WO2022217713A1

WO2022217713A1 - Syndrome monitoring and early warning method and apparatus, computer device, and storage medium

Info

Publication number: WO2022217713A1
Application number: PCT/CN2021/097415
Authority: WO
Inventors: 唐蕊
Original assignee: 平安科技（深圳）有限公司
Priority date: 2021-04-16
Filing date: 2021-05-31
Publication date: 2022-10-20
Also published as: CN112951443A; CN112951443B

Abstract

A syndrome monitoring and early warning method and apparatus, a computer device, and a readable storage medium, which relate to smart decision-making technology, and relate in particular to a prediction model. The method comprises: extracting symptom keywords from an electronic medical record of a patient (S110); according to the matching situation of the symptom keywords and keywords of a plurality of preset syndromes, determining whether the patient belongs to one of the syndrome cases therein (S120); if determined that the patient belongs to one of the syndrome cases therein, determining the syndrome as a target syndrome (S130); acquiring the number of cases of the target syndrome within a preset time unit (S140); according to the number of cases of the target syndrome within a current preset time unit, calculating a plurality of statistical indexes of the target syndrome (S150); inputting the plurality of statistical indexes of the target syndrome into a risk prediction model, so as to acquire a risk level of the target syndrome (S160); and issuing early warning information of the target syndrome according to the risk level of the target syndrome (S170). The present invention further relates to blockchain technology, and the obtained risk level may be stored in a blockchain.

Description

Syndrome monitoring and early warning method, device, computer equipment and storage medium

This application claims the priority of the Chinese patent application filed on April 16, 2021 with the application number 202110412510.X and the invention titled "Method, Device, Computer Equipment and Storage Medium for Syndrome Monitoring and Early Warning", the entire contents of which are Incorporated herein by reference.

technical field

The present application relates to the field of computer technology, and in particular, to a method, device, computer equipment and storage medium for monitoring and early warning of symptoms.

Background technique

Public health is a public undertaking related to the public health of the people in a country or a region. One of the important tasks of public health is the prevention, monitoring and treatment of major diseases, especially infectious diseases (such as tuberculosis, AIDS, SARS, new coronary pneumonia, etc.). Among them, syndrome monitoring and early warning is a kind of public health abnormal event monitoring and early warning mechanism ahead of infectious disease monitoring and early warning. Syndrome monitoring and early warning can actively monitor the budding state of public health events and give early warnings, providing a basis for public health personnel to take effective prevention and control strategies.

Syndrome monitoring can detect abnormalities or characteristics of disease outbreak signals in time. Syndrome early warning based on syndrome monitoring Through the capture of abnormal syndromes, early warning signals can be issued in advance of specific disease warnings.

The inventor realized that the traditional monitoring and early warning of symptoms mainly analyzes historical data, and then judges the risk level according to the threshold value set manually, which requires a lot of manpower, and the early warning effect is limited by the threshold value set based on experience, and the risk level is judged by the threshold value. It is easy to cause "one size fits all", and it is easy to cause false alarms or omissions of risk levels.

SUMMARY OF THE INVENTION

The present application provides a syndrome monitoring and early warning method, device, computer equipment and storage medium, which can efficiently and effectively provide a syndrome early warning risk level through an intelligent method, so as to provide a basis for public health personnel to take effective prevention and treatment strategies.

In a first aspect, the present application provides a method for monitoring and early warning of symptoms, the method comprising:

Extract symptom keywords from the patient's electronic medical record;

Judging whether the patient belongs to a case of one of the syndromes according to the matching situation of the symptom keywords and the keywords of several preset syndromes;

If it is determined that the patient belongs to a case of one of the syndromes, determine the syndrome as the target syndrome;

Obtain the number of cases of the target syndrome within a preset time unit;

Calculate several statistical indicators of the target syndrome according to the number of cases of the target syndrome within the current preset time unit;

Inputting several statistical indicators of the target syndrome into a risk prediction model to obtain the risk level of the target syndrome;

According to the risk level of the target syndrome, the early warning information of the target syndrome is released.

In a second aspect, the present application provides a syndrome monitoring and early warning device, including:

The symptom extraction module is used to extract symptom keywords from the electronic medical record of the patient;

The symptom matching module is used to judge whether the patient belongs to a case of one of the target syndromes according to the matching situation of the symptom keywords and the keywords of several preset syndromes;

a target syndrome determination module, configured to determine the syndrome as the target syndrome if it is determined that the patient belongs to one of the syndrome cases;

A case statistics module, used to obtain the number of cases of the target syndrome within a preset time unit;

The index calculation module is used to calculate some statistical indexes of the target syndrome according to the number of cases of the target syndrome in the current preset time unit;

a risk prediction module for inputting several statistical indicators of the target syndrome into a risk prediction model to obtain the risk level of the target syndrome;

An early warning module, configured to issue early warning information of the target syndrome according to the risk level of the target syndrome.

In a third aspect, the present application provides a computer device, the computer device includes a memory and a processor; the memory is used for storing a computer program; the processor is used for executing the computer program and executing the computer During the program, the above-mentioned symptom monitoring and early warning method is realized.

In a fourth aspect, the present application provides a computer-readable storage medium, where a computer program is stored in the computer-readable storage medium, and if the computer program is executed by a processor, the above-mentioned method for monitoring and early warning of a syndrome is implemented.

The present application discloses a syndrome monitoring method, device, computer equipment and storage medium. By extracting symptom keywords from an electronic medical record of a patient; according to the matching situation of the symptom keywords and the keywords of a target syndrome, the patient is judged. Whether it belongs to the case of the target syndrome; count the number of cases of the target syndrome according to the current preset time unit; calculate some statistical indicators of the target syndrome according to the number of cases of the target syndrome; input the statistical indicators into the risk prediction model, In order to obtain the predicted risk level, artificial intelligence-based syndrome monitoring and early warning can be realized, and the efficiency and accuracy of syndrome monitoring and early warning can be improved.

