State evaluation system for chronic kidney disease patient, training method of state detection model, and computer program product

Abstract

The embodiment of the invention provides a state evaluation system of a chronic kidney disease patient and a training method of a state detection model. The state evaluation system for the chronic kidney disease patient comprises a data acquisition device used for acquiring target medical data of a target object and inputting the target medical data into a pre-trained state detection model, and the target medical data comprises treatment data of hemodialysis treatment of the target object; the state detection model is used for outputting a target risk level corresponding to the target object according to the target medical data, and a training sample of the state detection model comprises sample medical data of a sample object and a sample risk level corresponding to the sample object; the rehabilitation information determining device is used for determining target auxiliary rehabilitation information corresponding to the target risk level; the rehabilitation information pushing device is used for pushing the target auxiliary rehabilitation information corresponding to the target risk level to the target object and / or the medical staff, and the accuracy of state evaluation of the chronic kidney disease patient is improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of medical data processing technology, and in particular to a system for evaluating the status of patients with chronic kidney disease, a method for training a status detection model, a training device for the status detection model, a computing device, and a computer program product. Background Art

[0002] In the treatment of chronic kidney disease (CKD), hemodialysis becomes a critical means of maintaining patients' lives when the disease progresses to the terminal stage. With the increasing incidence of CKD and the growing number of hemodialysis patients, the need for accurate patient status assessment for effective management has become increasingly urgent. Accurately assessing a patient's status provides crucial information for adjusting treatment plans and preventing complications, significantly improving patients' quality of life and prolonging their survival.

[0003] In related technologies, the assessment of hemodialysis patients' condition primarily relies on manual operations by medical staff. Medical staff collect the patient's physiological indicators and treatment information, and then, based on this knowledge and experience, manually judge the patient's condition, treatment effectiveness, and potential risks.

[0004] In the process of realizing the concept of this application, the inventors found that, on the one hand, the physiological indicators and treatment information of hemodialysis patients are complicated and changeable. The patient's dialysis situation is different each time, and the physiological indicators are affected by multiple factors such as diet, work and rest, and mood, and fluctuate significantly at different time points. When medical staff deal with large amounts of data, it is easy to omit or misjudge. For example, in the face of complex trends in changes in renal function indicators, if manual analysis alone is relied upon, subtle but important changes may not be detected in time, thereby delaying the diagnosis of the condition. On the other hand, manual evaluation is greatly affected by differences in the personal experience and professional level of medical staff. Different medical staff may have deviations in their understanding and judgment of the same data. Experienced medical staff may be more sensitive to abnormal situations, while those with less experience may find it difficult to make accurate assessments.

[0005] Therefore, how to accurately assess the status of patients with chronic kidney disease has become a technical problem that needs to be solved urgently. Summary of the Invention

[0006] The embodiments of the present application provide a chronic kidney disease patient status assessment system, a status detection model training method, a status detection model training device, a computing device, and a computer program product to address the technical problem in related technologies that the hemodialysis treatment process mainly relies on manual record-keeping and regular follow-up visits, resulting in delayed and incomplete information on the treatment process and low work efficiency of medical staff.

[0007] In the first aspect, an embodiment of the present application provides a state assessment system for patients with chronic kidney disease, comprising: a data acquisition device, a state detection model, a rehabilitation information determination device and a rehabilitation information push device, wherein the data acquisition device is used to acquire the target medical data of the target object and input the target medical data into a pre-trained state detection model, wherein the target medical data includes treatment data of the target object undergoing hemodialysis treatment; the state detection model is used to output the target risk level corresponding to the target object based on the target medical data, wherein the state detection model is obtained by training the original AI model with training samples, and the training samples include sample medical data of the sample object, and the sample risk level corresponding to the sample object; the rehabilitation information determination device is used to determine the target auxiliary rehabilitation information corresponding to the target risk level; the rehabilitation information push device is used to push the target auxiliary rehabilitation information corresponding to the target risk level to the target object and / or medical staff.

[0008] In the second aspect, an embodiment of the present application provides a training method for a state detection model, including: obtaining training samples, wherein the training samples include sample medical data of the sample objects, and the sample risk level corresponding to the sample objects; dividing the training samples into a training set, a validation set, and a test set; training the original AI model through the training set, and using the validation set to adjust the hyperparameters of the trained original AI model, and after the accuracy of the test set is tested to meet predetermined conditions, obtaining a state detection model for chronic kidney disease patient status assessment.

[0009] In the third aspect, an embodiment of the present application provides a training device for a state detection model, comprising: a second acquisition module for acquiring training samples, wherein the training samples include sample medical data of the sample objects and the sample risk levels corresponding to the sample objects; a sample division module for dividing the training samples into a training set, a validation set and a test set; a model training module for training the original AI model through the training set, and using the validation set to adjust the hyperparameters of the original AI model after training, and after the accuracy of the test set is tested to meet the predetermined conditions, a state detection model for chronic kidney disease patient status assessment is obtained.

[0010] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processing component, it implements any of the methods described above.

[0011] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program / instruction, which implements any of the methods described above when executed by a processing component.

[0012] In an embodiment of the present application, a state detection model suitable for hemodialysis patients is obtained by: obtaining target medical data of a target subject, wherein the target medical data includes treatment data of the target subject undergoing hemodialysis treatment; inputting the target medical data into a pre-trained state detection model, and having the state detection model output a target risk level corresponding to the target subject based on the target medical data. The state detection model is obtained by training an original AI model using training samples, wherein the training samples include sample medical data of the sample subject and the sample risk level corresponding to the sample subject; determining target-assisted rehabilitation information corresponding to the target risk level; and pushing the target-assisted rehabilitation information corresponding to the target risk level to the target subject and / or medical staff. By collecting sample medical data of a large number of sample subjects and their corresponding sample risk levels, constructing a training sample set, and training the original AI model, a state detection model suitable for hemodialysis patients is obtained. The state detection model can output the corresponding target risk level based on the input target medical data of the target subject, greatly improving the accuracy and efficiency of patient state assessment compared to manual judgment in related technologies. Based on the accurate risk level, corresponding target-assisted rehabilitation information can be further determined. The target-assisted rehabilitation information can include dietary adjustment suggestions, exercise guidance, medication reminders, and medical guidance in emergency situations. Finally, by pushing these target-assisted rehabilitation information to target subjects and / or medical staff, personalized and precise management of hemodialysis patients can be achieved, effectively improving the patients' treatment effects and quality of life. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0014] Figure 1 A flow chart showing a method for evaluating the status of a chronic kidney disease patient provided by the present application is shown;

[0015] Figure 2 A flow chart showing a method for training a state detection model provided by the present application is shown;

[0016] Figure 3 A schematic diagram of the structure of a chronic kidney disease patient status assessment system provided by the present application is shown;

[0017] Figure 4 This is an example diagram of an interactive interface for patients to upload vital signs, provided in an optional embodiment of the present application;

