A medical record-based disease anomaly early warning method, device and medium

Abstract

The application discloses a medical record-based disease anomaly early warning method, device and medium, the method comprising: acquiring current medical record information, encoding disease representation in the medical record information to obtain a structured first medical record vector; dimensionally compressing the first medical record vector through a pre-trained medical record vectorization model to obtain a low-dimensional second medical record vector; inputting the second medical record vector into a pre-trained disease trend prediction model to perform anomaly prediction on a specified disease appearing in the current medical record information, and outputting an anomaly expectation corresponding to the current medical record information according to a prediction result; and determining that a specified individual or a specified group is abnormal if the deviation level between the output result of the medical record information of the specified individual or the specified group obtained through the disease trend prediction model and the anomaly expectation reaches a preset level.

Description

Technical Field

[0001] This application relates to the field of anomaly early warning, specifically to a method, device, and medium for disease anomaly early warning based on medical records. Background Technology

[0002] With the continuous emergence of new diseases and the resurgence of existing diseases, current disease surveillance methods are no longer adequate for the demands of social emergency response. To ensure timely and accurate prediction of abnormal disease situations and to take effective prevention and control measures, disease early warning is generally achieved through data processing and data mining of patient symptom information and medical records. However, this approach requires extensive pre-processing of data, resulting in a large workload, low processing efficiency, and unsatisfactory early warning effects. Summary of the Invention

[0003] To address the aforementioned problems, this application proposes a disease anomaly early warning method based on medical records, comprising:

[0004] This application provides a disease anomaly early warning device based on medical records. The method includes: acquiring current medical record information, encoding disease representations in the medical record information to obtain a structured first medical record vector; compressing the first medical record vector to obtain a low-dimensional second medical record vector using a pre-trained medical record vectorization model; using a pre-trained disease trend prediction model, taking the second medical record vector as input, predicting anomalies of specified diseases appearing in the current medical record information, and outputting the expected anomalies corresponding to the current medical record information based on the prediction results; if the deviation level between the output result obtained by the disease trend prediction model and the expected anomalies of the medical record information of a specified individual or a specified group reaches a preset level, then it is determined that the specified individual or the specified group has anomalies.

[0005] In one implementation of this application, the prediction result includes the abnormal probability of a specified disease corresponding to the second medical record vector; after outputting the abnormal expectation corresponding to the current medical record information based on the prediction result, the method further includes: dividing individuals into multiple specified groups based on the individual characteristics corresponding to the individuals; the individual characteristics include at least one or more of the following: age, geographical location, and gender; calculating the first deviation degree between the abnormal probability and the abnormal expectation of different individuals in the specified groups, and the mean of the first deviation degree corresponding to different individuals; determining the deviation degree interval corresponding to the mean of the first deviation degree based on the pre-divided deviation degree interval, so as to determine the corresponding deviation level.

[0006] In one implementation of this application, after outputting the expected abnormality corresponding to the current medical record information based on the prediction result, the method further includes: calculating a second deviation degree between the abnormal probability of a specified disease for a specified individual and the expected abnormality; determining the deviation degree interval corresponding to the second deviation degree based on a pre-divided deviation degree interval, so as to determine the corresponding deviation level; and, if the deviation level is determined to reach a preset level, calculating the historical incidence rate of the disease suffered by the individual corresponding to the medical record information, so as to determine the disease type corresponding to the specified individual based on the historical incidence rate; the disease type includes common diseases and rare diseases.

[0007] In one implementation of this application, after determining that a specified individual or group has an abnormality, the method further includes: when a specified group has an abnormality, predicting the target incidence trend of the group in a specified time period after the current period based on a preset regression model; obtaining the historical incidence trend of the specified group in the specified time period over the years based on medical record information; and prompting the disease control department to take corresponding prevention and control measures for the specified group based on the target incidence trend and the historical incidence trend.

[0008] In one implementation of this application, after determining that a specified individual or group is abnormal, the method further includes: acquiring the transmission speed, transmission mode, and mortality rate of a specified disease at preset time intervals to determine the weight corresponding to the specified disease; determining the geographical location of individuals in the specified group; and marking the specified disease on a preset electronic map using corresponding display colors based on the weight and geographical location of the specified disease; the display colors correspond to the weight of the specified disease; and marking the radiation range corresponding to the specified disease on the electronic map based on the target incidence trend of the specified group.

