CN118098461A

CN118098461A - Clinical patient state monitoring method and system based on artificial intelligence

Info

Publication number: CN118098461A
Application number: CN202410180215.XA
Authority: CN
Inventors: 裴萌; 王茜; 翟建伟
Original assignee: Beijing Capton Pharmaceutical Technology Development Co ltd
Current assignee: Beijing Capton Pharmaceutical Technology Development Co ltd
Priority date: 2024-02-18
Filing date: 2024-02-18
Publication date: 2024-05-28

Abstract

The invention discloses a clinical patient state monitoring method and system based on artificial intelligence, comprising the following steps: firstly, acquiring real-time patient physiological state data, and then, using a self-attention model in an integrated model to decode the data in parallel to obtain a first clinical patient state evaluation result. And then, a second clinical patient state evaluation result is obtained from the privacy server, wherein the second clinical patient state evaluation result is obtained by calling a self-attention model corresponding to the privacy server to decode the comparison monitoring data set in parallel. And finally, carrying out integrated prediction based on the first and second clinical patient state evaluation results to obtain a target clinical patient state evaluation result of the current patient. By the design, the health condition of the patient can be estimated in real time, efficiently and accurately, the quality and efficiency of medical service are improved, and the privacy of the patient is protected.

Description

Clinical patient state monitoring method and system based on artificial intelligence

Technical Field

The invention relates to the technical field of artificial intelligence, in particular to a clinical patient state monitoring method and system based on artificial intelligence.

Background

Traditional clinical patient condition monitoring relies primarily on regular examination and observation of the patient by a doctor or medical personnel. However, this approach may be inefficient and may even be erroneous in cases where a large number of patients, complex diseases or long time monitoring is required. In recent years, with the development of artificial intelligence technology, the application of the artificial intelligence technology in the medical field is also becoming wider and wider. Particularly advanced machine learning techniques such as deep learning and self-attention models, can process a large amount of data and provide accurate prediction results, thereby significantly improving the efficiency and accuracy of clinical patient state monitoring. However, these methods typically require a large amount of computing resources and may raise privacy concerns because they typically require processing sensitive medical data on a central server. Therefore, how to effectively use artificial intelligence for clinical patient condition monitoring while preserving patient privacy is an important challenge.

Disclosure of Invention

The invention aims to provide an artificial intelligence-based clinical patient state monitoring method and system.

In a first aspect, an embodiment of the present invention provides an artificial intelligence-based clinical patient status monitoring method, including:

acquiring a current monitoring data set, wherein the current monitoring data set is composed of physiological state data generated by different monitoring nodes of a current patient in a real-time monitoring state;

invoking a self-attention model corresponding to the current patient in an integrated model to decode the current monitoring data set in parallel to obtain a first clinical patient state evaluation result of the current patient; the integrated model comprises a self-attention model corresponding to a privacy server of the current patient;

Obtaining a second clinical patient status assessment result for the current patient from the privacy server; the second clinical patient state evaluation result is a comparison monitoring data set acquired by the privacy server, and is obtained by parallel decoding of the comparison monitoring data set based on calling a self-attention model corresponding to the privacy server;

and carrying out integrated prediction based on the first clinical patient state evaluation result and the second clinical patient state evaluation result of the current patient to obtain a target clinical patient state evaluation result of the current patient.

In a second aspect, an embodiment of the present invention provides a server system, including a server, where the server is configured to perform the method described in the first aspect.

Compared with the prior art, the invention has the beneficial effects that: by adopting the clinical patient state monitoring method and system based on artificial intelligence, disclosed by the invention, the first clinical patient state evaluation result is obtained by acquiring real-time patient physiological state data and then using the self-attention model in the integrated model to decode the data in parallel. And then, a second clinical patient state evaluation result is obtained from the privacy server, wherein the second clinical patient state evaluation result is obtained by calling a self-attention model corresponding to the privacy server to decode the comparison monitoring data set in parallel. And finally, carrying out integrated prediction based on the first and second clinical patient state evaluation results to obtain a target clinical patient state evaluation result of the current patient. By the design, the health condition of the patient can be estimated in real time, efficiently and accurately, the quality and efficiency of medical service are improved, and the privacy of the patient is protected.

Drawings

In order to more clearly illustrate the technical solution of the embodiments of the present invention, the drawings that are required to be used in the embodiments will be briefly described. It is appreciated that the following drawings depict only certain embodiments of the invention and are therefore not to be considered limiting of its scope. Other relevant drawings may be made by those of ordinary skill in the art without undue burden from these drawings.

FIG. 1 is a schematic flow chart of steps of an artificial intelligence-based clinical patient state monitoring method according to an embodiment of the present invention;

fig. 2 is a schematic block diagram of a computer device according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. It will be apparent that the described embodiments are some, but not all, embodiments of the invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.

The following describes specific embodiments of the present invention in detail with reference to the drawings.

In order to solve the technical problems in the foregoing background, fig. 1 is a schematic flow chart of an artificial intelligence-based clinical patient status monitoring method according to an embodiment of the present disclosure, and the detailed description of the artificial intelligence-based clinical patient status monitoring method is provided below.

Step S201, a current monitoring data set is obtained, wherein the current monitoring data set is composed of physiological state data generated by different monitoring nodes of a current patient in a real-time monitoring state;

Step S202, calling a self-attention model corresponding to the current patient in an integrated model to decode the current monitoring data set in parallel to obtain a first clinical patient state evaluation result of the current patient; the integrated model comprises a self-attention model corresponding to a privacy server of the current patient;

step S203, a second clinical patient state evaluation result aiming at the current patient is obtained from the privacy server; the second clinical patient state evaluation result is a comparison monitoring data set acquired by the privacy server, and is obtained by parallel decoding of the comparison monitoring data set based on calling a self-attention model corresponding to the privacy server;

step S204, carrying out integrated prediction based on the first clinical patient state evaluation result and the second clinical patient state evaluation result of the current patient to obtain a target clinical patient state evaluation result of the current patient.

In an exemplary embodiment of the invention, an artificial intelligence based clinical patient condition monitoring system is provided for monitoring the physiological condition of a cardiac patient. The system collects physiological parameters of a patient in real time through biological sensors connected to the patient, such as electrocardiographs, blood pressure meters, oximeters, and the like. These parameters include heart rate, blood pressure, blood oxygen level, etc., and are stored in the form of a data set. In the cardiac patient monitoring system described above, the system will input current patient physiological state data into a trained self-attention model. The model uses neural network techniques to analyze physiological parameters of the patient and generate a first clinical patient state assessment. For example, for heart rate and blood pressure data, the model may detect whether abnormal or abnormal fluctuations exist and give corresponding assessment results. The system also uses a privacy server that stores large amounts of patient data and models. When a new patient is added, the system acquires historical data similar to the current patient from the privacy server and takes the historical data as a comparison monitoring data set. And then, calling the corresponding self-attention model on the privacy server to perform parallel decoding to obtain a second clinical patient state evaluation result. This result is derived based on patient data similar to the current patient. In the cardiac patient monitoring system, the target clinical patient state evaluation result of the current patient can be obtained by carrying out integrated prediction on the first clinical patient state evaluation result and the second clinical patient state evaluation result. For example, if a first clinical patient status assessment indicates that a patient may be at risk for a blood pressure increase, and a second clinical patient status assessment indicates that the patient has a very high rate of blood pressure increase in a patient similar to the past, then the integrated prediction may give a more accurate prediction of blood pressure increase. By the design, the method can provide accurate clinical patient state assessment results by acquiring the monitoring data set of the current patient, calling the self-attention model for decoding, acquiring the comparison monitoring data set from the privacy server and carrying out integrated prediction.

In another implementation of the present embodiment, for example, a diabetic patient is being monitored in real time. The patient is equipped with various physiological monitoring devices including blood glucose monitors, electrocardiographs, blood pressure monitors, and the like. These devices collect physiological state data of the patient, such as blood glucose level, heart rate, blood pressure, etc., at various time nodes (i.e., monitoring nodes). All of these collected data constitute the "current monitoring data set". This data set is processed and analyzed using artificial intelligence algorithms of the self-attention model. This model can fully take into account the correlation between the various data and decode them in parallel. After the decoding is completed, a first clinical patient state evaluation result, that is, the health condition of the patient obtained through AI analysis, is obtained. In addition, there is a privacy server that stores another type of data, referred to as a "contrast monitoring data set". This data set may be historical data or similar data for other patients for comparison with the current patient's data. The server also has a corresponding self-attention model that decodes the contrast data in parallel and generates a second clinical patient state assessment. Finally, the evaluation results obtained in the two steps are integrated, i.e. they are combined to obtain a more comprehensive and accurate prediction. The result of this integrated prediction is the final target clinical patient state assessment, which can help the physician to determine the patient's health and make corresponding treatment decisions.

In the embodiment of the present invention, the aforementioned step S201 may be implemented by performing the following manner.

(1) Acquiring a physiological monitoring data set of the current patient and a treatment record data set aiming at the current patient, and acquiring disease information of the current patient, wherein the disease information comprises monitoring starting time which is used for representing a time node of abnormal monitoring data of the current patient;

(2) Binding the disease information with the physiological monitoring data set and the treatment record data set respectively based on the disease information and monitoring starting time included in the disease information;

(3) And generating the current monitoring data set according to the physiological monitoring data set bound with the illness information and the treatment record data set bound with the illness information.

In an exemplary embodiment of the invention, a medical system is provided for monitoring the physiological status and treatment history of a cancer patient. The system collects physiological parameter data and treatment record data of the patient via sensors and electronic medical record systems connected to the patient. The physiological monitoring data set includes blood measurements, imaging results, and other physiological parameters (e.g., body temperature, heart rate, etc.). The treatment record data set includes chemotherapy regimen, surgical record, and drug use, among others. Furthermore, patient disease information, such as monitoring start time, is obtained for determining time nodes of abnormal monitoring data. In the above cancer patient monitoring system, it is bound to the physiological monitoring data set and the treatment record data set according to the disease information. For example, if the start time of monitoring in the patient information is 1 month and 1 day 2020, the system will bind this start time to the previously and later collected physiological monitoring data set and treatment record data set. In this way, each data point can be associated with a time node in the disease information and ensure the integrity and consistency of the data. The current monitoring data set may be generated by binding the patient information with the physiological monitoring data set and the treatment record data set. For example, the system may find physiological monitoring data and therapy record data before and after the point in time based on the monitoring start time in the disease information and combine them into one current monitoring data set. This data set includes all relevant data within a specific time frame that can be used for analysis and prediction in subsequent steps.

In another implementation of the embodiment of the present invention, illustratively, in the example of a diabetic patient described above, in addition to physiological monitoring data (e.g., blood glucose level, heart rate, blood pressure, etc.), treatment log data of the patient, such as medication, surgery, etc., that he received in the past, is collected. At the same time, it is also necessary to collect patient information such as when he is diagnosed with diabetes (i.e., monitoring the start time). These information are then correlated with physiological monitoring data and treatment record data based on patient disease information. For example, all physiological monitoring data after a patient begins to receive a particular treatment may be marked, or all treatment records after a patient is diagnosed with diabetes. And finally, integrating the two groups of associated data together to generate a brand new 'current monitoring data set'. This data set contains not only information about the physiological state, but also information about the course of treatment and the disease of the patient. In this way, the patient's health can be assessed from a more comprehensive perspective and more accurate treatment advice provided thereto.

