Artificial intelligence-based methods, systems, and storage media for monitoring hypothermia-related infections.

Abstract

This application discloses an artificial intelligence-based method, system, and storage medium for monitoring hypothermia-related infections. The method includes: extracting the type and occurrence time of infection events from comprehensive medical information based on body temperature ranges; extracting first sample data from all comprehensive medical information; generating second sample data corresponding to the first sample data based on time resolution; preprocessing the second sample data based on data type to extract data features; training a monitoring model based on the data features; collecting monitoring data from patients to be monitored in real time based on data tags; inputting the preprocessed monitoring data into the monitoring model and outputting predicted events and their predicted occurrence times; summarizing all predicted events and their predicted occurrence times to generate a monitoring report; and outputting feedback measures based on the monitoring report. This application improves the timeliness and accuracy of hypothermia-related infection monitoring.

Description

Technical Field

[0001] This application relates to the field of intelligent medical data monitoring technology, and in particular to a method, system and storage medium for monitoring hypothermic infection based on artificial intelligence. Background Technology

[0002] Hypothermia refers to a body temperature below 36°C. It commonly occurs during surgery due to factors such as anesthesia, low operating room temperatures, prolonged surgery, and fluid loss. Hypothermia adversely affects multiple physiological systems, including coagulation, immune function, and drug metabolism. Studies have shown a significant association between hypothermia and surgical site infection. Specifically, hypothermia suppresses the immune system and reduces white blood cell activity, thus increasing the risk of surgical site infection. Traditional thermometers typically only provide intermittent measurements, failing to monitor patient temperature changes continuously in real time. They also lack the ability to comprehensively analyze temperature data and correlate it with other physiological parameters (such as heart rate, blood pressure, and blood oxygen saturation), making it difficult to fully assess the impact of hypothermia on surgical site infection. Therefore, it is necessary to leverage the advantages of artificial intelligence in data analysis and pattern recognition to overcome the limitations of traditional monitoring methods.

[0003] A similar prior art is disclosed in Chinese patent application CN115331816A, which discloses an information decision support system for perioperative body temperature management in children with congenital heart disease. This system includes an information acquisition module, an automatic risk assessment module, a hypothermia protection module, and a nursing record generation module. The information acquisition module acquires patient data and transmits the relevant data to the automatic risk assessment module. The automatic risk assessment module has a calculation model. The relevant data is substituted into the calculation model for calculation. If the calculated value is ≥0.3387, the hypothermia protection module is automatically activated. This module uses an automatic knowledge push system for body temperature management of high-risk groups to push relevant nursing measures to medical staff and formulate a personalized nursing plan. If the calculated value is <0.3387, the personalized nursing plan ends, the termination time of the nursing measures is calculated, and medical staff are reminded to conduct a nursing outcome evaluation, while the evaluation time is recorded. Finally, the nursing record generation module generates nursing documentation. The beneficial effects of this invention are: it has good predictive efficacy and decision support functions. Another Chinese patent application, CN117476213A, discloses a method and system for early warning of hypothermia based on non-invasive parameters. The method includes: acquiring non-invasive biometrics of hypothermic patients and non-invasive biometrics of the medical environment; preprocessing the non-invasive biometrics of hypothermic patients to obtain a dataset; inputting the dataset into a real-time alert model to obtain a first probability value; inputting the non-invasive biometrics of the medical environment into a data maintenance and processing module to obtain a second probability value; calculating the probability of needing rewarming treatment based on the first and second probability values; determining whether the hypothermic patient needs rewarming treatment based on the probability of needing rewarming treatment to obtain a judgment result; and determining the early warning result for the hypothermic patient based on the judgment result.

[0004] The shortcomings of existing technologies are mainly reflected in the fact that they only predict the risk values ​​corresponding to low body temperature and generate decision management, and only conduct early warning analysis on non-invasive biomarkers of patients with hypothermia. They do not predict or analyze the types of risks associated with low body temperature. In practice, it is necessary to accurately predict and analyze infection events caused by low body temperature and generate timely feedback measures. Summary of the Invention

[0005] This application provides an artificial intelligence-based method, system, and storage medium for monitoring hypothermic infection, which can improve the timeliness and accuracy of hypothermic infection monitoring.

