Collection and analysis system for multi-source medical information

Abstract

The invention discloses a multi-source medical information acquisition and analysis system, which is characterized in that heart rate and body temperature data are collected in real time by using intelligent wearable equipment worn on a monitored old person to form first-source and second-source medical information; a time sequence encoder based on a forward LSTM (Long Short Term Memory) model is adopted to respectively extract hidden feature vectors associated with a time sequence from heart rate and body temperature data, a remarkable interaction relationship between the heart rate and the body temperature is further captured through interaction fusion among features, and whether the physiological state of the monitored old person is abnormal or not is determined based on the remarkable interaction relationship. Through the mode, the efficiency and accuracy of health monitoring of the old people are improved, and powerful support is provided for health management of the old people.

Description

Technical Field

[0001] The present application relates to the field of intelligent analysis, and more specifically, to a system for collecting and analyzing multi-source medical information. Background Art

[0002] With the aging of the population, how to efficiently and accurately monitor the health status of the elderly has become a focus of social attention. Heart rate and body temperature, as important physiological indicators, play a key role in health monitoring. For example, a simultaneous increase in heart rate and body temperature may indicate the presence of infection or inflammation, while an increase in heart rate but normal body temperature may indicate psychological stress or anxiety. By monitoring these two indicators in real time and continuously, an individual's health status can be assessed more accurately.

[0003] Most existing systems only focus on the collection and analysis of a certain type of physiological data, such as heart rate or body temperature, and lack the ability to comprehensively analyze multi-source physiological data. This single-source data collection method cannot fully reflect the health status of an individual, especially in the early diagnosis of complex diseases, where a single data source often cannot provide sufficient information support. Even if existing health analysis systems are able to collect a variety of physiological data, they often lack effective comprehensive analysis methods and cannot fully utilize the correlation between different data sources, thus affecting the accuracy and reliability of health assessments.

[0004] Therefore, a multi-source medical information collection and analysis system is expected. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a multi-source medical information collection and analysis system, which collects heart rate and body temperature data in real time by using a smart wearable device worn on the monitored elderly subject, forming the first source and the second source medical information, and using a time series encoder based on a forward LSTM model to extract time-series-related implicit feature vectors from the heart rate and body temperature data, respectively, and further capture the significant interactive relationship between heart rate and body temperature through interactive fusion between features, and based on this, determine whether the physiological state of the monitored elderly subject is abnormal. In this way, the efficiency and accuracy of health monitoring for the elderly are improved, providing strong support for elderly health management.

[0006] According to one aspect of the present application, a multi-source medical information collection and analysis system is provided, comprising:

[0007] A multi-source medical information collection module, used to collect a time queue of real-time heart rate and a time queue of real-time body temperature through a smart wearable medical device worn on a monitored elderly subject, wherein the time queue of real-time heart rate is used as the first source medical information, and the time queue of real-time body temperature is used as the second source medical information;

[0008] A data transmission module, used for transmitting the first source medical information and the second source medical information to a medical information collection and analysis server through a wireless communication module;

[0009] A medical information time series analysis module, used for performing forward time series coding on the first source medical information and the second source medical information, respectively, on the medical information collection and analysis server to obtain a heart rate value time series associated implicit feature vector and a body temperature time series associated implicit feature vector;

[0010] A multi-source medical information interactive fusion module is used to interactively fuse the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector on the medical information collection and analysis server to obtain a heart rate-body temperature time series feature significant interactive fusion representation vector;

[0011] A physiological state analysis module is used to determine whether the physiological state of the monitored elderly object is abnormal based on the heart rate-body temperature time series feature significant interaction fusion representation vector in the medical information collection and analysis server;

[0012] Among them, the multi-source medical information interactive fusion module includes: a conditional information construction unit, which is used to extract the potential correlation features between the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector as a conditional feature vector; a significance guided interactive fusion unit, which is used to guide the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector to perform cross-domain interactive fusion encoding based on the attention mechanism based on the conditional feature vector to obtain the heart rate-body temperature time series feature significant interactive fusion representation vector.