Description of drawings

In order to explain the technical solutions of the embodiments of the present application more clearly, the following briefly introduces the accompanying drawings used in the description of the embodiments. For those of ordinary skill, other drawings can also be obtained from these drawings without any creative effort.

1 is a schematic flowchart of a syndrome monitoring and early warning method provided by an embodiment of the present application;

2 is a schematic block diagram of the structure of a syndrome monitoring and early warning device provided by an embodiment of the present application;

FIG. 3 is a schematic structural block diagram of a computer device provided by an embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

The flowcharts shown in the figures are for illustration only, and do not necessarily include all contents and operations/steps, nor do they have to be performed in the order described. For example, some operations/steps can also be decomposed, combined or partially combined, so the actual execution order may be changed according to the actual situation. In addition, although the functional modules are divided in the schematic diagram of the device, in some cases, the modules may be divided differently from the schematic diagram of the device.

Embodiments of the present application provide a syndrome monitoring and early warning method, apparatus, computer device, and computer-readable storage medium. It is used to improve the efficiency and accuracy of syndrome monitoring and early warning based on artificial intelligence technology. Exemplarily, in the monitoring and early warning of the syndrome, if the historical data is analyzed and then the risk level is judged according to the manually set threshold, it requires a lot of manpower, and the early warning effect is limited by the threshold set based on experience. According to the syndrome early warning method of the embodiment of the present application, the early warning risk level is efficiently provided by the method based on artificial intelligence, and the problem of false positives or false negatives caused by "one size fits all" according to the threshold value in the traditional method when the risk level is judged by relying on the manual setting of the threshold value is avoided, Improves the accuracy of risk level predictions.

Wherein, the syndrome monitoring and early warning method can be applied to a server, and of course can also be applied to a terminal, wherein the terminal can be an electronic device such as a mobile phone, a tablet computer, a notebook computer, a desktop computer, etc.; the server can be, for example, a separate server or a server cluster. However, in order to facilitate understanding, the following embodiments will be described in detail with a method for monitoring and early warning of symptoms applied to a server.

Some embodiments of the present application will be described in detail below with reference to the accompanying drawings. The embodiments described below and features in the embodiments may be combined with each other without conflict.

Please refer to FIG. 1. FIG. 1 is a schematic flowchart of a method for monitoring and early warning of a syndrome provided by an embodiment of the present application.

As shown in FIG. 1 , the syndrome monitoring and early warning method may include the following steps S110-S170.

Step S110, extracting symptom keywords from the electronic medical record of the patient.

The symptom keywords may include symptoms and related attributes of the symptoms, and the related attributes may be the time, duration, and severity of the symptoms. ", the corresponding symptom keyword is "fever for three days", where "three days" is the manifestation of symptoms, and "three days" is one of the related attributes of symptoms. In specific implementation, the composition of the symptom keyword can be determined according to the composition of the keyword of the target syndrome. For example, the keyword of the target syndrome only has symptoms, then the symptom keyword also only retains symptoms such as "fever". It is sufficient to express. If the keyword of the target syndrome includes the manifestation of the symptom and the related attributes of the symptom, the symptom keyword also includes the manifestation of the symptom and the related attribute of the symptom.

Exemplarily, step S110 specifically includes steps S111-S112:

Step S111, through named entity recognition, extract the initial keywords of the symptoms of the patient from the text of the electronic medical record;

Named Entity Recognition (NER, Named Entity Recognition) is a natural language processing technology that recognizes entities with specific meanings in text. Through named entity recognition, the desired entity type can be extracted from the text. For example, the symptom of the patient is the type of entity that is expected to be obtained. The tool developed through the named entity recognition technology can extract specific symptoms such as "fever" and "nausea" in the sentence from the electronic medical record.

Step S112 , based on a preset standardized vocabulary, replace the alias vocabulary in the symptom initial keyword with a standard vocabulary to obtain the symptom keyword.

For example, if the initial keyword of the symptom is "nausea", and the corresponding standard medical term should be "diarrhea", then "nausea" is an alias word, and the symptom keyword obtained by replacing the standard word is "diarrhea".

The standardized vocabulary can be formed by collecting and arranging symptom-related alias words and standard words.

Exemplarily, the symptom keywords are organized into structured data. Structured data, also known as row data, is data that is logically expressed and implemented by a two-dimensional table structure. It strictly follows the data format and length specifications, and is mainly stored and managed through relational databases. Data management can be facilitated through structuring. , use, display.

Step S120 , according to the matching situation of the symptom keyword and the keywords of several preset syndrome groups, determine whether the patient belongs to a case of one of the syndrome groups.

Syndrome refers to multiple clinical symptoms that often appear at the same time when certain possible diseases appear. For example, as shown in Table 1, there are five preset syndromes and keywords corresponding to their symptoms in an embodiment. During specific implementation, new syndromes can be added or defined as needed.

Table 1: Syndromes and their corresponding keywords

In some embodiments, step S120 specifically includes an exact matching step and/or a fuzzy matching step.

Exemplarily, the exact matching step includes:

Compare the symptom keywords with each keyword of each syndrome, and if all the symptom keywords have corresponding keywords in one of the syndromes, then determine that the patient belongs to the case of the one of the syndromes .