[0018] Figure 5This is an example diagram of an interactive interface for patients to upload their daily behaviors, provided in an optional embodiment of the present application;

[0019] Figure 6 This is an example diagram of an interactive interface for drug records provided in an optional embodiment of the present application;

[0020] Figure 7 This is an example diagram of the interactive interface of the test sheet provided in an optional embodiment of the present application;

[0021] Figure 8 A structural diagram of a training device for a state detection model provided in this application is shown. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0023] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0024] For ease of reference, some terms used in this description are defined below. The terms and their respective definitions set forth herein are not strictly limited to these definitions—a term may be further defined by its use in this disclosure. As used herein, the term "exemplary" means serving as an example, instance, or illustration. Any aspect or design described herein as "exemplary" should not necessarily be construed as preferred over other aspects or designs. Rather, the use of the word "exemplary" is intended to present concepts in a concrete manner. In this application and the appended claims, the term "or" is intended to mean an inclusive "or," not an exclusive "or." That is, unless otherwise specified or clear from the context, "X employs A or B" is intended to mean any of the natural inclusive permutations. That is, if X employs A, X employs B, or X employs both A and B, then "X employs A or B" is satisfied in any of the above instances. As used herein, at least one of A or B means at least one of A, or at least one of B, or at least one of A and B. In other words, the phrase is disjunctive. The articles "a" and "an" as used in this application and the appended claims should generally be construed to mean "one or more" unless specified otherwise or clear from context to be in the singular.

[0025] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0026] In order to facilitate a detailed description of the solution involved in the present invention, the status assessment system is first described using a status assessment method implemented by the status assessment system for chronic kidney disease patients. Figure 1 A flow chart of a method for evaluating the status of a chronic kidney disease patient provided by the present application is shown in FIG. Figure 1 As shown, the method may include the following steps:

[0027] S101, acquiring target medical data of a target object, wherein the target medical data includes treatment data of hemodialysis treatment of the target object.

[0028] The target object may refer to a patient who needs hemodialysis treatment.

[0029] The target medical data may refer to all data related to the patient's hemodialysis treatment, including but not limited to treatment data, physiological indicator data, health behavior data, etc.

[0030] This step can accurately acquire the target medical data of the target subject, namely the patient's hemodialysis treatment data, by comprehensively integrating multi-source data. Specifically, this can include information such as the patient's medical history and allergy history, as well as the patient's previous hemodialysis treatment records. These records can include detailed data such as dialysis time (i.e., the specific duration of each dialysis treatment from start to finish), ultrafiltration volume (the amount of ultrafiltration fluid removed during the dialysis process), and anticoagulant dosage. Laboratory test results can also be included, such as blood test results (such as red blood cell, white blood cell, and platelet counts), renal function indicators (such as creatinine and urea nitrogen), and electrolyte indicators (such as electrolyte balance). These data can reflect the patient's physical condition from different perspectives. Additionally, patients can upload daily physiological data via wearable devices, such as blood pressure, heart rate, and weight data collected by smart bracelets, which can provide real-time insights into changes in the patient's physical condition during daily life. Furthermore, the target medical data can also include unstructured data, including diet, sleep, and exercise, that patients enter voluntarily via mobile applications. Diet includes the types and amounts of daily food intake, sleep records information such as bedtime, sleep duration, sleep quality, etc. Exercise includes data such as exercise type, exercise duration and intensity.

[0031] The integration of multi-source data can provide a more comprehensive picture of a patient's health status and treatment outcomes, helping to improve the accuracy of risk assessments and the effectiveness of assisted rehabilitation information. By integrating data from different sources, we can gain a more comprehensive understanding of how factors such as changes in a patient's physiological indicators, treatment compliance, and lifestyle habits influence their condition.

[0032] S102, input the target medical data into a pre-trained state detection model, and the state detection model outputs the target risk level corresponding to the target object based on the target medical data, wherein the state detection model is obtained by training the original AI model through training samples, and the training samples include sample medical data of the sample object and the sample risk level corresponding to the sample object.

[0033] Among them, the state detection model refers to a model obtained by training the original AI (Artificial Intelligence) model through training samples, which can be used for hemodialysis risk assessment.

[0034] The training samples refer to the sample data used to train the state detection model, including the sample medical data of the sample objects and the corresponding sample risk levels.

[0035] Among them, the target risk level refers to the current risk level of the target object output by the state detection model based on the target medical data, which can be divided into low risk, medium risk, high risk, etc.

[0036] In this step, the acquired target medical data is input into a pre-trained state detection model. The model analyzes and processes the input data and outputs the target risk level corresponding to the target object. The state detection model is constructed based on a large number of training samples. The training samples include sample medical data of the sample objects and the corresponding sample risk levels. Through learning from the training samples, the model can learn the correlation between different medical data and risk levels, thereby achieving accurate risk assessment of new input data. This data-driven risk assessment method is more accurate and efficient than traditional manual assessment methods. Therefore, it can improve the accuracy and efficiency of risk assessment and provide a scientific basis for the formulation of personalized assisted rehabilitation information. At the same time, through the automated evaluation of the model, the workload of medical staff is reduced and work efficiency is improved.

[0037] S103: Determine target assisted rehabilitation information corresponding to the target risk level.

[0038] Among them, target-assisted rehabilitation information refers to personalized diagnosis, treatment and care recommendations based on the risk level of the target object, or self-management reminders that help patients recover.

[0039] This step can develop targeted assisted rehabilitation information based on the target risk level output by the state detection model, combined with medical guidelines and clinical experience. For low-risk patients, basic health management and lifestyle advice may be sufficient; for medium-risk patients, increased monitoring of key physiological indicators and personalized dietary adjustment recommendations may be required; for high-risk patients, urgent reminders are required to contact medical staff immediately, and adjustments to dialysis plans or hospitalization may be required.

[0040] Patients at different risk levels require different assisted rehabilitation information to achieve the best treatment outcomes. By combining medical guidelines with clinical experience, more scientific and reasonable diagnosis and treatment recommendations can be developed, helping to improve patients' treatment outcomes and quality of life.

[0041] S104: Push target-assisted rehabilitation information corresponding to the target risk level to the target subject and / or medical staff.

[0042] Timely information transmission and reception are key to ensuring the effective implementation of assisted rehabilitation information. This step can push target assisted rehabilitation information corresponding to the target risk level to the target subject and / or medical staff through multiple channels (such as mobile applications, text messages, voice calls, etc.). This ensures that patients and medical staff can keep up to date with the latest diagnosis and treatment and nursing recommendations, as well as information that helps with rehabilitation, improves the efficiency of information transmission and reception, and ensures the effective implementation of assisted rehabilitation information. At the same time, timely reminders and suggestions can also help improve patients' treatment compliance and self-management capabilities, thereby further improving the effectiveness of long-term treatment of chronic diseases caused by hemodialysis.

