different types of exercise, such as running, walking, or cycling, and can also identify the intensity and frequency of different activities. Furthermore, some advanced wearable devices are equipped with heart rate monitoring functions, monitoring heart rate changes in real time during exercise to help assess whether the exercise intensity is appropriate, thereby avoiding the risks of overexertion.

[0027] Next, the acquisition of dietary data primarily relies on mobile applications. With the development of artificial intelligence technology, image recognition algorithms are now highly sophisticated. Users can upload photos of food to the application, which automatically identifies the food type and approximate portion size, thereby estimating nutritional components. This method greatly simplifies the traditional manual data entry process, improving efficiency and reducing the possibility of errors. Of course, manually entering daily dietary information is also an essential option, including selecting predefined food items, filling in specific intake amounts, and noting any special dietary habits (such as vegetarian or gluten-free).

[0028] Finally, the collection of psychological data is a crucial part of the entire health management process. Mobile applications play a key role in this regard. Regularly assessing patients' psychological state through questionnaires is a common practice. Standardized tools such as the Hospital Anxiety and Depression Scale (HADS) and the Symptom Checklist-90 (SCL-90) can help understand patients' mood fluctuations in a timely manner. Mood logs are another effective method, allowing users to record their feelings daily, which helps to identify trends in mood fluctuations. In addition, patient self-descriptions and observation records from healthcare professionals can also serve as sources of psychological data.

[0029] In this embodiment of the application, the risk identification module 120 is used to identify potential health risks of the cancer patient based on the patient's physiological data, exercise data, and dietary data to obtain identification results.

[0030] It is understandable that health risk identification plays a crucial role in the integrated traditional Chinese and Western medicine nursing management system for cancer patients. Identifying potential health risks enhances early warning capabilities, provides a scientific basis for developing personalized care plans, and promotes patient self-management and overall health management. However, traditional methods often rely on single types of data (such as physiological indicators), neglecting information from other important dimensions such as exercise and diet. This approach fails to comprehensively reflect a patient's health status and easily overlooks key risk factors. Furthermore, traditional methods often employ simple statistical analysis techniques, failing to delve into complex patterns and potential correlations within the data, making it difficult to capture subtle but important health changes.

[0031] Based on this, in the risk identification module, the technical concept of this application is to use deep learning-based data analysis and processing technology to extract features from the physiological data, exercise data, and dietary data of the tumor patients, respectively. Then, the extracted temporal implicit features of the exercise features and the temporal implicit features of the dietary features are fused. Based on the principal component compensation interaction representation between the temporal implicit features of the exercise-diet physiological regulation features and the temporal implicit features of the physiological features, the identification result is intelligently generated. By integrating multi-dimensional data such as physiological, exercise, and dietary data, this application can more comprehensively reflect the patient's health status, providing a solid foundation for developing personalized care plans. Simultaneously, the system can capture subtle but important health change signals, improving the sensitivity and specificity of risk identification.

[0032] Specifically, Figure 2 This is a block diagram of the risk identification module in the integrated traditional Chinese and Western medicine nursing management system for cancer patients according to an embodiment of this application. Figure 2As shown, the risk identification module 120 includes: a physiological feature extraction unit 121, used to extract physiological features from the physiological data of the tumor patient to obtain temporal implicit features of physiological features; a movement feature extraction unit 122, used to extract movement features from the movement data of the tumor patient to obtain temporal implicit features of movement features; a diet feature extraction unit 123, used to extract diet features from the diet data of the tumor patient to obtain temporal implicit features of diet features; a fusion unit 124, used to fuse the temporal implicit features of movement features and the temporal implicit features of diet features to obtain temporal implicit features of movement-diet physiological regulation features; and an interaction unit 125, used to perform principal component feature adaptive compensation interaction on the temporal implicit features of movement-diet physiological regulation features and the temporal implicit features of physiological features to obtain physiological-physiological regulatory variable compensation interaction features, and obtain the identification result based on the physiological-physiological regulatory variable compensation interaction features.

[0033] In this embodiment of the application, the physiological feature extraction unit 121 is used to extract physiological features from the physiological data of the tumor patient to obtain temporal implicit features of physiological features. Specifically, Figure 3 This is a block diagram of the physiological feature extraction unit in the integrated traditional Chinese and Western medicine nursing management system for cancer patients according to an embodiment of this application. Figure 3 As shown, the physiological feature extraction unit 121 includes: a physiological data temporal correlation feature extraction subunit 1211, used to process the time series of the heart rate value, the time series of the blood pressure value, and the time series of the weight using a temporal encoder based on an RNN-LSTM hybrid model to obtain a heart rate temporal correlation feature vector, a blood pressure temporal correlation feature vector, and a weight temporal correlation feature vector; and a physiological feature temporal hidden feature generation subunit 1212, used to fuse the heart rate temporal correlation feature vector, the blood pressure temporal correlation feature vector, and the weight temporal correlation feature vector to obtain a physiological feature temporal hidden encoding vector as the physiological feature temporal hidden feature.