For example, a number of preset syndromes include the first syndrome, the keyword of the first syndrome is {A, B, C, D}, the symptom keyword of the patient A is {A, B, C, D}, and the patient B's symptom keywords are {C,D,E,F,G}. One-to-one comparison is made between the symptom keywords of patient A and each keyword of the first syndrome, and the formula is calculated according to the matching degree: matching degree=the number of the keywords of the syndrome being compared that are consistent with the symptom keywords / The number of the symptom keywords, the matching degree is 4/4=1, indicating that each symptom keyword of the patient A has the same keyword of the first syndrome, so the patient A is the first symptom keyword. A case of a syndrome; one-to-one comparison between the symptom keywords of the patient B and each keyword of the first syndrome, the matching degree obtained according to the matching degree calculation formula is 2/5=0.4, and the matching degree is less than 1 , indicating that there are some keywords that do not have the first syndrome in the symptom keywords of the patient B, and it is currently impossible to determine that the patient B is a case of the first syndrome.

Exemplarily, the fuzzy matching step includes steps S122a-step S122e:

Step S122a, construct a multi-hot encoding of the symptom keyword and a multi-hot encoding of the keyword of a certain syndrome according to the symptom keyword and the keyword of a certain syndrome.

For example, the symptom keywords of patient B are {C, D, E, F, G}, the keywords of the first syndrome are {A, B, C, D}, and the symptom keywords of patient B correspond to the multiple fever The encoding is {0, 0, 1, 1, 1, 1, 1}, the elements in the multi-hot encoding respectively represent the absence of keyword A, the absence of keyword B, the presence of keyword C..., and the key of the first syndrome The corresponding multi-hot encoding of words is {1, 1, 1, 1, 0, 0, 0}.

Step S122b: Input the multi-hot encoding of the symptom keyword into an autoencoder, and obtain an autoencoder vector output by the autoencoder.

Exemplarily, the method further includes a training step S210 of acquiring the autoencoder, specifically including steps S211-S215.

Step S211: Acquire training data, where the training data includes multi-hot encoding.

Exemplarily, the multi-hot encoding of the symptom keyword or the multi-hot encoding of the keyword of a certain syndrome may be used as the training data, or the multi-hot encoding as the training data may be obtained by random generation.

Step S212: Obtain an autoencoder to be trained, where the autoencoder includes a first network and a second network.

Exemplarily, the structures of the first network and the second network are symmetrical; both the first network and the second network include multiple fully connected layers. A fully-connected layer, that is, each node of this layer is connected to all nodes of the previous layer, and the nonlinear expression capability of the autoencoder can be improved by using multiple fully-connected layers.

Step S213: Input the multi-hot encoding in the training data into the first network to obtain an auto-encoding vector output by the first network, where the dimension of the self-encoding vector is lower than the dimension of the multi-hot encoding.

The first network constitutes an encoder to compress the high-dimensional sparse multi-hot encoding into the low-dimensional dense self-encoding vector.

Step S214, inputting the self-encoding vector output by the first network into the second network to obtain a vector output by the second network, and the dimension of the vector output by the second network is equal to the dimension of the multi-hot encoding;

The second network constitutes a decoder that reconstructs the low-dimensional dense self-encoding vector into a high-dimensional sparse multi-hot encoding.

Step S215: Train the first network and the second network according to the vector output by the second network and the multi-hot encoding.

Exemplarily, according to the vector output by the second network and the error of the multi-hot encoding, the network parameters of the first network and the second network are adjusted, so that the vector output by the second network and the Multi-hot encoding tends to be consistent.

Exemplarily, after the training of the autoencoder is completed, the multi-hot encoding is input into the autoencoder, and the autoencoder vector is obtained at the output of the first network.

Step S122c, according to the multi-hot encoding of the keyword of a certain syndrome, obtain the self-encoding vector of the keyword of the certain syndrome by the autoencoder;

Exemplarily, the multi-hot encoding of the keywords of the certain syndrome is input into the autoencoder, and the autoencoder vector of the certain syndrome is obtained at the output of the first network.

Step S122d, calculating the similarity of the self-encoding vector of the symptom keyword and the self-encoding vector of the keyword of a certain syndrome;

Exemplarily, the similarity is calculated by cosine similarity, and the calculation formula of cosine similarity is:

Wherein, X is the self-encoding vector of the symptom keyword, Y is the self-encoding vector of the keyword of the certain syndrome, and n is the dimension of the self-encoding vector.

In other embodiments, the similarity may also be calculated by methods such as Euclidean distance, Manhattan distance, and Minkowski distance.

Step S122e, judging whether the patient belongs to a case of the certain syndrome according to the similarity.

Exemplarily, a similarity threshold is preset, for example, the similarity threshold is set to 0.8. If the similarity exceeds the similarity threshold, the corresponding patient is a case of the certain syndrome; otherwise, the corresponding patient is not a case of the certain syndrome.

Exemplarily, the precise matching step and the fuzzy matching step are used in combination, for example, the precise matching step is performed first, and it is not possible to determine whether the patient belongs to the certain syndrome after the precise matching step. , and then perform the fuzzy matching step to further determine whether the patient belongs to the case of the certain syndrome.

Step S130 , if it is determined that the patient belongs to a case of one of the syndromes, determine that the syndrome is the target syndrome.

Exemplarily, the electronic medical records are processed one by one through steps S110 to S130 to obtain all the target syndromes; if after all the electronic medical records are processed, the number of the target syndromes is still 0, it means that there is currently no forecast. No risk warning is required for cases of the designated syndromes.

Step S140: Acquire the number of cases of the target syndrome within a preset time unit.

For example, the preset time unit is days, and the number of cases of the target syndrome in days is obtained.

Exemplarily, the number of cases of the target syndrome may be counted during the execution of step S120 or step S130. For example, if it is determined that a patient belongs to one of the syndrome cases, the number of cases of the syndrome corresponding to the visit date is increased by one.