[0013] Compared with the prior art, the present application provides a multi-source medical information collection and analysis system, which collects heart rate and body temperature data in real time by using smart wearable devices worn on the monitored elderly subjects, forming the first source and the second source medical information, and using a time series encoder based on a forward LSTM model to extract time-series-related implicit feature vectors from the heart rate and body temperature data, and further capture the significant interactive relationship between heart rate and body temperature through interactive fusion between features, and based on this, determine whether the physiological state of the monitored elderly subject is abnormal. In this way, the efficiency and accuracy of health monitoring for the elderly are improved, providing strong support for elderly health management. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0015] Figure 1 is a block diagram of a system for collecting and analyzing multi-source medical information according to an embodiment of the present application;

[0016] Figure 2 A schematic diagram of data flow of a multi-source medical information collection and analysis system according to an embodiment of the present application;

[0017] Figure 3 It is a block diagram of a multi-source medical information interactive fusion module in a multi-source medical information collection and analysis system according to an embodiment of the present application. DETAILED DESCRIPTION

[0018] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0019] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0020] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0021] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0022] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0023] Most existing systems only focus on the collection and analysis of a certain type of physiological data, such as heart rate or body temperature, and lack the ability to comprehensively analyze multi-source physiological data. This single-source data collection method cannot fully reflect the health status of an individual, especially in the early diagnosis of complex diseases, where a single data source often cannot provide sufficient information support. Even if the existing health analysis system is able to collect a variety of physiological data, it often lacks effective comprehensive analysis methods and cannot fully utilize the correlation between different data sources, thus affecting the accuracy and reliability of health assessments. Therefore, a multi-source medical information collection and analysis system is expected.

[0024] In the technical solution of the present application, a multi-source medical information collection and analysis system is proposed. Figure 1 It is a block diagram of a system for collecting and analyzing multi-source medical information according to an embodiment of the present application. Figure 2 FIG. 1 is a data flow diagram of a multi-source medical information collection and analysis system according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the embodiment of the present application, the multi-source medical information collection and analysis system 300 includes: a multi-source medical information collection module 310, which is used to collect the time queue of real-time heart rate and the time queue of real-time body temperature through the smart wearable medical device worn on the monitored elderly object, wherein the time queue of real-time heart rate is used as the first source medical information, and the time queue of real-time body temperature is used as the second source medical information; a data transmission module 320, which is used to transmit the first source medical information and the second source medical information to the medical information collection and analysis server through the wireless communication module; a medical information timing analysis module 330, which is used to analyze the medical information on the medical information collection and analysis server. The first source medical information and the second source medical information are respectively forward time-series encoded to obtain a heart rate value time-series associated implicit feature vector and a body temperature time-series associated implicit feature vector; a multi-source medical information interactive fusion module 340 is used to interactively fuse the heart rate value time-series associated implicit feature vector and the body temperature time-series associated implicit feature vector in the medical information collection and analysis server to obtain a heart rate-body temperature time-series feature significant interactive fusion representation vector; a physiological state analysis module 350 is used to determine whether the physiological state of the monitored elderly object is abnormal based on the heart rate-body temperature time-series feature significant interactive fusion representation vector in the medical information collection and analysis server.

[0025] In particular, the multi-source medical information acquisition module 310 is used to collect the time queue of real-time heart rate and the time queue of real-time body temperature through the smart wearable medical device worn on the monitored elderly object, wherein the time queue of real-time heart rate is used as the first source medical information, and the time queue of real-time body temperature is used as the second source medical information. It should be understood that heart rate and body temperature are physiological indicators that change over time. By collecting the time queue of real-time heart rate and the time queue of real-time body temperature, the dynamic change trend of these indicators can be captured, and the changes in the physiological state of the elderly can be reflected in time. As two important physiological indicators, heart rate and body temperature can be comprehensively monitored to obtain more comprehensive health information. For example, a sudden increase in heart rate may indicate an increase in cardiac load, and an increase in body temperature may indicate infection or inflammation. Therefore, in the technical solution of the present application, it is expected to improve the efficiency and accuracy of health monitoring of the elderly through comprehensive analysis of multi-source data.