[0009] In one implementation of this application, the radiation range corresponding to a specified disease is marked on an electronic map based on the target disease trend corresponding to a specified group. Specifically, this includes: predicting the incidence rate corresponding to each radiation cycle based on the target disease trend corresponding to the specified group; the time interval corresponds to the radiation cycle; determining the total number of people in the specified group, and determining the number of patients within the radiation cycle based on the total number of people and the incidence rate; determining the corresponding risk level based on the ratio between the number of patients and the total number of people, and determining the corresponding first radiation radius based on the risk level; taking an individual in the specified group as the first center, and determining the first radiation range corresponding to each individual based on the first center and the first radiation radius; taking the center of the union region of each first radiation range as the second center, and the distance between the farthest point in each first radiation range from the second center and the second center as the second radiation radius, and determining the second radiation range corresponding to the specified disease.

[0010] In one implementation of this application, encoding disease representations in medical record information specifically includes: segmenting the medical record information into words, and selecting disease representations from the segmented medical record information according to a preset disease dictionary; the disease dictionary includes multiple disease representations; encoding each disease representation in the medical record information using the disease dictionary to obtain a one-bit encoded disease representation vector; deduplicating the disease representation vectors and integrating the deduplicated disease representation vectors to obtain a structured first medical record vector composed of disease representation vectors.

[0011] In one implementation of this application, after determining the disease type corresponding to a specified individual, the method further includes: if the disease type corresponding to the specified individual is a rare disease, obtaining the target medical record vector corresponding to the specified patient; specifying that the patient suffers from the rare disease; calculating the similarity between the target medical record vector and the first medical record vector corresponding to the individual, and selecting the specified patient corresponding to the target medical record vector with the highest similarity to the first medical record vector from multiple specified patients as the target patient; determining the special disease characteristics of the specified individual based on the target medical record vector corresponding to the target patient and the first medical record vector corresponding to the specified individual, and adding it to the clinical case library.

[0012] This application provides a disease anomaly early warning device based on medical records, including:

[0013] At least one processor; and,

[0014] A memory that is communicatively connected to at least one processor; wherein,

[0015] The memory stores instructions that can be executed by at least one processor, and the instructions, when executed by at least one processor, enable at least one processor to:

[0016] Obtain the current medical record information, encode the disease representation in the medical record information to obtain a structured first medical record vector;

[0017] By using a pre-trained medical record vectorization model, the first medical record vector is dimensionally compressed to obtain a low-dimensional second medical record vector.

[0018] Using a pre-trained disease trend prediction model, the second medical record vector is used as input to predict anomalies of specified diseases appearing in the current medical record information, and the expected anomalies corresponding to the current medical record information are output based on the prediction results.

[0019] If the deviation level between the medical record information of a specified individual or a specified group and the output result obtained by the disease trend prediction model reaches a preset level, then it is determined that there is an anomaly in the specified individual or the specified group.

[0020] This application provides a non-volatile computer storage medium storing computer-executable instructions, which are configured as follows:

[0021] Obtain the current medical record information, encode the disease representation in the medical record information to obtain a structured first medical record vector;

[0022] By using a pre-trained medical record vectorization model, the first medical record vector is dimensionally compressed to obtain a low-dimensional second medical record vector.

[0023] Using a pre-trained disease trend prediction model, the second medical record vector is used as input to predict anomalies of specified diseases appearing in the current medical record information, and the expected anomalies corresponding to the current medical record information are output based on the prediction results.

[0024] If the deviation level between the medical record information of a specified individual or a specified group and the output result obtained by the disease trend prediction model reaches a preset level, then it is determined that there is an anomaly in the specified individual or the specified group.

