CN118098625B

CN118098625B - Intelligent medical monitoring and early warning method and system based on disease diagnosis data

Info

Publication number: CN118098625B
Application number: CN202311679085.6A
Authority: CN
Inventors: 蒋天宏; 周正元; 张宁; 吕敏; 蒋烈; 龚利强; 徐里强; 周犇; 顾淑君; 张秋伊; 肖焕; 徐晓燕; 王辉; 张燕
Original assignee: CHANGSHU CENTER FOR DISEASE CONTROL AND PREVENTION
Current assignee: CHANGSHU CENTER FOR DISEASE CONTROL AND PREVENTION
Priority date: 2023-12-08
Filing date: 2023-12-08
Publication date: 2024-08-09
Anticipated expiration: 2043-12-08
Also published as: CN118098625A

Abstract

According to the intelligent medical monitoring and early warning method and system based on the disease diagnosis data, the intelligent medical monitoring and early warning system queries the corresponding trend thermal embedding vector from the target knowledge vector set aiming at the trend thermal matching vector of each case streaming event in the streaming trend thermal image stream, and considers that the target knowledge vector set is obtained by optimizing the initial knowledge vector set, the occupation of the storage space of the intelligent medical monitoring and early warning system is smaller than that of the initial knowledge vector set, and the corresponding streaming trend thermal image block is the same as that of the initial knowledge vector set, so that the traversal timeliness aiming at the knowledge vector set can be improved, the corresponding trend thermal embedding vector can be accurately and rapidly determined, and the corresponding streaming trend thermal image block can be accurately and timely determined aiming at each case streaming event through the trend thermal embedding vector, so that the streaming thermal map can be accurately and efficiently generated, and an analysis basis can be provided for epidemiological monitoring and early warning treatment accurately and reasonably in time.

Description

Intelligent medical monitoring and early warning method and system based on disease diagnosis data

Technical Field

The application relates to the technical field of data processing, in particular to an intelligent medical monitoring and early warning method and system based on disease diagnosis data.

Background

Big data visualization plays a vital role in smart medicine, and it presents complex data in an intuitive way, such as graphics and images, helping medical professionals, researchers, decision makers, and patients to better understand and analyze medical information. Applications of big data visualization in the smart medical field include visualization of Electronic Health Records (EHR), epidemic monitoring and early warning systems, equipment and asset management, etc.

First, the electronic health record system collects and stores health information for a large number of patients. Through visualization tools, doctors can quickly look into historical cases of patients, laboratory detection results, imaging examination pictures and the like. Secondly, by using big data visualization technology, the disease transmission model and the real-time monitoring data can be combined to generate an epidemic map and a trend chart, so that support is provided for public health decision (the treatment of the disease transmission model and the real-time monitoring data in the prior art CN111524611B, CN111524612B, CN113658713B and the like). For example, through hot spot display on a map, epidemic outbreak areas can be rapidly identified, resources can be allocated in time, and intervention measures can be taken. Then, the big data visualization can help a hospital management layer analyze medical service cost, identify areas with high cost and optimize resource allocation. For example, by visualizing the cost of different surgical types, drug use, and treatment regimens, hospitals can formulate more cost-effective treatment routes. It can be seen that big data visualization limits smart medicine not only to automated processing and analysis of medical data, but also includes the ability to provide deep holes and aid in decision making.

With the rapid development of smart medicine, the role of monitoring and early warning systems in epidemiology fields is becoming more important. The traditional disease monitoring method relies on manual collection and analysis of data, which is not only inefficient, but also prone to false alarms due to information delays or errors. In order to overcome the limitations, intelligent medical monitoring and early warning systems are proposed to realize automatic monitoring and timely early warning of disease transmission. In such a system, the generation of a flow-regulated trend thermal image flow is particularly critical, and the flow-regulated trend thermal image flow can convert complex epidemic data into visual images so as to help a decision maker to rapidly identify and respond to epidemic change.

Disclosure of Invention

The application provides at least an intelligent medical monitoring and early warning method and system based on disease diagnosis data.

The application provides a disease diagnosis data-based intelligent medical monitoring and early warning method, which is applied to an intelligent medical monitoring and early warning system, and comprises the following steps:

Combining the initial case streaming image record and streaming trend thermal image blocks of the previous x-1 case streaming events in the streaming trend thermal image stream to determine a trend thermal matching vector for the xth case streaming event in the streaming trend thermal image stream; the xth case streaming event is a case streaming event which needs streaming thermodynamic diagram mapping, and x is a positive integer;

aiming at trend thermal matching vectors of the xth case stream adjustment event, screening P trend thermal embedding vectors from a target knowledge vector set; wherein P is more than or equal to 2, and the target knowledge vector set is obtained by optimizing the initial knowledge vector set;

According to the characteristic difference between the P trend thermal embedding vectors and trend thermal matching vectors of the xth case streaming event, determining streaming trend thermal image blocks of the xth case streaming event;

When a streaming trend thermal image block is obtained for the case streaming event at the tail of the streaming trend thermal image stream, completing streaming thermodynamic diagram generation of the initial case streaming image record; the flow regulating trend thermal image flow is used for epidemiological monitoring and early warning treatment.

In some aspects, before the determining the trend thermal match vector for the x-th case streaming event in the streaming trend thermal image stream in combination with the streaming trend thermal image block of the first x-1 case streaming events in the initial case streaming image record and the streaming trend thermal image stream, the method further comprises:

generating a streaming track relation network aiming at each case streaming track in the streaming trend thermal image streaming sample by utilizing the time sequence characteristics of each case streaming track in the initial case streaming image recording sample and the streaming trend thermal image streaming sample; the initial case streaming image record sample and the streaming trend thermal image streaming sample are image sample doublets with different output modes but the same case label;

Generating a feature migration matrix aiming at the streaming track description vector corresponding to the streaming track relation network and each case streaming track, and creating the initial knowledge vector set based on the feature migration matrix of each case streaming track;

And carrying out optimization processing on the initial knowledge vector set to obtain the target knowledge vector set.

In some aspects, the optimizing the initial knowledge vector set to obtain the target knowledge vector set includes:

Integrating the knowledge vectors in the input knowledge vector set of the y-th round based on the feature differences among the knowledge vectors in the input knowledge vector set of the y-th round, and generating an output knowledge vector set of the y-th round based on the obtained feature integration result; wherein the incoming knowledge vector set of round 1 is the initial knowledge vector set; y is a positive integer;

determining an output knowledge vector set of a y-th round of rotation as an input knowledge vector set of a y+1th round of rotation, and starting the y+1th round of rotation;

When the cycle termination requirement is met, taking the output knowledge vector set of the last cycle as the target knowledge vector set, and completing the optimization processing on the initial knowledge vector set; wherein the cycle termination requirement includes at least one of the following conditions: and the downsampling coefficient between the output knowledge vector set and the initial knowledge vector set reaches a downsampling threshold, and y reaches the maximum cycle.

In some aspects, the feature differences between knowledge vectors in the set of knowledge vectors based on the y-th round of rotation integrate knowledge vectors in the set of knowledge vectors for the y-th round of rotation, and generate an output set of knowledge vectors for the y-th round of rotation based on the obtained feature integration result, including:

Determining an integration priority of each of the knowledge vectors in the incoming knowledge vector set of the y-th round of rotation based on a feature difference between each of the knowledge vectors in the incoming knowledge vector set of the y-th round of rotation and a y-th associated knowledge vector of each of the knowledge vectors;

According to the integration priority, carrying out consistency analysis on the case flow dispatching track corresponding to each knowledge vector and the case flow dispatching track corresponding to the y-th associated knowledge vector to obtain a consistency analysis result;

When the consistency analysis result represents that the case flow dispatching track corresponding to each knowledge vector is the same as the case flow dispatching track corresponding to the y-th associated knowledge vector, integrating each knowledge vector with the y-th associated knowledge vector to obtain the feature integration result of the y-th round of circulation;

And generating an output knowledge vector set of the y-th round by using the feature integration result of the y-th round and the case flow adjustment track corresponding to each knowledge vector.

In some aspects, the integrating each knowledge vector with the y-th associated knowledge vector to obtain the feature integration result of the y-th round of circulation includes:

Determining an integration index for each of the knowledge vectors and the y-th associated knowledge vector based on the first query frequency of each of the knowledge vectors and the second query frequency of the y-th associated knowledge vector; the first query frequency represents the queried times of each knowledge vector, and the second query frequency represents the queried times of the y-th associated knowledge vector;

And carrying out knowledge reinforcement on each knowledge vector and the y-th associated knowledge vector according to the integration index of each knowledge vector and the integration index of the y-th associated knowledge vector to obtain the feature integration result of the y-th round of circulation.