In the embodiment of the present invention, the physiological monitoring data set is composed of physiological monitoring data in a monitoring period, the treatment record data set is composed of treatment record data in the monitoring period, and if the monitoring start time of the disease information is a target monitoring start time, the target monitoring start time is located in the monitoring period; the foregoing step of binding the disease information with the physiological monitoring data set and the treatment record data set, respectively, based on the disease information and a monitoring start time included in the disease information may be performed by the following example.

(1) Binding the disease information with a physiological monitoring data set in a monitoring period according to the target monitoring starting time of the disease information, using the disease information as the physiological monitoring data set after binding the disease information, binding the disease information with a treatment record data set in the monitoring period according to the target monitoring starting time of the disease information, and using the disease information as the treatment record data set after binding the disease information.

In an exemplary embodiment of the invention, an artificial intelligence based cardiac patient monitoring system is provided for collecting physiological monitoring data and treatment record data of a patient during a monitoring period. The physiological monitoring data set includes various physiological parameters, such as blood pressure, heart rate, blood oxygen level, etc., measured by the patient during the monitoring period. The treatment record data set includes different treatments accepted by the patient, such as drug use, surgical records, and rehabilitation programs. In the cardiac patient monitoring system described above, the patient information is bound to the physiological monitoring data set and the treatment record data set within the monitoring period according to the target monitoring start time in the patient information. For example, if the target monitoring start time in the disease information is 2022, 1, then the system would bind this target monitoring start time to the physiological monitoring data set and the treatment record data set during 2022, 1, to 2022, 1, 31. And binding the illness information with the physiological monitoring data set and the treatment record data set according to the target monitoring starting time to obtain the physiological monitoring data set and the treatment record data set after binding the illness information. These data sets include physiological parameter data and treatment record data associated with the patient information over a particular monitoring period. For example, if the target monitoring start time is 2022, 1 month, 1 day, then the bound data set will include relevant physiological monitoring data and treatment record data for that period of time.

In another implementation of the embodiment of the invention, a specific monitoring period is set, for example three months, in the case of diabetics. This means that both the collected physiological monitoring data (e.g. blood glucose level, heart rate, blood pressure, etc.) and the treatment record data (e.g. drug usage record, surgical record, etc.) are within these three months. During this time, assuming that the patient was diagnosed with diabetes two months ago, this time is the "target monitoring start time" because it is the point in time at which the patient's condition begins to be of close concern. This time point was also within the set three month monitoring period. The target monitoring start time and all the physiological monitoring data after the target monitoring start time are correlated to form a new physiological monitoring data set after binding the illness information. Similarly, the target monitoring start time and all the treatment record data after the target monitoring start time are correlated to form a new treatment record data set after binding the disease information. Thus, the patient can clearly see the change of the patient after the diagnosis of diabetes, both from the physiological data and the therapeutic data.

In the embodiment of the present invention, the step of generating the current monitoring data set of the current patient according to the physiological monitoring data set after binding the disease information and the treatment record data set after binding the disease information may be implemented by the following example.

(1) Performing data cutting operation on the physiological monitoring data set bound with the disease information and performing data cutting operation on the treatment record data set bound with the disease information respectively to obtain at least one physiological monitoring sample data set and at least one treatment record sample data set;

(2) Taking any physiological monitoring sample data in the at least one physiological monitoring sample data set and any treatment record sample data in the at least one treatment record sample data set as a current monitoring data set of the current patient.

In an exemplary embodiment of the present invention, a medical system is provided for monitoring blood glucose levels and treatment notes of diabetics. Based on the previously described method, a current monitoring data set of the current patient may be generated from the physiological monitoring data set and the treatment record data set after binding the patient information. This data set includes the cut physiological monitoring sample data set and the treatment record sample data set. In the above-described diabetes patient monitoring system, the data cutting operation is performed with respect to the physiological monitoring data set and the treatment record data set after binding the disease information. For example, the physiological monitoring data set within a continuous monitoring period is divided into a plurality of sample data sets, each sample data set representing physiological monitoring data over a specific period of time. Likewise, the treatment record data sets in the continuous monitoring period are also cut to obtain a plurality of treatment record sample data sets. In this way, at least one physiological monitoring sample data set and at least one treatment record sample data set can be obtained. In the above-described diabetes patient monitoring system, one physiological monitoring sample data is selected from the at least one physiological monitoring sample data set, while one treatment record sample data is selected from the at least one treatment record sample data set, which are combined into the current monitoring data set of the current patient. For example, blood glucose data during 2022, 1.1.month, 1.month, and 7.years, 2022, is selected from a certain physiological monitoring sample data set, medication data during 2022, 1.1.month, and 7.1.month, 2022, is selected from a certain treatment record sample data set, and these data are combined into the current monitoring data set of the current patient. By the design, the physiological monitoring data and the treatment record data in a specific time period can be extracted, and finer monitoring and analysis contents can be provided for medical professionals so as to better evaluate the current state of a patient and formulate a personalized treatment scheme.

In another implementation of the embodiment of the present invention, for example, in the case of a diabetic patient, data cutting is first performed on the "physiological monitoring data set after binding of disease information" and the "treatment record data set after binding of disease information". It may involve cutting a continuous stream of data into several time windows of data, or dividing the data into several subsets according to some rule. For example, daily physiological monitoring data may be considered as one sample, or a record of each treatment may be considered as one sample. By the above described cutting operation, a series of physiological monitoring sample data sets and treatment record sample data sets are obtained. For example, if daily data is considered as one sample, multiple physiological monitoring samples and treatment record samples representing different dates may be obtained. Finally, these physiological monitoring sample data and treatment record sample data are integrated together to form a "current monitoring data set" for the current patient. For example, a physiological monitoring sample for a day and a treatment record sample for that day may be combined as the "current monitoring data set" for that day. Such a data set can provide richer, more time-efficient information that helps to more accurately assess the health status of a patient.

In an embodiment of the present invention, the self-attention model corresponding to the current patient in the integrated model includes a physiological monitoring evaluation model corresponding to the current patient and a treatment prognosis evaluation model corresponding to the current patient, and the current monitoring data set includes a physiological monitoring sample data set and a treatment record sample data set; the aforementioned step S202 may be implemented by the following example execution.

(1) Invoking a physiological monitoring evaluation model corresponding to the current patient to decode the physiological monitoring sample data set in parallel to obtain a physiological monitoring state evaluation result of the current patient;

(2) Invoking a treatment prognosis evaluation model corresponding to the current patient to decode the treatment record sample data set in parallel to obtain a treatment record evaluation result of the current patient;

(3) And taking the physiological monitoring state evaluation result and the treatment record evaluation result as a first clinical patient state evaluation result of the current patient.

In an exemplary embodiment of the present invention, an artificial intelligence based lung cancer patient diagnostic system is provided. In this system, the integrated model is made up of a plurality of sub-models, including the current patient self-attention model. This self-attention model consists of two parts: physiological monitoring assessment model and treatment prognosis assessment model. The physiological monitoring evaluation model is used for analyzing the physiological monitoring data of the current patient, extracting the characteristic information related to the lung cancer and evaluating the physiological state of the patient. The treatment prognosis evaluation model is used for analyzing the treatment record data of the current patient and predicting the treatment effect and prognosis condition of the patient. In the lung cancer patient diagnosis system, the physiological monitoring sample data set and the treatment record sample data set included in the current monitoring data set are input into the current patient self-attention model for parallel decoding. Firstly, a physiological monitoring sample data set is decoded by calling a physiological monitoring evaluation model corresponding to the current patient, so that a physiological monitoring state evaluation result of the current patient is obtained. For example, the system may analyze blood biochemical indicators and lung function test data over a continuous week of the patient, and evaluate the physiological state of the patient based on such data. And then, the treatment record sample data set is decoded by calling a treatment prognosis evaluation model corresponding to the current patient, so that a treatment record evaluation result of the current patient is obtained. For example, the system can analyze the patient's various treatment regimens, drug use, and surgical records received over the past three months, and evaluate the patient's treatment efficacy and prognosis from these data. In the above lung cancer patient diagnosis system, the physiological monitoring state evaluation result and the treatment record evaluation result are combined as the first clinical patient state evaluation result of the current patient. For example, the system may represent the physiological monitoring state assessment as the patient's current pulmonary function state and blood biochemical indicator health, and the treatment record assessment as the patient's efficacy and prognosis after receiving treatment. By combining these two partial evaluation results, a first clinical patient state evaluation result for the current patient can be obtained, providing more comprehensive patient state information for medical professionals to aid diagnosis and formulate treatment regimens.

In another implementation of the present embodiment, by way of example, in the case of diabetics, the "self-attention model" is actually composed of two sub-models: one is a "physiological monitoring assessment model" for analyzing physiological monitoring sample data; the other is a "treatment prognosis evaluation model" for analyzing treatment record sample data. The physiological monitor assessment model is first invoked to process physiological monitor sample data. For example, the model can judge whether the diabetes control condition of the patient is good or not according to parameters such as the blood sugar level, the heart rate and the like in the last days, so as to obtain a physiological monitoring state evaluation result. Then, a "treatment prognosis evaluation model" is invoked to process the treatment record sample data. For example, the model may analyze the patient's past several weeks of drug use, surgical recovery, etc., to determine how effective the treatment is, and thereby obtain "treatment record assessment results". Finally, the two assessment results are combined together to form a first clinical patient state assessment result. The result contains information on both the physiological state and the therapeutic effect of the patient, and can more comprehensively reflect the health condition of the patient.

In an embodiment of the present invention, the privacy server of the current patient includes at least one of an environmental monitoring sensor and a drug tracking component, and the self-attention model corresponding to the privacy server includes at least one of a medical environmental assessment model and a drug condition assessment model; the contrast monitoring data set comprises at least one of a medical environment data set acquired by the environment monitoring sensor and a medication condition data set acquired by the medication tracking assembly; the aforementioned step S203 may be implemented by the following example execution.

(1) At least one of medical environment assessment results obtained by parallel decoding of the medical environment assessment model based on the medical environment data set and drug condition assessment results obtained by parallel decoding of the drug condition assessment model based on the drug condition data set.

In the present embodiment, an intelligent medical system is assumed for monitoring the health of the elderly, for example. In this system, the privacy server of the current patient is the device that collects data through the environmental monitoring sensors and/or the medication tracking component. The environment monitoring sensor can monitor environmental factors such as indoor temperature, humidity, illumination and the like, so that the medical environment of the patient is estimated. The medication tracking assembly may record medication of the patient, including medication name, dosage, time, etc. In the above intelligent medical system, the privacy server may include a medical environment assessment model and/or a drug condition assessment model according to the need. The medical environment assessment model is used for analyzing a medical environment data set, such as environment monitoring data of temperature, humidity and the like, so as to assess whether the medical environment of the current patient is suitable. The drug condition assessment model is used for analyzing the drug condition data set, such as the data of drug names, doses and the like, so as to assess whether the drug condition of the patient is reasonable. The contrast monitoring data set is made up of data collected by the environmental monitoring sensor and the drug tracking assembly. The medical environment data set acquired by the environment monitoring sensor can comprise information such as indoor temperature, humidity, illumination and the like. The medication intake data set acquired by the medication tracking component may include information on the patient's medication name, dosage, time of administration, etc. These data sets will be used for subsequent evaluation and analysis. A second clinical patient status assessment result for the current patient is obtained from the privacy server. These assessment results are obtained by parallel decoding of the medical environment assessment model and/or the drug condition assessment model. For example, by invoking a medical environment assessment model to decode based on the medical environment data set, a medical environment assessment result may be obtained for assessing whether the medical environment in which the current patient is located is appropriate. In addition, if a medication condition assessment model is present, decoding may also be based on the medication condition data set.