[0025] The disease anomaly early warning method, device, and medium based on medical records provided in this application have at least the following beneficial effects:

[0026] Medical record information is structured and compressed into vector dimensions to obtain medical record vectors. These vectors are then used to predict whether anomalies exist in a specified individual or group. This natural language text-based disease anomaly prediction method allows data to be directly used in the final disease prediction process without the need for preparatory work such as cleaning, organizing, and labeling. This significantly reduces the workload of data preprocessing and improves prediction efficiency. Furthermore, the disease trend prediction model can predict the expected anomalies corresponding to the current medical record information. By comparing the expected anomalies with the output of the prediction model, disease anomaly detection can be achieved, making it more efficient and accurate. Attached Figure Description

[0027] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0028] Figure 1 A flowchart illustrating a disease anomaly early warning method based on medical records, provided as an embodiment of this application;

[0029] Figure 2 A flowchart illustrating another disease anomaly early warning method based on medical records provided in this application embodiment;

[0030] Figure 3This is a schematic diagram of a disease anomaly early warning device based on medical records, provided as an embodiment of this application. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0032] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0033] like Figure 1 As shown in the embodiment of this application, a disease anomaly early warning method based on medical records is provided, including:

[0034] S101: Obtain the current medical record information, encode the disease representation in the medical record information to obtain a structured first medical record vector.

[0035] Medical records are records kept by medical staff of the diagnosis, treatment, and other medical activities related to a patient's illness. They can exist in handwritten or electronic form. Medical records originate from healthcare systems such as hospitals and community clinics and include patient identification information, symptom information, and diagnostic information, reflecting the patient's condition.

[0036] This application's embodiment uses individual medical records for disease anomaly early warning. It requires obtaining individual medical record information from the medical system and structuring the natural text format of the medical record information, thus storing it in a programmatic form for easy subsequent predictive analysis. It should be noted that the individual medical record information here refers to electronic text format; handwritten text format medical record information needs to be uploaded to the medical system beforehand using an electronic scanner or other upload tool.

[0037] The server's structuring of medical record information is essentially a process of encoding vocabulary. The effective information within the medical record information that can be used for disease-related predictions constitutes symptom information. Therefore, the first step is to segment the medical record information into words, obtaining a set of medical record vocabulary specific to each individual's medical record. Then, based on a pre-defined symptom dictionary, disease characteristics recorded in the symptom dictionary are selected from the medical record vocabulary. This symptom dictionary includes words describing disease characteristics, such as chest tightness, headache, and cough.

[0038] After screening the disease representations, they can be encoded using a disease dictionary, so that each disease representation exists in vector form. Assuming the disease dictionary contains 1000 diseases, the disease representations in the medical records can be represented by 1000-dimensional vectors, with each vector component representing the detection result of a disease representation. After encoding the disease representations, the detection results can be quantified into the range [-1, 1], where 0 represents normal or no detection of the disease representation. For example, if the disease dictionary includes four disease representation terms: headache, body temperature, heart rate, and blood pressure, then the disease representations can be described by four-dimensional vectors. The components of each vector in (0, 0, 0.4, -0.3), (1, 0.2, 0, 0.3), and (0, 0, -0.1, 0) correspond to the detection results of headache, body temperature, heart rate, and blood pressure, respectively, and each vector corresponds to a different medical record. Based on this method, individual medical record vectors with the same dimensions as the disease dictionary are obtained, realizing a structured representation of the text medical records.

[0039] S102: By using a pre-trained medical record vectorization model, the first medical record vector is dimensionally compressed to obtain a low-dimensional second medical record vector.

[0040] The encoded structured medical record vector no longer exists in the form of natural text, but in the form of vectors. However, its dimensionality is relatively high (equivalent to the number of disease representations in a disease dictionary), making it difficult to predict and learn subsequently. Therefore, the server can collect structured medical record vectors from each individual to form a structured medical record dataset, which will then participate in the training process of the medical record vectorization model. This medical record vectorization model can be an unsupervised learning neural network model, or a vectorization model trained using PCA (Principal Component Analysis). By inputting the high-dimensional structured medical record vectors into the model, the dimensionality of the individual's structured medical record vectors is compressed, resulting in a low-dimensional second medical record vector. For example, if the first medical record vector has a dimension of 1000, the vectorization model can compress its dimension to 5. This low-dimensional second medical record vector is the feature vector that has the greatest impact on the prediction result. It should be noted that "high-dimensional" and "low-dimensional" are only relative terms. For example, less than 5 is considered low-dimensional, and more than 100 is considered high-dimensional. The values ​​listed here are only examples.