[0043] In an embodiment of the present application, a state detection model suitable for hemodialysis patients is obtained by: obtaining target medical data of a target subject, wherein the target medical data includes treatment data of the target subject undergoing hemodialysis treatment; inputting the target medical data into a pre-trained state detection model, and having the state detection model output a target risk level corresponding to the target subject based on the target medical data. The state detection model is obtained by training an original AI model using training samples, wherein the training samples include sample medical data of the sample subject and the sample risk level corresponding to the sample subject; determining target-assisted rehabilitation information corresponding to the target risk level; and pushing the target-assisted rehabilitation information corresponding to the target risk level to the target subject and / or medical staff. By collecting sample medical data of a large number of sample subjects and their corresponding sample risk levels, constructing a training sample set, and training the original AI model, a state detection model suitable for hemodialysis patients is obtained. The state detection model can output the corresponding target risk level based on the input target medical data of the target subject, greatly improving the accuracy and efficiency of patient state assessment compared to manual judgment in related technologies. Based on the accurate risk level, corresponding target-assisted rehabilitation information can be further determined. The target-assisted rehabilitation information can include dietary adjustment suggestions, exercise guidance, medication reminders, and medical guidance in emergency situations. Finally, by pushing these target-assisted rehabilitation information to target subjects and / or medical staff, personalized and precise management of hemodialysis patients can be achieved, effectively improving the patients' treatment effects and quality of life.

[0044] As an optional embodiment, determining the target-assisted rehabilitation information corresponding to the target risk level includes: outputting the target-assisted rehabilitation information corresponding to the target risk level by the state detection model, wherein the training sample also includes sample-assisted rehabilitation information corresponding to the sample risk level.

[0045] Alternatively, the state detection model may directly output target-assisted rehabilitation information corresponding to the target risk level. In this embodiment, the training sample includes not only the sample medical data of the sample subject and the corresponding sample risk level, but also the sample-assisted rehabilitation information corresponding to the sample risk level.

[0046] By adding sample auxiliary rehabilitation information data to the training samples, the state detection model not only learns how to assess risk levels during the training process, but also learns how to convert risk levels into specific auxiliary rehabilitation information. Since the model directly learns the mapping relationship between risk levels and auxiliary rehabilitation information, when the model is actually running, it can directly output the corresponding target auxiliary rehabilitation information based on the input target medical data without the need for additional rules or logic for mapping. It can also directly link the auxiliary rehabilitation information with the medical data, making the output auxiliary rehabilitation information more accurate and personalized, which helps to improve the patient's treatment effect and quality of life. In addition, the model directly outputs auxiliary rehabilitation information, which reduces the workload of medical staff in formulating auxiliary rehabilitation information, allowing them to devote more energy to the actual care of patients.

[0047] Take a specific hemodialysis patient as an example. The patient's blood pressure has fluctuated greatly recently, and treatment compliance has decreased. In the traditional diagnosis and treatment process, medical staff may need to manually analyze the patient's medical data first, assess their risk level, and then formulate corresponding auxiliary rehabilitation information based on the risk level. In this optional embodiment, the state detection model can directly output the corresponding risk level and auxiliary rehabilitation information based on the patient's medical data. For example, the model may determine that the patient is at a medium risk level and output corresponding auxiliary rehabilitation information, such as increasing the frequency of blood pressure monitoring, adjusting dietary recommendations, etc. These auxiliary rehabilitation information can be pushed to patients and medical staff in a timely manner through mobile applications, text messages or voice calls, ensuring that patients can take timely measures to improve their health status, while also facilitating medical staff to remotely monitor and guide patients.

[0048] As an optional embodiment, the target medical data is input into a pre-trained state detection model, and the state detection model outputs the target risk level corresponding to the target object based on the target medical data, including: inputting the target medical data into the state detection model, extracting temporal features by the long short-term memory network module in the state detection model, and extracting spatial features by the convolutional network module in the state detection model, wherein the temporal features are used to describe the characteristics of the target medical data that change over time, and the spatial features are used to describe the local correlation features in the target medical data; the classification module in the state detection model outputs the target risk level based on the temporal features and the spatial features.

[0049] Among them, the Long Short-Term Memory Network (LSTM) is a special recurrent neural network that is good at processing and predicting time intervals and long-term dependencies in sequence data.

[0050] Among them, convolutional network (CNN) is a feedforward neural network that is particularly good at processing data with grid structures, such as images, and can automatically extract local correlation features of data.

[0051] Among them, time series features: describe the characteristics of target medical data changing over time, such as the changing trends of physiological indicators such as blood pressure and heart rate over a period of time.

[0052] Among them, spatial features: describe the local correlation features in the target medical data, such as the correlation between different physiological indicators.

[0053] In this optional embodiment, when determining the target risk level corresponding to the target object, the target medical data can be input into a pre-trained state detection model. The state detection model can use a long short-term memory network (LSTM) module to extract temporal features and a convolutional network (CNN) module to extract spatial features. Finally, the classification module outputs the target risk level based on the extracted features. The LSTM module can capture complex patterns in the data that change over time, while the CNN module can extract local correlation features in the data. By combining these two features, the classification module can more comprehensively understand the target medical data and thus more accurately output the target risk level.

[0054] As an optional embodiment, the above method also includes: obtaining feedback data of the target object to which the target-assisted rehabilitation information is applied; inputting the feedback data into a reinforcement learning model constructed based on the state detection model, and the reinforcement learning model adjusting internal parameters according to the feedback data, wherein the reinforcement learning model after adjusting the parameters is a personalized model for the target object.

[0055] Among them, reinforcement learning is a machine learning method that learns how to take actions to maximize a certain cumulative reward by interacting with the environment. In this model, the agent (Agent) learns to take the best action in a specific environment by trying different actions and adjusting its strategy based on the results (rewards or penalties).

[0056] In the reinforcement learning model, internal parameters refer to the parameters used to determine how the agent chooses actions based on the current state, such as the weights of the neural network.

[0057] In this optional embodiment, during the application stage of the state detection model, a reinforcement learning mechanism can be introduced to further optimize the AI hemodialysis chronic disease model, so that the popular AI hemodialysis model can be adjusted in each judgment, and can be more suitable for a certain individual, thereby realizing personalized diagnosis and treatment for patients. Specifically, by obtaining feedback data of the target object to which the target assisted rehabilitation information is applied (such as the patient's response to self-management reminders, treatment effects, etc.), the feedback data can be input into the reinforcement learning model constructed based on the state detection model. The reinforcement learning model will adjust its internal parameters according to these feedback data to generate a personalized model for the target object. This adjustment is based on the goal of maximizing long-term cumulative rewards, so that the reminders and assisted rehabilitation information generated by the model can more effectively improve the patient's health status, and while maintaining the original universality, it is more adapted to the uniqueness of a certain patient.