In another implementation of the embodiment of the invention, for example, in the case of a heart patient, a privacy server is installed at the patient's home. The server is provided with a plurality of environment monitoring sensors for monitoring indoor temperature, humidity and other environmental factors. Meanwhile, a medicine tracking component is arranged on the server and can record the medicine use condition of a patient. In addition, this server also loads two self-attention models: a medical environment assessment model and a drug condition assessment model. The privacy server will collect data from the environmental monitoring sensors and the medication tracking component to form a "contrast monitoring data set". This data set may include some information about the indoor environment (e.g., temperature, humidity, etc.), as well as a patient's medication usage record. The two self-attention models will then process their respective data separately. The medical environment assessment model may evaluate a "medical environment assessment result" based on the environmental data, for example, it may determine whether the room temperature and humidity are suitable for the patient to live. Also, the drug condition assessment model may assess a "drug condition assessment result" based on the drug usage record, for example, it may determine whether the patient is taking the drug on time, whether the drug dosage is appropriate, etc. These assessment results will ultimately be referred to as "second clinical patient status assessment results" to provide more comprehensive patient health information.

In the embodiment of the present invention, the foregoing manner of parallel decoding of sample data sets based on corresponding evaluation models may be implemented by the following example implementation.

(1) Performing initial value processing on the characteristic coefficients, connection weights and network parameters of the corresponding evaluation models;

(2) Based on the characteristic coefficient of the initial value and the connection weight of the initial value, performing weighted accumulation operation on the first sample instance of the sample data set and the network parameter of the initial value to obtain a first clinical patient state evaluation result;

(3) The first clinical patient state assessment is taken as a network parameter for a subsequent cycle and combined with a second sample instance of the sample dataset for parallel decoding.

In an exemplary embodiment of the invention, an artificial intelligence based cardiac diagnostic system is provided. In this system, the aforementioned method is used, and corresponding evaluation models, such as a physiological monitoring evaluation model, a medical environment evaluation model, and the like, are prepared. In order to perform parallel decoding, the feature coefficients, the connection weights and the network parameters of the evaluation models need to be initialized for subsequent calculation. In the above-described heart disease diagnosis system, it is necessary to perform initial value processing on the feature coefficients, the connection weights, and the network parameters of the physiological monitoring evaluation model and the medical environment evaluation model. This can be achieved by random initialization, initial values preset from domain knowledge, or parameters using a pre-trained model. And carrying out weighted accumulation operation on the first sample instance in the sample data set according to the characteristic coefficient of the initial value, the connection weight and the network parameter. For example, using a physiological monitoring assessment model, the physiological monitoring data of the first sample instance is calculated with the network parameters of the initialization values to obtain a first clinical patient state assessment result. This evaluation may be a prediction or classification of the current heart condition of the patient. The first clinical patient status assessment is taken as the network parameter for the subsequent cycle. The second sample instance is then combined with the updated network parameters and decoded by the corresponding evaluation model. For example, using a medical environment assessment model, medical environment data for the second sample instance is calculated with updated network parameters to obtain a second clinical patient state assessment result. This assessment may further refine the prediction or classification of the patient's heart condition. By continually cycling through the updating of network parameters and combining with different sample instances in the sample dataset, parallel decoding can be performed, resulting in a more accurate clinical patient state assessment result.

In another implementation of the embodiment of the present invention, for example, in cardiac patient cases, whether a physiological monitoring assessment model, a treatment prognosis assessment model, or a medical environment assessment model and a drug condition assessment model, an initialization operation is first required. This means that the characteristic coefficients, connection weights and network parameters of these models are given a set of initial values. These initial values are then used to process the first sample instance of the sample dataset. Specifically, the value of each feature may be multiplied by its corresponding feature coefficient and all the results added to form a weighted sum. This weighted sum is the "first clinical patient state assessment result". After the first clinical patient state assessment is obtained, it is used as a network parameter for processing the next sample instance of the sample dataset. This is a typical round robin operation that can efficiently process the entire sample data set.

In the embodiment of the present invention, the foregoing step S204 may be implemented by the following example execution.

(1) Acquiring a first model weight preset for the first clinical patient state evaluation result and a second model weight preset for the second clinical patient state evaluation result in the integrated model;

(2) And respectively carrying out weighted accumulation operation on the first clinical patient state evaluation result and the second clinical patient state evaluation result based on the first model weight and the second model weight to obtain a target clinical patient state evaluation result of the current patient, wherein the target clinical patient state evaluation result is used for representing the confidence level of the high risk event of the current patient.

In an exemplary embodiment of the present invention, an intelligent medical monitoring system is provided that includes a plurality of models for assessing the health of a patient. In the system, an integrated prediction is made for a first clinical patient state assessment result and a second clinical patient state assessment result for a current patient according to the method described previously. For integrated prediction, a first model weight preset for a first clinical patient state evaluation result and a second model weight preset for a second clinical patient state evaluation result need to be acquired from an integrated model. These weights may be set by training, tuning, or experience. And carrying out weighted accumulation operation on the first clinical patient state evaluation result and the second clinical patient state evaluation result by using a preset first model weight and a preset second model weight. For example, a first clinical patient state assessment result is multiplied by a first model weight, a second clinical patient state assessment result is multiplied by a second model weight, and then the two results are added to obtain a target clinical patient state assessment result for the current patient. This target assessment result may be a comprehensive prediction or classification result regarding the current patient's health status. The target clinical patient state assessment results obtained through integrated prediction can be used to represent the confidence that the current patient is at a high risk event. For example, if the target evaluation result is closer to a certain marker (e.g., 1), it indicates that the current patient is at higher risk. Such confidence may help a physician or system make more accurate decisions, such as taking precautions ahead of time, adjusting treatment regimens, etc. By integrating the evaluation results of the multiple models, the judgment and prediction capability of the current patient on the occurrence of the high-risk event can be improved, so that the accuracy of medical decision can be improved.

In another implementation of the embodiment of the invention, for example, in the case of heart disease patients, integrated predictions of results from the various assessment models are required. Before starting this operation, the weights of the models are first obtained. For example, it may be found that physiological monitoring data more accurately reflects the health condition of the patient, and thus a higher weight is given to the "first clinical patient status assessment result" (i.e., the assessment result based on physiological monitoring data). At the same time, a weight is set for the "second clinical patient status assessment result" (i.e., an assessment result based on treatment record data or privacy server data). Then, based on these weights, a weighted accumulation operation is performed on the "first clinical patient state evaluation result" and the "second clinical patient state evaluation result". Specifically, each evaluation result may be multiplied by its corresponding weight, and then all the results may be added to form a comprehensive evaluation result, i.e., a "target clinical patient state evaluation result". This "target clinical patient status assessment" may help to more accurately determine the health status of the patient. For example, if the value of this result is high, then the patient may be considered to have a high likelihood of a high risk event, such as a heart attack or the like.

In the embodiment of the present invention, the following implementation manner is also provided.

(1) Obtaining patient information bound to the current patient;

(2) Performing a comparison operation on the patient information bound to the current patient and a target clinical patient state evaluation result of the current patient;

(3) And adjusting the integrated model based on the comparison condition until the integrated model with the adjusted and optimized is obtained.

In an embodiment of the present invention, an electronic medical record system of a hospital is considered as an example. In this system, according to the foregoing method, for patient state assessment and integrated prediction, it is first necessary to acquire disease information bound to the current patient. Such information may include patient diagnostic results, medical history, symptom descriptions, laboratory test results, and the like. In the electronic medical record system of the hospital, the previously acquired illness information bound with the current patient is compared with the obtained target clinical patient state evaluation result of the current patient. For example, the actual diagnosis of a patient is compared to a predicted disease condition, or a symptom description of the patient is compared to a predicted clinical symptom. By means of comparison operation, accuracy of integrated prediction can be judged, and possible differences or errors can be identified. Based on the results of the comparison operation, the performance and accuracy of the integrated model can be assessed. If the integrated prediction is found to be biased or inaccurate, tuning is required to improve the performance of the model. For example, tuning may be performed by adjusting weights, increasing or decreasing the number of models, modifying feature selection algorithms, and the like. The process can be iterated for a plurality of times until an integrated model with optimized functions is obtained, namely, the current patient state can be predicted more accurately and is matched with the disease information bound with the current patient state. By continuously optimizing the integrated model, the accuracy and reliability of patient state evaluation can be improved, thereby providing more accurate and effective treatment advice and decision basis for doctors.

In another implementation of the embodiment of the invention, illustratively, in the case of a heart patient, patient information bound to the patient is first acquired. Such information may include patient history, family history, allergic conditions, and the like. These disease information are then compared to the "target clinical patient status assessment results". For example, if a patient has a family history of diabetes, but the results of the assessment show that his blood glucose control is good, then problems may exist. And finally, adjusting the optimized integrated model according to the comparison result. For example, if the model is found to be inaccurate in evaluating certain types of patients (e.g., patients with a family history of diabetes), the parameters of the model may be adjusted to provide more accurate results when processing data for such patients.

In the embodiment of the present invention, the step of tuning the integrated model based on the comparison case may be implemented by the following example implementation.

(1) If the comparison condition indicates that the disease information is different from the target clinical patient state evaluation result, calculating to obtain a model cost parameter of the current patient according to the current monitoring data set, and executing adjustment operation on a first model weight which is preset by the integrated model and aims at the first clinical patient state evaluation result based on an Adam algorithm; and;

(2) And sending the comparison situation to the privacy server so that the privacy server determines a corresponding model cost parameter according to the comparison situation and the comparison monitoring data set, and executing adjustment operation on a second model weight preset by the integrated model and aiming at the second clinical patient state evaluation result based on an Adam algorithm and the corresponding model cost parameter so as to adjust the integrated model.

In an embodiment of the present invention, an exemplary cardiac risk prediction system is considered. In this system, the comparison operation shows that there is a difference between the disease information of the current patient and the target clinical patient status evaluation result according to the aforementioned method. To tune the integrated model, the current patient model cost parameters first need to be calculated using the current monitoring data set (e.g., blood pressure, heart rate, blood test results, etc.). This model cost parameter can be measured by comparing the actual disease condition with the difference of the model predictions. Then, based on Adam's algorithm (a commonly used gradient descent optimization algorithm), an adjustment operation is performed on the preset first model weights for the first clinical patient state evaluation results. And (3) fine-tuning the model weight towards the direction of reducing the model cost parameter by a gradient descent algorithm. This process can be iterated through multiple iterations to gradually optimize the integrated model to better adapt to the current patient's condition and characteristics. In the heart disease risk prediction system, the comparison condition is sent to an ad hoc privacy server. The privacy server has powerful computing power and data protection mechanisms to ensure the security of patient privacy. And after receiving the comparison condition and the comparison monitoring data set, the privacy server utilizes the information to determine the corresponding model cost parameters of the current patient. In this example, the model cost parameter is assumed to indicate that the integrated model has some error in predicting the risk of heart disease in the patient. And executing adjustment operation on a preset second model weight aiming at the second clinical patient state evaluation result by using the Adam algorithm and a corresponding model cost parameter by the privacy server. Through this tuning operation, the integrated model can gradually optimize and improve the ability to accurately predict the risk of heart disease in the patient. The privacy server has the advantage that it can analyze the monitoring data against multiple medical institutions to obtain more comprehensive information to adjust the model. Such collaboration and analysis may help to further refine the integrated model to better fit and accuracy in different patient populations.