Step S150: Calculate several statistical indexes of the target syndrome according to the number of cases of the target syndrome within the current preset time unit.

Exemplarily, according to the number of historical cases of the target syndrome in the current preset time unit, and the number of historical cases of the target syndrome, calculate several statistical indicators of the target syndrome, and the statistical indicators include the chain growth rate, One or more of year-over-year growth rate, historical percentile. The chain growth rate represents the change ratio of the statistics in two consecutive statistical periods. During the specific implementation, the chain ratio corresponding to multiple statistical periods can be calculated. If an early warning is carried out on September 17, 2020 (the number of cases of the target syndrome in the current preset time unit), and the number of cases of the target syndrome on September 16, 2020 (the number of historical cases of the target syndrome) to calculate the daily-monthly growth rate, based on the number of cases of the target syndrome from September 11, 2020 to September 17, 2020 and the number of cases of the target syndrome from September 4, 2020 to September 2020. Week-to-month growth rate calculated for the number of cases of the target syndrome described on 10th August 2020, based on the number of cases of the target syndrome described in the period from 18th August 2020 to 17th September 2020 and the The number of cases of the target syndrome mentioned on the 17th is calculated to calculate the month-on-month growth rate; the year-on-year growth rate generally refers to the change ratio of the statistics in a certain statistical period this year relative to the amount in the same period last year. Similarly, the corresponding statistics can be calculated. The year-on-year period of the cycle, for example, is calculated based on the number of cases of the target syndrome from September 11, 2020 to September 17, 2020 and the number of cases of the target syndrome from September 11, 2019 to September 17, 2019- Year-on-year growth rate, calculated based on the number of cases of the target syndrome from August 18, 2020 to September 17, 2020 and the number of cases of the target syndrome from August 18, 2019 to September 17, 2020 month-year-on-year Growth rate; historical percentile, which refers to the position at which the statistic in a certain period is quantified by percentage in the historical reference data, for example, the number of said cases in the past four days were 100, 500, 300, 200, and today's cases If the number of cases is 400, and the number of cases in the past 4 days is lower than the number of cases today, then the number of cases in the past four days is used as the historical reference data, and the historical percentile of the number of cases today is 3/4=75% ; Similarly, historical percentiles corresponding to multiple statistical periods can be calculated. For example, the number of cases of the target syndrome from January 1, 2019 to December 31, 2019 is used as historical reference data. The number of cases of the target syndrome on the 17th and the number of cases of the target syndrome from January 1, 2019 to December 31, 2019 Calculation date-historical percentile, based on September 11, 2020 to September 2020 The number of cases of the target syndrome described on the 17th and the number of cases of the target syndrome described in January 1, 2019 to December 31, 2019 Calculation week-historical percentile, based on August 18, 2020 to September 2020 Month-historical percentiles were calculated for the number of cases of the target syndrome on the 17th and the number of cases of the target syndrome from January 1, 2019 to December 31, 2019. Of course, during the specific implementation, the statistical indicators adopted in this method are not limited to the above-mentioned ones. The growth rate of the number of cases relative to the number of cases of the target syndrome described on September 14, 2020, the number of cases of the target syndrome described on September 15, 2020, and other statistical indicators that can reflect changes in data such as the fixed-base ratio can also be used in this method.

In this embodiment, the number of cases of the target syndrome is used as a statistic to calculate the statistical index. In other embodiments, the data obtained according to the number of cases of the target syndrome can be used as a statistic to calculate the statistical index. , for example, the statistical index is calculated by taking the ratio of the number of cases of the target syndrome to the patient as a statistic.

Step S160: Input several statistical indicators of the target syndrome into a risk prediction model to obtain a predicted risk level.

Exemplarily, the statistical indicators and risk factors are input into a risk prediction model to obtain the risk level of the target syndrome, and the risk factors include weather data and/or environmental data. For example, the weather data includes the current temperature, humidity, and whether it is a sunny, cloudy or rainy day, and the environmental data includes the current air quality, water quality level, and the like. By introducing the risk factor, the risk level model can predict the risk level from more abundant data dimensions, so that the prediction of the risk level is more accurate. The risk prediction model can be trained by the following methods: respectively inputting the statistical indicators and the risk factors corresponding to a certain time in the past into the initial model, the initial model in this embodiment is the XGboost model, and the specific implementation can also be Use feedforward neural network, support vector machine and other machine learning models; use the risk level at a certain time as the label of the statistical index, and use the label as the expected output of the model to train the model. The risk level at a certain time in the past, if there are corresponding query channels, you can directly query through these channels. For example, the target syndrome is fever respiratory syndrome, and the society has the corresponding risk level of the fever respiratory tract announced, you can use these channels to query. For the risk level of the fever respiratory tract published in the past, obtain the risk level of the fever respiratory syndrome at a certain time in the past; if there is no corresponding query channel, it can be judged according to the risk level of the main related diseases that can be queried for the target syndrome. For example, the target syndrome is febrile respiratory syndrome, there is currently no corresponding risk level of febrile respiratory tract, and the main disease related to febrile respiratory syndrome is the new crown epidemic, then the fever respiratory tract can be obtained according to the risk level of the new crown epidemic at that time. The risk level of the syndrome at that time.

Exemplarily, the risk prediction model is explained by an explanation model (SHAP, SHapley Additive exPlanations). By interpreting the model, the impact of the statistical indicators in the input of the risk model or the risk factors in obtaining the risk prediction result can be evaluated, so as to know which data has an important impact on the risk level, Subsequent focus on data with significant impact improves the usability of the risk prediction model.