In some aspects, the determining the integration index for each of the knowledge vectors and the y-th associated knowledge vector based on the first query frequency for each of the knowledge vectors and the second query frequency for the y-th associated knowledge vector includes:

when the first query frequency and the second query frequency are both 0, determining a first set index variable as an integrated index of each knowledge vector, and determining a second set index variable as an integrated index of a y-th associated knowledge vector;

when at least one of the first query frequency and the second query frequency is greater than 0, determining the ratio of the first query frequency to the sum of the first query frequency and the second query frequency as an integrated index of each knowledge vector, and determining the difference value of the integrated index of 1 and each knowledge vector as an integrated index of the y-th associated knowledge vector.

In some aspects, the y-th associated knowledge vector is a y-th of the Q associated knowledge vectors corresponding to each of the knowledge vectors; q associated knowledge vectors are obtained by digging before the 1 st round of circulation is started, and the Q associated knowledge vectors are sorted according to the ascending order of the characteristic difference of each knowledge vector; wherein Q is the maximum cycle.

Determining an integration priority of each of the knowledge vectors in the input knowledge vector set of the y-th round based on a feature difference between each of the knowledge vectors in the input knowledge vector set of the y-th round and a core associated knowledge vector of each of the knowledge vectors;

according to the integration priority, carrying out consistency analysis on the case flow dispatching track corresponding to each knowledge vector and the case flow dispatching track corresponding to the core associated knowledge vector to obtain a consistency analysis result;

When the consistency analysis result represents that the case flow dispatching track corresponding to each knowledge vector is identical to the case flow dispatching track corresponding to the core associated knowledge vector, integrating each knowledge vector with the core associated knowledge vector to obtain the characteristic integration result of the y-th round of circulation;

and generating the output knowledge vector set of the y-th round by utilizing the feature integration result of the y-th round and the case flow adjustment track corresponding to each knowledge vector.

In some aspects, the determining a trend thermal match vector for an x-th case streaming event in the streaming trend thermal image stream in combination with the streaming trend thermal image block of the first x-1 case streaming events in the initial case streaming image record and the streaming trend thermal image stream comprises:

Generating a streaming track relation network of the x-th case streaming event by utilizing the initial case streaming image record and streaming trend thermal image blocks of the previous x-1 case streaming events in the streaming trend thermal image stream;

and determining an streaming track description vector corresponding to the streaming track relation network as the trend thermal matching vector of the x-th case streaming event in the streaming trend thermal image stream.

In some aspects, the determining the streaming trend thermal image block of the x-th case streaming event according to the feature differences between the P trend thermal embedding vectors and the trend thermal matching vectors of the x-th case streaming event, respectively, includes:

According to the characteristic difference between the P trend thermal embedding vectors and the trend thermal matching vectors of the x-th case flow adjustment event respectively and the flow adjustment trend thermal judgment result generated by the pre-configured visual simulation network aiming at the trend thermal matching vectors of the x-th case flow adjustment event, determining the target flow adjustment trend thermal judgment result of the trend thermal matching vectors of the x-th case flow adjustment event in the flow adjustment region;

And determining the flow regulation trend thermal image block of the x-th case flow regulation event from the flow regulation region contained in the flow regulation region according to the target flow regulation trend thermal discrimination result.

In some aspects, the determining the target streaming trend thermal discrimination result of the trend thermal match vector of the x-th case streaming event in the streaming area according to the feature difference between the P trend thermal embedding vectors and the trend thermal match vector of the x-th case streaming event, and the streaming trend thermal discrimination result generated by the pre-configured visual simulation network for the trend thermal match vector of the x-th case streaming event, includes:

According to the characteristic difference between the P trend thermal embedding vectors and trend thermal matching vectors of the xth case stream adjustment event, determining the position characteristics corresponding to the P trend thermal embedding vectors respectively;

updating the position features by utilizing the integrated computing power of the target knowledge vector set to obtain updated position features;

generating a flow adjustment trend thermal discrimination result of the trend thermal matching vector of the x-th case flow adjustment event in the flow adjustment region by using updated position features of the P trend thermal embedding vectors;

And determining the target flow regulation trend thermal discrimination result of the trend thermal match vector of the x-th case flow regulation event in the flow regulation region by utilizing the flow regulation trend thermal discrimination result of the trend thermal match vector of the x-th case flow regulation event in the flow regulation region and the aggregation information of the flow regulation trend thermal discrimination result generated by the pre-allocation visual simulation network aiming at the trend thermal match vector of the x-th case flow regulation event.

The application also provides an intelligent medical monitoring and early warning system, which comprises a processor and a memory; the processor is in communication with the memory, and the processor is configured to read and execute a computer program from the memory to implement the method described above.

The present application also provides a computer readable storage medium having stored thereon a computer program which, when run, implements the method described above.

The technical scheme provided by the embodiment of the application can comprise the following beneficial effects: aiming at the trend thermal matching vector of each case flow adjustment event in the flow of the flow adjustment trend thermal image, the intelligent medical monitoring and early warning system inquires the corresponding trend thermal embedding vector from the target knowledge vector set, considers that the target knowledge vector set is obtained by optimizing the initial knowledge vector set, and the occupation of the storage space is smaller than that of the initial knowledge vector set, and the corresponding flow adjustment trend thermal image block is the same as that of the initial knowledge vector set, so that the traversing timeliness aiming at the knowledge vector set can be improved, the corresponding trend thermal embedding vector is accurately and rapidly determined, and the corresponding flow adjustment trend thermal image block is accurately and timely determined for each case flow adjustment event through the trend thermal embedding vector, so that the flow adjustment thermal map can be accurately and efficiently generated, and an analysis basis is provided for epidemiological monitoring and early warning processing timely, accurately and reasonably.

For a description of the effects of the intelligent medical monitoring and early warning system, the computer readable storage medium, see the description of the method.

In order to make the above objects, features and advantages of the present application more comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are necessary for the embodiments to be used are briefly described below, the drawings being incorporated in and forming a part of the description, these drawings showing embodiments according to the present application and together with the description serve to illustrate the technical solutions of the present application. It is to be understood that the following drawings illustrate only certain embodiments of the application and are therefore not to be considered limiting of its scope, for the person of ordinary skill in the art may admit to other equally relevant drawings without inventive effort.

Fig. 1 is a block diagram of an intelligent medical monitoring and early warning system according to an embodiment of the present application.

Fig. 2 is a flow chart of an intelligent medical monitoring and early warning method based on disease diagnosis data according to an embodiment of the application.

Detailed Description

Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples do not represent all implementations consistent with the application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the application.

Fig. 1 is a schematic diagram of a smart medical monitoring and early warning system 10 according to an embodiment of the present application, which includes a processor 102, a memory 104, and a bus 106. The memory 104 is used for storing execution instructions, including a memory and an external memory, where the memory may also be understood as an internal memory, and is used for temporarily storing operation data in the processor 102 and data exchanged with the external memory such as a hard disk, where the processor 102 exchanges data with the external memory through the memory, and when the intelligent medical monitoring and early-warning system 10 operates, the processor 102 and the memory 104 communicate through the bus 106, so that the processor 102 executes the intelligent medical monitoring and early-warning method based on disease diagnosis data according to the embodiment of the present application.

Referring to fig. 2, fig. 2 is a flow chart of a smart medical monitoring and early warning method based on disease diagnosis data, which is applied to a smart medical monitoring and early warning system and can include the following steps 110-140.

Step 110, determining a trend thermal matching vector for an xth case streaming event in the streaming trend thermal image stream by combining the initial case streaming image record and the streaming trend thermal image block of the previous x-1 case streaming events in the streaming trend thermal image stream; the x-th case streaming event is a case streaming event which needs streaming thermodynamic diagram mapping, and x is a positive integer.

Step 120, aiming at the trend thermal matching vector of the x-th case flow adjustment event, screening P trend thermal embedding vectors from a target knowledge vector set; and P is more than or equal to 2, and the target knowledge vector set is obtained by optimizing the initial knowledge vector set.

And 130, determining the streaming trend thermal image block of the x-th case streaming event according to the characteristic difference between the P streaming thermal embedding vectors and the streaming thermal matching vectors of the x-th case streaming event.