[0041] S103: Using a pre-trained disease trend prediction model, the second medical record vector is used as input to predict anomalies of specified diseases appearing in the current medical record information, and the expected anomalies corresponding to the current medical record information are output based on the prediction results.

[0042] After structuring and vectorizing individual medical record information, the server can sample at preset time intervals, collecting a medical record vector dataset composed of multiple second medical record vectors and using it as data samples. After continuously sampling medical record vector samples over N time periods, the first N-1 samples can be used as model input, and the Nth sample as a label to train the corresponding disease trend model. This disease trend model can predict anomalies of specified diseases appearing in the current period's medical record information and output the expected anomaly corresponding to the current period's medical record information based on the prediction results. The specified diseases can be diseases actually diagnosed in the medical record information or all diseases that have occurred. When performing anomaly detection, it is possible to detect only the diseases present in the medical record information or to detect all diseases. The final output results are the anomaly probabilities for each specified disease.

[0043] S104: If the deviation level between the medical record information of a specified individual or a specified group and the output result obtained by the disease trend prediction model reaches a preset level, then it is determined that there is an anomaly in the specified individual or the specified group.

[0044] An abnormal expectation represents the standard value that a specified individual or group can maintain when facing a specified disease within the current period. By comparing the deviation between the predicted result and the actual abnormal expectation, it can be determined whether the current sick individual or group has a disease abnormality. Specifically, the server first needs to determine the prediction result corresponding to the medical record information. For a specified individual, it is necessary to determine the second deviation degree between the predicted result of the specified disease for the specified individual and the abnormal expectation. According to the pre-divided deviation degree interval, the deviation degree interval corresponding to the second deviation degree is determined, thereby determining the corresponding deviation level. When the deviation level reaches the preset level, the specified individual has an abnormality. For example, if the predicted result of a certain individual shows an abnormal probability of 50%, while the abnormal expectation is 30%, then the deviation degree between the two is 20%. The preset deviation degree intervals are divided into three types: [0,10), [10,50), and [50,100]. The deviation levels corresponding to each interval are level one, level two, and level three, respectively. When the deviation level is greater than or equal to level two, it can be considered that the specified individual does not meet the expected abnormal expectation, and the individual has an abnormality.

[0045] If a specified individual exhibits abnormalities, further analysis is required. Specifically, medical data can be collected from multiple hospitals, medical forums, community healthcare information systems, statistical agencies, etc. Based on this multi-source data, the historical incidence rate of the specified individual's disease can be calculated. Then, based on the historical incidence rate, it can be determined whether the individual's disease type is common or rare. If the historical incidence rate is below a preset threshold, the specified individual is considered to have a rare disease.

[0046] At this point, clinical monitoring of the designated individual is necessary to identify abnormal disease manifestations, thereby continuously expanding the medical record database during treatment. First, the target medical record vectors corresponding to designated patients with the same rare disease are obtained. Of course, before obtaining this data, the medical record information of the designated patients should be structured. Then, the similarity between the target medical record vector and the first medical record vector of the designated individual is calculated using cosine similarity. After obtaining the similarity of the medical record vectors between the designated individual and multiple designated patients, the designated patient with the highest similarity is selected as the target patient. The target patient is used for more in-depth comparative analysis of the disease manifestations of the designated individual; therefore, the designated patient with the highest similarity should be selected as the standard for analysis. After identifying the target patient, the target medical record vector is compared with the first medical record vector of the designated individual to determine the specific disease manifestations of the designated individual with abnormalities. These specific disease manifestations are then added to the clinical case database, which is beneficial for the subsequent diagnosis and treatment of other patients.

[0047] In this embodiment, individuals can be divided into multiple groups based on their individual characteristics. These characteristics include at least one or more of the following: age, geographical location, and gender. Of course, individual characteristics can also include marital status, presence or absence of certain lifestyle habits, etc. Any characteristic that can distinguish categories can be used as an individual characteristic, and this application will not elaborate further. For a specified group, the first deviation between the abnormal probability and the expected abnormality for different individuals within the specified group is calculated, and the mean of these first deviations is calculated. This mean is the group deviation corresponding to the specified group. Based on the pre-defined deviation intervals, the deviation interval corresponding to the mean of the first deviation is determined, thereby determining the corresponding deviation level. If the deviation level between the predicted result and the expected abnormality reaches a preset level, it indicates that the disease development of the current specified group does not meet the expected outcome and may be abnormal. The setting of the deviation interval and the division of the deviation level can be set according to actual needs, and this application will not limit them further.