[0034] In this embodiment, the physiological data temporal correlation feature extraction subunit 1211 is used to process the time series of the heart rate value, the time series of the blood pressure value, and the time series of the weight using a time-series encoder based on an RNN-LSTM hybrid model to obtain heart rate temporal correlation feature vectors, blood pressure temporal correlation feature vectors, and weight temporal correlation feature vectors. Correspondingly, considering that the physiological data of cancer patients is often large and complex, containing various vital signs (such as weight, blood pressure, heart rate, etc.), these data are constantly changing with the passage of time, such as temporal fluctuations and periodic changes. Based on this, in the technical solution of this application, features reflecting changes in the patient's physiological state are extracted from the physiological data of the cancer patient to obtain physiological feature temporal implicit encoding vectors. Specifically, in one example of this application, a time-series encoder based on an RNN-LSTM hybrid model is used to process the time series of the heart rate value, the time series of the blood pressure value, and the time series of the weight to obtain heart rate temporal correlation feature vectors, blood pressure temporal correlation feature vectors, and weight temporal correlation feature vectors. It's understandable that traditional RNN units are capable of processing sequential data, capturing local features and short-term dynamic changes. They are suitable for processing relatively short-term, direct changes in heart rate, blood pressure, and weight time series data, such as instantaneous changes in heart rate during a particular activity. LSTM units, on the other hand, are better able to capture deeper time dependencies and long-term trends. LSTM can effectively mine and model long-term patterns reflected in heart rate, blood pressure, and weight data, such as the potential relationship between long-term blood pressure trends and cardiovascular disease.

[0035] In this embodiment of the application, the physiological feature temporal implicit feature generation subunit 1212 is used to fuse the heart rate temporal associated feature vector, the blood pressure temporal associated feature vector, and the weight temporal associated feature vector to obtain a physiological feature temporal implicit encoding vector as the physiological feature temporal implicit feature. It should be understood that data such as heart rate, blood pressure, and weight can reflect different aspects of the physiological state of cancer patients. To gain a more comprehensive understanding of the true physiological state of cancer patients, this application can fuse the heart rate temporal associated feature vector, the blood pressure temporal associated feature vector, and the weight temporal associated feature vector. The resulting physiological feature temporal implicit encoding vector can carry the complex information of the physiological state of cancer patients in a more comprehensive and integrated way. Compared to using a feature vector of a single physiological data point, the fused vector enables the system to more sensitively capture subtle changes and abnormal trends in the patient's physiological state.

[0036] In this embodiment, the motion feature extraction unit 122 is used to extract motion features from the motion data of the tumor patient to obtain temporal latent features of motion features. Specifically, in this embodiment, the motion feature extraction unit is used to process the motion data at each time point in the motion data of the tumor patient using the temporal encoder based on the RNN-LSTM hybrid model to obtain a temporal latent encoding vector of motion features as the temporal latent features of motion features. It should be understood that the motion data of tumor patients contains rich information, such as intensity, duration, frequency, etc., and these data are usually presented in time series form, exhibiting dynamism and complexity. Therefore, in order to mine the potential patterns and temporal regularities in the data, in the technical solution of this application, motion features are extracted from the motion data of the tumor patient to obtain a temporal latent encoding vector of motion features. In one embodiment of this application, the temporal encoder based on the RNN-LSTM hybrid model is used to process the motion data at each time point in the motion data of the tumor patient to obtain a temporal latent encoding vector of motion features. In particular, motion behavior is often dynamic and periodic, and data at certain points in time may reflect specific activity patterns or anomalies. Time-by-time processing can help identify these key time points and extract valuable features, thereby better capturing the correlation and dynamic changes of motion data across different time points.

[0037] In this embodiment, the dietary feature extraction unit 123 is used to extract dietary features from the dietary data of the cancer patient to obtain temporal implicit features of dietary features. Specifically, in this embodiment, the dietary feature extraction unit is used to: use a BERT-based dietary data semantic encoder to semantically encode the dietary data at each time point in the dietary data of the cancer patient to obtain a time series of dietary feature semantic codes; and concatenate the time series of dietary feature semantic codes to obtain a temporal implicit encoding vector of dietary features as the temporal implicit features of dietary features. It should be understood that the dietary data of cancer patients contains rich information, such as food types, intake, eating frequency, and eating time. These data have many dimensions and are interrelated. Based on this, in the technical solution of this application, dietary features are extracted from the dietary data of the cancer patient to extract and capture important dietary feature information, and a temporal implicit encoding vector of dietary features is obtained. Specifically, in one embodiment of this application, a BERT-based semantic encoder is used to semantically encode the dietary data of the cancer patient at various time points to obtain a time series of dietary feature semantic codes. That is, the BERT model performs well in natural language processing tasks and has powerful semantic understanding capabilities. The dietary data of cancer patients contains rich semantic information, such as food names and dietary descriptions. The BERT model employs a bidirectional Transformer architecture, which can simultaneously consider the contextual information before and after the text, making the obtained dietary feature semantic codes more accurate and complete. Dietary behavior is often dynamic and periodic; data at certain time points may reflect specific dietary habits or abnormal situations. Time-by-time processing can help identify these key time points and extract valuable semantic features. Then, the time series of dietary feature semantic codes are concatenated to integrate dietary conditions at different times, including dietary trends and periodic patterns, into a single vector, obtaining a temporal implicit encoding vector of dietary features. This constructs a vector representation that comprehensively and holistically describes the dietary characteristics of cancer patients.

[0038] In this embodiment, the fusion unit 124 is used to fuse the temporal implicit features of the exercise characteristics and the temporal implicit features of the diet characteristics to obtain temporal implicit features of exercise-diet physiological regulation characteristics. Specifically, in this embodiment, the fusion unit is used to concatenate the temporal implicit encoding vector of the exercise characteristics and the temporal implicit encoding vector of the diet characteristics to obtain a concatenated temporal implicit encoding vector of exercise-diet physiological regulation characteristics as the temporal implicit features of exercise-diet physiological regulation characteristics. Accordingly, considering that exercise and diet are important factors affecting health, there is a complex interaction between the two. For example, exercise can regulate appetite, while diet may affect physical fitness and recovery speed. Therefore, in order to capture the potential correlation between the two and provide more comprehensive physiological regulation information, this application concatenates the temporal implicit encoding vector of the exercise characteristics and the temporal implicit encoding vector of the diet characteristics to obtain a concatenated temporal implicit encoding vector of exercise-diet physiological regulation characteristics. This can better describe the comprehensive physiological regulation of patients at different time points and provide a more comprehensive perspective for a deeper understanding of the patient's health status.