Step 140, when a flow adjustment trend thermal image block is obtained for the case flow adjustment event at the tail of the flow adjustment trend thermal image flow, completing flow adjustment thermodynamic diagram generation about the initial case flow adjustment image record; the flow regulating trend thermal image flow is used for epidemiological monitoring and early warning treatment.

By applying the steps 110-140, the intelligent medical monitoring and early warning system queries the corresponding trend thermal embedding vector from the target knowledge vector set aiming at the trend thermal matching vector of each case streaming event in the streaming trend thermal image stream, considers that the target knowledge vector set is obtained by optimizing the initial knowledge vector set, and the occupation of the storage space of the intelligent medical monitoring and early warning system is smaller than that of the initial knowledge vector set, and the corresponding streaming trend thermal image block is the same as that of the initial knowledge vector set, so that the traversal timeliness aiming at the knowledge vector set can be improved, the corresponding trend thermal embedding vector can be accurately and quickly determined, and the corresponding streaming trend thermal image block can be accurately and timely determined aiming at each case streaming event through the trend thermal embedding vector, so that the streaming thermal map can be accurately and efficiently generated, and analysis basis can be provided for epidemiological monitoring and early warning treatment timely, accurately and reasonably.

In step 110, the initial case streaming image record is a series of initial or early case survey data recorded in image form. These images may contain information about geographic location, time stamp, number of patients, etc., which are used as the basis for subsequent analysis.

Streaming trend thermal image streaming is a dynamic data set, each data point representing a case streaming event, showing the trend of disease progression over time. These data may be represented in the form of a thermodynamic diagram in which the shades of color represent the intensity or frequency of certain variables (e.g., number of cases).

Case-flow events refer to epidemiological investigation activities related to a particular case, including the identification, recording, analysis, and tracking of cases, etc.

A streaming trend thermal image block refers to a specific region or "block" of a streaming event for a single case in the overall thermal image stream that reflects the temporal and spatial characteristics of the event.

The trend thermal match vector is a mathematical tool for describing and comparing trends of case-stream events. Each element in the vector may represent a different aspect of the case signature, such as morbidity, speed of propagation, etc.

The case-streaming event that is in need of streaming thermodynamic map mapping is a case event that is currently being analyzed and is ready for addition to the thermodynamic image stream.

Taking respiratory infections as an example, thermodynamic diagrams are now used to monitor and predict the trend of the transmission of the respiratory infections being transmitted.

With respect to step 110, the initial case data is collected and converted to an image format (initial case streaming image record). Existing x-1 case-stream events are analyzed, and the data of the events are converted into a series of thermal image blocks to form a continuous thermal image stream reflecting the spreading trend of the disease from the beginning to the present. For the x-th case streaming event, a trend thermal match vector is determined based on the data of the previous x-1 events. This vector captures the trend characteristics of the case currently being analyzed and compares it to known trends. In this process, the system may analyze the case-streaming image records using a machine learning algorithm and generate trend thermal match vectors. These vectors may then be used to predict future trends in the disease and as part of a real-time monitoring and early warning system.

In other possible examples, an example trend thermal match vector may be constructed, which contains hypothetical values that characterize different cases of an infectious disease (e.g., pneumonia) over a particular period of time.

To construct this vector, it is first defined what case features each element will represent. For example, there may be the following elements:

morbidity (number of new cases per day divided by total population);

propagation speed (rate of increase of number of new cases per day compared to the previous day);

Geographic spread (change in the number of affected regions);

average length of stay (average number of stay per patient);

Disease severity indicators (e.g., the proportion of inpatients in the new case);

Now, this vector is filled with some specific values. Assuming that the x-th case-stream event is being observed, the following data may be obtained:

incidence of disease: 0.00002 (assuming 100 additional cases are in a 50-ten thousand population city);

propagation speed: 1.05 (indicating that the number of cases increased by 5% over the previous day);

Geographical diffusion range: 2 (indicating that the affected area has increased from 10 to 12 from the previous day);

Average length of stay: 7 days (average patient hospitalization for one week);

disease severity index: 0.2 (indicating that 20% of the new cases require hospitalization)

Based on the above information, the trend thermal match vector may be:

[ \text { vector } = (0.00002,1.05,2,7,0.2) ]

This vector captures a number of important trends and features of the epidemiological event being observed at specific points in time. In practice, this vector may be used to compare with similar vectors at other points in time or in other regions to identify patterns, predict trends, or evaluate progression of epidemics. By using machine learning algorithms, the vectors can be used as inputs to train the model and help make more accurate predictions and response measures.

In step 120, the set of target knowledge vectors refers to a set of knowledge vectors, each of which represents a disease propagation pattern, a trend of popularity, or other important information related thereto. These vectors may be derived from historical data, scientific research, or existing epidemiological models and used to compare with new case-stream events.

The trending thermal embedding vector is a vector selected from a set of target knowledge vectors for comparison with the trending thermal matching vector of the particular case-stream event. They represent the knowledge and trends that the system considers to be most relevant to the current event.

The initial set of knowledge vectors is the original set of vectors prior to being optimized to form the target set of knowledge vectors. These vectors may be generated based on previous case data, epidemiological studies or public health records as a preliminary compilation of knowledge of disease transmission.

Optimization refers to the process of processing an initial set of knowledge vectors using algorithms to improve their relevance and accuracy to current epidemic data. Optimization may involve operations such as deleting vectors that are no longer relevant, merging similar vectors, updating vectors to reflect new data, etc.

For example, respiratory infectious disease, the specific application in step 120 is as follows: it is assumed that the system is monitoring for an outbreak of respiratory infections, such as pneumonia. A series of initial case-flow image records have been collected and a flow trend thermal image stream has been generated from the first x-1 cases. For the x-th case being analyzed (e.g., a newly reported case), the system has created a trend thermal match vector from the existing thermal image stream.

Now, in step 120, a set of already optimized target knowledge vectors is used, which may include case-like vectors, propagation mode vectors, prevalence change vectors, etc. From this optimized set, the system will pick P trending thermodynamic embedding vectors that are most similar to the trending thermodynamic match vector of the x-th case or most likely reveal useful information. This selection process may rely on machine learning algorithms that can identify which knowledge vectors best match the features of the current case.

In this way, the background of each new case can be better understood and its possible propagation trend predicted to help in the formulation of preventive measures.

In other examples, the trending thermal embedding vector is used to compare the trending thermal matching vector to a related vector in the set of target knowledge vectors to find a vector that is most similar or most representative of the current case-stream event characteristics. In practice, these embedded vectors may be derived through various machine learning techniques, such as by training a model to identify different trends and patterns.

The following is a numerical example provided for two hypothetical trend thermal embedding vectors based on the previously constructed trend thermal matching vector.

The trend thermal match vector (for the x-th case-stream tuning event) is:

[ \text { matching vector } = (0.00002,1.05,2,7,0.2) ]

Now, it is assumed that the set of target knowledge vectors has a series of trend thermal embedding vectors representing disease propagation trends at different points in time or in different regions. The following are the values of two of the vectors:

trend thermal embedding vector a (which may represent a past event similar to the current case): [ \text { embedded vector a } = (0.000018,1.04,3,6,0.18) ].

Trend thermal embedding vector B (which may represent another less similar past event):

[ \text { embedding vector B } = (0.00003,1.10,1,5,0.25) ]

The elements contained in each embedded vector are consistent with the matching vector and represent the same case characteristics, namely morbidity, transmission speed, geographic spread range, average hospitalization duration and disease severity index.

By calculating the similarity between these embedded vectors and the matching vectors (e.g., using metrics such as euclidean distance, cosine similarity, etc.), it is possible to evaluate which embedded vector is closer to the characteristics of the case-streaming event currently being analyzed. In this example, if only A and B are compared, it may be found that vector A is numerically closer to the matching vector, and thus it may be chosen to better represent the trend of the xth event.

In epidemiological monitoring and early warning systems, selection of the embedded vector that best matches the current event may better understand the disease's pattern of transmission and take appropriate countermeasures accordingly.

In step 130, in the context of data analysis and machine learning, feature differences refer to differences in the respective feature dimensions between different data entities (trend thermal embedding vectors and trend thermal matching vectors in this example). These differences can be quantified in a number of ways, for example by calculating the euclidean distance, manhattan distance, or cosine similarity between the two vectors, etc. The feature differences help identify which factors differ significantly from known events in new case-streaming events.