[0026] In particular, the data transmission module 320 is used to transmit the first source medical information and the second source medical information to the medical information collection and analysis server through the wireless communication module. The wireless communication module ensures the timeliness of the medical information collection and analysis server receiving and analyzing the data by transmitting the collected data to the server in real time. In addition, through the remote monitoring function of the wireless communication module, medical staff can check the health status of the elderly in real time at different locations, which is very important for timely discovery and treatment of health problems.

[0027] In particular, the medical information time series analysis module 330 is used to perform forward time series encoding on the first source medical information and the second source medical information in the medical information collection and analysis server to obtain the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector. Considering that it is difficult to capture subtle physiological changes by analyzing only a single heart rate or body temperature data, in order to obtain the complex interactive relationship between heart rate and body temperature, so as to realize the comprehensive analysis of multimodal data, in the technical solution of the present application, the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector are interactively fused to obtain the heart rate-body temperature time series feature significant interactive fusion representation vector. In a specific example of the present application, the first source medical information and the second source medical information are respectively input into a time series encoder based on a forward LSTM model to obtain the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector. The LSTM (Long Short-Term Memory) model is a special recurrent neural network that is particularly good at processing time-dependent sequence data. By inputting the time series of heart rate and body temperature into the LSTM model, the implicit characteristics of the time series association between heart rate and body temperature can be effectively captured.

[0028] In particular, the multi-source medical information interactive fusion module 340 is used to interactively fuse the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector on the medical information collection and analysis server to obtain a heart rate-body temperature time series feature significant interactive fusion representation vector. In a specific example of the present application, Figure 3 As shown, the multi-source medical information interactive fusion module 340 includes: a conditional information construction unit 341, which is used to extract the potential correlation features between the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector as a conditional feature vector; a significance guided interactive fusion unit 342, which is used to guide the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector to perform cross-domain interactive fusion encoding based on the attention mechanism based on the conditional feature vector to obtain the heart rate-body temperature time series feature significant interactive fusion representation vector.

[0029] Specifically, the conditional information construction unit 341 is used to extract the potential correlation features between the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector as a conditional feature vector. Here, first, the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector are input into the implicit correlation feature capture network to obtain the heart rate-body temperature time series implicit correlation feature vector; by inputting the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector into the implicit correlation feature capture network to learn the intrinsic connection between the features, so as to obtain the heart rate-body temperature time series implicit correlation feature vector. Specifically, in the technical solution of the present application, the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector are fused according to the position points and then passed through a neural network layer based on the tanh function to obtain the heart rate-body temperature time series implicit correlation feature vector. Next, in order to further refine the implicit information of the features and assign appropriate weights, the joint implicit features are transformed nonlinearly using activation mechanisms such as the Sigmoid function to emphasize key information and suppress irrelevant noise. The Sigmoid function, with its smooth and continuous change characteristics, can effectively highlight the most critical part of the heart rate-body temperature time series implicit correlation feature. Specifically, the heart rate-body temperature time series implicit correlation feature vector is activated based on the Sigmoid function to obtain the conditional feature vector.

[0030] Specifically, the significance-guided interactive fusion unit 342 is used to guide the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector to perform cross-domain interactive fusion encoding based on the attention mechanism based on the conditional feature vector to obtain the heart rate-body temperature time series feature significant interactive fusion representation vector. In an embodiment of the present application, first, based on the semantic relevance of the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector relative to the conditional feature vector, the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector are feature modulated to obtain the modulated heart rate value time series associated implicit feature vector and the modulated body temperature time series associated implicit feature vector; then, the modulated heart rate value time series associated implicit feature vector, the modulated body temperature time series associated implicit feature vector and the conditional feature vector are cross-domain interactively encoded based on the converter structure to obtain the heart rate-body temperature time series feature significant interactive fusion representation vector.