[0048] When monitoring individuals or groups for anomalies, servers typically set specific time intervals and periodically report monitoring results. When anomalies are detected in a designated group, the server can predict the target disease incidence trend for that group within a specified time period based on a pre-defined regression model. For example, if anomaly analysis is performed on a group every month, and the disease incidence of a resident group on a particular street was abnormal in October, a pre-defined regression model, such as the Autoregressive Integrated Moving Average (ARIMA) model, can be used to predict the disease incidence trend for the next monitoring period, November. The server also determines the historical disease incidence trend of the designated group over the years within the specified time period based on the group's medical records and compares this historical trend with the predicted target disease incidence trend. If the difference between the two trends exceeds a certain range, it indicates that the current abnormal group is highly likely to have abnormal future disease incidence, requiring the disease control department to take appropriate prevention and control measures to prevent a sudden outbreak. For example, the incidence rate can be used to reflect the disease trend. A preset regression model can be used to obtain the target incidence rate and historical incidence rate of the disease within a specified time period. When the difference between the two exceeds the preset value, it indicates that the current incidence trend of the specified group is abnormal, and the intervention of the disease control department is required to avoid a group outbreak of the disease.

[0049] In one embodiment, when a server determines that an anomaly exists in a specified group, it can generate an electronic early warning map to provide real-time dynamic early warning of related diseases.

[0050] Specifically, the server can collect medical data centrally at preset time intervals, determine the transmission speed, transmission mode, and mortality rate of a specified disease based on the medical data, and then determine the weight corresponding to that specified disease. The weight of a specified disease can be calculated by weighted summation of the transmission speed, transmission mode, and mortality rate, and the weighting coefficient can be set by the user. The weight is directly proportional to the transmission speed and mortality rate; that is, the faster the transmission speed and the higher the mortality rate, the greater the weight of the specified disease.

[0051] Furthermore, the server needs to determine the geographical location of each individual within the specified group. Then, based on the weight of the disease type and the geographical location, it marks the disease type on a pre-set electronic map using corresponding display colors. The display color corresponds to the weight of the disease type; the higher the color contrast of a disease type on the map, the more prominent it is, and the greater its weight. Simultaneously, the server can also predict the incidence rate for each radiation cycle based on the predicted target disease trend. Here, the radiation cycle refers to the medical data collection cycle, corresponding to a preset time interval. Further, the total number of people in the specified group is determined, and based on the total number of people and the incidence rate, the number of infected individuals within the radiation cycle is determined. The corresponding risk level is determined based on the ratio between the number of infected individuals and the total number of people. The risk level indicates the speed of disease spread; the higher the risk level, the larger the radiation range of the specified disease, and the larger the corresponding first radius. For each individual in the specified group, the individual's location is used as the center, and the first radiation radius is used as the radius to determine the corresponding first radiation range. After obtaining the first radiation ranges corresponding to multiple individuals, the center of the union region of the first radiation ranges is taken as the second center, and the distance between the farthest point of each first radiation range from the second center and the second center is taken as the second radiation radius. This determines the second radiation range corresponding to the specified disease. The second radiation range can cover all individuals in the specified group suffering from the specified disease.

[0052] It should be noted that during the clinical treatment of abnormal groups, the server can establish a monitoring network by obtaining the geographical location of the medical institutions where each individual in the group is located and the real-time uploaded individual treatment progress. In this way, when a medical institution has a high cure rate and available medical resources, real-time scheduling of medical resources and relevant interventions in treatment methods can be carried out.