[0039] In this embodiment of the application, the interaction unit 125 is used to perform principal component feature adaptive compensation interaction on the temporal implicit features of the exercise-diet physiological regulation features and the temporal implicit features of the physiological features to obtain physiological-physiological regulation variable compensation interaction features, and to obtain the recognition result based on the physiological-physiological regulation variable compensation interaction features. Specifically, Figure 4 This is a block diagram of the interactive units in the integrated traditional Chinese and Western medicine nursing management system for cancer patients according to an embodiment of this application. Figure 4As shown, the interaction unit 125 includes: a feature principal component analysis subunit 1251, used to perform feature principal component analysis on the temporal implicit cascaded encoding vector of the exercise-diet physiological regulation feature and the temporal implicit encoding vector of the physiological feature to obtain a set of principal component encoding vectors of the exercise-diet physiological regulation feature and a set of principal component encoding vectors of the physiological feature; and a difference compensation weight calculation subunit 1252, used to calculate the difference embedding compensation of physiological-physiological regulation variables between the set of principal component encoding vectors of the exercise-diet physiological regulation feature and the set of principal component encoding vectors of the physiological feature. The code weight vector; the compensation interaction subunit 1253, used to embed the compensation coding weight vector based on the physiological-physiological regulatory variable difference, and to perform mean calculation and aggregation interaction on the set of principal component coding vectors of the exercise-diet physiological regulatory features and the set of principal component coding vectors of the physiological features time series to obtain the physiological-physiological regulatory variable compensation interaction feature vector as the physiological-physiological regulatory variable compensation interaction feature; the health risk identification result generation subunit 1254, used to input the physiological-physiological regulatory variable compensation interaction feature vector into the classifier-based health risk identifier to obtain the identification result.

[0040] It is understandable that exercise, diet, and physiological state are closely linked. Exercise affects physiological indicators such as metabolism and cardiovascular function, while diet provides the body with energy and nutrients, directly or indirectly affecting physiological functions. For example, long-term regular exercise combined with a reasonable diet can improve cardiovascular function, lower blood pressure, and optimize heart rate. These characteristics exhibit complex nonlinear relationships, each carrying unique and complementary information. Based on this, this application obtains physiological-physiological regulatory variable compensation interaction features by performing principal component feature adaptive compensation interaction on the temporal implicit features of the exercise-diet physiological regulation features and the temporal implicit features of the physiological features. In particular, principal component analysis can extract key principal components, reduce data dimensionality, and remove redundant information. Adaptive compensation interaction can dynamically adjust feature weights according to the inherent structure and changes of the data, ensuring that important information is not lost during dimensionality reduction, thereby accurately reflecting the health status of cancer patients.

[0041] In detail, the first step is to perform principal component analysis on the temporal hidden cascaded encoding vectors of the exercise-diet physiological regulation features and the temporal hidden encoding vectors of the physiological features to obtain the set of principal component encoding vectors of the exercise-diet physiological regulation features and the set of principal component encoding vectors of the physiological features. This process can be represented as:

[0042]

[0043]

[0044] in, It is the temporal hidden concatenated encoding vector of the exercise-diet physiological regulation features. For eigenprincipal component analysis, It was through The calculated covariance matrix of the exercise-diet physiological regulation features sample It is the principal component orthogonal matrix of the physiological regulatory characteristics of exercise and diet. For each principal component encoding vector of the exercise-diet physiological regulation feature in the set of principal component encoding vectors, The diagonal matrix represents the physiological regulatory characteristics of exercise and diet. , They are respectively and The corresponding weight value, for The transpose of the matrix, It is the temporal hidden encoding vector of the physiological feature. It was through The calculated covariance matrix of the physiological characteristics time series samples It is a principal component orthogonal matrix of physiological and temporal features. This refers to the set of temporal principal component encoding vectors for each physiological feature. This is a diagonal matrix representing the time series of physiological characteristics. , They are respectively and The corresponding weight value, for The transpose of .

[0045] It is understandable that the original temporal implicit cascade encoding vectors of exercise-diet physiological regulation features and physiological features contain a large amount of redundant information. For example, in the actual collected physiological data, indicators such as heart rate and blood pressure measured at different time points may have certain correlations. This related information is repeatedly recorded in the original data, increasing the complexity of data processing. In order to remove this redundant information and make the features more compact and representative, the technical solution of this application requires performing feature principal component analysis on the temporal implicit cascade encoding vectors of exercise-diet physiological regulation features and physiological features. The set of principal component encoding vectors of exercise-diet physiological regulation features and the set of principal component encoding vectors of physiological features obtained after feature principal component analysis removes the redundant information of the original data and emphasizes the most important features, which is beneficial to improving the accuracy of health risk assessment for cancer patients.