[0058] Traditional management models are often based on general medical knowledge and clinical experience, making it difficult to fully account for the individual differences of each patient. However, by introducing reinforcement learning mechanisms, management strategies can be customized based on specific patient feedback, improving personalization. Furthermore, as a patient's health status changes and treatment progresses, the initial AI model may no longer be fully applicable. By continuously collecting feedback and adjusting model parameters, the model's adaptability can be enhanced, allowing it to more accurately reflect the patient's actual condition.

[0059] Therefore, the personalized auxiliary rehabilitation information in the embodiments provided by the present invention is more in line with the actual needs of patients, helps to improve patient satisfaction and trust, and can achieve more accurate management and reminders, helping patients to better follow the treatment plan, thereby improving treatment effects, and further reducing the burden on medical staff and improving work efficiency.

[0060] As an optional embodiment, obtaining feedback data of the target object to which the target-assisted rehabilitation information is applied includes at least one of the following: obtaining examination result data of the target object in the hospital to which the target-assisted rehabilitation information is applied; obtaining daily physical data filled in by the target object through an application; obtaining consultation data obtained by medical staff during a consultation with the target object to which the target-assisted rehabilitation information is applied.

[0061] Among them, consultation data refers to the relevant information generated during the discussion and diagnosis of a case by multiple doctors or experts, including the doctor's diagnostic opinions, treatment plan recommendations, etc.

[0062] In this optional embodiment, feedback data from hemodialysis patients is collected through multiple channels to more comprehensively understand the patient's health status and treatment response. The feedback data includes but is not limited to: hospital examination results, daily physical data entered in the application, and consultation data.

[0063] Among them, the hospital examination result data directly reflects the patient's physiological indicators and changes in condition, and is an important basis for evaluating treatment effects and adjusting auxiliary rehabilitation information; daily physical data, such as blood pressure, heart rate, weight, etc., can monitor the patient's physical condition in real time and promptly detect potential health risks; and consultation data brings together the opinions and suggestions of multiple doctors or experts, providing a more comprehensive reference for formulating personalized auxiliary rehabilitation information.

[0064] Inputting this data into the AI hemodialysis chronic disease model can help the model more accurately determine a patient's risk level and generate more precise self-management reminders and treatment recommendations. With more comprehensive data input, the AI hemodialysis chronic disease model can more accurately determine changes in a patient's condition and treatment effectiveness, allowing for timely adjustments to auxiliary rehabilitation information and improved treatment outcomes.

[0065] As an optional embodiment, the state detection model is also used to provide doctors with data analysis reports on the target object and treatment plan recommendations based on the target medical data, wherein the training samples include: sample data analysis reports corresponding to the sample medical data, and sample treatment plan recommendations corresponding to the sample risk level; the state detection model is also used to provide nurses with nursing recommendations on the target object based on the target medical data, wherein the training samples include sample nursing recommendations corresponding to the sample risk level.

[0066] Among them, data analysis report refers to a report generated based on the collected medical data through statistical analysis, data mining and other methods. It aims to reveal the patterns, trends or anomalies in the data and provide a basis for medical decision-making.

[0067] Among them, treatment plan recommendations refer to treatment plans or suggestions proposed by doctors or AI models based on the patient's specific condition, physiological indicators, and medical knowledge.

[0068] Among them, nursing recommendations refer to nursing measures or suggestions proposed by nurses or AI models based on the patient's specific condition and nursing needs, aiming to promote patient recovery, prevent complications, etc.

[0069] This optional embodiment can utilize the trained state detection model to provide doctors and nurses with personalized data analysis reports, treatment plan recommendations, and nursing recommendations based on target medical data (such as the patient's physiological indicators, examination results, historical medical records, etc.).

[0070] Specifically, training samples include sample data analysis reports corresponding to sample medical data, as well as sample treatment and nursing recommendations corresponding to sample risk levels. During training, the AI model continuously attempts to match the input sample medical data with the corresponding analysis reports, treatment plans, and nursing recommendations, optimizing the matching results by adjusting model parameters. Once training is complete, the model will automatically generate corresponding analysis reports, treatment recommendations, and nursing recommendations based on the new target medical data.

[0071] Traditionally, doctors and nurses have spent considerable time manually analyzing patient medical data to develop treatment plans and care plans. AI models, however, can automate this process, significantly improving diagnosis and treatment efficiency. AI models can learn from large amounts of sample data, identifying underlying patterns and trends within the data, and providing doctors and nurses with more accurate and comprehensive diagnostic and treatment recommendations. This helps improve the quality of diagnosis and treatment and reduce misdiagnoses and missed diagnoses.

[0072] As an optional embodiment, when the target medical data includes structured data and unstructured data, the data processing module in the state detection model is also used to process the unstructured data and convert the unstructured data into structured data.

[0073] Among them, unstructured data refers to data without a fixed format or pattern, such as text, images, audio, etc. These data are difficult to store and analyze using traditional relational databases.

[0074] Among them, structured data refers to data with a clear format and predefined model, such as tabular data in a relational database, which can be stored and queried through fixed fields and types.

[0075] When the target medical data includes structured data (such as basic patient information and laboratory test results in hospital information systems) and unstructured data (such as text descriptions of diet, sleep, exercise, etc. entered by patients through mobile applications), the data processing module in the state detection model can first clean all data to remove useless information, noise and outliers to ensure the accuracy and reliability of the data; then, it can use technologies such as natural language processing (NLP) to convert unstructured data into structured data. For example, text parsing technology can be used to extract key nutritional information from dietary records and convert it into quantifiable numerical values or classification labels; finally, the converted structured data can be standardized to ensure the consistency of the data format and facilitate subsequent model analysis.

[0076] Unstructured data often contains rich information, but it's difficult to directly extract key features. Converting it to structured data can better capture the characteristics and patterns within the data, thereby improving the accuracy and predictive capabilities of the model. Furthermore, converting data from different sources and formats into structured data facilitates data integration and sharing.

[0077] Figure 2 A flow chart of a training method for a state detection model provided by the present application is shown in FIG. Figure 2 As shown, the method may include the following steps:

[0078] S201: Acquire a training sample, wherein the training sample includes sample medical data of a sample object and a sample risk level corresponding to the sample object.

[0079] Before model training, a large amount of sample data must be collected, including sample medical data of the sample subjects and their corresponding sample risk levels. Sample medical data can include physiological indicators (such as blood pressure, heart rate, weight, etc.), laboratory test results (such as blood routine, renal function, electrolytes, etc.), and historical medical records. The sample risk level is an assessment of the severity of the sample subject's current condition based on medical expertise and clinical experience.

[0080] The learning process of machine learning models is based on large amounts of data. Through continuous learning and adjustment, the model can gradually grasp the patterns and regularities in the data. During the data collection phase, sufficient, diverse, and representative training data is required. The collected training samples will serve as the foundation for AI model learning. Only in this way can the model learn the characteristic manifestations of different conditions, help the model understand the relationship between different medical data and risk levels, and make accurate judgments when faced with new data.

[0081] S202, dividing the training samples into a training set, a validation set and a test set.