[0006] In a first aspect, this application provides an artificial intelligence-based method for monitoring hypothermia-related infections, the method comprising:

[0007] Collect comprehensive medical information of surgical patients, set body temperature ranges, extract the type and time of infection events from the comprehensive medical information based on the body temperature ranges, extract first sample data from all the comprehensive medical information based on the type and time of infection, set time resolution, and generate second sample data corresponding to the first sample data based on the time resolution;

[0008] Obtain the data type and data label of the second sample data, preprocess the second sample data based on the data type and extract data features, and train and generate a monitoring model based on the data features;

[0009] Based on the data tags, the monitoring data of the patients to be monitored are collected in real time. After the monitoring data is preprocessed, it is input into the monitoring model and the predicted event and the predicted occurrence time of the predicted event are output.

[0010] Summarize all predicted events and their predicted occurrence times to generate a monitoring report, and output feedback measures based on the monitoring report.

[0011] In conjunction with the first aspect, generating the second sample data corresponding to the first sample data based on the time resolution includes:

[0012] Obtain basic information about the surgical patient, and based on the basic information, obtain subsample data with the same category and a time resolution higher than daily units from all the first sample data;

[0013] Medical data is extracted from the subsample data, specific changes in any of the medical data are obtained, and time information corresponding to the specific changes is marked. Based on the time information, an estimated event and the estimated occurrence time corresponding to the estimated event are generated. The specific changes include changes in data values, changes in collection frequency, and changes in data items.

[0014] The estimated event and the estimated occurrence time are added to the first sample data to generate the second sample data.

[0015] In conjunction with the first aspect, generating an estimated event and the estimated occurrence time corresponding to the estimated event based on the time information includes:

[0016] A medical event database is established, supplementary events associated with the infection event are obtained from the medical event database, related events are extracted from the supplementary events based on the specific changes, and the related events are set as the estimated events;

[0017] Output a change timestamp based on the change type of the specific change, output a lag time based on the change type, and combine the change timestamp and the lag time to set the estimated occurrence time.

[0018] In conjunction with the first aspect, the step of training and generating a monitoring model based on the data features includes:

[0019] Based on the data labels, the data features are labeled with category labels, the category labels corresponding to the infection events are set as target variables, and all the data features are encoded based on the label encoding and then converted into a numerical sample set;

[0020] A neural network model is constructed by standardizing the numerical sample set and inputting it into the neural network model. The neural network model is trained based on the target variable by setting a training set and a validation set. The hyperparameters of the neural network model are adjusted based on the grid search method, and the adjusted neural network model is set as the monitoring model.

[0021] In conjunction with the first aspect, training the neural network model based on the target variable by setting training and validation sets includes:

[0022] The neural network model includes an input layer, an intermediate layer, and an output layer. The input layer and a portion of the first neurons contained in the intermediate layer are designated as a first training unit, and the output layer and the remaining second neurons contained in the intermediate layer are designated as a second training unit.

[0023] The numerical sample set is divided into the training set, the validation set, and the test set. Training data matching the category label and the label encoding is obtained in the training set. The training data is input into the first training unit to generate a first analysis result. The second training unit receives the first analysis result and generates a second analysis result.

[0024] The second analysis result is compared with the category label based on the target variable, and the model error is calculated based on the loss function. This step is repeated until the model error is less than the first preset value.

[0025] In conjunction with the first aspect, the real-time collection of monitoring data of the patient to be monitored based on the data tags includes:

[0026] Based on the data tags, a sensor device is deployed to collect vital sign data of the patient to be monitored. Based on the operating status of the sensor device, it is determined whether all the vital sign data are valid. If all are valid, the vital sign data is set as the monitoring data. Otherwise, all the vital sign data is divided into a valid dataset and an invalid dataset. The encoder and decoder of the autoencoder are trained using the valid dataset, and the encoder is trained using the invalid dataset. The encoder maps the invalid dataset to the latent space of the valid dataset. The trained encoder encodes and decodes the invalid data in the invalid dataset, recovers the invalid data, and sets it as the monitoring data.