Step S170: Publish early warning information of the target syndrome according to the risk level of the target syndrome.

For example, the risk level of the target syndrome includes high, medium and low, and the target syndrome is febrile respiratory syndrome. If the risk level of the target syndrome is high, the early warning information of the target syndrome to be released includes: high risk of fever and respiratory syndrome, heightened vigilance, and implementation of high-level prevention and control measures; if the risk level of the target syndrome is medium, all published The early warning information of the target syndrome includes: the risk of febrile respiratory syndrome, pay close attention to the situation of the disease, and implement general prevention and control measures; if the risk level of the target syndrome is low, the published early warning information of the target syndrome includes: fever respiratory syndrome low risk.

The present application predicts the risk level by the method based on the risk prediction model, which avoids the defects of labor-intensive and low efficiency in the manual analysis. At the same time, the risk level is obtained by the risk prediction model through nonlinear calculation, which avoids judging the risk level by the threshold value. It is easy to cause "one size fits all", which improves the accuracy of risk level prediction.

It should be emphasized that, in order to further ensure the privacy and security of the above symptom keywords, statistical indicators, risk prediction models and risk levels, the above symptom keywords, statistical indicators, the risk prediction model and risk levels can also be stored in one area. in the nodes of the blockchain.

The blockchain referred to in this application is a new application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm. Blockchain, essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information to verify its Validity of information (anti-counterfeiting) and generation of the next block. The blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.

As shown in FIG. 2 , the syndrome monitoring and early warning device includes: a symptom extraction module 110 , a symptom matching module 120 , a target syndrome determination module 130 , a case statistics module 140 , an index calculation module 150 , a risk prediction module 160 and an early warning module 170 .

The symptom extraction module 110 is used for extracting symptom keywords from the electronic medical record of the patient;

Exemplarily, the symptom extraction module 110 includes an initial keyword extraction module and a normalization module.

an initial keyword extraction module, used for extracting the initial keywords of the symptoms of the patient from the text of the electronic medical record through named entity recognition;

The standardization module is configured to replace the alias words in the symptom initial keywords with standard words based on a preset tagging vocabulary, so as to obtain the symptom keywords. The standardized vocabulary can be formed by collecting and arranging symptom-related alias words and standard words.

The symptom matching module 120 is configured to judge whether the patient belongs to a case of one of the syndromes according to the matching situation of the symptom keyword and the keywords of several preset syndromes.

Exemplary symptom matching modules 120 include exact matching modules and fuzzy matching modules.

Exemplarily, an exact matching module is used to compare the symptom keywords with each keyword of each syndrome, and if all the symptom keywords have corresponding keywords in one of the syndromes, determine the The patient belongs to the case of one of the syndromes described.

Exemplarily, the fuzzy matching module includes a multi-hot encoding building module, a symptom keyword encoding module, a target syndrome keyword encoding module, a similarity calculation module, and a case judgment module.

A multi-hot encoding building module is configured to construct a multi-hot encoding of the symptom keyword and a multi-hot encoding of the keyword of a certain syndrome according to the symptom keyword and the keyword of a certain syndrome.

The symptom keyword encoding module is configured to input the multi-hot encoding of the symptom keyword into the autoencoder, and obtain the autoencoder vector output by the autoencoder.

The syndrome keyword encoding module is configured to obtain the autoencoding vector of the keyword of the certain syndrome through the autoencoder according to the multi-hot encoding of the keyword of the certain syndrome.

The similarity calculation module is used to calculate the self-encoding vector of the symptom keyword and the similarity of the keyword of a certain syndrome.

Similarity of autoencoding vectors.

A case judgment module, configured to judge whether the patient belongs to a case of the certain syndrome group according to the similarity.

Exemplarily, the apparatus further includes an autoencoder training module, and the autoencoder training module includes a training data acquisition module, an autoencoder initial module, a compression module, a decompression module, and an autoencoder training module.

a training data acquisition module for acquiring training data, the training data including multi-hot encoding;

The autoencoder initial module is used to obtain the autoencoder to be trained, and the autoencoder includes a first network and a second network;

a compression module, configured to input the multi-hot encoding in the training data into the first network to obtain an auto-encoding vector output by the first network, the dimension of the self-encoding vector is lower than the dimension of the multi-hot encoding ;

A decompression module, configured to input the self-encoding vector output by the first network into the second network, to obtain a vector output by the second network, the dimension of the vector output by the second network is equal to the multi-hot encoding dimension;

An autoencoder training module, configured to train the first network and the second network according to the vector output by the second network and the multi-hot encoding.

The target syndrome determination module 130 is configured to determine the syndrome as the target syndrome if it is determined that the patient belongs to one of the syndrome cases.

A case statistics module 140, configured to acquire the number of cases of the target syndrome within a preset time unit;

The index calculation module 150 is configured to calculate several statistical indexes of the target syndrome according to the number of cases of the target syndrome in a preset time unit;

Exemplarily, according to the number of cases of the target syndrome in the current preset time unit, and the number of historical cases of the target syndrome, calculate several statistical indicators of the target syndrome, and the statistical indicators include a chain growth rate, a year-on-year growth rate, One or more of growth rate, historical percentile.

The risk prediction module 160 is configured to input several statistical indicators of the target syndrome into a risk prediction model to obtain the risk level of the target syndrome.

Exemplarily, the statistical indicators and risk factors are input into a risk prediction model to obtain the risk level of the target syndrome, and the risk factors include weather data and/or environmental data. For example, the weather data includes the current temperature, humidity, and whether it is a sunny, cloudy or rainy day, and the environmental data includes the current air quality, water quality level, and the like. By introducing the risk factor, the risk level model can predict the risk level from more abundant data dimensions, so that the prediction of the risk level is more accurate.