[0082] In an embodiment of the present application, after obtaining the training sample, it can be divided into three independent parts: a training set, a validation set, and a test set. The training set is used to train the original AI model so that it can gradually learn the rules and patterns in the data; the validation set is used to adjust the model's hyperparameters (such as learning rate, batch size, etc.) during the training process to improve the performance of the model; the test set is used to evaluate the model's accuracy and other performance indicators after the training is completed. The training samples are divided into different data sets to ensure the generalization ability of the model and the fairness of the performance evaluation. The training set is used to allow the model to learn the rules in the data, the validation set is used to adjust the model to avoid overfitting or underfitting, and the test set is used to finally evaluate the performance of the model to ensure the effectiveness of the model in practical applications.

[0083] By dividing the training samples into different datasets, we can better control the model training process and avoid overfitting or underfitting during the training process. At the same time, independent test sets can also more accurately evaluate the performance of the model, providing strong support for model optimization and improvement.

[0084] S203, train the original AI model through the training set, use the validation set to adjust the hyperparameters of the trained original AI model, and after the accuracy is tested through the test set to meet the predetermined conditions, obtain a state detection model for chronic kidney disease patient status assessment.

[0085] Among them, hyperparameters refer to the parameters that need to be set before model training. These parameters have a significant impact on the training process and performance of the model.

[0086] Among them, accuracy is an indicator to measure the consistency between the model prediction results and the actual results.

[0087] In the embodiment provided by the present invention, the original AI model can be trained using a training set first. During the training process, the model continuously adjusts its internal parameters to minimize the prediction error. Then, the validation set is used to adjust the hyperparameters of the model to optimize the performance of the model. Finally, the test set is used to evaluate the accuracy of the trained model. If the accuracy of the model meets the predetermined conditions (such as the accuracy rate, recall rate, etc. reaching a certain threshold), the model is considered to have been trained and can be used to assess the status of patients with chronic kidney disease.

[0088] It's important to note that in machine learning, model training involves continuously optimizing internal parameters to minimize prediction error. By continuously adjusting the model's parameters and hyperparameters, a model can be obtained that performs well on the training set. However, this does not necessarily mean that the model will perform well in real-world applications. Therefore, an independent test set is necessary to evaluate the model's generalization ability and ensure its effectiveness in real-world applications.

[0089] Through the above steps, data can be collected, training samples can be obtained, and a state detection model that can accurately assess the risk level of hemodialysis patients and provide effective support to medical staff can be obtained through training. This solves the technical problem in related technologies that the hemodialysis treatment process mainly relies on manual records and regular follow-up visits, resulting in delayed and incomplete information on the treatment process and low work efficiency of medical staff, thereby achieving the beneficial effect of improving the work efficiency of medical staff and reducing the labor intensity of medical staff.

[0090] As an optional embodiment, the above method also includes: constructing a reinforcement learning model based on the state detection model; obtaining feedback data corresponding to the target-assisted rehabilitation information output by the state detection model, wherein the feedback data is used to describe the feedback of the target object to which the target-assisted rehabilitation information is applied on the target-assisted rehabilitation information; training the reinforcement learning model according to the feedback data to obtain a personalized model for the target object.

[0091] Based on the state detection model, a reinforcement learning model can be further constructed. This reinforcement learning model aims to learn how to adjust auxiliary rehabilitation information based on feedback from patients and / or medical staff through interaction with the environment (i.e., the hemodialysis chronic disease diagnosis and treatment scenario) to achieve better treatment results.

[0092] The reinforcement learning model gradually learns the optimal behavior strategy by continuously trying different actions (i.e., risk levels and auxiliary rehabilitation information) and adjusting its strategy based on environmental feedback rewards (such as improved patient health and increased medical staff efficiency). During the hemodialysis treatment of chronic diseases, this mechanism can help the model dynamically adjust auxiliary rehabilitation information based on actual patient feedback, achieving personalized treatment results.

[0093] Traditional AI models are often trained and predicted based on static data, making them difficult to adapt to the complex and ever-changing environment of chronic disease management in hemodialysis. Reinforcement learning models, on the other hand, learn through interaction with the environment and can better adapt to dynamically changing circumstances, providing more precise diagnosis and treatment services for patients.

[0094] It should be noted that the above-mentioned reinforcement training process can occur during the actual application of the state detection model, that is, when the state detection model training is completed, a reinforcement learning model can be constructed on this basis. At this time, the reinforcement model can be directly used to diagnose and treat a patient. The reinforcement learning model can include a state detection model, or in other words, the state detection model is the intelligent part of the reinforcement learning model, which is used to select actions (i.e., the auxiliary rehabilitation information output by the state detection model) based on the current state (i.e., the patient's medical data). Then, the additional parameter adjustment network in the reinforcement learning model can judge the reward of this action selection based on the feedback from the environment (i.e., the feedback data provided by the patient) and adjust the parameters based on the reward, so that the entire reinforcement learning model can adjust and optimize the suggestions based on the patient's feedback. In addition, during the adjustment process, the parameters of the trained state detection model can be locked unchanged, and only the parameters of the newly added network when constructing the reinforcement learning model can be adjusted. This ensures that the knowledge already learned by the model will not be lost.

[0095] As an optional embodiment, obtaining feedback data corresponding to the target-assisted rehabilitation information output by the state detection model includes at least one of the following: obtaining examination result data of the target object to which the target-assisted rehabilitation information is applied in the hospital; obtaining daily physical data filled in by the target object to which the target-assisted rehabilitation information is applied through an application; obtaining consultation data obtained by medical staff during a consultation with the target object to which the target-assisted rehabilitation information is applied.

[0096] In this optional embodiment, feedback data may include the patient's hospital examination results, daily physical data entered by the patient via the app, and consultation data from medical staff. By acquiring and analyzing this feedback data, the reinforcement learning model's behavioral strategies can be continuously optimized to improve the effectiveness and safety of treatment. This also helps improve the efficiency of medical staff and reduce workload.

[0097] Feedback data is fundamental to reinforcement learning model training. By analyzing patient feedback, we can evaluate the effectiveness of different rehabilitation support information and adjust the reinforcement learning model's behavioral strategies accordingly. For example, if a patient's health significantly improves after adopting a specific rehabilitation support recommendation, the reinforcement learning model will be more inclined to adopt that recommendation. Conversely, if the recommendation is ineffective, the model will make adjustments.

[0098] As an optional embodiment, the original AI model is trained using a training set, and the hyperparameters of the trained original AI model are adjusted using a validation set. After the accuracy is tested on a test set to meet predetermined conditions, a state detection model for chronic kidney disease patient state assessment is obtained, including: when the training sample includes a sample data analysis report corresponding to the sample medical data, and a sample treatment plan recommendation corresponding to the sample risk level, the training sample is used to train the original AI model to obtain a state detection model for providing a data analysis report on the target object and a treatment plan recommendation to the doctor based on the target medical data; when the training sample includes a sample nursing recommendation corresponding to the sample risk level, the training sample is used to train the original AI model to obtain a state detection model for providing nursing recommendations on the target object to the nurse based on the target medical data.