[0046] Specifically, in this embodiment, the difference compensation weight calculation subunit is used to: construct the set of principal component encoding vectors of the exercise-diet physiological regulation features and the set of principal component encoding vectors of the physiological features into an aggregated encoding feature map of the principal components of the exercise-diet physiological regulation features and an aggregated encoding feature map of the principal components of the physiological features. This process can be represented as:

[0047]

[0048]

[0049] in, For shape reshaping operations, For each principal component encoding vector of the exercise-diet physiological regulation feature in the set of principal component encoding vectors, This refers to the set of temporal principal component encoding vectors for each physiological feature. This is a principal component aggregation coding feature map of exercise-diet physiological regulation characteristics. It is a temporal principal component aggregation encoding feature map of physiological characteristics;

[0050] Differential embedding is performed on the principal component aggregation coding feature map of the exercise-diet physiological regulation features and the temporal principal component aggregation coding feature map of the physiological features to obtain the exercise-diet physiological regulation branch weight vector and the temporal branch weight vector of the physiological features. This process can be represented as follows:

[0051]

[0052] in, This is a principal component aggregation coding feature map of exercise-diet physiological regulation characteristics. It is a temporal principal component aggregation encoding feature map of physiological characteristics. It is an average pooling operation. and They are The corresponding first weight matrix and second weight matrix, and They are The corresponding first weight matrix and second weight matrix, It is a non-linear activation function. To output the activation function, It is the weight vector of the exercise-diet physiological regulation branch. It is the physiological feature temporal branch weight vector;

[0053] The absolute value of the difference vector between the exercise-diet physiological regulation branch weight vector and the physiological feature time-series branch weight vector is calculated to obtain the physiological-physiological regulation variable difference embedding compensation encoding weight vector. This process can be expressed as:

[0054]

[0055] in, It is the weight vector of the exercise-diet physiological regulation branch. It is the physiological feature temporal branch weight vector. It is subtracted based on position. For the absolute value operation, It is a weight vector for embedding compensation encoding of physiological-physiological regulatory variable differences.

[0056] It is understandable that the sets of principal component encoding vectors for exercise-diet physiological regulation features and the sets of principal component encoding vectors for time-series physiological features reside in high-dimensional spaces, with complex data structures that are difficult to intuitively understand and analyze. By constructing them as feature maps and mapping them to three-dimensional space, the complex high-dimensional information can be projected into a three-dimensional space that is easier for the model to understand. Specifically, in this three-dimensional space, the relative positional relationships and local clustering patterns between the principal component encoding vectors can be more easily captured by the model, which can provide direction for further analysis of patients' health risk status.

[0057] Accordingly, while both the exercise-diet physiological regulation feature map and the physiological feature time-series feature map reflect information related to the health of cancer patients, they have different focuses. The exercise-diet physiological regulation feature map mainly reflects the comprehensive influence of exercise and diet on physiological regulation, while the physiological feature time-series map focuses more on the changes in physiological state over time. To delve deeper into the subtle differences between these two feature maps, especially the subtle differences containing information crucial for health risk identification, this application requires differential embedding processing of the exercise-diet physiological regulation feature map and the physiological feature time-series feature map. The generated exercise-diet physiological regulation branch weight vector and physiological feature time-series branch weight vector can quantify the importance of differences. Specifically, the larger the weight value, the higher the importance of the feature difference at that position for health risk identification; conversely, the lower the weight value, the lower the importance. This quantification effect provides a clear numerical basis for subsequent compensation mechanisms and information fusion, enabling more reasonable analysis based on these quantified weights when dealing with complex health risk identification tasks.

[0058] Specifically, in this embodiment, the compensation interaction subunit is used to: calculate the positional mean vector of the set of principal component encoding vectors of the exercise-diet physiological regulation features and the set of principal component encoding vectors of the physiological features to obtain the principal component representation encoding vectors of the exercise-diet physiological regulation features and the principal component representation encoding vectors of the physiological features. This process can be expressed as:

[0059]

[0060]

[0061] in, The set of principal component encoding vectors for the exercise-diet physiological regulation features is the first... Principal component encoding vectors of exercise-diet physiological regulation features The set of temporal principal component encoding vectors of the physiological features is the first... Temporal principal component encoding vectors of physiological features It is the set of principal component encoding vectors of exercise-diet physiological regulation features and the number of vectors in the set of principal component encoding vectors of physiological features over time. It is the principal component representation encoding vector of the physiological regulatory features of exercise-diet. It is the temporal principal component representation encoding vector of physiological characteristics;

[0062] Based on the physiological-physiological regulatory variable difference embedding compensation coding weight vector, the principal component representation coding vector of the exercise-diet physiological regulatory features and the temporal principal component representation coding vector of the physiological features are aggregated and interacted to obtain the physiological-physiological regulatory variable compensation interaction feature vector. This process can be expressed as:

[0063]

[0064] in, It is a non-linear activation function. It is the principal component representation encoding vector of the physiological regulatory features of exercise-diet. It is a temporal principal component representation encoding vector of physiological characteristics. It is a dot product by position. It is point convolutional coding. It is the physiological-physiological regulatory variable compensation interaction feature vector.

[0065] It is understandable that the sets of principal component encoding vectors for exercise-diet physiological regulation features and the sets of principal component encoding vectors for physiological features over time typically contain rich and complex data. These sets may have high dimensionality and a large number of vectors, making subsequent aggregation and interaction operations computationally intensive and significantly reducing system processing efficiency. By calculating the positional mean vector, the data can be compressed in a relatively simple and direct way, reducing the amount of data and making subsequent calculations more efficient. Moreover, calculating the positional mean vector can highlight the core features in the feature vector set. Specifically, since mean calculation can smooth out some short-term or random feature value changes, the resulting principal component encoding vectors for exercise-diet physiological regulation features and the principal component encoding vectors for physiological features over time can better reflect the core trends and stable characteristics of the data as a whole.