The early warning module 170 is configured to issue early warning information of the target syndrome according to the risk level.

It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, for the specific working process of the above-described device and each module and unit, reference may be made to the corresponding process in the foregoing method embodiments. No longer.

The methods and apparatus of the present application may be used in numerous general purpose or special purpose computing system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, including the above Distributed computing environment of any system or device, etc.

Exemplarily, the above-mentioned method and apparatus can be implemented in the form of a computer program, and the computer program can be executed on a computer device as shown in FIG. 3 .

Please refer to FIG. 3 , which is a schematic diagram of a computer device provided by an embodiment of the present application. The computer device can be a server or a terminal.

As shown in FIG. 3, the computer device includes a processor, a memory, and a network interface connected through a system bus, wherein the memory may include a non-volatile storage medium and an internal memory.

The nonvolatile storage medium can store operating systems and computer programs. The computer program includes program instructions, which, when executed, can cause the processor to execute any one of the syndrome monitoring and early warning methods.

The processor is used to provide computing and control capabilities to support the operation of the entire computer equipment.

The internal memory provides an environment for running the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any one of the syndrome monitoring and early warning methods.

The network interface is used for network communication, such as sending assigned tasks. Those skilled in the art can understand that the structure of the computer device is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. More or fewer components are shown in the figures, either in combination or with different arrangements of components.

It should be understood that the processor may be a central processing unit (Central Processing Unit, CPU), and the processor may also be other general-purpose processors, digital signal processors (Digital Signal Processors, DSP), application specific integrated circuits (Application Specific Integrated circuits) Circuit, ASIC), Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Wherein, the general-purpose processor can be a microprocessor or the processor can also be any conventional processor or the like.

Wherein, in some embodiments, the processor is configured to run a computer program stored in the memory, so as to realize the following steps: extracting symptom keywords from the electronic medical record of the patient; If it is determined that the patient belongs to a case of one of the syndrome groups according to the matching of the keywords of the syndrome, it is determined whether the patient belongs to a case of one of the syndrome groups, and the syndrome is determined as the target syndrome; number of cases; according to the number of cases of the target syndrome in the current preset time unit, calculate some statistical indicators of the target syndrome; input some statistical indicators of the target syndrome into a risk prediction model to obtain the risk level of the target syndrome ; According to the risk level of the target syndrome, the early warning information of the target syndrome is released.

Exemplarily, the processor is configured to extract symptom keywords from an electronic medical record of a patient, to achieve: extracting initial keywords of symptoms of a patient from the text of the medical record through named entity recognition; based on a preset standardized vocabulary , and replace the alias words in the symptom initial keywords with standard words to obtain the symptom keywords.

Exemplarily, the processor is configured to realize that when judging whether the patient belongs to a case of the target syndrome according to the matching situation of the symptom keyword and the keywords of a number of preset syndromes, realize: compare the symptom keyword with each syndrome Each keyword is compared, and if all the symptom keywords have corresponding keywords in one of the syndromes, it is determined that the patient belongs to the case of the one of the syndromes; or/and realize: according to the Symptom keywords and keywords of a certain syndrome, construct the multi-hot encoding of the symptom keywords and the multi-hot encoding of the keywords of a certain syndrome; input the multi-hot encoding of the symptom keywords into the self-encoder, Obtain the self-encoding vector of the symptom keyword output by the autoencoder; according to the multi-hot encoding of the keyword of the certain syndrome, obtain the self-encoding of the keyword of the certain syndrome through the autoencoder vector; calculate the similarity between the self-encoding vector of the symptom keyword and the self-encoding vector of the keyword of a certain syndrome; determine whether the patient belongs to a case of the certain syndrome according to the similarity.

Exemplarily, when the processor is configured to acquire the autoencoder, it is implemented: acquiring training data, where the training data includes multi-hot encoding; acquiring an autoencoder to be trained, where the autoencoder includes a first network and a second network; input the multi-hot encoding in the training data into the first network, and obtain the self-encoding vector output by the first network, the dimension of the self-encoding vector is lower than the dimension of the multi-hot encoding; The self-encoding vector output by the first network is input into the second network to obtain a vector output by the second network, and the dimension of the vector output by the second network is equal to the dimension of the multi-hot encoding; according to the second network A vector of network outputs and the multi-hot encoding to train the first network and the second network.

Exemplarily, the processor is used for the number of cases of the target syndrome in the current preset time unit, and when calculating several statistical indicators of the target syndrome, it is realized: according to the number of cases of the target syndrome in the current preset time unit, and The number of historical cases of the target syndrome is calculated, and several historical statistical indicators of the target syndrome are calculated, and the statistical indicators include one or more of a month-on-month growth rate, a year-on-year growth rate, and a historical percentile.

Exemplarily, when the processor is configured to input several statistical indicators of the target syndrome into a risk prediction model to obtain a predicted risk level, it is achieved: inputting the statistical indicators and risk factors into a risk prediction model to obtain the target syndrome. , the risk factors include weather data and/or environmental data. For example, the weather data includes the current temperature, humidity, and whether it is a sunny, cloudy or rainy day, and the environmental data includes the current air quality, water quality level, and the like.

A computer-readable storage medium, where the computer-readable storage medium stores a computer program, the computer program includes program instructions, and the processor executes the program instructions to implement any one of the syndromes provided in the embodiments of the present application Monitoring and early warning methods.

The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as a hard disk or a memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk equipped on the computer device, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) ) card, Flash Card, etc. The computer-readable storage medium may be non-volatile or volatile.