[0027] In conjunction with the first aspect, the generation of the monitoring report includes:

[0028] The monitoring model acquires the feature quantities of the monitoring data, performs predictive analysis on the feature quantities based on the data labels, and outputs the predicted event and the predicted occurrence time of the predicted event;

[0029] Based on the predicted occurrence time, all the predicted events are arranged sequentially to generate the monitoring report.

[0030] In conjunction with the first aspect, a medical knowledge base is constructed, and a medical method corresponding to the monitoring report is retrieved from the medical knowledge base based on a natural language processing algorithm. The feedback measures are then generated based on the medical method.

[0031] Secondly, this application provides an artificial intelligence-based hypothermia infection monitoring system, the artificial intelligence-based hypothermia infection monitoring system comprising:

[0032] The collection module is used to collect comprehensive medical information of surgical patients, set a body temperature range, extract the type and time of infection events from the comprehensive medical information based on the body temperature range, extract first sample data from all the comprehensive medical information based on the type and time of infection, set a time resolution, and generate second sample data corresponding to the first sample data based on the time resolution.

[0033] The training module is used to obtain the data type and data label of the second sample data, preprocess the second sample data based on the data type, extract data features, and train and generate a monitoring model based on the data features.

[0034] The prediction module is used to collect monitoring data of the patients to be monitored in real time according to the data tags, input the monitoring data into the monitoring model after preprocessing the monitoring data, and output the predicted event and the predicted occurrence time of the predicted event.

[0035] The feedback module is used to summarize all the predicted events and the predicted occurrence time, generate a monitoring report, and output feedback measures based on the monitoring report.

[0036] A third aspect of this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the aforementioned artificial intelligence-based low-temperature infection monitoring method.

[0037] The technical solution provided in this application firstly involves collecting and filtering comprehensive medical information to obtain first sample data. This first sample data is then supplemented and expanded based on temporal resolution to generate second sample data, thereby improving the completeness and accuracy of the sample data. Next, a neural network model is trained based on the second sample data. The second analysis result is generated based on the first analysis result until the model error is less than a first preset value, and a monitoring model is generated. This improves the generalization ability and predictive accuracy of the monitoring model while ensuring the security of the second sample data. Finally, the monitoring model performs predictive analysis on the monitoring data of the patients to be monitored, outputs predicted events, and generates monitoring reports and corresponding feedback measures. This improves the accuracy of detecting hypothermia infection in the monitored patients, and enhances the efficiency and reliability of hypothermia infection monitoring.

[0038] This application also uses an automatic encoder to perform validity analysis on vital sign data and generate monitoring data, which can improve the completeness and accuracy of the monitoring data. Attached Figure Description

[0039] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a schematic diagram of one embodiment of the artificial intelligence-based hypothermia infection monitoring method in this application.

[0041] Figure 2 This is a schematic diagram of one embodiment of the neural network model training structure in this application.

[0042] Figure 3 This is a schematic diagram of one embodiment of the detailed process of an automatic encoder processing invalid data in the present application.

[0043] Figure 4 This is a schematic diagram of one embodiment of the artificial intelligence-based hypothermia infection monitoring system in this application. Detailed Implementation

[0044] This application provides an artificial intelligence-based method, system, and storage medium for monitoring hypothermia-related infections. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0045] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the artificial intelligence-based hypothermia infection monitoring method in this application includes:

[0046] Step S101: Collect comprehensive medical information of surgical patients, set body temperature ranges, extract the type and time of infection events from the comprehensive medical information based on the body temperature ranges, extract the first sample data from all comprehensive medical information based on the type and time of infection, set the time resolution, and generate the second sample data corresponding to the first sample data based on the time resolution.

[0047] It is understood that the executing entity of this application can be an AI-based low-temperature infection monitoring device, or it can be a terminal or a server; the specific implementation is not limited here. This application's embodiments use a server as an example for illustration.