[0066] Accordingly, considering that the principal component representation encoding vector of exercise-diet physiological regulation features and the temporal principal component representation encoding vector of physiological features reflect the patient's health-related information from different perspectives, exercise and diet influence physiological regulation, while physiological features directly reflect the body's immediate state. To comprehensively assess the patient's health risk, it is necessary to fuse these two sets of feature information for a more comprehensive understanding of the patient's health status. Based on this, this application uses the physiological-physiological regulation variable difference embedding compensation encoding weight vector to perform aggregation interaction processing on the principal component representation encoding vector of exercise-diet physiological regulation features and the temporal principal component representation encoding vector of physiological features. During the interaction process, if the two encoding vectors differ significantly in some aspects, the physiological-physiological regulation variable difference embedding compensation encoding weight vector will adjust its interaction method accordingly, processing the differing parts more meticulously. For example, if physiological features show abnormal blood pressure, but exercise-diet physiological regulation features indicate that the patient's recent exercise has improved cardiovascular function, the weight vector will highlight the correlation between the two during aggregation interaction, so that the final result reflects both the blood pressure problem and the positive effects of exercise, thereby achieving adaptability to the differences in the original features. Through this precise feature aggregation and interaction method, the resulting physiological-physiological regulatory variable compensation interaction feature vector can more accurately reflect the patient's actual health status.

[0067] Subsequently, the physiological-physiological regulatory variable compensation interaction feature vector is input into a classifier-based health risk identifier to obtain the identification result. That is, the physiological-physiological regulatory variable compensation interaction feature obtained through adaptive compensation interaction between the exercise-diet physiological regulatory feature temporal implicit cascade encoding vector and the physiological feature temporal implicit encoding vector is used for classification processing to intelligently generate the identification result. Specifically, the identification result here can be the risk identification result for various complications that patients are prone to during treatment. For example, when a patient's blood pressure is unstable for a long period, and insufficient exercise leads to poor blood circulation, and an unbalanced diet affects immunity, then based on the physiological-physiological regulatory variable compensation interaction feature vector encoded from this patient information, the health risk identifier will analyze it and ultimately output an identification result indicating an increased risk of cardiovascular and infectious complications.

[0068] Here, when the temporal implicit cascaded encoding vector of the exercise-diet physiological regulation features and the temporal implicit encoding vector of the physiological features represent the spliced ​​encoding features of exercise data and diet data and the individual encoding features of physiological data of tumor patients, respectively, during the feature principal component compensation interaction, the insufficient principal component compensation correlation caused by the commonality difference in the source data encoding will cause the sparse correlation interaction performance of the physiological-physiological regulation variable compensation interaction feature vector, thereby reducing the accuracy of the identification results obtained by the health risk identifier based on the classifier due to the predictability of classification inference.

[0069] In a preferred example, inputting the physiological-physiological regulatory variable compensation interaction feature vector into a classifier-based health risk identifier to obtain the identification result includes:

[0070] Based on the eigenvalues ​​of the physiological-physiological regulatory variable compensation interaction feature vector Distance and Distance yields physiological-physiological regulatory variable compensation interaction-distance matrix and physiological-physiological regulatory variable compensation interaction two-distance matrix ;

[0071] The number of overdistributed eigenvalues ​​in the physiological-physiological regulatory variable compensation interaction eigenvector where the difference between the mean and the eigenvalues ​​is greater than the variance of the eigenvalues ​​is determined. The reciprocal of the base-2 logarithm of the number of overdistributed eigenvalues ​​and the exponent of the base-natural constant of the reciprocal of the number of overdistributed eigenvalues ​​are then calculated to obtain the full-network interaction representation value of the first physiological-physiological regulatory variable compensation interaction. Second physiological-physiological regulatory variable compensation interaction full network interactive representation value :

[0072]

[0073]

[0074]

[0075] in, This represents the feature vector of the physiological-physiological regulatory variable compensation interaction. The first characteristic vector representing the physiological-physiological regulatory variable compensation interaction feature vector is... 1 eigenvalue, Indicates calculation The number, It is the physiological-physiological regulatory variable compensation interaction feature vector The number of hyperdistributed eigenvalues ​​in the dataset, and and It is the physiological-physiological regulatory variable compensation interaction feature vector The mean and variance of all eigenvalues;

[0076] Calculate the hidden basic feature vectors of the physiological-physiological regulatory variable compensation interaction, using the reciprocal of the full-network interaction representation value of the first physiological-physiological regulatory variable compensation interaction and the reciprocal of the full-network interaction representation value of the second physiological-physiological regulatory variable compensation interaction as exponents:

[0077]

[0078]

[0079] in, This represents the feature vector calculated using the reciprocal of the first physiological-physiological regulatory variable compensation interaction value across the entire network as its exponent. This indicates that the feature vector calculated using the reciprocal of the value of the second physiological-physiological regulatory variable compensation interaction across the entire network is the exponent. This represents the hidden basic feature vector of the first physiological-physiological regulatory variable compensation interaction. This represents the hidden basic feature vector representing the compensation interaction between the second physiological-physiological regulatory variables;

[0080] The first physiological-physiological regulator compensation interaction hidden basic feature vector, which serves as the row vector, is multiplied by the physiological-physiological regulator compensation interaction distance matrix to obtain the physiological-physiological regulator compensation interaction distance query vector. ,in, Represents matrix multiplication;

[0081] The physiological-physiological regulator compensation interaction binary distance matrix is ​​multiplied by the second physiological-physiological regulator compensation interaction hidden basic feature vector, which is a column vector, to obtain the physiological-physiological regulator compensation interaction binary distance response vector. ;

[0082] The weighted sum of the distance query vector of the physiological-physiological regulatory variable compensation interaction one and the distance response vector of the physiological-physiological regulatory variable compensation interaction two is calculated to obtain the optimized physiological-physiological regulatory variable compensation interaction feature vector;

[0083] The optimized physiological-physiological regulatory variable compensation interaction feature vector is input into the classifier-based health risk identifier to obtain the identification result.