[0053] Figure 2 This is a flowchart illustrating another disease anomaly early warning method based on medical records provided in an embodiment of this application. Figure 2As shown, firstly, individual medical record information is acquired and structured to obtain a structured first medical record vector. This structured first medical record vector is then collected to form a structured medical record dataset. Based on this dataset, a medical record vectorization model for dimensionality reduction is trained. Using this model, individual medical record information is vectorized to obtain a low-dimensional second medical record vector. This second medical record vector is then collected, and a disease trend prediction model is trained using this sample dataset. The disease trend prediction model is then used to detect anomalies in the second medical record vector. Anomaly detection is divided into individual-based and group-based anomaly detection. For abnormal individuals, timely warnings are required; for abnormal groups, disease development trends need to be predicted, and warnings issued for abnormal disease development trends. Furthermore, during the anomaly detection process for groups, an electronic warning map can be dynamically generated in real time.

[0054] The above are embodiments of the methods proposed in this application. Based on the same idea, some embodiments of this application also provide devices and non-volatile computer storage media corresponding to the above methods.

[0055] Figure 3 This is a schematic diagram of a disease anomaly early warning device based on medical records, provided as an embodiment of this application. Figure 3 As shown, it includes:

[0056] At least one processor; and,

[0057] At least one processor-communication-connected memory; wherein,

[0058] The memory stores instructions that can be executed by at least one processor, and the instructions, when executed by at least one processor, enable at least one processor to:

[0059] Obtain the current medical record information, encode the disease representation in the medical record information to obtain a structured first medical record vector;

[0060] By using a pre-trained medical record vectorization model, the first medical record vector is dimensionally compressed to obtain a low-dimensional second medical record vector.

[0061] Using a pre-trained disease trend prediction model, the second medical record vector is used as input to predict anomalies of specified diseases appearing in the current medical record information, and the expected anomalies corresponding to the current medical record information are output based on the prediction results.

[0062] If the deviation level between the medical record information of a specified individual or a specified group and the output result obtained by the disease trend prediction model reaches a preset level, then it is determined that there is an anomaly in the specified individual or the specified group.

[0063] This application provides a non-volatile computer storage medium storing computer-executable instructions, which are configured as follows:

[0064] Obtain the current medical record information, encode the disease representation in the medical record information to obtain a structured first medical record vector;

[0065] By using a pre-trained medical record vectorization model, the first medical record vector is dimensionally compressed to obtain a low-dimensional second medical record vector.

[0066] Using a pre-trained disease trend prediction model, the second medical record vector is used as input to predict anomalies of specified diseases appearing in the current medical record information, and the expected anomalies corresponding to the current medical record information are output based on the prediction results.

[0067] If the deviation level between the medical record information of a specified individual or a specified group and the output result obtained by the disease trend prediction model reaches a preset level, then it is determined that there is an anomaly in the specified individual or the specified group.

[0068] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device and medium embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the description of the method embodiments.

[0069] The devices and media provided in this application are one-to-one with the methods. Therefore, the devices and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0070] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0071] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0072] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0073] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0074] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0075] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0076] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0077] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0078] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A disease anomaly early warning method based on medical records, characterized in that, The method includes: Obtain the current medical record information, encode the disease representation in the medical record information to obtain a structured first medical record vector; By using a pre-trained medical record vectorization model, the first medical record vector is dimensionally compressed to obtain a low-dimensional second medical record vector. Using a pre-trained disease trend prediction model, the second medical record vector is used as input to predict anomalies in the specified diseases appearing in the current medical record information, and the expected anomalies corresponding to the current medical record information are output based on the prediction results. If the deviation level between the medical record information of a specified individual or a specified group and the output result obtained by the disease trend prediction model reaches a preset level, then it is determined that the specified individual or the specified group has an anomaly. The prediction result includes the abnormal probability of the specified disease corresponding to the second medical record vector; After outputting the expected abnormality corresponding to the current period's medical record information based on the prediction results, the method further includes: Based on the individual characteristics of each individual, the individuals are divided into multiple designated groups; the individual characteristics include at least one or more of the following: age, geographical location, and gender; Calculate the first deviation between the abnormal probability and the abnormal expectation for different individuals in the specified group, and the mean of the first deviations for different individuals; Based on the pre-defined deviation ranges, determine the deviation range corresponding to the mean of the first deviation, and thus determine the corresponding deviation level; After outputting the expected abnormality corresponding to the current period's medical record information based on the prediction results, the method further includes: Calculate the second deviation between the abnormal probability of the specified disease for the specified individual and the expected abnormality; Based on the pre-defined deviation ranges, determine the deviation range corresponding to the second deviation, and thus determine the corresponding deviation level; If the deviation level is determined to reach a preset level, the historical incidence rate of the disease suffered by the individual corresponding to the medical record information is calculated, so as to determine the disease type corresponding to the specified individual based on the historical incidence rate; the disease type includes common diseases and rare diseases; After determining the disease type corresponding to the specified individual, the method further includes: If the disease type corresponding to the specified individual is a rare disease, obtain the target medical record vector corresponding to the specified patient; the specified patient suffers from the rare disease. Calculate the similarity between the target medical record vector and the first medical record vector corresponding to the individual, and select the designated patient corresponding to the target medical record vector with the highest similarity to the first medical record vector from multiple designated patients as the target patient; Based on the target medical record vector corresponding to the target patient and the first medical record vector corresponding to the specified individual, the specific disease characteristics of the specified individual are determined and added to the clinical case library.