Taking respiratory tract infections (e.g., pneumonia) as an example, the operation of performing step 130 may include the following: collecting data: a trend thermal match vector is generated from the xth case-stream event being analyzed. This vector may contain information about the case's morbidity, spread rate, geographical spread, average length of stay, disease severity index, etc.; selecting a reference vector: several trend thermal embedding vectors are selected from the P trend thermal embedding vectors screened in step 120 for comparison. These vectors may represent historically similar cases or predicted results from other epidemiological models; calculating the difference: for each trend thermal embedding vector, the feature differences between it and the trend thermal matching vector of the current case are calculated. This may involve subtracting the values of the corresponding features in the vector, or using more complex statistical methods to evaluate the differences; determining a thermal image block: based on these differences, it is decided how to integrate the data of the current case into the existing flow trend thermal image stream. If the current case differs little from some trend thermal embedding vector, it may be considered consistent with previous patterns and mapped into thermodynamic diagrams accordingly. If the differences are large, it may be necessary to create a new thermal image block or to make adjustments to existing image blocks to reflect these new findings.

For example, assume that the trend thermal match vector for the x-th case is (0.00002,1.05,2,7,0.2), and that the two trend thermal insert vectors selected from the set of target knowledge vectors are (0.000015,1.03,3,5,0.15) and (0.00003,1.08,1,9,0.25), respectively. By calculating the feature differences, it can be found that the xth case is closer to the first vector in terms of propagation velocity and geographic spread, but closer to the second vector in terms of disease severity. Thus, the output of step 130 may be a new thermal image block that combines the information of the two trending thermal embedding vectors to describe the x-th case.

In step 140, the last of the case-streaming events refers to the last case recorded in a certain time period, or in a certain dataset. In real-time monitoring systems, the so-called "end" may be updated continuously, as new cases continue to be reported and recorded.

Flow-mediated thermodynamic diagram generation involves creating a complete thermodynamic diagram based on all case-mediated events that have been analyzed. This thermodynamic diagram shows the trend of the disease over time and place, for example, it can show which areas show an increase in the number of cases and which areas remain stable or decrease.

The flow trend thermal image flow is used for epidemiological monitoring and early warning processing to indicate that the generated flow trend thermal image flow is not only used for recording past case information, but also used for analyzing the trend and mode of disease transmission so as to provide early warning for future epidemic development.

Taking respiratory infections (e.g., pneumonia) as an example, the specific operations in step 140 may include the following: it is assumed that each new case of pneumonia has been tracked and recorded over a period of time, and that the first x-1 cases have been analyzed for flow trend thermal image flow. When the x-th case, the last case in the current dataset, is analyzed, a streaming trend thermal image block for that case is generated based on the previous steps 110, 120, and 130. Once the image block of the x-th case is determined, it can be added to the existing flow trend thermal image stream, thus completing thermodynamic diagram generation over the entire time period.

The complete thermodynamic diagram provides an intuitive tool by which the development trend of epidemic situation can be monitored, the possible hot spot areas can be identified, and corresponding public health strategies and intervention measures can be formulated accordingly. Finally, the process forms a dynamic monitoring and early warning system which can be updated with the continuous entry of new cases and provides real-time epidemic information.

The scheme described in step 140 emphasizes that the generation and updating of the flow-regulated trend thermal image flow is a continuous process, and the thermodynamic diagram continuously reflects the latest epidemiological situation as new data is added. By the above method, data-based decisions can be made faster.

In this way, a highly dynamic and gradually optimized epidemic monitoring system can be built. Trending thermodynamic match vectors are first created from historical data and then compared with an optimized knowledge vector set using these vectors to screen out trending thermodynamic embedding vectors most relevant to the current case event. This process not only enhances the depth of understanding of existing data, but also enhances the ability to predict new occurrence trends. Through continuous iteration and optimization, the disease transmission mode can be mapped more accurately along with the time and data accumulation, and can be visualized into a thermodynamic diagram, so that public health decision making and emergency response are effectively supported. The method improves the real-time performance and accuracy of epidemic situation analysis, makes epidemiological early warning treatment more scientific and systematic, provides data support for prevention and control measures, and finally aims to slow down disease transmission speed and reduce public health risks.

For the embodiment of the present application, it should be noted that the technical solutions described in steps 110 to 140 are a method for epidemiological monitoring and early warning processing, and relate to technical steps of data analysis, pattern recognition and information image generation. Specifically, the scheme includes: determining a trend thermal matching vector of the next case event through the initial case streaming image record and the analysis of the previous x-1 cases; screening out related trend thermal embedded vectors from the optimized target knowledge vector set based on the trend thermal matching vectors; deciding how to represent the selected vector in the flow trend thermal image block according to the characteristic difference between the selected vector and the current case; and the generation of the whole flow regulating trend thermal image flow is completed and is used for epidemiological monitoring and early warning. The above steps are all technical activities of data processing and image generation, which do not involve diagnostic or therapeutic activities performed directly on the human or animal body. These steps are computer-implemented automated processes aimed at providing insight into the trend of disease transmission and assisting in formulating public health response measures, rather than diagnosing or treating individual patients directly. Therefore, these steps involve more technical problems of software algorithms, data analysis, image processing, etc., than clinical medical practice.

Still further, from another perspective, the solution described in steps 110-140 above describes an intelligent medical monitoring and pre-warning system that generates a thermodynamic diagram for epidemiological monitoring by analyzing data in a flow-regulated trend thermodynamic image stream. The technical scheme focuses on data processing, optimizes the use of storage space, improves timeliness of traversing the knowledge vector set, and finally achieves the aim of quickly and accurately determining the flow adjustment trend thermal image block. These steps are information processing procedures for monitoring and analyzing the trend of disease transmission and do not directly relate to diagnostic or therapeutic procedures performed on the human or animal body.

Therefore, the technical scheme is mainly applied to data collection, analysis and visualization, and is not directly applied to medical behaviors of patients. It provides a decision-making aid that helps health departments and epidemiologists better understand and predict the pattern of disease transmission, thereby making a more effective public health response. Such a system can be seen as an advanced data processing and pattern recognition tool, the outcome of which is a reference for guiding but not replacing the diagnosis and treatment decisions made by medical professionals.

Further, the technical solutions described in steps 110-140 are not directly applied to the diagnosis or treatment process in medical practice, but are a tool and method for data analysis and trend prediction, which can be obtained by comprehensive analysis in combination with the prior art such as CN111524611B, CN111524612B, CN 113658713B.

In some possible embodiments, before determining a trend thermal match vector for the x-th case pitch event in the streaming trend thermal image stream, in conjunction with the streaming trend thermal image block of the first x-1 case pitch events in the initial case pitch image record and streaming trend thermal image stream described in step 110, the method further comprises steps 210-230.

Step 210, using the time sequence characteristics of each case streaming track in the initial case streaming image record sample and the streaming trend thermal image streaming sample to generate a streaming track relation network for each case streaming track in the streaming trend thermal image streaming sample; the initial case streaming image record sample and the streaming trend thermal image streaming sample are image sample doublets with different output modes and the same case label.

Step 220, generating a feature migration matrix according to the streaming track description vector corresponding to the streaming track relation network and each case streaming track, and creating the initial knowledge vector set based on the feature migration matrix of each case streaming track.

And 230, performing optimization processing on the initial knowledge vector set to obtain the target knowledge vector set.

Steps 210 through 230 described above involve preparation before generating the trending thermodynamic match vectors, including building a streaming track relationship network, generating feature migration matrices, and optimizing the initial knowledge vector set. In the following it is illustrated in connection with an example of an infectious respiratory disease,

In step 210, a relationship net is created using actual case data of respiratory infections (e.g., pneumonia). For example, assume that there is a set of initial case-streaming image record samples, each of which may contain information about the patient's age, sex, geographic location, contact history, etc. At the same time, there are corresponding streaming trend thermal image streaming samples that show the spread trend and range of the same set of cases over time. Using the timing characteristics in these samples (e.g., the time interval for the disease to spread from one person to another), a streaming track relationship network can be constructed that describes the connections between cases, which helps reveal how the disease spreads among the population.

In step 220, a feature migration matrix is generated, one feature migration matrix being generated for each case flow trace based on the flow trace relationship network in step 210. This matrix may contain the transition probabilities or intensities of different features on the trajectory, such as the probability of passing from one community to another, or the average time of the disease from asymptomatic to complication. Using these matrices, an initial set of knowledge vectors can be created that reflects the preliminary knowledge of the disease propagation mode.