[0084] In other words, to address the issue of insufficient overall active query encoding performance caused by long distances exceeding local spacing limits in the physiological-physiological regulatory variable compensation interaction feature vector under a fixed numerical sequence layout, a multi-dimensional attribute base form based on the self-distance matrix dot product of the physiological-physiological regulatory variable compensation interaction feature vector is adopted to capture its complex architecture of value domain interaction across the entire network. Furthermore, the performance of the live-stream sales temporal semantic query response encoding vector is reconstructed by simulating the large-scale multi-dimensional attribute hiding basis of the physiological-physiological regulatory variable compensation interaction feature vector. This achieves the reconstruction of the active search performance relationship of the physiological-physiological regulatory variable compensation interaction feature vector, thereby reflecting the predictable reconstruction under the actual series arrangement action of the physiological-physiological regulatory variable compensation interaction feature vector, and improving the accuracy of the recognition results obtained by inputting the physiological-physiological regulatory variable compensation interaction feature vector into a classifier-based health risk identifier.

[0085] In summary, the risk identification module 120 is clearly described. It uses deep learning-based data analysis and processing technology to extract features from the physiological, exercise, and dietary data of cancer patients. Then, it fuses the extracted temporal implicit features of exercise and diet features, intelligently generating identification results based on the principal component compensation interaction representation between the temporal implicit features of exercise-diet physiological regulation and the temporal implicit features of physiological features. By integrating multi-dimensional data such as physiological, exercise, and dietary data, it can more comprehensively reflect the patient's health status, providing a solid foundation for developing personalized care plans. Simultaneously, the system can capture subtle but important health change signals, improving the sensitivity and specificity of risk identification.

[0086] In this embodiment, the assessment module 130 is used to assess the level of positivity of the cancer patient based on the psychological status data to obtain an assessment result. It should be understood that mental health directly affects a patient's physical health and treatment adherence. A positive mindset helps enhance the immune system's function and improve the body's ability to fight disease. Furthermore, patients with stable emotions and a positive mindset are more likely to follow medical advice and receive treatment on time, thereby improving overall treatment effectiveness. By assessing the level of positivity of the cancer patient based on the psychological status data, it is possible to identify patients who may have psychological problems. For these patients, more personalized care plans can be provided, and care programs can be optimized, thereby improving the targeting and effectiveness of care. The assessment result can be a positivity level, such as high positivity, medium positivity, and low positivity. Specifically, patients with high positivity have a good emotional state, rarely experiencing negative emotions such as anxiety or depression. Although psychological pressure exists, they can cope effectively and it does not affect their attitude towards treatment and life. Moreover, they adopt proactive coping methods, actively cooperate with treatment, actively participate in rehabilitation activities, and are full of confidence in the future. Patients with moderate positivity have generally stable emotions, but occasionally experience anxiety or depression. Psychological pressure has some impact on their emotions and behaviors, but it has not seriously interfered with treatment and life. Their coping methods include both positive and negative coping behaviors. Patients with low positivity are in a state of anxiety, depression, and other negative emotions for a long time. They have greater psychological pressure, which seriously affects their confidence in treatment and their cooperation. Moreover, they often adopt negative coping methods, such as avoiding treatment and refusing to communicate.

[0087] The following is a detailed explanation of a specific implementation process for "assessing the level of positivity of the cancer patients based on the aforementioned psychological data to obtain assessment results":

[0088] Given that raw psychological data often contains various issues, data preprocessing is necessary. Data cleaning is the primary task of preprocessing, which means carefully identifying and removing duplicate data to avoid data redundancy interfering with subsequent analysis. Data with obvious errors, such as patients misreporting their age or scale scores exceeding the normal range, needs correction. For data with missing values, appropriate supplementation should be made through further communication with the patient or by referencing data from similar patients. Regarding data standardization, since different scales and data formats may differ, they need to be standardized to the same units and value ranges. For example, HADS scale scores (0-21 points) and SCL-90 scale scores (1-5 points) can be mapped to the 0-1 interval using a normalization formula. This makes data from different sources comparable and can more accurately reflect the patient's psychological state in subsequent analysis.

[0089] Preprocessed data provides a solid foundation for feature extraction. Key emotional features can be extracted from scale scores. Taking the HADS scale as an example, anxiety scores directly reflect the degree of tension, anxiety, and excessive worry in patients, while depression scores reflect low mood and loss of interest in things. The higher the score, the more pronounced the negative emotions of the patient, and correspondingly, the lower the psychological positivity. Patient self-descriptions and medical staff observation records also contain rich cognitive and attitudinal characteristics. When patients express positive thoughts such as "I believe the treatment will have good results" or "I hope to recover and return to normal life as soon as possible," or when medical staff observe patients actively participating in rehabilitation discussions and actively cooperating with treatment, these represent positive cognitions and attitudes. Conversely, a skeptical or resistant attitude towards treatment, or exhibiting negative behaviors towards treatment and rehabilitation, are considered negative characteristics.

[0090] After extracting features, in-depth analysis is necessary. Statistical analysis plays a crucial role in this process. By calculating statistics such as the mean and standard deviation of various features, we can understand the central tendency and dispersion of the patient group's psychological state. Simultaneously, comparing the feature data of patients at different treatment stages allows us to clearly observe the dynamic changes in psychological positivity, such as changes in patients' anxiety and confidence in treatment before and after chemotherapy. Building models using machine learning algorithms is also an effective analytical method. Algorithms such as Support Vector Machines (SVM) and Random Forests can use extracted features as input to classify and predict patients' psychological positivity. Through model training and optimization, we can identify the feature combinations that have the greatest impact on positivity levels, such as specific combinations of features like anxiety scores, attitudes towards treatment, and participation in social activities. Furthermore, correlation analysis can explore the relationships between various features; for example, there may be a negative correlation between anxiety and sleep quality. Understanding these relationships helps in a more comprehensive assessment of the patient's psychological state.