[0031] Among them, based on the semantic relevance of the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector relative to the conditional feature vector, the process of feature modulating the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector to obtain the modulated heart rate value time series associated implicit feature vector and the modulated body temperature time series associated implicit feature vector includes: calculating the first semantic relevance factor of the heart rate value time series associated implicit feature vector relative to the conditional feature vector; calculating the second semantic relevance factor of the body temperature time series associated implicit feature vector for the conditional feature vector; normalizing the first semantic relevance factor and the second semantic relevance factor, and using the normalized first semantic relevance factor and the second semantic relevance factor to weighted modulate the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector to obtain the modulated heart rate value time series associated implicit feature vector and the modulated body temperature time series associated implicit feature vector. That is, the correlation between the heart rate features and the body temperature features is evaluated by calculating the semantic correlation factor between each original feature vector and the conditional feature vector to determine the importance of each feature for the final representation. It is worth mentioning that in order to obtain a more accurate weight distribution, in other implementation schemes of the present application, it is possible to consider introducing advanced technologies such as mutual information estimation and canonical correlation analysis (CCA), so as to not only improve the quality of the final representation, but also enhance the model's ability to resist noise interference. Then, the above-mentioned semantic correlation factors are normalized and weighted modulated on the original features, and the weights of the heart rate value time series associated implicit feature vectors and the body temperature time series associated implicit feature vectors are dynamically adjusted according to the size of the semantic correlation factors to highlight the features that have a greater impact on the health status. Specifically, the process of calculating the first semantic relevance factor of the implicit feature vector associated with the heart rate value time series relative to the conditional feature vector includes: calculating the dot division vector between the implicit feature vector associated with the heart rate value time series and the conditional feature vector, and calculating the base-two logarithm of the absolute value of each eigenvalue in the dot division vector to obtain a semantic relevance representation vector; calculating the dot product vector between the semantic relevance representation vector and the implicit feature vector associated with the heart rate value time series, and calculating an exponential function value with base e and the sum of the eigenvalues ​​of the dot product vector as the exponent to obtain the first semantic relevance factor.

[0032] More specifically, the process of performing cross-domain interactive encoding based on the converter structure on the modulated heart rate value temporal association implicit feature vector, the modulated body temperature temporal association implicit feature vector and the conditional feature vector to obtain the heart rate-body temperature temporal feature significant interactive fusion representation vector includes: using the modulated heart rate value temporal association implicit feature vector as a query vector, the modulated body temperature temporal association implicit feature vector as a key vector and the conditional feature vector as a value vector, inputting the modulated heart rate value temporal association implicit feature vector, the modulated body temperature temporal association implicit feature vector and the conditional feature vector into the feature significant guided interaction module based on the converter structure to obtain the heart rate-body temperature temporal feature significant interactive fusion representation vector. That is, the modulated feature vector is input into the feature significant guided interaction module based on the converter structure, so as to utilize the design of the Transformer architecture to perform significant guided interaction between features, so as to realize adaptive fusion between different features. In particular, in this process, through the converter structure mechanism, the model can simultaneously examine the interactions between different features from multiple perspectives to discover the most valuable information flow path, so that the system can adaptively focus on the most important feature combination mode, thereby improving the overall expression ability. Specifically, first, the modulated heart rate value time series associated implicit feature vector is multiplied by the transposed vector of the modulated body temperature time series associated implicit feature vector and then divided by the square root of the length of the body temperature time series associated implicit feature vector to obtain the heart rate-body temperature time series feature query attention score matrix; further, the heart rate-body temperature time series feature query attention score matrix is ​​passed through the softmax function and multiplied by the conditional feature vector to obtain the heart rate-body temperature time series feature significant interaction fusion representation vector.

[0033] In summary, in the above embodiment, in the medical information collection and analysis server, the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector are interactively fused to obtain a heart rate-body temperature time series feature significant interactive fusion representation vector, including: processing the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector using the following feature interactive fusion formula to obtain the heart rate-body temperature time series feature significant interactive fusion representation vector, wherein the feature interactive fusion formula is:

[0034]

[0035] X c =sigmoid(X J )

[0036]

[0037] X1=X1w1

[0038] X2=X2w2

[0039]