[0048] Specifically, comprehensive medical information refers to the data obtained from the hospital's electronic medical record (EMR), surgical record, anesthesia record, and nursing record systems. This includes, but is not limited to, basic patient information (age, gender, weight, height, past medical history, allergy history, etc.), surgery-related information (surgery type, surgery time, anesthesia method, intraoperative medications, etc.), intraoperative monitoring data (body temperature, heart rate, blood pressure, blood oxygen saturation, respiratory rate, etc.), and postoperative nursing records (body temperature changes, wound condition, drainage fluid status, laboratory test results, etc.). Based on clinical experience and literature, a temperature range for hypothermia is determined. For example, a body temperature below 36.0°C is defined as hypothermia, with mild hypothermia defined as 34°C to 36°C, moderate hypothermia as 32°C to 34°C, and severe hypothermia as below 32°C. Intraoperative and postoperative body temperature data are extracted from the comprehensive medical information and categorized according to the defined temperature ranges. An infection event refers to an event with symptoms of infection that occurs from the start to the end of surgery. Based on clinical diagnostic criteria, the type of infection event is determined, such as surgical site infection (SSI), pneumonia, urinary tract infection, etc. Extracting the occurrence time of infection events from comprehensive medical information and matching it with body temperature data can be achieved using semantic recognition algorithms. The first sample data refers to data information containing low body temperature and infection events extracted based on body temperature ranges and infection events; therefore, there are multiple first sample data sets. Time resolution refers to the time scale, such as every hour, every 30 minutes, or every 15 minutes. Because the first sample data may contain data with high time resolution, it can affect the accuracy of the infection event occurrence time. For example, if body temperature data is recorded daily, the influencing factors related to low body temperature data are controversial. Therefore, it is necessary to reduce the time resolution and increase the data information corresponding to the time scale in the first sample data to generate the second sample data. The second sample data supplements and expands upon the first sample data.

[0049] Step S102: Obtain the data type and data label of the second sample data, preprocess the second sample data based on the data type, extract data features, and train and generate a monitoring model based on the data features.

[0050] Specifically, the data type in the second sample data is identified, such as numerical, categorical, or time-based. Data labels refer to the category of any data information in the second sample data, such as body temperature, heart rate, and blood pressure. Preprocessing includes handling missing values, outliers, and duplicate values; for example, filling missing values ​​with the mean or removing outliers. Numerical data is standardized using methods such as Z-score standardization or Min-Max standardization. Based on the data type, features are extracted to generate data features, such as time series features (mean, variance, trend, etc.), statistical features (maximum, minimum, median, etc.), and physiological indicator features (heart rate, blood pressure, blood oxygen saturation, etc.). A suitable machine learning model is selected, such as logistic regression, decision trees, random forests, support vector machines (SVM), or neural networks. The preprocessed second sample data is then input into the model for training to generate the monitoring model.

[0051] Step S103: Collect monitoring data of the patients to be monitored in real time based on data tags, input the preprocessed monitoring data into the monitoring model, and output the predicted events and the predicted occurrence time of the predicted events.

[0052] Specifically, based on data tags, monitoring data such as body temperature, heart rate, and blood pressure of the patients to be monitored are collected in real time. The real-time collected data is preprocessed, such as standardized and normalized. The preprocessed data is then input into a trained monitoring model, which outputs a predicted event (such as whether an infection has occurred) and the predicted time of occurrence.

[0053] Step S104: Summarize all predicted events and their predicted occurrence times, generate a monitoring report, and output feedback measures based on the monitoring report.

[0054] Specifically, all predicted events and their predicted occurrence times are compiled to generate a monitoring report. The report includes patient information, monitoring data, prediction results, and predicted occurrence times. Based on the prediction results and clinical guidelines, corresponding feedback measures are developed, such as adjusting treatment plans, strengthening monitoring, and administering prophylactic medication. These feedback measures are then communicated to healthcare professionals to guide clinical decision-making.

[0055] In one specific embodiment, generating second sample data corresponding to the first sample data based on temporal resolution includes:

[0056] (1) Obtain basic information of surgical patients, and based on the basic information, obtain subsample data with the same category and time resolution higher than day units from all first sample data.