[0091] Finally, the assessment results are determined. Based on the feature analysis, scientific and reasonable assessment criteria need to be developed to clarify the patient's level of psychological positivity. Score thresholds can be set to classify different levels of positivity. For example, a total score of 80 or above, calculated from all features, can be defined as high positivity, indicating that the patient is optimistic and actively cooperates with treatment; 60-79 points indicate moderate positivity, suggesting that while the patient has some positive attitude, some negative factors still need attention; and below 60 points indicates low positivity, indicating that the patient has relatively serious psychological problems, with negative emotions dominating. Simultaneously, the assessment process should be adjusted to fully consider the patient's individual circumstances, taking into account factors such as the severity of the disease and the treatment cycle. For example, patients with advanced cancer may generally have lower psychological positivity scores than early-stage patients due to the more severe condition, requiring appropriate adjustments to the assessment criteria. Ultimately, the assessment results can be presented in the form of a report. This report not only includes the patient's score and positivity level but also details the main influencing factors. For example, for a patient with a psychological positivity score of 65, which falls into the moderate positivity category, the report might indicate that the main influencing factors are concerns about the disease prognosis and discomfort caused by chemotherapy side effects.

[0092] In this embodiment, the suggestion generation module 140 is used to input the physiological data, exercise data, dietary data, and psychological status data of the tumor patient, as well as the identification results and the evaluation results, into a traditional Chinese and Western medicine integrated nursing health management suggestion engine based on a large language model to obtain integrated traditional Chinese and Western medicine integrated nursing health management suggestions. It should be understood that a large language model is an artificial intelligence model based on deep learning. Through unsupervised learning on massive amounts of text data, it learns the structure, semantics, and grammar of language, thereby generating natural, fluent, and input-relevant text content. It possesses powerful language understanding and generation capabilities, can process and analyze various types of text information, and generate reasonable responses based on the given context. Considering the physiological, exercise, dietary, psychological, identification, and assessment data of cancer patients, a comprehensive picture of their health status and needs is created from different dimensions. To generate more comprehensive and accurate nursing recommendations based on this rich and detailed information, this application requires inputting the physiological, exercise, dietary, and psychological data of the cancer patients, along with the identification and assessment results, into a large language model-based integrated traditional Chinese and Western medicine nursing health management recommendation engine for deep understanding and analysis. This will generate suitable integrated traditional Chinese and Western medicine nursing health management recommendations for the cancer patients. For example, a cancer patient's physiological data shows high blood pressure (150 / 95 mmHg), a slightly fast heart rate (95 beats / minute), and a weight loss of 5 kg in the past three months; exercise data indicates only two exercise sessions per week, each lasting about 20 minutes, mostly walking; dietary data shows a preference for high-salt foods and low intake of vegetables and fruits; the assessment result is moderately positive, with the main stress stemming from concerns about disease recurrence; the identification results suggest a risk of cardiovascular disease and malnutrition. Based on this data, the integrated traditional Chinese and Western medicine nursing and health management recommendations generated by the large language model may be as follows: In terms of Western medicine, emphasis should be placed on medication management. Patients are advised to regularly measure their blood pressure and heart rate, and, based on the doctor's assessment, consider using antihypertensive drugs to control blood pressure. Simultaneously, close monitoring of heart rate changes is crucial, and medication adjustments should be made as necessary to maintain cardiovascular stability. Regarding nutritional support, a detailed dietary plan should be developed, increasing the intake of high-quality protein such as lean meat, fish, and legumes to improve malnutrition. The intake of high-salt foods should be reduced, keeping daily salt intake below 5 grams, and the consumption of vegetables and fruits should be increased to ensure an adequate supply of vitamins and minerals. In terms of exercise guidance, the frequency and intensity of exercise should be gradually increased, with at least 5 exercise sessions per week, each lasting 30-45 minutes. In addition to walking, aerobic exercises such as jogging and cycling can be added to enhance cardiopulmonary function. Attention should be paid to gradually increasing exercise intensity to avoid overexertion, and proper warm-up and stretching before and after exercise should be performed to prevent injury.In Traditional Chinese Medicine (TCM), herbal remedies are used based on TCM syndrome differentiation and treatment. Given the patient's potential for poor blood circulation and spleen / stomach weakness, modified Bazhen Tang (Eight-Treasure Decoction) can be prescribed, which has the effects of promoting blood circulation, removing blood stasis, strengthening the spleen and stomach, improving digestion, and enhancing the body's resistance. Traditional therapies include moxibustion on acupoints such as Zusanli (ST36), Neiguan (PC6), and Shenque (CV8), 2-3 times a week, 15-20 minutes per acupoint each time. This warms the meridians, regulates blood circulation, relieves anxiety, and enhances the body's immunity. Simultaneously, massaging acupoints such as Baihui (GV20) and Shenting (GV24) on the head for 10-15 minutes each morning and evening can relax the mind and body and improve sleep quality. For emotional regulation, patients are advised to regulate their emotions through listening to soothing music and reading to maintain a cheerful mood. They are also encouraged to participate in patient support activities, share their anti-cancer experiences, and strengthen their confidence in overcoming the disease.

[0093] In summary, the integrated traditional Chinese and Western medicine nursing management system 100 for cancer patients, based on embodiments of this application, is explained. It first collects data on the physiological, exercise, dietary, and psychological conditions of cancer patients. Then, it identifies potential health risks of cancer patients by analyzing the physiological, exercise, and dietary data, and assesses the patient's motivation based on psychological data. Subsequently, all collected data and the identification and assessment results are input into an integrated traditional Chinese and Western medicine nursing health management suggestion engine based on a large language model to generate integrated traditional Chinese and Western medicine nursing health management suggestions tailored to the actual situation of the cancer patient. This provides comprehensive and personalized health management services for cancer patients.