[0099] When training a state detection model, the data content of the training samples can be customized based on the desired output data. For example, if you want a model that can provide data analysis reports and treatment recommendations to doctors, the training samples should include sample data analysis reports corresponding to the sample medical data and sample treatment recommendations corresponding to the sample risk level. Similarly, if you want to train a model that can provide nursing advice to nurses, the training samples should include sample nursing recommendations corresponding to the sample risk level.

[0100] By customizing the model's output, we can meet the diverse needs of medical professionals in their work. For example, doctors may require detailed data analysis reports and treatment recommendations to develop more precise diagnosis and treatment plans, while nurses may require specific nursing advice to perform their nursing work. This helps improve the practicality and relevance of AI models.

[0101] Figure 3 FIG. 1 shows a schematic diagram of a system for evaluating the status of a chronic kidney disease patient provided by the present application. Figure 3 As shown, the status assessment system for chronic kidney disease patients includes: a data acquisition device 301, a status detection model 302, a rehabilitation information determination device 303 and a rehabilitation information push device 304, wherein:

[0102] The data acquisition device is used to acquire the target medical data of the target object and input the target medical data into a pre-trained state detection model, wherein the target medical data includes the treatment data of the target object undergoing hemodialysis treatment; the state detection model is used to output the target risk level corresponding to the target object based on the target medical data, wherein the state detection model is obtained by training the original AI model through training samples, and the training samples include the sample medical data of the sample object and the sample risk level corresponding to the sample object; the rehabilitation information determination device is used to determine the target assisted rehabilitation information corresponding to the target risk level; the rehabilitation information push device is used to push the target assisted rehabilitation information corresponding to the target risk level to the target object and / or medical staff.

[0103] Optionally, when the training sample further includes sample-assisted rehabilitation information corresponding to the sample risk level, the state detection model outputs target-assisted rehabilitation information corresponding to the target risk level.

[0104] Optionally, the state detection model includes: a long short-term memory network module, used to extract the temporal features of the target medical data, wherein the temporal features are used to describe the characteristics of the target medical data that change over time; a convolutional network module, used to extract the spatial features of the target medical data, wherein the spatial features are used to describe the local correlation features in the target medical data; and a classification module, used to output the target risk level based on the temporal features and spatial features.

[0105] Optionally, the system also includes: a data acquisition module, used to obtain feedback data of the target object to which the target-assisted rehabilitation information is applied, and input the feedback data into a reinforcement learning model constructed based on the state detection model; a reinforcement learning module, used to adjust internal parameters according to the feedback data, wherein the reinforcement learning model after adjusting the parameters is a personalized model for the target object.

[0106] Optionally, the feedback data acquired by the data acquisition module includes at least one of the following: examination result data of the target object applying the target-assisted rehabilitation information in the hospital; daily physical data filled in by the target object applying the target-assisted rehabilitation information through the application; consultation data obtained by medical staff during consultation with the target object applying the target-assisted rehabilitation information.

[0107] Optionally, when the training samples include: a sample data analysis report corresponding to the sample medical data, and a sample treatment plan recommendation corresponding to the sample risk level, the state detection model is also used to provide the doctor with a data analysis report and treatment plan recommendation on the target object based on the target medical data; when the training samples include sample nursing recommendations corresponding to the sample risk level, the state detection model is also used to provide the nurse with nursing recommendations on the target object based on the target medical data.

[0108] Optionally, in the case that the target medical data includes structured data and unstructured data, the state detection model further includes: a data processing module, which is used to process the unstructured data and convert the unstructured data into structured data. The system obtains target medical data of a target subject, wherein the target medical data includes treatment data of the target subject undergoing hemodialysis treatment. The target medical data is input into a pre-trained state detection model, and the state detection model outputs a target risk level corresponding to the target subject based on the target medical data. The state detection model is obtained by training an original AI model using training samples, wherein the training samples include sample medical data of the sample subject and the sample risk level corresponding to the sample subject. Target assisted rehabilitation information corresponding to the target risk level is determined. The target assisted rehabilitation information corresponding to the target risk level is pushed to the target subject and / or medical staff. That is, by organizing various data of the patient undergoing hemodialysis treatment, on the one hand, medical staff can quickly obtain comprehensive and accurate information. On the other hand, a state detection model is constructed and used to perform preliminary analysis of the data, providing corresponding personalized diagnosis and treatment recommendations and analysis reports, providing guidance to medical staff. This solves the technical problem in related technologies that the hemodialysis treatment process mainly relies on manual record-keeping and regular follow-up visits, resulting in delayed and incomplete treatment process information and low work efficiency of medical staff, thereby achieving the beneficial effect of improving the work efficiency of medical staff and reducing the labor intensity of medical staff.

[0109] It should be noted that the aforementioned data acquisition device 301, state detection model 302, rehabilitation information determination device 303, and rehabilitation information push device 304 correspond to steps S101 to S104 in the embodiments. The examples and application scenarios implemented by these modules and corresponding steps are the same, and their implementation principles and technical effects are not further elaborated. The specific manner in which each module and unit of the apparatus in the aforementioned embodiments performs operations has been described in detail in the embodiments of the method and will not be further elaborated here.

[0110] Based on the above embodiment and optional embodiment, an optional implementation manner is provided, which is described in detail below.

[0111] In an optional embodiment of the present application, a specific implementation of a method for establishing an AI hemodialysis chronic disease model and generating patient self-management reminders is provided. This method aims to solve problems such as delayed and incomplete information caused by reliance on manual records and regular follow-up visits during hemodialysis treatment, as well as low work efficiency of medical staff.

[0112] First, it is necessary to establish an AI hemodialysis chronic disease model.

[0113] During the data collection phase, data from hospital information systems, patients' wearable devices, and patients' mobile application data can be collected. Hospital information system data can integrate basic patient information, previous hemodialysis treatment records (such as dialysis time, ultrafiltration volume, anticoagulant dosage, etc.), and laboratory test results (blood routine, renal function, electrolytes, and other indicators); wearable device data can include patients' daily physiological data, such as blood pressure, heart rate, and weight; and mobile application data can include unstructured data such as diet, sleep, and exercise entered by patients.

[0114] Then, the various types of collected data can be cleaned, erroneous values and missing values can be removed, and outliers can be corrected or eliminated using statistical methods; unstructured data can be converted into structured form, such as extracting key nutritional information from dietary text descriptions through natural language processing technology; data of different formats and magnitudes can be normalized to make them suitable for subsequent model training.

[0115] Then, based on medical expertise and data analysis results, we can screen out features that have a significant impact on the management of chronic diseases in hemodialysis patients, such as serum creatinine trends, blood pressure circadian rhythms, and ultrafiltration rate stability. These screened features can be combined and transformed to construct new feature dimensions, thereby improving the model's resolution.