Finally, in step 230, by analyzing the initial knowledge vector sets obtained in step 220 and utilizing machine learning and data analysis techniques, these vector sets can be optimized, noise removed, inaccurate information updated, and data quality improved. The target knowledge vector set obtained in this way is refined and can be more effectively used for the generation of the flow adjustment trend thermal image flow and the analysis of the case flow adjustment event in the subsequent steps.

Through the steps, a solid data base can be provided for the epidemiological monitoring and early warning processing system, which can enable the system to be more accurate in trend analysis and prediction.

In a preferred embodiment, the optimization process performed on the initial knowledge vector set in step 230 obtains the target knowledge vector set, including steps 231-233.

Step 231, integrating knowledge vectors in the input knowledge vector set of the y-th round based on feature differences among knowledge vectors in the input knowledge vector set of the y-th round, and generating an output knowledge vector set of the y-th round based on the obtained feature integration result; wherein the incoming knowledge vector set of round 1 is the initial knowledge vector set; y is a positive integer.

Step 232, determining the output knowledge vector set of the y-th round of loops as the input knowledge vector set of the y+1st round of loops, and starting the y+1st round of loops.

Step 233, when the cycle termination requirement is met, taking the output knowledge vector set of the last cycle as the target knowledge vector set, and completing the optimization processing on the initial knowledge vector set; wherein the cycle termination requirement includes at least one of the following conditions: and the downsampling coefficient between the output knowledge vector set and the initial knowledge vector set reaches a downsampling threshold, and y reaches the maximum cycle.

The above embodiments describe an iterative process for optimizing an initial knowledge vector set to generate a target knowledge vector set. This process involves feature difference analysis of the loop, integration of knowledge vectors, and stopping the loop according to set termination conditions. This technique can be specified using data management of respiratory infections (e.g., pneumonia) as an example.

For example, there is a large initial set of knowledge vectors that contain thousands to tens of thousands of multidimensional feature vectors for cases of pneumonia. These characteristics may include information on the location, time of case findings, age and sex of the patient, virus variant information, clinical manifestations of the patient, etc.

In round 1, starting from this initial knowledge vector set. First, feature differences between vectors are calculated, such as by Euclidean distance or other similarity metric methods. Based on these differences, knowledge vector integration is performed, for example, using clustering algorithms to merge similar cases together, or using dimensionality reduction techniques such as Principal Component Analysis (PCA) to reduce the dimensionality of the feature space. Finally, a set of output knowledge vectors for round 1 is generated that is more compact than the input set, while preserving as much important information as possible of the original data.

The set of output knowledge vectors obtained in step 231 is taken as the set of input knowledge vectors for the 2 nd round of loops, and then the process of step 231 is repeated. Each iteration may further refine the data, extract more meaningful feature representations, or classify cases more effectively.

This process continues until the preset cycle termination requirements are met. For example, if after several iterations, the size of the output knowledge vector set has been scaled down to a certain proportion of the initial set (i.e., the downsampling factor reaches a threshold), or a set maximum round of looping has been performed, then the iteration will stop. The set of output knowledge vectors at this time is determined as the set of target knowledge vectors.

In practical situations, the optimization process can help public health professionals to quickly identify key trends and modes of epidemic situations, and simultaneously reduces resources required for data storage and processing. The optimized set of target knowledge vectors may be more efficiently used in a monitoring system to support epidemiological pre-warning and decision making. For example, when a new case-stream event occurs, the system can more quickly find the historical case features related to the new case-stream event from the target knowledge vector set, thereby accelerating epidemic analysis and response speed.

In some exemplary embodiments, feature differences between knowledge vectors in the set of input knowledge vectors based on the y-th round of rotation in step 231 are integrated for knowledge vectors in the set of input knowledge vectors of the y-th round of rotation and based on the resulting feature integration results, an output set of knowledge vectors of the y-th round of rotation is generated, comprising steps 2311-2314.

Step 2311, determining an integration priority of each of the knowledge vectors in the input knowledge vector set of the y-th round based on a feature difference between each of the knowledge vectors in the input knowledge vector set of the y-th round and a y-th associated knowledge vector of each of the knowledge vectors.

And 2312, according to the integration priority, performing consistency analysis on the case flow dispatching track corresponding to each knowledge vector and the case flow dispatching track corresponding to the y-th associated knowledge vector to obtain a consistency analysis result.

And 2313, when the consistency analysis result represents that the case flow dispatching track corresponding to each knowledge vector is the same as the case flow dispatching track corresponding to the y-th associated knowledge vector, integrating each knowledge vector with the y-th associated knowledge vector to obtain the feature integration result of the y-th round of circulation.

Step 2314, generating an output knowledge vector set of the y-th round by using the feature integration result of the y-th round and the case flow adjustment track corresponding to each knowledge vector.

The above exemplary embodiments describe an iterative optimization process aimed at improving disease prevalence trend prediction models by integrating relevant knowledge vectors. This process may be applied in epidemiological data analysis of respiratory infections such as pneumonia. The following description will be made in connection with this example.

First, there is an initial set of knowledge vectors that contain various features about the pneumonia transmission trace, such as infection rate, time series of symptoms appearance, etc. These vectors may be generated from historical data or epidemiological models.

At the beginning of the y-th round, the system will evaluate the feature differences between each knowledge vector and its associated vector. These differences may be based on a variety of criteria, such as similarity, distance measure, or correlation. Based on these differences, the system determines the integration priority, i.e., decides which knowledge vectors are more likely to provide valuable information, and should prioritize integration.

And carrying out consistency analysis on the case flow adjustment track corresponding to each knowledge vector and the track corresponding to the associated knowledge vector according to the integration priority. This analysis checks whether both traces show similar disease transmission patterns or trends. The resulting consistency analysis results will indicate which knowledge vectors are coordinated with each other.

If the consistency analysis results show that the trajectories corresponding to certain knowledge vectors are identical or highly consistent, these vectors will be integrated. The integration operation may involve merging, averaging or employing more complex mathematical models to integrate the information of these vectors to enhance the system's understanding of disease propagation trends.

And generating a new output knowledge vector set by using the results obtained from the feature integration and the case flow tuning track corresponding to each knowledge vector. This updated vector set will reflect more accurate and comprehensive disease transmission information after integration.

In practical applications, such as monitoring pneumonia epidemic situation, the iterative optimization process can continuously improve accuracy of disease propagation models. With the completion of each cycle, the predictive model becomes more reliable, can better delineate the spread trend of the disease, helping public health authorities to formulate more effective intervention strategies and emergency measures. The technical scheme ensures that the epidemiological monitoring and early warning system is more efficient and accurate in processing large-scale health data and is vital for quick response to public health events.

Under some exemplary design considerations, each knowledge vector is integrated with the y-th associated knowledge vector in step 2313 to obtain the feature integration result of the y-th round of loops, including steps 23131-23132.

Step 23131, determining an integration index for each knowledge vector and the y-th associated knowledge vector based on the first query frequency of each knowledge vector and the second query frequency of the y-th associated knowledge vector; the first query frequency represents the number of times each knowledge vector is queried, and the second query frequency represents the number of times the y-th associated knowledge vector is queried.

And 23132, performing knowledge enhancement on each knowledge vector and the y-th associated knowledge vector according to the integration index of each knowledge vector and the integration index of the y-th associated knowledge vector to obtain the feature integration result of the y-th cycle.

The knowledge vector integration referred to in step 2313 is an important element in the optimization process, which involves integrating a single knowledge vector with the associated knowledge vector to form an integrated result with enhanced feature representation. Steps 23131 and 23132 describe how to determine an integration index based on query frequency and conduct knowledge enhancement accordingly. The operation of these steps in practical applications can be illustrated by the following examples.

Taking epidemic data of respiratory infections (e.g., pneumonia) as an example, each knowledge vector may represent characteristics of a set of cases over a range of times and places, such as morbidity, speed of transmission, patient demographics, etc. The associated knowledge vector may refer to other data points directly related to the cases, such as contact history, travel history, hospital visit records, and the like.

The above step 23131 involves determining an integration index, at which point the system considers the query frequency of each knowledge vector, i.e., the number of times that vector is referenced in the data analysis. This index reflects the relative importance or usage of the vector. Likewise, query frequency of the y-th associated knowledge vector is also taken into account. According to the two frequency indexes, an integration index is respectively determined for each knowledge vector and the associated knowledge vector. This integration index may be a weighting factor that will be used in the next step to guide the degree of knowledge reinforcement.

Step 23132 involves knowledge enrichment, which the system performs on each knowledge vector and its associated knowledge vector according to the determined integrated index. This process may involve merging similar features, strengthening the representation strength of a particular feature, or readjusting feature weights. Ultimately, these enhanced vectors form the feature integration result of the y-th round of rotation.