[0040] Wherein, X1 represents the heart rate value time series associated implicit feature vector, X2 represents the body temperature time series associated implicit feature vector, Indicates adding by position point, X J represents the implicit correlation feature vector of heart rate-body temperature time series, W VT and b VT Represent the weight matrix and bias parameter of the neural network layer, tanh represents the hyperbolic tangent function, sigmoid represents the sigmoid activation function, and X c represents the conditional eigenvector, and represents the i-th eigenvalue in the heart rate value time series associated implicit feature vector, the i-th eigenvalue in the body temperature time series associated implicit feature vector and the i-th eigenvalue in the conditional feature vector, |·| is the absolute value symbol, log represents the logarithmic function with base 2, exp(·) represents the exponential function with base e, n represents the length of the heart rate value time series associated implicit feature vector, and the heart rate value time series associated implicit feature vector, the body temperature time series associated implicit feature vector and the conditional feature vector have the same length, CI1 and CI2 represent the first semantic relevance factor and the second semantic relevance factor, respectively, w1 and w2 represent the normalized first semantic relevance factor and the second semantic relevance factor, respectively, X1' and X2' represent the modulated heart rate value time series associated implicit feature vector and the modulated body temperature time series associated implicit feature vector, (·) T represents the transpose of a vector, represents matrix multiplication operation, softmax is the normalized exponential function, X r A representation vector representing the significant interaction fusion of the heart rate-body temperature time series features.

[0041] In particular, the physiological state analysis module 350 is used to determine whether the physiological state of the monitored elderly object is abnormal based on the heart rate-body temperature time series feature significant interaction fusion representation vector in the medical information collection and analysis server. In a specific example of the present application, the heart rate-body temperature time series feature significant interaction fusion representation vector is input into a classifier-based analysis result generator to obtain an analysis result, and the analysis result is used to indicate whether the physiological state of the monitored elderly object is abnormal. That is, by utilizing the powerful classification mapping ability of the classifier model, by learning the heart rate-body temperature time series feature significant interaction fusion information contained in the heart rate-body temperature time series feature significant interaction fusion representation vector, the characteristic patterns of normal and abnormal states are identified, thereby achieving real-time monitoring of health conditions.

[0042] Preferably, in an example of the present application, inputting the heart rate-body temperature time series feature significant interaction fusion representation vector into a classifier-based analysis result generator to obtain an analysis result includes:

[0043] Calculate the absolute value sum of all eigenvalues ​​of the heart rate-body temperature time series feature significant interaction fusion representation vector and the square root of the square sum to obtain the first heart rate-body temperature time series feature significant interaction fusion representation space structure value and the second heart rate-body temperature time series feature significant interaction fusion representation space structure value, that is:

[0044] w1=∑ i |v i |

[0045]

[0046] Among them, v i represents the ith eigenvalue of the heart rate-body temperature time series feature significant interactive fusion representation vector, w1 represents the first heart rate-body temperature time series feature significant interactive fusion representation space structure value, and w2 represents the second heart rate-body temperature time series feature significant interactive fusion representation space structure value;

[0047] Determine the number of eigenvalues ​​L of all eigenvalues ​​of the heart rate-body temperature time series feature significant interaction fusion representation vector, that is, the length of the heart rate-body temperature time series feature significant interaction fusion representation vector;

[0048] For each eigenvalue of the heart rate-body temperature time series feature significant interaction fusion representation vector, the first heart rate-body temperature time series feature significant interaction fusion representation spatial structure value minus the product of the eigenvalue and the number of eigenvalues ​​is calculated to obtain the first heart rate-body temperature time series feature significant interaction fusion representation long-range dependency value x i =w1-v i ×L, where w1 represents the significant interaction fusion of the first heart rate-body temperature time series feature and v represents the spatial structure value. i represents the ith eigenvalue of the heart rate-body temperature time series feature significant interaction fusion representation vector, L represents the number of eigenvalues ​​of all eigenvalues ​​of the heart rate-body temperature time series feature significant interaction fusion representation vector, x i Indicates that the significant interaction fusion of the first heart rate-body temperature time series features indicates a long-range dependency value;

[0049] Calculate the square root of the number of eigenvalues ​​multiplied by the product of the eigenvalues ​​minus the second heart rate-body temperature time series feature significant interaction fusion representation spatial structure value to obtain the second heart rate-body temperature time series feature significant interaction fusion representation long-range dependency value Among them, w2 represents the significant interactive fusion of the second heart rate-body temperature time series features to represent the spatial structure value, vi represents the ith eigenvalue of the heart rate-body temperature time series feature significant interaction fusion representation vector, L represents the number of eigenvalues ​​of all eigenvalues ​​of the heart rate-body temperature time series feature significant interaction fusion representation vector, y i Indicates that the significant interaction fusion of the second heart rate-body temperature time series features indicates a long-range dependency value;