[0116] Then, appropriate AI algorithms, such as the long short-term memory (LSTM) network from deep learning combined with the convolutional neural network (CNN) architecture, can be selected to build an AI model. The LSTM is used to capture the temporal changes in the patient's physiological indicators, while the CNN efficiently convolutions the feature space to extract local correlation features.

[0117] Finally, the preprocessed dataset can be divided into training set, validation set and test set in a certain proportion. The training set is used for model training, the model hyperparameters are adjusted through the validation set, and the model performance is evaluated on the test set.

[0118] Secondly, the trained AI model can be put into practical application and reinforcement learning technology can be introduced so that the entire reinforcement learning model can further adjust the model parameters according to the feedback from the environment, making the output auxiliary rehabilitation information more professional, accurate and effective.

[0119] Agent: refers to the intelligent interactive module configured in the medical management system, which integrates natural language processing, machine learning and knowledge graph technology, and can receive, understand and process multi-dimensional information from patients, nurses and doctors, and make decisions based on this information.

[0120] Environment: includes all relevant factors in the entire hemodialysis chronic disease management scenario, such as the hospital's medical resource environment, the patient's individual condition, the interaction environment between doctors and patients, and the external policy environment.

[0121] State: A comprehensive description of the current status of the hemodialysis chronic disease management system, including the patient's physiological indicator data, treatment status, health behavior data, the working status of medical staff, and the patient's response to previous self-management reminders.

[0122] Action a: The decision-making behavior taken by the agent in a specific state s, such as generating self-management reminders, providing nursing advice to nurses, generating condition analysis reports for doctors, etc.

[0123] Reward r (Reward): The feedback value used to evaluate the effectiveness of the agent's action a, integrating the feedback from the patient, nurse, and doctor.

[0124] Based on the trained AI hemodialysis chronic disease model, the patient's latest monitoring data is input in real time, and the model outputs the patient's current risk level (low risk, medium risk, high risk).

[0125] Based on different risk levels, corresponding self-management reminder rules can be formulated in combination with medical guidelines and clinical experience. Specifically, for low-risk patients: reminders to have a regular schedule, moderate exercise, and maintain a balanced diet, and popular science articles can be pushed on a weekly basis. For medium-risk patients: in addition to basic life reminders, reminders for monitoring key physiological indicators are increased, such as daily reminders to measure blood pressure and weight and upload data, and personalized dietary adjustment suggestions are pushed. For high-risk patients: emergency reminders are issued, requiring immediate contact with medical staff, and at the same time, some daily activities that may aggravate the condition, such as high-intensity exercise, are suspended, and detailed medical guidance is pushed.

[0126] Secondly, reminders can be sent to patients through various channels, such as pop-up notifications on mobile apps, text messages, voice calls, etc. A reminder review and feedback mechanism can be set up within the mobile app to facilitate medical staff to track patients' responses to reminders.

[0127] If the patient does not respond in time, the reminder method will be automatically upgraded or manual follow-up will be arranged.

[0128] This specific implementation method can automatically collect, process and analyze large amounts of patient data, reduce the workload of manual recording and follow-up, solve the problems of asynchronous information storage and prone to data errors, realize unified recording and archiving of patient information, and improve information circulation and utilization efficiency; and the AI model generates personalized hemodialysis plans to assist in decision-making, reducing the workload of manual data analysis and comprehensive consideration of multiple factors; in addition, the risk assessment and reminder mechanism based on the AI model and intelligent agent can timely detect health risks and take measures to reduce the risk of adverse reactions to dialysis; the intelligent agent generates personalized self-management reminders, which can help patients follow the treatment plan and improve the treatment effect.

[0129] Figure 4 This is an example diagram of an interactive interface for patients to upload vital signs, provided in an optional embodiment of the present application; Figure 5 This is an example diagram of an interactive interface for patients to upload their daily behaviors, provided in an optional embodiment of the present application; Figure 6 This is an example diagram of an interactive interface for drug records provided in an optional embodiment of the present application; Figure 7 This is an example diagram of the interactive interface of the test sheet provided in the optional embodiment of the present application; Figure 4 、 Figure 5 、 Figure 6 and Figure 7 As shown, patients can use a mobile app to fill in and upload their vital signs, daily behavior data, and medication records, and can also promptly access hospital-issued laboratory test reports. Simultaneously, the AI model can read the data in the app and, based on a comprehensive understanding of the patient's condition, provide a corresponding risk level and treatment recommendations.

[0130] Through the above specific implementation steps, the present invention realizes the establishment of an AI hemodialysis chronic disease model and the generation of patient self-management reminders, effectively solving the problems caused by reliance on manual records and regular follow-up visits during hemodialysis treatment, improving the work efficiency of medical staff, reducing the labor intensity of medical staff, improving the safety and quality of treatment, and enhancing patients' self-management ability.

[0131] According to an embodiment of the present application, a training device for a state detection model is provided. Figure 8 A schematic diagram of the structure of a training device for a state detection model provided by the present application is shown in FIG. Figure 8 As shown, it includes: a second acquisition module 801, a sample division module 802 and a model training module 803. The device is described below.

[0132] The second acquisition module 801 is configured to acquire a training sample, wherein the training sample includes sample medical data of a sample object and a sample risk level corresponding to the sample object.

[0133] The sample division module 802 is connected to the second acquisition module 801 and is used to divide the training samples into a training set, a validation set and a test set.

[0134] The model training module 803 is connected to the above-mentioned sample division module 802, and is used to train the original AI model through the training set, and use the verification set to adjust the hyperparameters of the trained original AI model. After the accuracy of the test set is tested to meet the predetermined conditions, a state detection model for chronic kidney disease patient status assessment is obtained.

[0135] It should be noted that the second acquisition module 801, sample partitioning module 802, and model training module 803 correspond to steps S201 to S203 in the embodiment. The examples and application scenarios implemented by the multiple modules and corresponding steps are the same, and their implementation principles and technical effects are not repeated here. The specific manner in which each module and unit performs operations in the apparatus in the above embodiment has been described in detail in the embodiment of the method and will not be elaborated on here.

[0136] The embodiment of the present application further provides a computer device, which may include a storage component and a processing component;

[0137] The storage component stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component to implement any one of the above methods.

[0138] Of course, the computer device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0139] The input / output interface provides an interface between the processing component and peripheral interface modules, which may be output devices, input devices, etc. The communication component is configured to facilitate wired or wireless communication between the computing device and other devices.

[0140] The processing component may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0141] The storage component is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0142] The display component may be an electroluminescent (EL) element, a liquid crystal display or a micro display having a similar structure, or a direct retinal display or a similar laser scanning display.

[0143] It should be noted that when implementing the above-mentioned computing device method or processing method, it can be a physical device or an elastic computing host provided by a cloud computing platform. It can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or as a single server or a single terminal device.