[0057] (2) Extract medical data from the subsample data, obtain specific changes in any medical data, and mark the time information corresponding to the specific changes. Generate estimated events and estimated occurrence times corresponding to the estimated events based on the time information. The specific changes include changes in data values, changes in collection frequency, and changes in data items.

[0058] (3) Add the estimated events and their estimated occurrence times to the first sample data to generate the second sample data.

[0059] Specifically, basic information about surgical patients, especially age, is extracted from the first sample data. Based on clinical experience and research, patients' ages are divided into different ranges, such as children (0-12 years), adolescents (13-18 years), young adults (19-40 years), middle-aged (41-65 years), and elderly (66 years and above). Subsample data is then selected from all the first sample data, identifying data of the same category as the current patients, such as surgical site infections (SSI), pneumonia, and urinary tract infections. Data with a time resolution higher than daily units, such as hourly, 30-minute, or 15-minute data, are selected.

[0060] The selected subsample data is stored in a structured format, ensuring that each data point includes a timestamp and corresponding medical data. Medical data includes surgery-related information, intraoperative monitoring data, and postoperative care records. Specific changes include changes in data values ​​(e.g., sudden increase in body temperature, increased heart rate, decrease in blood pressure), changes in collection frequency (e.g., changes in the frequency of vital sign monitoring, changes in the frequency of drug administration), and changes in data items (e.g., new laboratory tests, or changes in imaging results). Statistical methods (e.g., Z-score) or machine learning methods (e.g., anomaly detection models) can be used to identify changes in data values, compare collection frequencies across different time periods, compare data items at different time points, and identify new or changed data items. Identified specific changes are marked, and the time information of the changes is recorded. The time information of specific changes is used as the estimated occurrence time of estimated events. For example, in the first sample data, at timetamp T1, the body temperature was 36°C, heart rate was 85°C, blood pressure was 118 / 82 mmHg, antibiotics were administered, and the white blood cell count was 10,000. After identifying characteristic changes, a decrease in body temperature was found. The second sample data could then be supplemented as follows: at timetamp T1, body temperature was 36°C, heart rate was 85°C, blood pressure was 118 / 82 mmHg, antibiotics were administered, white blood cell count was 10,000, and a hypothermia event occurred. Through the above steps, the process of generating second sample data based on temporal resolution can effectively capture and label specific changes in medical data. These changes not only enrich the content of the sample data but also provide more comprehensive information for subsequent monitoring model training, thereby improving the accuracy and reliability of hypothermia infection monitoring.

[0061] In one specific embodiment, generating an estimated event and its corresponding estimated occurrence time based on time information includes:

[0062] (1) Establish a medical event database, obtain supplementary events related to infection events from the medical event database, extract related events from the supplementary events based on specific changes, and set the related events as estimated events.

[0063] (2) Output change timestamps based on the change type of a specific change, output lag time based on the change type, and set the change timestamps and lag time as the estimated occurrence time.

[0064] Specifically, a medical event database refers to a database of possible medical symptoms based on medical knowledge, including the correlations between various medical symptoms. Supplementary events are other medical events identified in the medical event database that have a strong correlation with the infection event based on correlations; therefore, there are multiple supplementary events. Among all supplementary events, medical events that only occur due to specific changes are considered associated events; therefore, these associated events are designated as estimated events.

[0065] The specific types of change are the data value changes, sampling frequency changes, and data item changes mentioned above. For data value changes, a first change curve can be predicted, and a change timestamp is generated based on the predicted time of the first change curve. For sampling frequency changes, a change timestamp is generated based on the time interval of the sampling frequency changes. For data item changes, the correlation between all data items can be obtained, and other data items corresponding to the strongest correlation with the data item change can be selected. A second change curve is then generated based on the prediction time of the second change curve, and a change timestamp is generated based on the predicted time of the second change curve. Lag time refers to the time period during which a medical event is delayed after a specific change occurs. For example, if the time period required from a decrease in body temperature to the occurrence of the infection event "coagulation dysfunction" is T2, then the lag time is T2. Therefore, the change timestamp and lag time need to be combined and set as the estimated occurrence time.