For example, in processing pneumonia epidemic data, if a certain knowledge vector represents a high risk case population in a specific area, and the query frequency of this vector is high, it is given a high integration index when integrating. This means that the information in this vector is important in subsequent analysis and decision making and therefore needs to be emphasized in the feature integration results. In this way, the optimization process not only improves the accuracy and reliability of the target knowledge vector set, but also ensures that the system can quickly respond to important epidemiological changes, thereby better serving public health monitoring and early warning tasks.

Under some possible design considerations, the first query frequency based on each of the knowledge vectors in step 23131, and the second query frequency of the y-th associated knowledge vector, determine an integration index for each of the knowledge vectors and the y-th associated knowledge vector, respectively, including steps 231311-231312.

And 231311, when the first query frequency and the second query frequency are both 0, determining a first set index variable as an integrated index of each knowledge vector, and determining a second set index variable as an integrated index of a y-th associated knowledge vector.

And 231312, when at least one of the first query frequency and the second query frequency is greater than 0, determining the ratio of the first query frequency to the sum of the first query frequency and the second query frequency as an integrated index of each knowledge vector, and determining the difference value of the integrated index of 1 and each knowledge vector as an integrated index of the y-th associated knowledge vector.

In step 23131, an integration index is mentioned that determines a knowledge vector based on the query frequency. This process means that the system will evaluate the importance and relevance of the individual knowledge vectors in the previous queries and decide their weights in the integration process accordingly. Query frequency reflects the number of times a certain knowledge vector is retrieved in a historical query, which is typically closely related to the value that the vector contributes to the predictive model.

Taking pneumonia as an example, a large amount of data about different cases, each corresponding to a knowledge vector, is collected in the past, and contains information about virus type, infection area, patient age, etc. Every time a new case is queried, the system retrieves relevant historical case information from the knowledge vector set according to similarity. In this context, query frequency may represent popularity or importance of a particular case type or combination of features.

When the query frequency is 0: if a certain knowledge vector and its associated vector are never referenced in the historical queries (i.e., both query frequencies are 0), they are given a preset fixed value on the integration index. For example, the first set index variable may be a lower value, indicating that these vectors contribute less to the current analysis; and the second set index variable is also a predefined value representing the integration weight of the associated vector.

When at least one of the query frequencies is greater than 0: in this case, the integrated index for each knowledge vector will be determined based on the ratio of its query frequency to the sum of the query frequencies of the associated vectors. Thus, if a vector is referenced more frequently in a historical query, it gets higher weight when integrated reflecting its greater value. Meanwhile, the integration index of the association vector is 1 minus the integration index of the knowledge vector, so that the sum of the two indexes is ensured to be 1, namely normalization processing.

In practice, this means that knowledge vectors that provide more useful information (e.g., that contribute more insight into disease propagation patterns) will take more important positions in the integration process. This helps ensure that the optimized knowledge vector set can more accurately reflect information most critical to epidemiological analysis, thereby improving the predictive power of the overall system and the quality of decision support.

In the above embodiment, the y-th associated knowledge vector is a y-th associated knowledge vector of Q associated knowledge vectors corresponding to each knowledge vector; q associated knowledge vectors are obtained by digging before the 1 st round of circulation is started, and the Q associated knowledge vectors are sorted according to the ascending order of the characteristic difference of each knowledge vector; wherein Q is the maximum cycle.

In the above-described embodiment, each knowledge vector has been associated with a set of associated knowledge vectors before the optimization process begins. The associated knowledge vectors are mined from the initial data based on feature differences and sorted by feature difference size from each knowledge vector. In each round of rotation, the knowledge vector for the current consideration is integrated into its corresponding one of the associated knowledge vectors. The following is an example of how this technique can be applied to the analysis of epidemic data of respiratory infections (e.g. pneumonia).

(1) Pretreatment stage

And (3) data mining: first, the initial disease propagation data is mined to identify knowledge vectors for each case or region. At the same time, for each vector, a set of Q associated knowledge vectors with obvious relevance or similarity thereto is identified. For example, if one knowledge vector represents epidemic data for a city, its associated knowledge vector may include epidemic data for surrounding cities or data for other cities of similar population structure and traffic connection.

Feature difference ordering: each knowledge vector is then ranked in ascending order according to the characteristic differences (e.g., geographic distance, rate of case growth, vaccination rate, etc.) between its associated vector. The association vectors with small feature differences are ranked in front, meaning that they have a higher similarity to the original knowledge vector.

(2) Iterative optimization process

Cycle 1: in the 1 st round of circulation, the system selects the associated knowledge vector with the smallest characteristic difference (namely, the associated knowledge vector arranged at the forefront) from the associated vector list corresponding to each knowledge vector for integration.

Subsequent rounds: as the loop progresses, each knowledge vector will integrate the next associated knowledge vector in its list in turn. For example, in round 2, each knowledge vector will integrate the second associated vector in its list, and so on.

Through such a process, the knowledge vector set is gradually optimized, and each round may improve the quality of the data and the accuracy of the analysis. For example, with a pneumonia epidemic monitoring system, an initial round may integrate data related to a primary transportation hub, and subsequent rounds may gradually introduce data of different population densities or socioeconomic status areas, further refining an epidemic model. Such a system can provide a more comprehensive view to epidemiologists, helping them better understand the disease transmission dynamics and developing effective preventive and control measures.

The whole iterative process continues until a loop termination condition is met, such as reaching a preset downsampling coefficient threshold or completing all Q rounds of loops. After completion, the resulting set of target knowledge vectors can be used to support rapid and accurate epidemiological analysis and decision making.

In some exemplary embodiments, feature differences between knowledge vectors in the set of input knowledge vectors based on the y-th round of rotation in step 231 are integrated for knowledge vectors in the set of input knowledge vectors of the y-th round of rotation and based on the resulting feature integration results, an output set of knowledge vectors of the y-th round of rotation is generated, comprising steps 231 a-231 d.

Step 231a, determining an integration priority of each of the knowledge vectors in the input knowledge vector set of the y-th round based on a feature difference between each of the knowledge vectors in the input knowledge vector set of the y-th round and a core associated knowledge vector of each of the knowledge vectors.

And 231b, according to the integration priority, carrying out consistency analysis on the case streaming track corresponding to each knowledge vector and the case streaming track corresponding to the core associated knowledge vector to obtain a consistency analysis result.

And 231c, integrating each knowledge vector with the core associated knowledge vector when the consistency analysis result represents that the case streaming track corresponding to each knowledge vector is identical to the case streaming track corresponding to the core associated knowledge vector, and obtaining the feature integration result of the y-th round of circulation.

And 231d, generating the output knowledge vector set of the y-th round by utilizing the feature integration result of the y-th round and the case flow adjustment track corresponding to each knowledge vector.

In the above exemplary embodiments, it is involved how to optimize disease-modifying information by considering feature differences and consistency analysis between knowledge vectors. Steps 231a through 231d describe an iterative process in which the goal of integrating knowledge vectors is to improve the accuracy of the predictive model. This process can be applied to improve the analysis and processing of epidemic data for respiratory infections (e.g., pneumonia).

Step 231a described above involves determining an integration priority, in which the integration priority is determined based on the feature differences between each knowledge vector and its core associated knowledge vector. For example, in studying pneumonic cases, a knowledge vector may contain case data for a particular region, while its core associated knowledge vector may represent another set of cases or other important features (e.g., shared vehicles, aggregation sites, etc.) within the same region. By comparing the differences between these vectors, it can be decided which vectors should be prioritized in the subsequent integration.

Step 231b described above involves a consistency analysis, followed by a consistency analysis of the case-stream-tuned trajectories represented by each knowledge vector and the trajectories represented by its core-associated knowledge vector according to the integration priority. This analysis helps determine whether the different data points exhibit similar disease transmission patterns, thereby evaluating whether they are suitable for integration.

Step 231c described above involves performing feature integration, and if the consistency analysis results show that the two trajectories are the same or very similar, then performing an integration operation. This may involve merging the data of the two vectors to form a more comprehensive and representative representation of the feature. This integration is advantageous in reducing redundancy of data and improving the efficiency of the overall model.

Step 231d above involves generating an output knowledge vector set, and finally, using the obtained feature integration result and the relevant case-stream alignment information, generating a new output knowledge vector set. The updated vector set contains refined data with higher quality, and can provide better support for monitoring and predicting the disease transmission trend.