[0050] The index value calculated by taking the first heart rate-body temperature time series feature significant interaction fusion indicating long-range dependency value as the exponent of the natural constant and the inverse of the second heart rate-body temperature time series feature significant interaction fusion indicating long-range dependency value are weighted summed to obtain the optimized eigenvalue corresponding to each eigenvalue Among them, x i Indicates that the first heart rate-body temperature time series features are significantly interactively fused to represent the long-range dependency value, y i Indicates the significant interaction fusion of the second heart rate-body temperature time series features, represents the long-range dependency value, e represents the natural constant, α and β represent weighted hyperparameters, and f' i Indicates the optimized eigenvalue corresponding to each eigenvalue;

[0051] The optimized feature values ​​are combined into an optimized heart rate-body temperature time series feature significant interaction fusion representation vector;

[0052] The optimized heart rate-body temperature time series feature significant interaction fusion representation vector is input into the classifier-based analysis result generator to obtain the analysis result.

[0053] That is, in view of the possible spatial structure loss in the high-dimensional space of the feature set of the heart rate-body temperature time series feature significant interaction fusion representation vector, which causes the weight of the classifier to implicitly infer the spatial structure information based on the features, resulting in inconsistent convergence. A long-distance feature dependency relationship is established based on the spatial structure representation of the heart rate-body temperature time series feature significant interaction fusion representation vector relative to the overall feature scale of the heart rate-body temperature time series feature significant interaction fusion representation vector, so as to establish the feature local connectivity of the heart rate-body temperature time series feature significant interaction fusion representation vector, and to capture the spatial ambiguous information of the object feature value through the unstructured feature value point prediction of the heart rate-body temperature time series feature significant interaction fusion representation vector, thereby improving the spatial inductive bias perception ability of the feature set of the heart rate-body temperature time series feature significant interaction fusion representation vector, improving the convergence consistency of the classifier, thereby improving the convergence and generalization effect of the feature set of the heart rate-body temperature time series feature significant interaction fusion representation vector for classification regression under complex spatial structure representation, and improving the accuracy of the analysis result obtained by the analysis result generator based on the classifier.

[0054] As described above, the multi-source medical information collection and analysis system 300 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a multi-source medical information collection and analysis algorithm. In one possible implementation, the multi-source medical information collection and analysis system 300 according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the multi-source medical information collection and analysis system 300 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the multi-source medical information collection and analysis system 300 can also be one of the many hardware modules of the wireless terminal.

[0055] Alternatively, in another example, the multi-source medical information collection and analysis system 300 and the wireless terminal may also be separate devices, and the multi-source medical information collection and analysis system 300 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0056] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A multi-source medical information collection and analysis system, characterized in that: include: A multi-source medical information collection module, used to collect a time queue of real-time heart rate and a time queue of real-time body temperature through a smart wearable medical device worn on a monitored elderly subject, wherein the time queue of real-time heart rate is used as the first source medical information, and the time queue of real-time body temperature is used as the second source medical information; A data transmission module, used for transmitting the first source medical information and the second source medical information to a medical information collection and analysis server through a wireless communication module; A medical information time series analysis module, used for performing forward time series coding on the first source medical information and the second source medical information, respectively, on the medical information collection and analysis server to obtain a heart rate value time series associated implicit feature vector and a body temperature time series associated implicit feature vector; A multi-source medical information interactive fusion module is used to interactively fuse the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector on the medical information collection and analysis server to obtain a heart rate-body temperature time series feature significant interactive fusion representation vector; A physiological state analysis module is used to determine whether the physiological state of the monitored elderly object is abnormal based on the heart rate-body temperature time series feature significant interaction fusion representation vector in the medical information collection and analysis server; Among them, the multi-source medical information interactive fusion module includes: a conditional information construction unit, which is used to extract the potential correlation features between the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector as a conditional feature vector; a significance guided interactive fusion unit, which is used to guide the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector to perform cross-domain interactive fusion encoding based on the attention mechanism based on the conditional feature vector to obtain the heart rate-body temperature time series feature significant interactive fusion representation vector.