2. The method for early warning of disease anomalies based on medical records according to claim 1, characterized in that, After determining that the specified individual or the specified group is abnormal, the method further includes: In the event of anomalies in the specified group, the target incidence trend of the group in a specified time period after the current period is predicted based on a preset regression model. Based on the medical record information, obtain the historical incidence trend of the specified group within the specified time period over the years; Based on the target incidence trend and the historical incidence trend, the disease control department is advised to take corresponding prevention and control measures for the designated group.

3. The method for early warning of disease anomalies based on medical records according to claim 2, characterized in that, After determining that the specified individual or the specified group is abnormal, the method further includes: At preset time intervals, the transmission speed, transmission mode, and mortality rate of the specified disease are obtained to determine the weight corresponding to the specified disease; The geographical locations of individuals within the specified group are determined, and the specified diseases are marked on a preset electronic map using corresponding display colors based on the weights corresponding to the specified diseases and the geographical locations; the display colors correspond to the weights corresponding to the specified diseases. Based on the target incidence trend corresponding to the specified group, the radiation range corresponding to the specified disease is marked on the electronic map.

4. The method for early warning of disease anomalies based on medical records according to claim 3, characterized in that, Based on the target incidence trend corresponding to the specified group, the radiation range corresponding to the specified disease is marked on the electronic map, specifically including: Based on the target disease trend corresponding to the specified group, the incidence rate corresponding to each radiation cycle is predicted; the time interval corresponds to the radiation cycle. Determine the total number of people in the designated group, and based on the total number of people and the incidence rate, determine the number of patients within the radiation cycle; Based on the ratio between the number of patients and the total number of people, the corresponding risk level is determined, and the corresponding first radiation radius is determined based on the risk level. Using an individual in the specified group as the first center, the first radiation range corresponding to each individual is determined based on the first center and the first radiation radius; The second radiation range corresponding to the specified disease is determined by taking the center of the union region of each of the first radiation ranges as the second center and the distance between the farthest point of each of the first radiation ranges from the second center and the second center as the second radiation radius.

5. The method for early warning of disease anomalies based on medical records according to claim 1, characterized in that, Encoding the disease characteristics in the medical record information specifically includes: The medical record information is segmented into words, and disease characteristics are filtered out from the segmented medical record information according to a preset disease dictionary; the disease dictionary includes multiple disease characteristics. The disease dictionary is used to encode each disease representation in the medical record information to obtain a one-bit encoded disease representation vector. The disease representation vectors are deduplicated, and the deduplicated disease representation vectors are integrated to obtain a structured first medical record vector composed of the disease representation vectors.