[0144] When the above-mentioned computing device implements the above-mentioned method, it can be specifically implemented as an electronic device. The electronic device can refer to a device used by the user, which has the computing, Internet access, communication and other functions required by the user, such as a mobile phone, tablet computer, personal computer, wearable device, etc.

[0145] It should be noted that the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. It can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or as a single server or a single terminal device.

[0146] The present application also provides a computer-readable storage medium storing a computer program that, when executed by a computer, implements the aforementioned method. The computer-readable medium may be included in the electronic device described in the aforementioned embodiment, or may exist independently and not be incorporated into the electronic device.

[0147] The present application also provides a computer program product comprising a computer program carried on a computer-readable storage medium, which, when executed by a computer, can implement the above-described method. In such an embodiment, the computer program can be downloaded and installed from a network and / or installed from a removable medium. When executed by a processor, the computer program performs the various functions defined in the system of the present application.

[0148] It should be noted that the embodiments of the present application may involve the use of user data. In actual applications, user-specific personal data can be used in the scheme described herein within the scope permitted by applicable laws and regulations, subject to the requirements of applicable laws and regulations of the country where the user is located (for example, with the user's explicit consent, effective notification to the user, etc.).

[0149] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0150] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0151] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A system for assessing the status of patients with chronic kidney disease, characterized in that: include: Data acquisition device, state detection model, rehabilitation information determination device and rehabilitation information push device, wherein, The data acquisition device is used to acquire target medical data of a target subject and input the target medical data into a pre-trained state detection model, wherein the target medical data includes treatment data of hemodialysis treatment of the target subject; The state detection model is used to output a target risk level corresponding to the target object based on the target medical data, wherein the state detection model is obtained by training the original AI model with training samples, and the training samples include sample medical data of the sample object and the sample risk level corresponding to the sample object; The rehabilitation information determining means is used to determine target assisted rehabilitation information corresponding to the target risk level; The rehabilitation information pushing device is used to push target-assisted rehabilitation information corresponding to the target risk level to the target subject and / or medical staff.

2. The system according to claim 1, wherein: In a case where the training sample further includes sample-assisted rehabilitation information corresponding to the sample risk level, the state detection model outputs target-assisted rehabilitation information corresponding to the target risk level.

3. The system according to claim 1, wherein: The state detection model includes: a long short-term memory network module, configured to extract time series features of the target medical data, wherein the time series features are used to describe characteristics of the target medical data that change over time; a convolutional network module, configured to extract spatial features of the target medical data, wherein the spatial features are used to describe local correlation features in the target medical data; A classification module is used to output the target risk level according to the temporal characteristics and the spatial characteristics.

4. The system according to claim 1, wherein: The system further comprises: a data acquisition module, configured to acquire feedback data of the target subject to which the target-assisted rehabilitation information is applied, and input the feedback data into a reinforcement learning model constructed based on the state detection model; A reinforcement learning module is used to adjust internal parameters according to the feedback data, wherein the reinforcement learning model after adjusting the parameters is a personalized model for the target object.

5. The system according to claim 4, characterized in that The feedback data acquired by the data acquisition module includes at least one of the following: Examination result data of the target subject in the hospital using the target-assisted rehabilitation information; Daily physical data filled in by the target subject through the application program to which the target-assisted rehabilitation information is applied; The medical staff consults the target object using the target-assisted rehabilitation information to obtain consultation data.

6. The system according to any one of claims 1 to 5, characterized in that When the training sample includes: a sample data analysis report corresponding to the sample medical data, and a sample treatment plan recommendation corresponding to the sample risk level, the state detection model is also used to provide the doctor with a data analysis report and treatment plan recommendation on the target object based on the target medical data; when the training sample includes a sample nursing recommendation corresponding to the sample risk level, the state detection model is also used to provide the nurse with nursing recommendations on the target object based on the target medical data.

7. The system according to any one of claims 1 to 5, characterized in that In the case where the target medical data includes structured data and unstructured data, the state detection model further includes: The data processing module is used to process the unstructured data and convert the unstructured data into structured data.

8. A training method for a state detection model, characterized in that: include: Acquire a training sample, wherein the training sample includes sample medical data of a sample subject and a sample risk level corresponding to the sample subject; Dividing the training samples into a training set, a validation set, and a test set; The original AI model is trained using the training set, and the hyperparameters of the trained original AI model are adjusted using the validation set. After the accuracy of the test set is tested to meet predetermined conditions, a state detection model for chronic kidney disease patient status assessment is obtained.

9. The method according to claim 8, characterized in that The method further comprises: Building a reinforcement learning model based on the state detection model; Acquiring feedback data corresponding to the target-assisted rehabilitation information output by the state detection model, wherein the feedback data is used to describe feedback from a target object to which the target-assisted rehabilitation information is applied on the target-assisted rehabilitation information; The reinforcement learning model is trained according to the feedback data to obtain a personalized model for the target object.

10. The method according to claim 9, characterized in that The obtaining of feedback data corresponding to the target-assisted rehabilitation information output by the state detection model includes at least one of the following: Obtaining examination result data of the target subject in the hospital to which the target-assisted rehabilitation information is applied; acquiring daily physical data filled in by the target subject using the target-assisted rehabilitation information through the application; Acquire consultation data obtained by medical personnel through consultation with the target object applying the target-assisted rehabilitation information.

11. The method according to claim 8, characterized in that The training of the original AI model using the training set, adjusting the hyperparameters of the trained original AI model using the validation set, and obtaining a status detection model for chronic kidney disease patient status assessment after the accuracy of the test set is tested to meet predetermined conditions include: In a case where the training samples include a sample data analysis report corresponding to the sample medical data and a sample treatment plan recommendation corresponding to the sample risk level, the original AI model is trained using the training samples to obtain a state detection model for providing a data analysis report on the target subject and a treatment plan recommendation to the doctor based on the target medical data; In the case where the training sample includes sample nursing recommendations corresponding to the sample risk level, the original AI model is trained using the training sample to obtain the state detection model for providing nursing recommendations about the target object to the nurse based on the target medical data.

12. A computer program product, characterized in that The method comprises a computer program / instruction, which, when executed by a processing component, implements the training method of the state detection model according to any one of claims 8 to 11.

13. A computer program product, characterized in that The method comprises a computer program / instruction, wherein when the computer program / instruction is executed by a processing component, a method for evaluating the status of a patient with chronic kidney disease is implemented, wherein the method for evaluating the status of a patient with chronic kidney disease comprises the following steps: Acquiring target medical data of a target subject, wherein the target medical data includes treatment data of hemodialysis treatment performed on the target subject; Inputting the target medical data into a pre-trained status detection model, and having the status detection model output a target risk level corresponding to the target subject based on the target medical data, wherein the status detection model is obtained by training the original AI model using training samples, and the training samples include sample medical data of the sample subject and the sample risk level corresponding to the sample subject; determining target assisted rehabilitation information corresponding to the target risk level; Push target-assisted rehabilitation information corresponding to the target risk level to the target subject and / or medical staff.

Images