In another implementation of the present embodiment, it is assumed, by way of example, that a patient with a family history of diabetes is being treated. However, from the physiological monitoring assessment model (i.e. "first clinical patient status assessment") an assessment was obtained indicating that his glycemic control was good. This is clearly a problem, as a patient with a family history of diabetes generally increases his risk of developing disease. In this case, a "model cost parameter" is first calculated, which may include calculating the difference between the predicted blood glucose level and the actual blood glucose level of the model. The Adam algorithm will then be used to adjust the "first model weights". Adam's algorithm is an optimization algorithm that dynamically adjusts the learning rate of each parameter based on the first moment estimate and the second moment unbiased estimate of the gradient. This comparison is then sent to the privacy server. The privacy server is a server that specifically processes patient treatment record data and environmental data. Here, it can be found that although the patient has not taken hypoglycemic agents recently, his blood sugar remains within the normal range. This also indicates that there are some problems. Thus, the privacy server calculates another "model cost parameter" that may include calculating the difference between the model predicted blood glucose level and the actual blood glucose level. It would then also use Adam's algorithm to adjust the "second model weights". In this way, parameters of the model may be adjusted to more accurately predict the health of the patient.

In the embodiment of the invention, the model cost parameters of the current patient comprise cost parameters of the current patient corresponding to a physiological monitoring and evaluating model, cost parameters of the current patient corresponding to a treatment prognosis evaluating model and mutual influence cost parameters between the physiological monitoring and evaluating model and the treatment prognosis evaluating model; the corresponding model cost parameters comprise a cost parameter of a medical environment assessment model, a cost parameter of a drug condition assessment model, a mutual influence cost parameter between the medical environment assessment model and the drug condition assessment model, a mutual influence cost parameter between the physiological monitoring assessment model and the medical environment assessment model, a mutual influence cost parameter between the physiological monitoring assessment model and the drug condition assessment model, a mutual influence cost parameter between the treatment prognosis assessment model and the medical environment assessment model, and a mutual influence cost parameter between the treatment prognosis assessment model and the drug condition assessment model.

In an embodiment of the present invention, a heart disease management system is assumed by way of example. In the system, the model cost parameter of the current patient is calculated through the method. Firstly, the physiological monitoring data (such as electrocardiogram, blood pressure and the like) of the current patient are used for evaluation, and the cost parameter of a physiological monitoring evaluation model is obtained. Then, according to the treatment prognosis condition (such as postoperative recovery condition, drug treatment effect and the like) of the current patient, calculating the cost of the treatment prognosis evaluation model. In addition, the interplay between the physiological monitoring assessment model and the treatment prognosis assessment model needs to be considered. For example, physiological monitoring data may be found to have some effect on the accuracy of the treatment prognosis evaluation model, or vice versa. In order to more fully assess the patient's condition and prognosis, these interacting cost parameters need to be calculated. In the above-mentioned heart disease management system, the corresponding model cost parameter relates to a medical environment evaluation model and a drug condition evaluation model. The medical environment assessment model may take into account the medical environment in which the patient is located (e.g., hospital, home care, etc.), while the drug condition assessment model may analyze the patient's current drug usage. Cost parameters of these models need to be calculated to evaluate their contribution and accuracy in model integration. Furthermore, interactions between these models need to be considered.

In another implementation of the embodiment of the present invention, for example, if the physiological monitor assessment model predicts that the patient's blood pressure should be normal, but in fact the patient's blood pressure is too high, then the cost parameter of the physiological monitor assessment model will be increased. Similarly, if the treatment prognosis evaluation model predicts that the patient should have a decrease in blood pressure after taking the drug according to the treatment regimen, but in fact the blood pressure does not change significantly, then the cost parameter of the treatment prognosis evaluation model will also increase. Meanwhile, if there is a conflict between the predictions of the two models (e.g., one model predicts an increase in blood pressure and the other model predicts a decrease in blood pressure), the "mutual influence cost parameter" will also increase. In the privacy server, a series of model cost parameters are also calculated. These cost parameters reflect the difference between the predicted results of the various models and the actual situation, and whether the predicted results are consistent between the different models. For example, if the medical environment assessment model predicts that the living environment of the patient has no negative impact on his blood pressure, but the medication condition assessment model predicts that the patient has an elevated blood pressure due to forgetting to take the medication, then there is a cost of interaction parameter between the two models.

In the embodiment of the invention, the first model weight comprises a model weight of a physiological monitoring and evaluating model and a model weight corresponding to a treatment prognosis evaluating model; the second model weight comprises a model weight of a medical environment evaluation model and a model weight corresponding to a drug condition evaluation model; wherein, any model weight is used for executing weighted accumulation operation on the clinical patient state evaluation result of the corresponding evaluation model.

In an exemplary embodiment of the present invention, consider an intelligent health monitoring system that uses a physiological monitoring assessment model and a treatment prognosis assessment model to assess the health status of a patient. The model weights are assigned to the two models according to the method described above. For example, a physiological monitoring assessment model may be responsible for assessing the health status of a patient based on physiological indicators such as blood pressure, heart rate, blood glucose, etc., while a treatment prognosis assessment model may take into account factors such as surgical effects, drug treatment effects, etc. Corresponding model weights are calculated for the two models, respectively. These model weights will be used to perform a weighted accumulation operation on the physiological monitoring assessment results and the treatment prognosis assessment results, resulting in a more comprehensive patient state assessment result. The system uses a medical environment assessment model and a drug condition assessment model to assess the treatment environment and drug condition of a patient. According to the foregoing method, model weights are assigned to the two models. For example, the medical environment assessment model may consider the effect of the medical environment condition in which the patient is located on the treatment outcome, while the drug condition assessment model may analyze the patient's current drug use contribution to the treatment outcome. Corresponding model weights are calculated for the two models. These model weights will be used to perform a weighted summation operation on the medical environment assessment results and the drug condition assessment results to provide a more accurate patient state assessment result. Wherein, any model weight is used for executing weighted accumulation operation on the clinical patient state evaluation result of the corresponding evaluation model. In the intelligent health monitoring system or the medical aid decision making system, after the evaluation results of the physiological monitoring evaluation model, the treatment prognosis evaluation model, the medical environment evaluation model and the drug condition evaluation model are obtained, corresponding model weights are assigned to the evaluation models according to the foregoing methods. Then, for each result of the evaluation model, a weighted accumulation operation is performed using the corresponding model weight.

In another implementation of the present embodiment, the "first model weight" is for a physiological monitoring assessment model and a treatment prognosis assessment model, for example, in the case of a cardiac patient. For example, the physiological monitoring evaluation model can be evaluated based on physiological indexes such as blood pressure and heart rate of a patient, and the treatment prognosis evaluation model can be evaluated according to the treatment scheme and the actual administration condition of the patient. Each model has a corresponding weight that is used to determine the duty cycle of the model's assessment results in the overall assessment results. Likewise, the "second model weight" is for the medical environment assessment model and the drug condition assessment model. The medical environment assessment model may be based on factors such as the living environment of the patient, eating habits, etc., while the medication condition assessment model may be based on the medication records of the patient. The two models also each have a corresponding weight. Either the "first model weight" or the "second model weight" is used to perform a weighted accumulation operation on the corresponding evaluation result. Specifically, each evaluation result is multiplied by its corresponding weight, and then all the results are added to form a comprehensive evaluation result. This result may more fully reflect the health of the patient.

In the embodiment of the present invention, the foregoing step S201 may be implemented by the following example execution.

(1) Acquiring identity tags of similar patients of the current patient, wherein the similar patients of the current patient comprise patients which are positioned in the same medical area as the current patient or have the same symptoms as the current patient;

(2) Based on the identity tag, acquiring a physiological monitoring data set and a treatment record data set of the same kind of patients;

(3) And taking the physiological monitoring data set of the same kind of patients and the treatment record data set of the same kind of patients as the current monitoring data set.

In an exemplary embodiment of the present invention, a cross-regional health information system is considered for collecting and storing patient health data. In this system, different screening conditions may be used in order to obtain homogeneous patient identity tags for the current patient. For example, the same type of patient may be determined based on the medical area in which the patient is located (e.g., like a hospital, the same clinic, etc.), or based on the symptoms of the patient (e.g., fever, cough, etc.). Through the identity tags, related data of the same type of patient can be further acquired so as to be compared and referenced with the current patient, and therefore the state of the current patient can be better estimated and a treatment scheme can be formulated. In the above health information system, the identity tags of the same type of patient may be used to retrieve their physiological monitoring data and treatment record data according to the method described above. For example, physiological monitoring data (e.g., blood pressure, heart rate, body temperature, etc.) and treatment records (e.g., surgery, medication, etc.) of patients with the same medical condition are obtained from a database by matching the patients with the same medical area or with the same symptoms. This results in a physiological monitoring data set and a treatment record data set for the same type of patient, which will be part of the current monitoring data set for more comprehensive assessment of the current patient's status and personalized treatment plan formulation. And combining the obtained physiological monitoring data set and the treatment record data set of the similar patients with the physiological monitoring data and the treatment record data of the current patients to form a complete current monitoring data set. Thus, by taking the data of the same kind of patients as a reference, the system can better analyze and compare the data of the current patient, provide more accurate health assessment results, and formulate a personalized treatment scheme for the current patient based on the treatment record data of the same kind of patients. This comprehensive analysis of the patient data of the same type helps to improve the accuracy and effectiveness of patient health management.

In another implementation of the present embodiment, for example, in the case of a heart patient, it is first necessary to determine which patients are of the same type as the current patient. This may be based on whether they are in the same medical area, e.g. both in the cardiology department, or whether they have the same symptoms, e.g. both chest pain. Once these homogeneous patients are identified, their identity tags can be obtained. These identity tags can then be used to obtain physiological monitoring data and treatment record data for the same type of patient. The physiological monitoring data may include information about their blood pressure, heart rate, etc., while the treatment record data may include information about their medication records, surgical records, etc. Finally, the physiological monitoring data and the treatment record data of the same kind of patients are combined to form a current monitoring data set. This data set will be used to train and optimize the model of the pair.

In the embodiment of the invention, the target clinical patient state evaluation result is used for representing the target confidence that the current patient has a high risk event; the embodiment of the invention also provides the following implementation modes.

(1) Performing a comparison operation on the target confidence coefficient of the high risk event of the current patient indicated by the target clinical patient state evaluation result and a preset confidence coefficient threshold value;

(2) And when the target confidence exceeds the preset confidence threshold, initiating a focused attention warning for the current patient.