Claims

1. A comprehensive nursing management system integrating traditional Chinese and Western medicine for cancer patients, characterized in that, include: The data acquisition module is used to acquire physiological data, exercise data, dietary data, and psychological status data of cancer patients. The risk identification module is used to identify potential health risks of the cancer patients based on their physiological data, exercise data, and dietary data to obtain identification results; the assessment module is used to assess the positive level of the cancer patients based on their psychological status data to obtain assessment results. The suggestion generation module is used to input the physiological data, exercise data, dietary data, and psychological status data of the cancer patient, as well as the identification results and the assessment results, into a traditional Chinese and Western medicine integrated nursing health management suggestion engine based on a large language model to obtain integrated traditional Chinese and Western medicine nursing health management suggestions; wherein, the risk identification module includes: A physiological feature extraction unit is used to extract physiological features from the physiological data of the tumor patient to obtain temporal implicit features of physiological features; A motion feature extraction unit is used to extract motion features from the motion data of the tumor patient to obtain temporal implicit features of motion features; A dietary feature extraction unit is used to extract dietary features from the dietary data of the tumor patients to obtain temporal implicit features of dietary features. The fusion unit is used to fuse the temporal implicit features of the exercise features and the temporal implicit features of the diet features to obtain the temporal implicit features of the exercise-diet physiological regulation features; An interaction unit is used to perform principal component feature adaptive compensation interaction on the temporal latent features of the exercise-diet physiological regulation features and the temporal latent features of the physiological features to obtain physiological-physiological regulation variable compensation interaction features, and to obtain the recognition result based on the physiological-physiological regulation variable compensation interaction features; The interactive unit includes: The feature principal component analysis subunit is used to perform feature principal component analysis on the temporal implicit cascaded coding vector of exercise-diet physiological regulation features and the temporal implicit coding vector of physiological features to obtain the set of principal component coding vectors of exercise-diet physiological regulation features and the set of principal component coding vectors of physiological features. The differential compensation weight calculation subunit is used to: construct the set of principal component encoding vectors of exercise-diet physiological regulation features and the set of principal component encoding vectors of physiological features time series as aggregated encoding feature maps of exercise-diet physiological regulation features and aggregated encoding feature maps of physiological features time series; perform differential embedding on the aggregated encoding feature maps of exercise-diet physiological regulation features and aggregated encoding feature maps of physiological features time series to obtain the weight vectors of exercise-diet physiological regulation branches and the weight vectors of physiological features time series; and calculate the absolute value of the difference vector between the weight vectors of exercise-diet physiological regulation branches and the weight vectors of physiological features time series to obtain the differential embedding compensation encoding weight vector of physiological-physiological regulation variables. The compensation interaction subunit is used to: calculate the positional mean vector of the set of principal component encoding vectors of exercise-diet physiological regulation features and the set of principal component encoding vectors of physiological features in time series to obtain the principal component representation encoding vectors of exercise-diet physiological regulation features and the principal component representation encoding vectors of physiological features in time series; and aggregate and interact the principal component representation encoding vectors of exercise-diet physiological regulation features and the principal component representation encoding vectors of physiological features in time series based on the difference embedding compensation encoding weight vector of physiological-physiological regulation variables to obtain the compensation interaction feature vector of physiological-physiological regulation variables. The health risk identification result generation subunit is used to input the physiological-physiological regulatory variable compensation interaction feature vector into the classifier-based health risk identifier to obtain the identification result.

2. The integrated traditional Chinese and Western medicine nursing management system for cancer patients according to claim 1, characterized in that, The physiological data includes time series of heart rate values, time series of blood pressure values, and time series of body weight.

3. The integrated traditional Chinese and Western medicine nursing management system for cancer patients according to claim 2, characterized in that, The physiological feature extraction unit includes: The physiological data temporal correlation feature extraction subunit is used to process the time series of the heart rate value, the time series of the blood pressure value, and the time series of the weight using a time encoder based on an RNN-LSTM hybrid model to obtain the heart rate temporal correlation feature vector, the blood pressure temporal correlation feature vector, and the weight temporal correlation feature vector, respectively. The physiological feature temporal latent feature generation subunit is used to fuse the heart rate temporal associated feature vector, the blood pressure temporal associated feature vector and the weight temporal associated feature vector to obtain the physiological feature temporal latent encoding vector as the physiological feature temporal latent feature.

4. The integrated traditional Chinese and Western medicine nursing management system for cancer patients according to claim 3, characterized in that, The motion feature extraction unit is used to: process the motion data of the tumor patient at each time point in the motion data using the time encoder based on the RNN-LSTM hybrid model to obtain the motion feature time-series latent encoding vector as the motion feature time-series latent feature.

5. The integrated traditional Chinese and Western medicine nursing management system for cancer patients according to claim 4, characterized in that, The dietary feature extraction unit is used for: A BERT-based semantic encoder was used to semantically encode the dietary data of the tumor patients at various time points to obtain a time series of semantically encoded dietary features. The time series of the semantic encoding of the dietary features are concatenated to obtain the temporal latent encoding vector of the dietary features as the temporal latent features of the dietary features.

6. The integrated traditional Chinese and Western medicine nursing management system for cancer patients according to claim 5, characterized in that, The fusion unit is used to: concatenate the temporal latent coding vector of the exercise feature and the temporal latent coding vector of the diet feature to obtain the temporal latent concatenated coding vector of the exercise-diet physiological regulation feature as the temporal latent feature of the exercise-diet physiological regulation feature.

Images