Applying this procedure to practical situations, this approach can be used to track and analyze the spread of pneumonia. Through intelligent integration of historical cases, a hidden transmission mode can be revealed, a decision maker is helped to take more targeted control measures, and finally the purpose of controlling epidemic situation spreading is achieved. In addition, the optimized data set can provide a more reliable basis for an early warning system of future epidemic situation, so that the early warning system can respond to potential health threats more quickly.

In some alternative embodiments, the streaming trend thermal image block described in step 110, which combines the initial case streaming image record and the previous x-1 case streaming events in the streaming trend thermal image stream, determines a trend thermal match vector for the xth case streaming event in the streaming trend thermal image stream, including steps 111-112.

And 111, generating a streaming track relation network of the x-th case streaming event by utilizing the initial case streaming image record and streaming trend thermal image blocks of the previous x-1 case streaming events in the streaming trend thermal image stream.

And 112, determining the streaming track description vector corresponding to the streaming track relation network as the trend thermal matching vector of the x-th case streaming event in the streaming trend thermal image stream.

In the above embodiments, a technical solution for epidemiological monitoring and analysis, in particular image analysis related to case-stream-planning (contact training), is discussed. This process includes using the initial case-streaming image record and the existing streaming trend thermal image stream to determine a trend thermal match vector for the new case-streaming event. The following are specific implementations of step 110, step 111, and step 112, and how they may be combined to support epidemiological analysis.

Step 110, among other things, describes how existing streaming data and thermal image streaming can be utilized to assist in the analysis of new case streaming events. In this context, each case-stream event has a corresponding image record, showing the trend of the event in time and space.

Wherein step 111 involves generating a streaming track relationship network: and constructing a streaming track relation network of the x-th case streaming event by using the initial case streaming image record and combining streaming trend thermal image blocks of the previous x-1 events. This network depicts potential connections between cases, such as person-to-person contact, or connections between places that people have visited. For example, if the initial image record shows the spread pattern of a disease in a city, new cases may be added to this network according to their geographic location and time stamp, and connected to existing cases according to contact history and other relevant parameters.

Wherein step 112 involves determining a trend thermal match vector: and converting the newly constructed streaming track relation network into a streaming track description vector, namely a trend thermal matching vector. This vector encodes the location and role of the xth case-streaming event in the network, including the interaction information with other cases. In epidemiological analysis, this vector may be used to identify possible propagation paths, high risk areas, or key propagators.

Consider, for example, a scene of a pneumonia epidemic. By analyzing the case-stream image records, researchers can identify which areas have infection hotspots and which traffic routes may promote virus spread. Every time a new case report exists, the system updates the streaming track relation network and generates a corresponding trend thermal matching vector. Thus, over time, the evolution of the epidemic can be monitored and data-based decisions made accordingly. In summary, by this approach, epidemic data can be systematically collected and analyzed to form a global view that helps understand and address trends and patterns of disease transmission.

In some possible embodiments, determining the streaming trend thermal image block of the x-th case streaming event in step 130 according to the feature differences between the P trend thermal embedding vectors and the trend thermal matching vectors of the x-th case streaming event, respectively, includes steps 131-132.

And 131, according to the characteristic difference between the P trend thermal embedded vectors and the trend thermal matching vectors of the x-th case streaming event, and the streaming trend thermal discrimination result generated by the pre-configured visual simulation network aiming at the trend thermal matching vectors of the x-th case streaming event, determining the target streaming trend thermal discrimination result of the trend thermal matching vectors of the x-th case streaming event in the streaming region.

And 132, determining the flow adjustment trend thermal image block of the x-th case flow adjustment event from the flow adjustment region contained in the flow adjustment region according to the target flow adjustment trend thermal discrimination result.

It can be seen that step 130 and its substeps 131 and 132 describe a procedure for associating trending thermal embedding vectors with specific case-stream events and generating corresponding trending thermal image blocks based thereon. This process can be used to create a visual representation of disease transmission to more intuitively understand how viruses propagate in different areas. The following is an example implementation of this solution.

Assume that there is a set of case-stream events, each of which is associated with data for a particular location and time point in the pneumonia epidemic. Such data may include information on the number of cases, speed of transmission, hospital visits, etc. The trend thermal embedding vector is the result of encoding the data and is used for displaying epidemic hot spots on a map.

Wherein step 131 involves determining a trend thermal match vector: at this step, the system calculates the feature differences between the P trending thermodynamic embedding vectors and the trending thermodynamic matching vector for the xth case-stream event. This may be accomplished by various distance measures (e.g., euclidean distance, cosine similarity). Meanwhile, a visual simulation network (possibly a deep learning model) is pre-configured, and trend thermal matching vectors of the x-th case flow adjustment event are analyzed to generate trend thermal discrimination results. This result can reflect the trend of the spread of the targeted case-stream events in the stream-regulated area, such as which areas show a trend of increasing or decreasing infection rate.

Wherein step 132 involves generating a trending thermal image block: according to the thermal discrimination result of the target flow adjustment trend determined in step 131, the system selects one or more representative areas from the plurality of flow adjustment areas contained in the flow adjustment area. Then, trend thermal image blocks are generated from these regions for the x-th case streaming event. These image tiles visually exhibit spatial distribution and temporal development trends of case-stream events, enabling users to intuitively identify and understand hot spot areas of disease transmission.

For example, if a new aggregate infection event is found in a commercial area of a city, the system will generate a trending thermodynamic image block by analyzing trending thermodynamic embedding vectors and matching vectors associated with the event, and by visually modeling the output of the network. This image patch may highlight the infection trends in the business and surrounding areas, helping the decision maker to formulate more targeted interventions.

By the technical scheme, a large amount of epidemic situation data can be effectively integrated and analyzed, a powerful vision auxiliary tool is provided, and support is provided for epidemic situation resistance. This also facilitates a data driven decision process, helps to more quickly identify disease propagation patterns, optimizes resource allocation, and improves emergency response strategies.

Under other exemplary design ideas, according to the feature difference between the P trend thermal embedding vectors and the trend thermal matching vector of the x-th case streaming event in step 131, and the streaming trend thermal discrimination result generated by the pre-configured visual simulation network for the trend thermal matching vector of the x-th case streaming event, determining the target streaming trend thermal discrimination result of the trend thermal matching vector of the x-th case streaming event in the streaming area, including steps 1311-1314.

Step 1311, determining the position features corresponding to the P trend thermal embedding vectors according to the feature differences between the P trend thermal embedding vectors and the trend thermal matching vectors of the x-th case flow adjustment event.

And 1312, updating the position features by utilizing the integrated calculation force of the target knowledge vector set to obtain updated position features.

Step 1313, generating a flow adjustment trend thermal discrimination result of the trend thermal match vector of the x-th case flow adjustment event in the flow adjustment region by using the updated position features of the P trend thermal embedding vectors.

Step 1314, determining the target flow adjustment trend thermal discrimination result of the trend thermal match vector of the x-th case flow adjustment event in the flow adjustment region by utilizing the flow adjustment trend thermal discrimination result of the trend thermal match vector of the x-th case flow adjustment event in the flow adjustment region and the aggregation information of the flow adjustment trend thermal discrimination result generated by the pre-configuration visual simulation network aiming at the trend thermal match vector of the x-th case flow adjustment event.

The design concept relates to the use of trend thermal embedding vectors to analyze and discriminate features of a particular case-stream event and combine it with the results of a pre-configured visual simulation network. By this method, the propagation trend and pattern of cases in a geographical area can be more accurately identified and understood. The following is a detailed description of how steps 1311-1314 are applied.

Firstly, determining the respective corresponding position features according to the feature difference between the P trend thermal embedding vectors and the trend thermal matching vectors of the xth case streaming event. In the context of epidemic monitoring, trend thermal embedding vectors may represent case trends in different regions, while location features can reflect the roles and importance of these regions in case propagation networks.

Next, the location features are updated with the integrated computing power of the set of target knowledge vectors. This process may involve re-evaluating and correcting the location features so that they better reflect the actual epidemic dynamics and propagation patterns.

And generating a thermal judgment result of the flow regulation trend in the flow regulation region for the x-th case flow regulation event by using the updated position characteristic. This result helps reveal the location of the event in the overall epidemic propagation network, as well as the relative strength and importance of its trend compared to other events.

And finally, determining the final trend thermal matching vector of the x-th case flow adjustment event by combining the judging result of the trend thermal matching vector in the flow adjustment area and the flow adjustment trend thermal judging result generated by the pre-configuration visual simulation network. The aggregated information may include multidimensional data points extracted from the simulated network that have reference value in assessing the impact of a case in regional propagation.