In an exemplary embodiment of the present invention, consider an intelligent health monitoring system that evaluates the health status of a current patient by analyzing his or her physiological monitoring data, medical history, diagnostic information, etc., and presents a target clinical patient status evaluation result. This result may be a number or an indicator that indicates the likelihood of the current patient experiencing a high risk event. In the intelligent health monitoring system, according to the method, the target confidence of the state evaluation result of the target clinical patient is compared with a preset confidence threshold value. The preset confidence threshold may be a threshold preset by the system, or may be a threshold set by a doctor or expert according to experience. For example, if the target confidence level exceeds a preset confidence threshold, i.e., the likelihood of the current patient having a high risk event is high, then the next processing operation will be triggered. When the target confidence exceeds a preset confidence threshold, the system will initiate a focused attention alert. This alert may be communicated to medical personnel, nurses or the patient himself/herself in the form of a system interface, mobile application, text message, etc. The purpose of this focused warning is to alert the relevant personnel to the current high risk status of the patient and take corresponding measures to avoid or mitigate the potential high risk event. For example, medical personnel may schedule more frequent monitoring, adjust treatment regimens, provide emergency assistance, and the like. By the method, according to the comparison operation of the target clinical patient state evaluation result and the confidence threshold, the intelligent health monitoring system can timely discover possible high-risk events of the current patient and initiate a focus attention warning so as to improve the safety and health management effect of the patient.

In another implementation of the present embodiment, for example, in the case of a cardiac patient, it is first necessary to compare the "target clinical patient state assessment" (i.e., the integrated model derived comprehensive assessment of patient health) to a preset confidence threshold. For example, it may be set that if the model predicts that the confidence level of a patient's heart attack exceeds 80%, then that patient is considered to be at high risk. Then, if the confidence level of the model predictions is found to exceed this threshold, a focused attention alert for the current patient is initiated. This alert may be sent to the relevant healthcare personnel in an email, cell phone push notification or other manner, giving them a high risk of knowing the health of the patient, requiring special attention.

In the embodiment of the present invention, after step S204, the embodiment of the present invention provides the following implementation manner.

(1) Receiving a medical staff coordination instruction sent by a scheduling server, wherein the medical staff coordination instruction is sent by the scheduling server in response to a medical staff coordination command triggered by the target clinical patient state evaluation result, and the target clinical patient state evaluation result comprises a plurality of clinical patient state evaluation sub-results and sub-result association relations corresponding to the plurality of clinical patient state evaluation sub-results;

(2) Based on the evaluation contents of the one-to-one matching of the plurality of clinical patient state evaluation sub-results, respectively determining the coordination results of medical staff of the one-to-one matching of the plurality of clinical patient state evaluation sub-results;

(3) According to the determined coordination results of the medical staff, coordinating the corresponding medical staff for the multiple clinical patient state evaluation sub-results;

(4) And respectively executing the plurality of clinical patient state evaluation sub-results based on the coordinated and determined medical staff and the sub-result association relationship to obtain an emergency task allocation result of the target clinical patient state evaluation result.

In the embodiment of the invention, an example is assumed that a hospital uses an artificial intelligence-based clinical patient state monitoring method, and in this scenario, a target clinical patient state evaluation result of a current patient is obtained. For example, based on the first clinical patient state evaluation result and the second clinical patient state evaluation result of the current patient, the results of the two evaluation methods are comprehensively considered, and the target clinical patient state evaluation result of the current patient is obtained as "needing close attention". In the above scenario, hospitals use dispatch servers to coordinate the work of healthcare workers. When the system judges that the target clinical patient state evaluation result of the current patient is that close attention is needed, the scheduling server can send a medical staff coordination instruction. For example, the dispatch server sends an instruction to the nurse requesting her dispatch of a nurse to the ward of the current patient. The target clinical patient state assessment results may comprise a plurality of clinical patient state assessment sub-results, each representing a different aspect of the assessment content. For example, one sub-result may be an abnormality in the heart rate of the patient and another sub-result may be an abnormality in the blood pressure of the patient. For each sub-result, the matching healthcare worker coordination result can be determined in advance according to the regulations of the hospital. For example, for sub-results of heart rate anomalies, the hospital prescribes that the nurse is responsible for monitoring and handling related tasks; while for sub-results of blood pressure abnormalities, hospitals prescribe that doctors be responsible for monitoring and handling related tasks. After determining the coordination results of the medical staff matched with each sub-result, the corresponding medical staff can be respectively assigned to each sub-result. For example, a heart rate abnormality sub-result is assigned to nurse a for monitoring and processing, and a blood pressure abnormality sub-result is assigned to doctor B for monitoring and processing. Based on the individual healthcare personnel and sub-outcome associations determined in the previous step, in this scenario, nurse a will monitor the patient for heart rate abnormalities and take appropriate action to handle, for example, administer medications or notify the doctor of further examination. Doctor B monitors the patient for abnormal blood pressure and decides whether or not to adjust the dosage of the drug or take other therapeutic measures. They execute corresponding tasks according to the association relation between the coordination result and the sub-result. Through the execution of the steps, the emergency task allocation result of the target clinical patient state evaluation result can be finally obtained. For example, nurse a and doctor B are assigned to coordinate and process, respectively, the assessment sub-results for heart rate abnormalities and blood pressure abnormalities of the current patient. They take appropriate action to deal with the abnormal situation of the patient according to the task allocation result, and ensure that the patient is timely nursed and treated.

In another implementation of the present embodiment, it is assumed, by way of example, that in the case of a heart patient, a target clinical patient condition assessment is obtained, which indicates that the patient may be at risk for a heart attack. This result may consist of multiple sub-results, e.g., physiological monitoring assessment sub-results (e.g., hypertension), treatment prognosis assessment sub-results (e.g., patients are not taking medications on time), etc. When the result is sent to the dispatch server, the server generates and sends a healthcare worker coordination indication. The corresponding healthcare worker reconciliation result then needs to be determined based on the specifics of each sub-result. For example, for physiological monitoring assessment sub-results (hypertension), it may be necessary to notify a doctor responsible for a heart patient; for prognosis of treatment, the sub-result (not taken on time) may need to be notified to the pharmacist or nurse. Then, the coordination of the health care worker can be started. Each healthcare worker will be told about the questions they need to pay attention to and how they should deal with. Finally, the association relation among all the sub-results is considered to determine the distribution result of the emergency task. For example, if it is known that the patient's blood pressure is too high, possibly because he is not taking a dose on time, the doctor and pharmacist responsible for the heart patient can be given the opportunity to deal with the problem.

In the embodiment of the present invention, the step of determining the healthcare worker coordination result with which the plurality of clinical patient state evaluation sub-results are matched one by one based on the evaluation contents with which the plurality of clinical patient state evaluation sub-results are matched one by one may be implemented by the following example.

For each of the plurality of clinical patient state assessment sub-results, performing the following steps, respectively:

(1) Determining a target healthcare professional class of a healthcare worker required for one clinical patient state assessment sub-result based on the assessment content of the one clinical patient state assessment sub-result for the one clinical patient state assessment sub-result;

(2) And determining a medical staff coordination result corresponding to the clinical patient state evaluation sub-result based on the current non-tired working staff number of the target medical professional class.

In the embodiment of the invention. Illustratively, assume a hospital uses an artificial intelligence based clinical patient condition monitoring method and obtains multiple clinical patient condition assessment sub-results according to previous procedures. For example, one of the sub-results may be an abnormality in the heart rate of the patient and the other sub-result may be an abnormality in the blood pressure of the patient. In the above scenario, taking heart rate abnormality as an example, according to the evaluation content of heart rate abnormality, it may be determined that the target healthcare professional class of the required healthcare staff is a nurse. This is because heart rate anomalies typically require nurses to have related knowledge and skills to monitor and handle. After determining that the target healthcare professional class is a nurse according to the regulations of the hospital and the current working condition, the number of nurses available in the current non-fatigue state needs to be considered. For example, if two nurses in a hospital are currently in a non-tired state, it may be determined that the medical staff coordination result corresponding to the clinical patient state assessment sub-result is to dispatch one nurse for the relevant task. Through the execution of the steps, the sub-results can be evaluated for each clinical patient state, the target healthcare professional class of the required healthcare personnel is determined based on the evaluation content, and the corresponding healthcare personnel coordination result is determined based on the current non-fatigue working population of the target healthcare professional class. This ensures that appropriate healthcare workers are assigned to handle different clinical patient condition assessment sub-results in a given situation, providing timely and effective care services.

In another implementation of the present embodiment, for example, in the case of heart disease patients, there is a clinical patient condition assessment sub-result that is hypertensive. From the evaluation of this sub-result, it may be determined that a cardiologist (target healthcare professional class) is needed to address this problem. Then, the number of cardiac doctors currently on duty and not in fatigue needs to be considered. If this number is sufficiently large, the task of dealing with the problem of hypertension may be assigned to one of the doctors. If this number is not sufficient, then the task allocation scheme may need to be reconsidered, or other viable solutions may be sought.

In an embodiment of the present invention, the step of determining the healthcare worker coordination result corresponding to the one clinical patient state evaluation sub-result based on the current non-tired worker population of the target healthcare professional class may be performed by the following example.

(1) If the number of the current non-tired operation people is greater than a non-tired operation people threshold, taking a first personnel coordination result as a medical personnel coordination result corresponding to the clinical patient state evaluation sub-result, wherein the first personnel coordination result comprises: when executing the one clinical patient state assessment sub-result, coordinating a desired target healthcare worker number for the one clinical patient state assessment sub-result;

(2) If the number of the current non-fatigue operation people is smaller than or equal to the threshold value of the number of the non-fatigue operation people, taking a second personnel coordination result as a medical personnel coordination result corresponding to the clinical patient state evaluation sub-result, wherein the second personnel coordination result comprises: a preparation time node prior to execution of the one clinical patient state assessment sub-result coordinates a desired target healthcare worker population for the one clinical patient state assessment sub-result.

In the present embodiment, illustratively, in the above scenario, the target healthcare professional category has been determined to be a nurse according to a previous procedure, and the number of nurses available in the current non-tired state is considered. Next, according to the threshold value of the number of non-tired workers, the following two cases are performed. Assume that the threshold number of non-tired workers set by the hospital is 3 nurses. If in the present case there are 4 nurses in a non-tired state, the first personnel coordination result will be selected as the healthcare personnel coordination result for the clinical patient state assessment sub-result. The first personnel coordination result comprises the target healthcare worker number required for coordination, namely, the task of assigning 3 nurses to perform the evaluation sub-result. In the present case, only 2 nurses are in a non-tired state, less than the non-tired worker population threshold. The second personnel coordination result is selected as a healthcare personnel coordination result of the clinical patient status assessment sub-result. The second person reconciles the results including the readiness time node prior to performing the assessment sub-result and the target healthcare worker population required for reconciliation. For example, 30 minutes before the evaluation sub-result is performed, an additional nurse needs to be mobilized to participate in the relevant task. According to the current non-tired working population of the target medical professional class, the medical personnel coordination result corresponding to the clinical patient state evaluation sub-result can be determined. According to different conditions, the first or second personnel are selected to coordinate the results, and the required target healthcare worker number or the preparation time node is designated accordingly. Therefore, medical staff can be reasonably distributed according to the resource condition and the workload of a hospital, and medical coordination work required by clinical patient state evaluation sub-results is ensured to be executed under different conditions.