6. A disease anomaly early warning device based on medical records, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enable the at least one processor to: Obtain the current medical record information, encode the disease representation in the medical record information to obtain a structured first medical record vector; By using a pre-trained medical record vectorization model, the first medical record vector is dimensionally compressed to obtain a low-dimensional second medical record vector. Using a pre-trained disease trend prediction model, the second medical record vector is used as input to predict anomalies in the specified diseases appearing in the current medical record information, and the expected anomalies corresponding to the current medical record information are output based on the prediction results. If the deviation level between the medical record information of a specified individual or a specified group and the output result obtained by the disease trend prediction model reaches a preset level, then it is determined that the specified individual or the specified group has an anomaly. The prediction result includes the abnormal probability of the specified disease corresponding to the second medical record vector; After outputting the expected abnormality corresponding to the current period's medical record information based on the prediction results, the method also includes: Based on the individual characteristics of each individual, the individuals are divided into multiple designated groups; the individual characteristics include at least one or more of the following: age, geographical location, and gender; Calculate the first deviation between the abnormal probability and the abnormal expectation for different individuals in the specified group, and the mean of the first deviations for different individuals; Based on the pre-defined deviation ranges, determine the deviation range corresponding to the mean of the first deviation, and thus determine the corresponding deviation level; After outputting the expected abnormality corresponding to the current period's medical record information based on the prediction results, the method also includes: Calculate the second deviation between the abnormal probability of the specified disease for the specified individual and the expected abnormality; Based on the pre-defined deviation ranges, determine the deviation range corresponding to the second deviation, and thus determine the corresponding deviation level; If the deviation level is determined to reach a preset level, the historical incidence rate of the disease suffered by the individual corresponding to the medical record information is calculated, so as to determine the disease type corresponding to the specified individual based on the historical incidence rate; the disease type includes common diseases and rare diseases; After determining the disease type corresponding to the specified individual, the method further includes: If the disease type corresponding to the specified individual is a rare disease, obtain the target medical record vector corresponding to the specified patient; the specified patient suffers from the rare disease. Calculate the similarity between the target medical record vector and the first medical record vector corresponding to the individual, and select the designated patient corresponding to the target medical record vector with the highest similarity to the first medical record vector from multiple designated patients as the target patient; Based on the target medical record vector corresponding to the target patient and the first medical record vector corresponding to the specified individual, the specific disease characteristics of the specified individual are determined and added to the clinical case library.

7. A non-volatile computer storage medium storing computer-executable instructions, characterized in that, The computer-executable instructions are set as follows: Obtain the current medical record information, encode the disease representation in the medical record information to obtain a structured first medical record vector; By using a pre-trained medical record vectorization model, the first medical record vector is dimensionally compressed to obtain a low-dimensional second medical record vector. Using a pre-trained disease trend prediction model, the second medical record vector is used as input to predict anomalies in the specified diseases appearing in the current medical record information, and the expected anomalies corresponding to the current medical record information are output based on the prediction results. If the deviation level between the medical record information of a specified individual or a specified group and the output result obtained by the disease trend prediction model reaches a preset level, then it is determined that the specified individual or the specified group has an anomaly. The prediction result includes the abnormal probability of the specified disease corresponding to the second medical record vector; After outputting the expected abnormality corresponding to the current period's medical record information based on the prediction results, the method also includes: Based on the individual characteristics of each individual, the individuals are divided into multiple designated groups; the individual characteristics include at least one or more of the following: age, geographical location, and gender; Calculate the first deviation between the abnormal probability and the abnormal expectation for different individuals in the specified group, and the mean of the first deviations for different individuals; Based on the pre-defined deviation ranges, determine the deviation range corresponding to the mean of the first deviation, and thus determine the corresponding deviation level; After outputting the expected abnormality corresponding to the current period's medical record information based on the prediction results, the method also includes: Calculate the second deviation between the abnormal probability of the specified disease for the specified individual and the expected abnormality; Based on the pre-defined deviation ranges, determine the deviation range corresponding to the second deviation, and thus determine the corresponding deviation level; If the deviation level is determined to reach a preset level, the historical incidence rate of the disease suffered by the individual corresponding to the medical record information is calculated, so as to determine the disease type corresponding to the specified individual based on the historical incidence rate; the disease type includes common diseases and rare diseases; After determining the disease type corresponding to the specified individual, the method further includes: If the disease type corresponding to the specified individual is a rare disease, obtain the target medical record vector corresponding to the specified patient; the specified patient suffers from the rare disease. Calculate the similarity between the target medical record vector and the first medical record vector corresponding to the individual, and select the designated patient corresponding to the target medical record vector with the highest similarity to the first medical record vector from multiple designated patients as the target patient; Based on the target medical record vector corresponding to the target patient and the first medical record vector corresponding to the specified individual, the specific disease characteristics of the specified individual are determined and added to the clinical case library.

Images