2. The multi-source medical information collection and analysis system according to claim 1 is characterized in that: The medical information time series analysis module is used to: The first source medical information and the second source medical information are respectively input into a time series encoder based on a forward LSTM model to obtain the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector.

3. The multi-source medical information collection and analysis system according to claim 2 is characterized in that: The condition information construction unit includes: An implicit correlation feature extraction subunit, used for inputting the heart rate value time series correlation implicit feature vector and the body temperature time series correlation implicit feature vector into an implicit correlation feature capture network to obtain a heart rate-body temperature time series implicit correlation feature vector; The feature activation subunit is used to perform feature activation based on the Sigmoid function on the heart rate-body temperature time series implicit correlation feature vector to obtain the conditional feature vector.

4. The multi-source medical information collection and analysis system according to claim 3 is characterized in that: The implicit association feature extraction subunit is used to: The heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector are fused according to the position points and then passed through a neural network layer based on the tanh function to obtain the heart rate-body temperature time series implicit associated feature vector.

5. The multi-source medical information collection and analysis system according to claim 4 is characterized in that: The saliency-guided interactive fusion unit comprises: a feature modulation subunit, configured to perform feature modulation on the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector based on the semantic relevance of the heart rate value time series associated implicit feature vector and the body temperature time series associated implicit feature vector relative to the conditional feature vector to obtain a modulated heart rate value time series associated implicit feature vector and a modulated body temperature time series associated implicit feature vector; The cross-domain interactive coding subunit is used to perform cross-domain interactive coding based on the converter structure on the modulated heart rate value time series associated implicit feature vector, the modulated body temperature time series associated implicit feature vector and the conditional feature vector to obtain the heart rate-body temperature time series feature significant interactive fusion representation vector.

6. The multi-source medical information collection and analysis system according to claim 5, characterized in that: The characteristic modulation subunit comprises: A first semantic relevance measurement secondary subunit, used to calculate a first semantic relevance factor of the heart rate value temporal association implicit feature vector relative to the conditional feature vector; A second semantic relevance measurement secondary subunit, used to calculate a second semantic relevance factor of the temperature time series associated implicit feature vector with respect to the conditional feature vector; The weighted modulation secondary subunit is used to normalize the first semantic relevance factor and the second semantic relevance factor, and use the normalized first semantic relevance factor and the second semantic relevance factor to weighted modulate the heart rate value timing-associated implicit feature vector and the body temperature timing-associated implicit feature vector to obtain the modulated heart rate value timing-associated implicit feature vector and the modulated body temperature timing-associated implicit feature vector.

7. The multi-source medical information collection and analysis system according to claim 6, characterized in that: The first semantic relevance measurement secondary subunit is used to: Calculating a dotted division vector between the heart rate value time series associated implicit feature vector and the conditional feature vector, and calculating a base-two logarithm of the absolute value of each feature value in the dotted division vector to obtain a semantic correlation representation vector; Calculate the dot product vector between the semantic relevance representation vector and the heart rate value time series associated implicit feature vector, and calculate an exponential function value with e as the base and the sum of the eigenvalues ​​of the dot product vector as the exponent to obtain the first semantic relevance factor.

8. The multi-source medical information collection and analysis system according to claim 7, characterized in that: The cross-domain interaction encoding subunit is used to: Taking the implicit feature vector associated with the modulated heart rate value time series as the query vector, the implicit feature vector associated with the modulated body temperature time series as the key vector and the conditional feature vector as the value vector, the implicit feature vector associated with the modulated heart rate value time series, the implicit feature vector associated with the modulated body temperature time series and the conditional feature vector are input into a significant guided interaction module between features based on a converter structure to obtain the significant interaction fusion representation vector of the heart rate-body temperature time series features.

9. The multi-source medical information collection and analysis system according to claim 8, characterized in that: The physiological state analysis module is used to: The heart rate-body temperature time series feature significant interaction fusion representation vector is input into a classifier-based analysis result generator to obtain an analysis result, and the analysis result is used to indicate whether the physiological state of the monitored elderly subject is abnormal.

Images