In another implementation of the present embodiment, it is assumed, by way of example, that in the case of a heart patient, there are a sufficient number of non-tiring cardiologists (i.e., a number greater than a set threshold). In this case, a part of the doctors can be directly allocated to deal with the problem of the hypertension. This is the so-called "first person coordination result". However, if there are not enough non-tiring cardiologists, then a different strategy needs to be adopted. This may mean that a period of time (i.e., a "readiness time node") is required before the problem is handled to begin coordination of the healthcare worker's work to ensure adequate healthcare worker participation in handling the problem. This is the so-called "second person reconciliation result".

(1) Determining a predicted non-tired job number for the target healthcare specialty category within an execution period of the one clinical patient status assessment sub-result based on the current non-tired job number for the target healthcare specialty category, historical resource usage data for the target healthcare specialty category, and newly added resource demand data for the target healthcare specialty category;

(2) And determining a medical staff coordination result corresponding to the one clinical patient state evaluation sub-result based on the predicted non-fatigue operation population.

In an embodiment of the present invention, it is assumed, by way of example, that a hospital uses an intelligent scheduling system to manage healthcare worker resources and has determined a target healthcare professional category as a nurse based on previous procedures. In this step, the system considers the current non-tiring job population of the target healthcare professional category, for example, 3 nurses in a non-tired state. At the same time, the system will analyze historical resource usage data, such as the work and needs of nurses of the specialty category over the past few days. In addition, the system may obtain additional resource demand data for the target healthcare professional category, such as where new patients are assigned to the professional category and corresponding care services are required. Taking these factors into account, the system can predict how many non-tired nurses will be expected to be available during the execution of the clinical patient status assessment sub-outcome period. In the event that 4 non-tired nurses are predicted to be available, the system will use this information as a basis to determine the healthcare worker coordination outcome of a clinical patient status assessment sub-outcome. For example, if multiple clinical patient status assessment sub-results need to be processed during the execution period and 3 nurses are required to coordinate work with the current sub-result, the system may assign 3 non-tired nurses to the sub-result based on the predicted number of non-tired nurses to coordinate work. By executing the steps, the predicted non-tired job number of one clinical patient state evaluation sub-result in the execution period can be determined based on the current non-tired job number of the target healthcare professional class, the historical resource usage data and the newly added resource demand data. And then, determining a medical staff coordination result corresponding to the sub-result according to the predicted non-fatigue working number, so as to ensure that proper quantity of medical staff resources can be reasonably distributed when the clinical patient state evaluation sub-result is executed. Thus, medical staff can be predicted and scheduled in advance to meet the requirements of different clinical patient state evaluation sub-results and optimize the utilization efficiency of medical resources.

In another implementation of the embodiment of the invention, for example, in the case of heart disease patients, first of all, the number of currently non-tiring cardiologists needs to be considered. Then, it is also necessary to view the workload of the cardiologist in the past (i.e. "historical resource usage data") and the workload that may be increased in the future (i.e. "newly added resource demand data"). Such information may help predict how many cardiologists are in a non-tired state and operational during the execution period that addresses the problem of hypertension. Then, it is determined how to distribute the work of the healthcare worker based on the predicted number of non-tired workers. For example, if the number of predicted non-tired tasks is sufficiently large, then the task of treating the hypertension problem may be assigned to some of the doctors. If this number is not sufficient, then the task allocation scheme may need to be reconsidered, or other viable solutions may be sought.

In the embodiment of the present invention, the step of determining the coordination result of the medical staff corresponding to the clinical patient status evaluation sub-result based on the predicted non-fatigue job population may be implemented by the following example.

(1) If the predicted non-tired job population is greater than a non-tired job population threshold, taking a first personnel coordination result as a medical personnel coordination result corresponding to the one clinical patient state evaluation sub-result, the first personnel coordination result comprising: when executing the one clinical patient state assessment sub-result, coordinating a desired target healthcare worker number for the one clinical patient state assessment sub-result;

(2) If the number of the predicted non-fatigue operation people is smaller than or equal to the threshold value of the number of the non-fatigue operation people, taking a second personnel coordination result as a medical personnel coordination result corresponding to the one clinical patient state evaluation sub-result, wherein the second personnel coordination result comprises: a preparation time node prior to execution of the one clinical patient state assessment sub-result coordinates a desired target healthcare worker population for the one clinical patient state assessment sub-result.

In an embodiment of the present invention, it is assumed, by way of example, that a hospital uses an intelligent scheduling system to predict the number of non-tired workers and that 5 non-tired nurses are available during the execution period, according to previous procedures. In this step, the system compares the magnitude of the threshold for predicting the number of non-tired workers and the threshold for the number of non-tired workers, and selects corresponding personnel coordination results according to different conditions. If 6 non-tired nurses were found to be available in the forecast, the non-tired worker population threshold was set to 5 nurses. At this point, the first personnel coordination result will be a healthcare personnel coordination result of the clinical patient state assessment sub-result. The first person reconciliation result includes the target healthcare worker number required for the evaluation sub-result to reconcile when the evaluation sub-result is performed. In this example, the system would assign 5 non-tired nurses to perform the assessment sub-result. Only 4 non-tired nurses were found to be available in the forecast, less than the non-tired worker population threshold. At this point, the second personnel coordination result will be a healthcare personnel coordination result of the clinical patient status assessment sub-result. The second person reconciles the results including the readiness time node prior to performing the assessment sub-result and the target healthcare worker population required for reconciliation. For example, 30 minutes before the evaluation sub-result is performed, an additional nurse needs to be mobilized to participate in the relevant task. By executing the steps, the coordination result of the medical staff corresponding to the clinical patient state evaluation sub-result can be determined according to the comparison based on the number of the predicted non-fatigue workers. And selecting a first or second personnel coordination result according to the predicted relationship between the number of non-tired operators and the threshold value of the number of non-tired operators, and correspondingly designating the required target medical care operator number or the preparation time node. Therefore, medical staff can be reasonably distributed under different conditions according to the resource conditions and the workload of a hospital, and enough non-tired medical staff can participate in the work when the clinical patient state evaluation sub-result is executed.

In another implementation of the present embodiment, it is assumed, by way of example, that in the case of a heart patient, the number of predicted non-tiring cardiologists exceeds a set threshold. In this case, a part of the doctors can be directly allocated to deal with the problem of the hypertension. This is the so-called "first person coordination result". However, if the number of predicted non-tiring cardiologists is insufficient, then a different strategy needs to be adopted. This may mean that a period of time (i.e., a "readiness time node") is required before the problem is handled to begin coordination of the healthcare worker's work to ensure adequate healthcare worker participation in handling the problem. This is the so-called "second person reconciliation result".

(1) Determining a target healthcare professional class and a target healthcare worker number of medical workers required by one clinical patient state evaluation sub-result based on the evaluation content of the one clinical patient state evaluation sub-result aiming at the one clinical patient state evaluation sub-result;

(2) If the number of the target medical care task persons is larger than the threshold value of the medical care task persons, determining a medical care person coordination result corresponding to the clinical patient state evaluation sub-result based on the person number floating interval corresponding to the target medical care professional class.

In an embodiment of the present invention, it is assumed, by way of example, that a hospital uses an intelligent scheduling system to manage clinical patient status assessment sub-results and that there are multiple sub-results to be processed simultaneously. In this step, the system will perform the following steps for each clinical patient status assessment sub-result to determine the corresponding healthcare worker coordination result. Consider a specific clinical patient state assessment sub-outcome, which may be a patient-specific condition assessment and treatment plan. In this step, the system will analyze the healthcare worker resources required for the sub-result based on the evaluation content. For example, if the evaluation indicates that the patient requires a cardiologist and two nurses to evaluate and treat, the system will determine the target healthcare professional category as a cardiologist and set the target healthcare worker number to 2. After analysis, the target healthcare worker count was found to be 3, and the healthcare worker count threshold was set to be 2. At this time, the system may determine a healthcare worker coordination outcome for the clinical patient status assessment sub-outcome based on the floating intervals of people corresponding to the target healthcare specialty category. For example, the number of cardiologists may float between ±1, i.e. 2 or 4 cardiologists may be scheduled. In this case, the system would choose to schedule 3 cardiologists to perform the assessment sub-results. By executing the steps, for each clinical patient state evaluation sub-result, according to the evaluation content, the target medical professional class of the required medical staff and the target medical staff number, the corresponding medical staff coordination result can be determined. If the number of the target healthcare workers is greater than the threshold value of the number of the healthcare workers and the floating interval of the number corresponding to the target healthcare professional class allows, the system can schedule the healthcare workers according to the reasonable number in the range of the floating interval so as to meet the requirement of the evaluation sub-result. This ensures that healthcare worker resources are reasonably allocated according to the specific needs of each sub-outcome when processing multiple clinical patient status assessment sub-outcomes.

In another implementation of the present embodiment, illustratively, first, the evaluation content of each clinical patient state evaluation sub-result needs to be reviewed. Taking the example of hypertension, it may be decided that the problem needs to be addressed by a cardiologist and how many doctors are needed to address the problem. If the number of doctors needed to address this problem of hypertension exceeds a set threshold, then consideration needs to be given to how to coordinate the healthcare staff. This may involve looking at the floating intervals of the number of cardiologists (e.g., how many cardiologists at least and at most may be working) and deciding how to assign tasks based on this information.

In the embodiment of the present invention, the step of determining the coordination result of the medical staff corresponding to the one clinical patient state evaluation sub-result based on the floating interval of the number of people corresponding to the target medical care professional category may be implemented by the following example.

(1) If the number of people floating interval of the target healthcare professional class in the execution period of the one clinical patient state evaluation sub-result is smaller than or equal to a number of people floating threshold, taking a first personnel coordination result as a medical personnel coordination result corresponding to the one clinical patient state evaluation sub-result, wherein the first personnel coordination result comprises: when executing the one clinical patient state assessment sub-result, coordinating a desired target healthcare worker number for the one clinical patient state assessment sub-result;

(2) If the number of people floating interval of the target healthcare professional class in the execution period of the one clinical patient state evaluation sub-result is greater than the number of people floating threshold, taking a second personnel coordination result as a medical personnel coordination result corresponding to the one clinical patient state evaluation sub-result, wherein the second personnel coordination result comprises: a preparation time node prior to execution of the one clinical patient state assessment sub-result coordinates a desired target healthcare worker population for the one clinical patient state assessment sub-result.