The above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of various equivalents within the technical scope disclosed in the present application. Modifications or substitutions shall be covered by the protection scope of this application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

A syndrome monitoring and early warning method, wherein the method comprises:

Extract symptom keywords from the patient's electronic medical record;

Judging whether the patient belongs to a case of one of the syndromes according to the matching situation of the symptom keywords and the keywords of several preset syndromes;

If it is determined that the patient belongs to a case of one of the syndromes, determine the syndrome as the target syndrome;

Obtain the number of cases of the target syndrome within a preset time unit;

Calculate several statistical indicators of the target syndrome according to the number of cases of the target syndrome within the current preset time unit;

Inputting several statistical indicators of the target syndrome into a risk prediction model to obtain the risk level of the target syndrome;

According to the risk level of the target syndrome, the early warning information of the target syndrome is released.
The syndrome monitoring and early warning method according to claim 1, wherein, according to the matching situation of the symptom keywords and the keywords of several preset syndromes, judging whether the patient belongs to a case of one of the syndromes includes:

Compare the symptom keywords with each keyword of each syndrome, and if all the symptom keywords have corresponding keywords in one of the syndromes, then determine that the patient belongs to the case of the one of the syndromes .
The syndrome monitoring and early warning method according to claim 1, wherein, according to the matching situation of the symptom keywords and the keywords of several preset syndromes, it is judged whether the patient belongs to a case of one of the syndromes, including:

According to the symptom keyword and the keyword of a certain syndrome, construct the multi-hot encoding of the symptom keyword and the multi-hot encoding of the keyword of the certain syndrome;

Input the multi-hot encoding of the symptom keyword into an autoencoder, and obtain the autoencoder vector of the symptom keyword output by the autoencoder;

According to the multi-hot encoding of the keyword of the certain syndrome, the autoencoder vector of the keyword of the certain syndrome is obtained by the autoencoder;

Calculate the similarity between the self-encoding vector of the symptom keyword and the self-encoding vector of the keyword of a certain syndrome;

According to the similarity, it is determined whether the patient belongs to a case of the certain syndrome.
The syndrome monitoring and early warning method according to claim 3, wherein: the method further comprises:

obtaining training data, the training data includes multi-hot encoding;

Obtain an autoencoder to be trained, the autoencoder includes a first network and a second network;

Inputting the multi-hot encoding in the training data into the first network to obtain an auto-encoding vector output by the first network, where the dimension of the self-encoding vector is lower than the dimension of the multi-hot encoding;

Inputting the self-encoding vector output by the first network into the second network, to obtain a vector output by the second network, the dimension of the vector output by the second network is equal to the dimension of the multi-hot encoding;

The first network and the second network are trained according to the vector output by the second network and the multi-hot encoding.
The syndrome monitoring and early warning method according to any one of claims 1-4, wherein: according to the number of cases of the target syndrome in the current preset time unit, calculate some statistical indicators of the target syndrome, including:

According to the number of cases of the target syndrome in the current preset time unit, and the number of historical cases of the target syndrome, calculate several statistical indicators of the target syndrome, and the statistical indicators include the chain growth rate, year-on-year growth rate, historical One or more of the percentiles.
The syndrome monitoring and early warning method according to any one of claims 1-4, wherein: inputting the statistical index into a risk prediction model to obtain the risk level of the target syndrome, comprising:

The statistical indicators and risk factors are input into a risk prediction model to obtain the risk level of the target syndrome, and the risk factors include weather data and/or environmental data.
The syndrome monitoring and early warning method according to claim 1, wherein the extracting symptom keywords from the electronic medical record of the patient comprises:

Through named entity recognition, the initial keywords of the symptoms of the patient are extracted from the text of the medical record;

According to a preset standardized vocabulary, the alias vocabulary in the symptom initial keyword is replaced with a standard vocabulary to obtain the symptom keyword.
A syndrome monitoring and early warning device, wherein the device comprises:

The symptom extraction module is used to extract symptom keywords from the electronic medical record of the patient;

The symptom matching module is used to judge whether the patient belongs to a case of one of the target syndromes according to the matching situation of the symptom keywords and the keywords of several preset syndromes;

a target syndrome determination module, configured to determine the syndrome as the target syndrome if it is determined that the patient belongs to one of the syndrome cases;

A case statistics module, used to obtain the number of cases of the target syndrome within a preset time unit;

an index calculation module, used for calculating several statistical indexes of the target syndrome according to the number of cases of the target syndrome in a preset time unit;

a risk prediction module for inputting several statistical indicators of the target syndrome into a risk prediction model to obtain the risk level of the target syndrome;

An early warning module, configured to issue early warning information of the target syndrome according to the risk level of the target syndrome.
A computer device, wherein the computer device includes a memory and a processor;

the memory for storing computer programs;

The processor is configured to execute the computer program and implement the following steps when executing the computer program:

Extract symptom keywords from the patient's electronic medical record;

Judging whether the patient belongs to a case of one of the syndromes according to the matching situation of the symptom keywords and the keywords of several preset syndromes;

If it is determined that the patient belongs to a case of one of the syndromes, determine the syndrome as the target syndrome;

Obtain the number of cases of the target syndrome within a preset time unit;

Calculate several statistical indicators of the target syndrome according to the number of cases of the target syndrome within the current preset time unit;

Inputting several statistical indicators of the target syndrome into a risk prediction model to obtain the risk level of the target syndrome;

According to the risk level of the target syndrome, the early warning information of the target syndrome is released.
The computer device according to claim 9, wherein, when the processor determines whether the patient belongs to a case of one of the syndromes according to the matching situation of the symptom keyword and the keywords of several preset syndromes , which is used to implement:

Compare the symptom keywords with each keyword of each syndrome, and if all the symptom keywords have corresponding keywords in one of the syndromes, then determine that the patient belongs to the case of the one of the syndromes .
The computer device according to claim 9, wherein, when the processor determines whether the patient belongs to a case of one of the syndromes according to the matching situation of the symptom keyword and the keywords of several preset syndromes , which is used to implement:

According to the symptom keyword and the keyword of a certain syndrome, construct the multi-hot encoding of the symptom keyword and the multi-hot encoding of the keyword of the certain syndrome;

Input the multi-hot encoding of the symptom keyword into an autoencoder, and obtain the autoencoder vector of the symptom keyword output by the autoencoder;

According to the multi-hot encoding of the keyword of the certain syndrome, the autoencoder vector of the keyword of the certain syndrome is obtained by the autoencoder;

Calculate the similarity between the self-encoding vector of the symptom keyword and the self-encoding vector of the keyword of a certain syndrome;

According to the similarity, it is determined whether the patient belongs to a case of the certain syndrome.
The computer device of claim 11, wherein the processor is further configured to implement:

obtaining training data, the training data includes multi-hot encoding;

Obtain an autoencoder to be trained, the autoencoder includes a first network and a second network;

Inputting the multi-hot encoding in the training data into the first network to obtain an auto-encoding vector output by the first network, where the dimension of the self-encoding vector is lower than the dimension of the multi-hot encoding;

Inputting the self-encoding vector output by the first network into the second network, to obtain a vector output by the second network, the dimension of the vector output by the second network is equal to the dimension of the multi-hot encoding;

The first network and the second network are trained according to the vector output by the second network and the multi-hot encoding.
The computer device according to any one of claims 9-12, wherein, when the processor calculates several statistical indicators of the target syndrome according to the number of cases of the target syndrome in the current preset time unit , which is used to implement:

According to the number of cases of the target syndrome in the current preset time unit, and the number of historical cases of the target syndrome, calculate several statistical indicators of the target syndrome, and the statistical indicators include the chain growth rate, year-on-year growth rate, historical One or more of the percentiles.
The computer device according to any one of claims 9-12, wherein, when the processor implements the inputting the statistical index into a risk prediction model to obtain the risk level of the target syndrome, the processor is configured to implement :

The statistical indicators and risk factors are input into a risk prediction model to obtain the risk level of the target syndrome, and the risk factors include weather data and/or environmental data.
A computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the following steps:

Extract symptom keywords from the patient's electronic medical record;

Judging whether the patient belongs to a case of one of the syndromes according to the matching situation of the symptom keywords and the keywords of several preset syndromes;

If it is determined that the patient belongs to a case of one of the syndromes, determine the syndrome as the target syndrome;

Obtain the number of cases of the target syndrome within a preset time unit;

Calculate several statistical indicators of the target syndrome according to the number of cases of the target syndrome within the current preset time unit;

Inputting several statistical indicators of the target syndrome into a risk prediction model to obtain the risk level of the target syndrome;

According to the risk level of the target syndrome, the early warning information of the target syndrome is released.
The computer-readable storage medium according to claim 15, wherein the processor determines whether the patient belongs to one of the syndromes according to the matching situation of the symptom keyword and the keywords of a plurality of preset syndromes. The case is used to achieve:

Compare the symptom keywords with each keyword of each syndrome, and if all the symptom keywords have corresponding keywords in one of the syndromes, then determine that the patient belongs to the case of the one of the syndromes .
The computer-readable storage medium according to claim 15, wherein the processor determines whether the patient belongs to one of the syndromes according to the matching situation of the symptom keyword and the keywords of a plurality of preset syndromes. The case is used to achieve:

According to the symptom keyword and the keyword of a certain syndrome, construct the multi-hot encoding of the symptom keyword and the multi-hot encoding of the keyword of the certain syndrome;

Input the multi-hot encoding of the symptom keyword into an autoencoder, and obtain the autoencoder vector of the symptom keyword output by the autoencoder;

According to the multi-hot encoding of the keyword of the certain syndrome, the autoencoder vector of the keyword of the certain syndrome is obtained by the autoencoder;

Calculate the similarity between the self-encoding vector of the symptom keyword and the self-encoding vector of the keyword of a certain syndrome;

According to the similarity, it is determined whether the patient belongs to a case of the certain syndrome.
The computer-readable storage medium of claim 17, wherein the processor is further configured to implement:

obtaining training data, the training data includes multi-hot encoding;

Obtain an autoencoder to be trained, the autoencoder includes a first network and a second network;

Inputting the multi-hot encoding in the training data into the first network to obtain an auto-encoding vector output by the first network, where the dimension of the self-encoding vector is lower than the dimension of the multi-hot encoding;

Inputting the self-encoding vector output by the first network into the second network, to obtain a vector output by the second network, the dimension of the vector output by the second network is equal to the dimension of the multi-hot encoding;

The first network and the second network are trained based on the vector output by the second network and the multi-hot encoding.
The computer-readable storage medium according to any one of claims 15-18, wherein the processor calculates several statistics of the target syndrome according to the number of cases of the target syndrome within the current preset time unit. When the indicator is used, it is used to achieve:

According to the number of cases of the target syndrome in the current preset time unit, and the number of historical cases of the target syndrome, calculate several statistical indicators of the target syndrome, and the statistical indicators include the chain growth rate, year-on-year growth rate, historical One or more of the percentiles.
The computer-readable storage medium according to any one of claims 15-18, wherein, when the processor implements the inputting the statistical index into a risk prediction model to obtain the risk level of the target syndrome, use To achieve:

The statistical indicators and risk factors are input into a risk prediction model to obtain the risk level of the target syndrome, and the risk factors include weather data and/or environmental data.