Applying this approach to a practical scenario, such as the management of a pneumonia epidemic, can thereby determine the location of a new occurrence of the case in the epidemic propagation map, understand its potential propagation trend, and how to correlate with a known case trend. By combining historical data with real-time analysis, this approach can provide a more dynamic and accurate view of the propagation trend, allowing the decision maker to make more informed response measures, such as targeted resource allocation or public health guidelines.

In some independent embodiments, after completing the flow pattern generation with respect to the initial case flow pattern record when a flow trend thermal image block is obtained for the case flow event at the end of the flow trend thermal image stream as described in step 140, the method further comprises step 150.

Step 150, generating an emergency plan report text aiming at epidemiological monitoring and early warning based on the flow-regulating trend thermal image flow; the emergency plan report text comprises an emergency material reserve plan, a medical resource allocation scheduling plan and a traffic road network adjustment plan.

In this embodiment, step 140 focuses on how to generate streaming trend thermal image blocks and use them to complete the analysis of the initial case streaming image records. Step 150, in turn, describes how the information obtained from steps 110 through 140 may be used to generate an emergency plan report text intended to support epidemiological monitoring and early warning systems. The following is an example of how these steps can be applied to a practical case.

It is assumed that corresponding thermal image blocks have been generated for all case-streaming events in the streaming trend thermal image stream, each representing the disease propagation of a particular region over a period of time. The trending thermal image block of the last case-stream event marks that the analysis of the entire dataset has been completed.

Based on the results of step 140, the system generates a comprehensive emergency plan report text. The purpose of this report is to provide a policy guideline for future possible epidemic situations. The report text includes the following key parts:

(1) Emergency material reserve protocol: according to epidemic situation hot spot areas displayed by the thermal image blocks, the proposal suggests that epidemic prevention materials such as masks, disinfectants, detection kits and the like are reserved in which areas. Meanwhile, the number and types of storage can be recommended according to the propagation speed and the potential risk;

(2) Medical resource allocation scheduling scheme: based on the current distribution of medical resources and epidemic trend revealed by the thermal image block, the plan guides how to reasonably allocate key medical resources such as medical staff, respirators, ICU beds and the like. For example, if a region shows signs of a rapid rise in infection rate, then it may be desirable to prioritize the allocation of more medical resources to that region.

(3) Traffic network adjustment scheme: when an epidemic outbreak occurs in some areas, controlling the flow of people is one of the key measures to slow down the spread of viruses. Traffic network adjustment plans will suggest whether certain traffic routes need to be restricted or quarantine sites set based on thermodynamic diagram data. For example, if there are consecutive case-stream events starting from a primary junction, it may be desirable to consider temporarily closing the junction or directing traffic around.

By combining Geographic Information Systems (GIS) and epidemiological models, the thermal image block can be transformed into specific countermeasures. The resulting emergency plan report text may direct how to react quickly and effectively in the face of a new case-stream event. The systematic plan report not only can help to cope with the current epidemic situation, but also provides an effective early warning and coping framework for similar situations which may occur in the future.

In the embodiment of the present application, it should be noted that the technical solution described in the above step 150 still does not belong to the disease diagnosis and treatment method.

First, the solution of step 150 is aimed at supporting epidemiological monitoring and early warning, in terms of scope and nature, rather than diagnosing or treating any disease directly. It is more biased towards an information management system that aims to assist decision makers in making strategic decisions about public health and safety by analyzing epidemic data.

Second, as for the operation object, the operation object of the technical solution of step 150 is data (such as a flow trend thermal image flow), and a plan report based on data analysis, rather than diagnosing or treating directly for the individual patient.

Further, with respect to the subject of implementation, the approach of step 150 is primarily for assisting a decision maker, rather than being implemented on a patient in a clinical setting by a doctor or other medical professional.

Finally, the solution of step 150 is not limited to the diagnostic and therapeutic activities of the medical professional, and the emergency plan report text produced thereby provides a reference for the prevention and control measures without intervention in specific medical practice. Thus, it serves a broader field of public health management.

Further, there is also provided a readable storage medium having stored thereon a program which when executed by a processor implements the above-described method.

It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described system and apparatus may refer to corresponding procedures in the foregoing method embodiments, which are not repeated herein. In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. The above-described apparatus embodiments are merely illustrative, for example, the division of the units is merely a logical function division, and there may be other manners of division in actual implementation, and for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some communication interface, device or unit indirect coupling or communication connection, which may be in electrical, mechanical or other form.

Claims

1. An intelligent medical monitoring and early warning method based on disease diagnosis data is characterized by being applied to an intelligent medical monitoring and early warning system, and comprises the following steps:

When a streaming trend thermal image block is obtained for the case streaming event at the tail of the streaming trend thermal image stream, completing streaming thermodynamic diagram generation of the initial case streaming image record; the flow-regulated trend thermal image flow is used for epidemiological monitoring and early warning treatment;

The determining the streaming trend thermal image block of the x-th case streaming event according to the feature differences between the P trend thermal embedding vectors and the trend thermal matching vectors of the x-th case streaming event, comprises:

Determining the flow regulation trend thermal image block of the x-th case flow regulation event from the flow regulation region contained in the flow regulation region according to the target flow regulation trend thermal discrimination result;

The determining a target streaming trend thermal discrimination result of the trend thermal match vector of the x-th case streaming event in the streaming area according to the feature difference between the P trend thermal embedding vectors and the trend thermal match vector of the x-th case streaming event and the streaming trend thermal discrimination result generated by the pre-configured visual simulation network aiming at the trend thermal match vector of the x-th case streaming event comprises the following steps:

2. The method of claim 1, wherein the streaming trend thermal image block combining the initial case streaming image record and the previous x-1 case streaming events in the streaming trend thermal image stream, before determining a trend thermal match vector for the xth case streaming event in the streaming trend thermal image stream, the method further comprises:

3. The method of claim 2, wherein the optimizing the initial set of knowledge vectors to obtain the target set of knowledge vectors comprises:

4. The method of claim 3, wherein the integrating knowledge vectors in the set of incoming knowledge vectors for the y-th round based on feature differences between knowledge vectors in the set of incoming knowledge vectors for the y-th round, and generating the set of output knowledge vectors for the y-th round based on the resulting feature integration results, comprises:

5. The method of claim 4, wherein said integrating each of said knowledge vectors with a y-th of said associated knowledge vector to obtain said feature integration result for a y-th round of rotation comprises:

Carrying out knowledge enhancement on each knowledge vector and the y-th associated knowledge vector according to the integration index of each knowledge vector and the integration index of the y-th associated knowledge vector to obtain the feature integration result of the y-th cycle;

wherein the determining an integration index for each of the knowledge vectors and the y-th associated knowledge vector based on the first query frequency of each of the knowledge vectors and the second query frequency of the y-th associated knowledge vector, respectively, comprises:

When at least one of the first query frequency and the second query frequency is greater than 0, determining the ratio of the first query frequency to the sum of the first query frequency and the second query frequency as an integrated index of each knowledge vector, and determining the difference value of the integrated index of 1 and each knowledge vector as an integrated index of the y-th associated knowledge vector;

The y-th associated knowledge vector is the y-th associated knowledge vector in Q associated knowledge vectors corresponding to each knowledge vector; q associated knowledge vectors are obtained by digging before the 1 st round of circulation is started, and the Q associated knowledge vectors are sorted according to the ascending order of the characteristic difference of each knowledge vector; wherein Q is the maximum cycle.

6. The method of claim 3, wherein the integrating knowledge vectors in the set of incoming knowledge vectors for the y-th round based on feature differences between knowledge vectors in the set of incoming knowledge vectors for the y-th round, and generating the set of output knowledge vectors for the y-th round based on the resulting feature integration results, comprises:

7. The method of claim 1, wherein the streaming trend thermal image block combining the initial case streaming image record and the first x-1 case streaming events in the streaming trend thermal image stream, determining a trend thermal match vector for the x-th case streaming event in the streaming trend thermal image stream, comprises:

8. An intelligent medical monitoring and early warning system is characterized by comprising a processor and a memory; the processor being communicatively connected to the memory, the processor being adapted to read a computer program from the memory and execute it to carry out the method of any of the preceding claims 1-7.

9. A computer readable storage medium, characterized in that it has stored thereon a computer program, which, when run, implements the method of any of the preceding claims 1-7.