In the embodiment of the invention, for example, in the foregoing intelligent scheduling system, according to the number of people floating interval corresponding to the target healthcare professional category, the system determines the coordination result of the healthcare personnel for each clinical patient state evaluation sub-result. For example, during the execution cycle of a particular clinical patient state assessment sub-result, the target healthcare professional category is cardiologists, and the number of cardiologists ranges from ±1. If the population float threshold is set to 2, the maximum number of floats allowed when this sub-result is performed is 2 cardiologists. It is assumed that at this point the system finds from the analysis of the number of target healthcare workers that only 1 cardiologist needs to be scheduled to perform this sub-result. Because the population float interval is less than or equal to the population float threshold, the system will select the first personnel coordination result, i.e., assign 1 cardiologist to participate in the execution of this sub-result. During the execution cycle of a clinical patient state assessment sub-result, the target healthcare professional category is cardiologists, and the number of cardiologists ranges from + -2. If the population float threshold is set to 1, the maximum number of floats allowed when this sub-result is performed is 1 cardiologist. Suppose at this point the system finds, from an analysis of the number of target healthcare workers, that 3 cardiologists need to be scheduled to perform the sub-results. Since the population float interval is greater than the population float threshold, the system will select a second personnel reconcile the results, i.e., a preparation time node prior to execution of the sub-result, and the system will designate that 3 cardiologists be scheduled to participate in the execution of the sub-result during this time node. Through the execution of the steps, the coordination result of the medical staff corresponding to the clinical patient state evaluation sub-result is determined according to the comparison of the head number floating interval corresponding to the target medical care professional class and the head number floating threshold value. If the number floating interval of the target medical care professional class is smaller than or equal to the number floating threshold, selecting a first personnel coordination result, namely setting the required number of target medical care tasks; if the head count floating interval is greater than the head count floating threshold, a second head count coordination result, i.e., a set preparation time node prior to execution of the assessment sub-result, is selected and an appropriate number of healthcare workers are scheduled to meet the execution needs. Therefore, the coordination strategy can be flexibly adjusted according to the resource change conditions of medical staff under different conditions, and the successful execution of the clinical patient state evaluation sub-result is ensured.

In another implementation of the present embodiment, by way of example, first, it is necessary to look at the floating intervals of the number of heart surgeons in the execution cycle that deal with the problem of hypertension (e.g., how many heart surgeons may be working at least and at most). If this floating interval is less than or equal to the set threshold, then a portion of the physician may be directly assigned to deal with this problem. This is the so-called "first person coordination result". However, if the number of people floating in the execution period for the cardiologist to deal with the problem of hypertension exceeds a set threshold, a different strategy needs to be adopted. This may mean that a period of time (i.e., a "readiness time node") is required before the problem is handled to begin coordination of the healthcare worker's work to ensure adequate healthcare worker participation in handling the problem. This is the so-called "second person reconciliation result".

In the embodiment of the present invention, the step of respectively executing the plurality of clinical patient state evaluation sub-results based on the respective medical staff and the sub-result association relationship determined by coordination to obtain the emergency task allocation result of the target clinical patient state evaluation result may be implemented by the following example execution.

(1) Constructing a clinical patient state evaluation graph structure of the target clinical patient state evaluation result based on the sub-result association relationship, wherein each evaluation entity in the clinical patient state evaluation graph structure represents one clinical patient state evaluation sub-result;

(2) Polling each evaluation entity in the clinical patient state evaluation graph structure, wherein each polling one evaluation entity, executing the clinical patient state evaluation sub-result corresponding to the one evaluation entity based on medical staff coordinated with the clinical patient state evaluation sub-result corresponding to the one evaluation entity, and obtaining a pending emergency task allocation result corresponding to the one evaluation entity;

(3) And determining the emergency task allocation result of the target clinical patient state evaluation result based on the determined various pending emergency task allocation results.

In an exemplary embodiment of the present invention, consider the scenario of an emergency center, where a team of multiple doctors and nurses are responsible for treating patients coming from the hospital. Each healthcare worker is assigned a specific clinical patient status assessment sub-outcome according to the coordination strategy and sub-outcome association in the system. For example, doctor a is responsible for electrocardiographic assessment, doctor B is responsible for blood testing, and nurse C is responsible for body temperature measurement, etc. They each perform a corresponding evaluation operation according to the coordination strategy and the assigned sub-result, and generate a pending emergency task assignment result. In the scenario of the emergency center, the system constructs a clinical patient state assessment graph structure according to the sub-result association. Each assessment entity in the figure represents a clinical patient state assessment sub-result, such as electrocardiogram assessment, blood detection, and body temperature measurement. Such a structure can conveniently manage and track the execution of each assessment entity. Based on the clinical patient status assessment graph structure, the system begins to poll each assessment entity. First, the system will examine the electrocardiogram to evaluate this evaluation entity and view the clinical patient status evaluation sub-results associated therewith. The system then determines the physician responsible for the electrocardiogram evaluation according to the coordination strategy and causes the physician to perform the electrocardiogram evaluation operation. Similarly, the system can poll other assessment entities in turn, and assign corresponding medical staff to execute corresponding assessment operations, so as to obtain the distribution result of the pending emergency task corresponding to each assessment entity. After polling of all assessment entities is completed, the system determines emergency task allocation results for the target clinical patient status assessment results based on the various pending emergency task allocation results. The system comprehensively considers the importance and the emergency degree of the allocation result of each undetermined task, and makes a decision according to the preset priority, the resource availability and other factors. Finally, the system generates the most appropriate emergency task assignment results, ensuring that each clinical patient status assessment sub-result is properly processed and executed. Through the execution of the steps, according to the medical staff and the sub-result association relation determined in a coordinated manner, the system can respectively execute a plurality of clinical patient state evaluation sub-results and obtain the emergency task allocation result of the target clinical patient state evaluation result. First, a clinical patient state evaluation graph structure is constructed according to the sub-result association relationship, wherein each evaluation entity represents one clinical patient state evaluation sub-result. And then, the system performs execution by polling the evaluation entities and the corresponding medical staff to obtain the distribution result of the undetermined first-aid task corresponding to each evaluation entity. Finally, based on the importance and urgency of the pending task assignment results, the system determines an emergency task assignment result for the target clinical patient status assessment result. Therefore, the processing and execution of the multiple clinical patient state evaluation sub-results can be efficiently completed according to factors such as coordination and priority, and successful distribution of emergency tasks is ensured.

In another implementation of the embodiment of the present invention, first, it is necessary to construct a clinical patient status evaluation graph structure that includes all the clinical patient status evaluation sub-results (i.e., hypertension and tachycardia) and reflects the association between these sub-results. Each assessment entity, i.e. each clinical patient status assessment sub-result, is then polled. Taking the example of hypertension, this sub-result may be performed by already coordinated medical personnel, for example, assigned to certain cardiologists to deal with the problem. After execution, a pending emergency task assignment result is obtained. And finally, integrating all the pending emergency task allocation results so as to obtain the emergency task allocation result of the target clinical patient state evaluation result.

In the embodiment of the present invention, the step of obtaining the pending emergency task allocation result corresponding to the one assessment entity may be implemented by the following example execution, based on the medical staff coordinating the clinical patient status assessment sub-result corresponding to the one assessment entity.

(1) And if the one assessment entity obtains the pending emergency task allocation result corresponding to the pre-entity, executing the pending emergency task allocation result corresponding to the one assessment entity based on the medical staff coordinated with the pending emergency task allocation result corresponding to the pre-entity and the pending emergency task allocation result corresponding to the pre-entity to obtain the pending emergency task allocation result corresponding to the one assessment entity.

In the present embodiment, consider, by way of example, an emergency department scenario in which a nurse is responsible for pain assessment tasks and needs to coordinate with a physician. The nurse and doctor are assigned to the healthcare worker of the assessment entity according to the coordination strategy in the system. Together they perform pain assessment operations and then generate pending emergency task assignment results corresponding to the assessment entity. In the emergency department setting, doctor a is responsible for performing the primary diagnostic evaluation task and nurse B is responsible for performing the medical history. Doctor a and nurse B are assigned to the healthcare worker corresponding to the lead entity according to the coordination policy in the system. Doctor a first performs a preliminary diagnostic evaluation operation and generates a pending emergency task assignment result. And then, the nurse B coordinates and executes the medical history recording operation based on the undetermined task allocation result of the doctor A, and generates undetermined emergency task allocation result corresponding to the evaluation entity. By executing the steps, the system can ensure that each assessment entity is properly processed and executed according to the coordination of medical staff and the undetermined task allocation result of the prepositioned entity. If one assessment entity does not have a pre-entity or the pre-entity does not have a pending task allocation result, the system directly executes the clinical patient status assessment sub-result corresponding to the assessment entity. If one assessment entity has a pre-entity and a corresponding pending task allocation result, the system executes a clinical patient state assessment sub-result of the assessment entity based on coordination of medical staff and the pending task allocation result of the pre-entity so as to obtain a pending emergency task allocation result corresponding to the assessment entity. Therefore, the execution of the evaluation entity can be flexibly adjusted according to the state of the preposed entity and the task allocation situation, and the smooth allocation of the emergency task is ensured.

In another implementation of the present embodiment, in the case of an exemplary cardiac patient, two clinical patient state assessment sub-outcomes, hypertension (pre-entity) and tachycardia (current assessment entity) are assumed. The following are the steps of processing these results: first, the allocation result of the pending emergency task corresponding to the pre-entity (hypertension) is obtained. The current assessment entity (e.g., heart beat too fast) may then be performed based on the already coordinated assignment of pending emergency tasks to the healthcare personnel (e.g., certain cardiologists) and pre-entities. For example, depending on the result of the treatment of hypertension, it may be necessary to adjust the method of treating the tachycardia problem or the number of medical personnel required. After the execution, the allocation result of the undetermined emergency task corresponding to the evaluation entity with the too fast heartbeat is obtained.

In an embodiment of the present invention, the foregoing step of determining the emergency task allocation result of the target clinical patient status assessment result based on the determined respective pending emergency task allocation results may be implemented by the following example execution.

(1) And if the one evaluation entity does not have a post-entity, taking the pending emergency task allocation result corresponding to the one evaluation entity as an emergency task allocation result of the target clinical patient state evaluation result.

In an exemplary embodiment of the present invention, consider the scenario of an emergency department, where a team of multiple doctors and nurses are responsible for treating patients coming from the hospital. According to the aforementioned method, each assessment entity generates a corresponding pending emergency task assignment result. In this step, the system comprehensively considers the importance and urgency of all the pending task allocation results, and determines the final emergency task allocation result based on the preset priority, resource availability, and other factors. This ensures that the outcome of the target clinical patient condition assessment is properly assigned emergency tasks.

In the emergency department scenario, consider a pain assessment task that has no post-entity. When the previous steps are executed, the system executes the clinical patient state evaluation sub-result corresponding to the evaluation entity according to the pending task allocation result of the coordination and prepositive entity, and generates a pending emergency task allocation result. Because there is no post-entity, the system will take the pending emergency task allocation result corresponding to the assessment entity as an emergency task allocation result for the target clinical patient status assessment result. This ensures that the pain assessment task is properly handled and performed and is assigned as a final emergency task. By performing the above steps, the system may determine an emergency task allocation result for the target clinical patient status assessment result based on the determined relationship of the pending emergency task allocation result and the assessment entity. If one assessment entity does not have a post-entity, the pending emergency task allocation result corresponding to the assessment entity can directly become an emergency task allocation result of the target clinical patient state assessment result. This ensures that the results of each assessment entity are correctly assigned to the process and used and recorded as the final result.

In another implementation of the present embodiment, for example, in the case of a heart patient, two clinical patient state assessment sub-results, hypertension (pre-entity) and tachycardia (current assessment entity) are assumed. In this case, it is assumed that the assessment entity is "fast heart beat" without a post-entity. That is, it is the last node in the clinical patient state assessment graph structure. Then the pending emergency task assignment result corresponding to the "too fast heart beat" assessment entity (e.g., how many cardiologists are needed to deal with the problem) is treated as an emergency task assignment result for the targeted clinical patient status assessment result.

An embodiment of the present invention provides a computer device 100, where the computer device 100 includes a processor and a nonvolatile memory storing computer instructions, and when the computer instructions are executed by the processor, the computer device 100 performs the aforementioned artificial intelligence-based clinical patient condition monitoring method. As shown in fig. 2, fig. 2 is a block diagram of a computer device 100 according to an embodiment of the present invention. The computer device 100 comprises a memory 111, a processor 112 and a communication unit 113. For data transmission or interaction, the memory 111, the processor 112 and the communication unit 113 are electrically connected to each other directly or indirectly. For example, the elements may be electrically connected to each other via one or more communication buses or signal lines.

The foregoing description, for purpose of explanation, has been presented with reference to particular embodiments. The illustrative discussions above are not intended to be exhaustive or to limit the disclosure to the precise forms disclosed. Many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the disclosure and its practical application, to thereby enable others skilled in the art to best utilize the disclosure and various embodiments with various modifications as are suited to the particular use contemplated.

Claims

1. An artificial intelligence based method for monitoring the status of a clinical patient, comprising:

2. The method of claim 1, wherein the obtaining the current monitoring data set comprises:

Acquiring a physiological monitoring data set of the current patient and a treatment record data set aiming at the current patient, and acquiring disease information of the current patient, wherein the disease information comprises monitoring starting time which is used for representing a time node of abnormal monitoring data of the current patient; the physiological monitoring data set is composed of physiological monitoring data in a monitoring period, the treatment record data set is composed of treatment record data in the monitoring period, and if the monitoring starting time of the illness information is a target monitoring starting time, the target monitoring starting time is located in the monitoring period;

Binding the disease information with a physiological monitoring data set in a monitoring period according to the target monitoring starting time of the disease information, and using the disease information as the physiological monitoring data set after binding the disease information;

performing data cutting operation on the physiological monitoring data set bound with the disease information and performing data cutting operation on the treatment record data set bound with the disease information respectively to obtain at least one physiological monitoring sample data set and at least one treatment record sample data set;

Taking any physiological monitoring sample data in the at least one physiological monitoring sample data set and any treatment record sample data in the at least one treatment record sample data set as a current monitoring data set of the current patient.

3. The method of claim 1, wherein the self-attention model corresponding to the current patient in the integrated model comprises a physiological monitoring assessment model corresponding to the current patient and a treatment prognosis assessment model corresponding to the current patient, the current monitoring data set comprising a physiological monitoring sample data set and a treatment record sample data set; the calling the self-attention model corresponding to the current patient in the integrated model to decode the current monitoring data set in parallel to obtain a first clinical patient state evaluation result of the current patient comprises the following steps:

Invoking a physiological monitoring evaluation model corresponding to the current patient to decode the physiological monitoring sample data set in parallel to obtain a physiological monitoring state evaluation result of the current patient;

invoking a treatment prognosis evaluation model corresponding to the current patient to decode the treatment record sample data set in parallel to obtain a treatment record evaluation result of the current patient;

and taking the physiological monitoring state evaluation result and the treatment record evaluation result as a first clinical patient state evaluation result of the current patient.

4. The method of claim 1, wherein the privacy server of the current patient comprises at least one of an environmental monitoring sensor and a medication tracking component, and the self-attention model corresponding to the privacy server comprises at least one of a medical environmental assessment model and a medication condition assessment model;

the contrast monitoring data set comprises at least one of a medical environment data set acquired by the environment monitoring sensor and a medication condition data set acquired by the medication tracking assembly;

The obtaining, from the privacy server, a second clinical patient status assessment result for the current patient, comprising:

At least one of medical environment assessment results obtained by parallel decoding of the medical environment assessment model based on the medical environment data set and drug condition assessment results obtained by parallel decoding of the drug condition assessment model based on the drug condition data set;

The method further includes the manner in which the sample data sets are decoded in parallel based on the respective evaluation models:

performing initial value processing on the characteristic coefficients, connection weights and network parameters of the corresponding evaluation models;

Based on the characteristic coefficient of the initial value and the connection weight of the initial value, performing weighted accumulation operation on the first sample instance of the sample data set and the network parameter of the initial value to obtain a first clinical patient state evaluation result;

The first clinical patient state assessment is taken as a network parameter for a subsequent cycle and combined with a second sample instance of the sample dataset for parallel decoding.

5. The method of claim 1, wherein the performing an integrated prediction based on the first clinical patient state assessment result and the second clinical patient state assessment result for the current patient to obtain a target clinical patient state assessment result for the current patient comprises:

Acquiring a first model weight preset for the first clinical patient state evaluation result and a second model weight preset for the second clinical patient state evaluation result in the integrated model;

And respectively carrying out weighted accumulation operation on the first clinical patient state evaluation result and the second clinical patient state evaluation result based on the first model weight and the second model weight to obtain a target clinical patient state evaluation result of the current patient, wherein the target clinical patient state evaluation result is used for representing the confidence level of the high risk event of the current patient.

6. The method according to claim 1, wherein the method further comprises:

obtaining patient information bound to the current patient;

performing a comparison operation on the patient information bound to the current patient and a target clinical patient state evaluation result of the current patient;

if the comparison condition indicates that the disease information is different from the target clinical patient state evaluation result, calculating to obtain a model cost parameter of the current patient according to the current monitoring data set, and executing adjustment operation on a first model weight which is preset by the integrated model and aims at the first clinical patient state evaluation result based on an Adam algorithm; and;

The comparison condition is sent to the privacy server, so that the privacy server determines corresponding model cost parameters according to the comparison condition and the comparison monitoring data set, and executes adjustment operation on a second model weight which is preset by the integrated model and aims at the second clinical patient state evaluation result based on an Adam algorithm and the corresponding model cost parameters, so as to perform optimization on the integrated model until an integrated model with the optimized condition is obtained, wherein the model cost parameters of the current patient comprise cost parameters of a physiological monitoring evaluation model corresponding to the current patient, cost parameters of a treatment prognosis evaluation model corresponding to the current patient and mutual influence cost parameters between the physiological monitoring evaluation model and the treatment prognosis evaluation model; the corresponding model cost parameters comprise a cost parameter of a medical environment assessment model, a cost parameter of a drug condition assessment model, a mutual influence cost parameter between the medical environment assessment model and the drug condition assessment model, a mutual influence cost parameter between the physiological monitoring assessment model and the medical environment assessment model, a mutual influence cost parameter between the physiological monitoring assessment model and the drug condition assessment model, a mutual influence cost parameter between the treatment prognosis assessment model and the medical environment assessment model, and a mutual influence cost parameter between the treatment prognosis assessment model and the drug condition assessment model; the first model weight comprises a model weight of a physiological monitoring and evaluating model and a model weight corresponding to a treatment prognosis evaluating model; the second model weight comprises a model weight of a medical environment evaluation model and a model weight corresponding to a drug condition evaluation model; any model weight is used to perform a weighted accumulation operation on the clinical patient state assessment results of the corresponding assessment model.

7. The method of claim 1, wherein after the integrated prediction based on the first clinical patient state assessment and the second clinical patient state assessment of the current patient, the method further comprises:

Receiving a medical staff coordination instruction sent by a scheduling server, wherein the medical staff coordination instruction is sent by the scheduling server in response to a medical staff coordination command triggered by the target clinical patient state evaluation result, and the target clinical patient state evaluation result comprises a plurality of clinical patient state evaluation sub-results and sub-result association relations corresponding to the plurality of clinical patient state evaluation sub-results;

based on the evaluation contents of the one-to-one matching of the plurality of clinical patient state evaluation sub-results, respectively determining the coordination results of medical staff of the one-to-one matching of the plurality of clinical patient state evaluation sub-results;

According to the determined coordination results of the medical staff, coordinating the corresponding medical staff for the multiple clinical patient state evaluation sub-results;

constructing a clinical patient state evaluation graph structure of the target clinical patient state evaluation result based on the sub-result association relationship, wherein each evaluation entity in the clinical patient state evaluation graph structure represents one clinical patient state evaluation sub-result;

Each evaluation entity in the clinical patient state evaluation graph structure is polled, wherein, if one evaluation entity obtains a pending emergency task allocation result corresponding to a pre-entity, the clinical patient state evaluation sub-result corresponding to the one evaluation entity is executed based on medical staff coordinated with the clinical patient state evaluation sub-result corresponding to the one evaluation entity and the pending emergency task allocation result corresponding to the pre-entity if the one evaluation entity obtains the pending emergency task allocation result corresponding to the one evaluation entity;

And if the one evaluation entity does not have a post-entity, taking the pending emergency task allocation result corresponding to the one evaluation entity as an emergency task allocation result of the target clinical patient state evaluation result.

8. The method of claim 7, wherein the determining healthcare worker coordination results for the one-to-one matching of the plurality of clinical patient state assessment sub-results based on the one-to-one matching of the plurality of clinical patient state assessment sub-results, respectively, comprises:

Determining a target healthcare professional class of a healthcare worker required for one clinical patient state assessment sub-result based on the assessment content of the one clinical patient state assessment sub-result for the one clinical patient state assessment sub-result;

Determining a predicted non-tired job number for the target healthcare specialty category within an execution period of the one clinical patient status assessment sub-result based on the current non-tired job number for the target healthcare specialty category, historical resource usage data for the target healthcare specialty category, and newly added resource demand data for the target healthcare specialty category;

If the predicted non-tired job population is greater than a non-tired job population threshold, taking a first personnel coordination result as a medical personnel coordination result corresponding to the one clinical patient state evaluation sub-result, the first personnel coordination result comprising: when executing the one clinical patient state assessment sub-result, coordinating a desired target healthcare worker number for the one clinical patient state assessment sub-result;

If the number of the predicted non-fatigue operation people is smaller than or equal to the threshold value of the number of the non-fatigue operation people, taking a second personnel coordination result as a medical personnel coordination result corresponding to the one clinical patient state evaluation sub-result, wherein the second personnel coordination result comprises: a preparation time node prior to execution of the one clinical patient state assessment sub-result coordinates a desired target healthcare worker population for the one clinical patient state assessment sub-result.

9. The method of claim 7, wherein the determining healthcare worker coordination results for the one-to-one matching of the plurality of clinical patient state assessment sub-results based on the one-to-one matching of the plurality of clinical patient state assessment sub-results, respectively, comprises:

determining a target healthcare professional class and a target healthcare worker number of medical workers required by one clinical patient state evaluation sub-result based on the evaluation content of the one clinical patient state evaluation sub-result aiming at the one clinical patient state evaluation sub-result;

If the target healthcare worker number is greater than a healthcare worker number threshold, when a number floating interval of the target healthcare professional category in an execution period of the one clinical patient state evaluation sub-result is less than or equal to the number floating threshold, taking a first personnel coordination result as a healthcare worker coordination result corresponding to the one clinical patient state evaluation sub-result, wherein the first personnel coordination result comprises: when executing the one clinical patient state assessment sub-result, coordinating a desired target healthcare worker number for the one clinical patient state assessment sub-result;

When the number of people floating interval of the target healthcare professional class in the execution period of the one clinical patient state evaluation sub-result is larger than the number of people floating threshold, a second person coordination result is used as a medical person coordination result corresponding to the one clinical patient state evaluation sub-result, and the second person coordination result comprises: a preparation time node prior to execution of the one clinical patient state assessment sub-result coordinates a desired target healthcare worker population for the one clinical patient state assessment sub-result.

10. A server system comprising a server for performing the method of any of claims 1-9.