[0066] In one specific embodiment, training and generating a monitoring model based on data features includes:

[0067] (1) Label the data features with category labels based on data labels, set the category labels corresponding to the infection events as target variables, and encode all data features based on label encoding and convert them into a numerical sample set.

[0068] (2) Construct a neural network model. After standardizing the numerical sample set, input it into the neural network model. Set the training set and validation set based on the target variable to train the neural network model. Adjust the hyperparameters of the neural network model based on the grid search method. Set the adjusted neural network model as the monitoring model.

[0069] Specifically, category labels are tagging information used to label data features. By setting the category labels corresponding to data features indicating infection events as the target variable, the target variable is multi-category label information. Label encoding is used to encode and transform data features, ensuring that all features are numerical, thus generating a numerical sample set.

[0070] The neural network model can be configured as a multilayer perceptron, effectively improving the accuracy of hypothermia infection monitoring. Hyperparameters refer to the values ​​of parameters such as learning rate, batch size, number of hidden layers, and number of neurons during the training process of the neural network model. The accuracy of the neural network model can be obtained using validation and test sets, thus generating the monitoring model.

[0071] In one specific embodiment, training a neural network model based on setting training and validation sets according to the target variable includes:

[0072] (1) The neural network model includes an input layer, an intermediate layer and an output layer. The first neurons contained in the input layer and part of the intermediate layer are set as the first training unit, and the second neurons contained in the output layer and the remaining intermediate layer are set as the second training unit.

[0073] (2) Divide the numerical sample set into a training set, a validation set and a test set. Obtain training data with category labels and label codes in the training set. Input the training data into the first training unit to generate the first analysis result. The second training unit receives the first analysis result and generates the second analysis result.

[0074] (3) Compare the second analysis results with the category labels based on the target variable, and calculate the model error based on the loss function. Repeat this step until the model error is less than the first preset value.

[0075] Specifically, such as Figure 2 As shown, the input layer L1 and part of the intermediate layer L2 form the first training unit, and the output layer L4 and the remaining intermediate layer L3 form the second training unit. The neural network model includes the first training unit and the second training unit. The first analysis result is output by part of the intermediate layer, and the second analysis result is output by the output layer. The second analysis result is the probability of the category of the infection event corresponding to the target variable.

[0076] By comparing the second analysis result with the category label, the difference between the predicted value and the true value can be calculated. Based on the calculated model error, the weights and biases of the model are adjusted using the backpropagation algorithm. This training process is repeated until the model error is less than the first preset value.

[0077] In one specific embodiment, monitoring data of the patient to be monitored is collected in real time based on data tags, including:

[0078] Sensor devices are deployed based on data tags to collect vital sign data of patients to be monitored. The validity of all vital sign data is determined based on the operating status of the sensor devices. If all data are valid, they are set as monitoring data. Otherwise, all vital sign data are divided into valid and invalid datasets. The encoder and decoder of the autoencoder are trained using the valid dataset, and the encoder is trained using the invalid dataset. The encoder maps the invalid dataset to the latent space of the valid dataset. The trained encoder encodes and decodes the invalid data in the invalid dataset, recovers the invalid data, and sets it as monitoring data.

[0079] Specifically, vital sign data corresponds to data tags. Operating status refers to whether the sensor equipment is operating normally. If the operating status is abnormal, it indicates the presence of invalid vital sign data; otherwise, it means all vital sign data is valid, and the vital sign data is set as monitoring data.

[0080] To improve the effectiveness and accuracy of monitoring data, invalid data needs to be recovered. The specific recovery process is as follows: Figure 3 As shown, an autoencoder is an artificial neural network used for unsupervised learning, primarily for data compression, reconstruction, and dimensionality reduction. It mainly consists of an encoder and a decoder. Real-time acquisition of monitoring data from patients based on data tags, followed by processing invalid data through an autoencoder, effectively ensures the quality and integrity of the monitoring data. This method combines sensor technology and data processing technology to improve the accuracy and reliability of hypothermia infection monitoring.

[0081] In one specific embodiment, generating a monitoring report includes:

[0082] (1) The monitoring model acquires the feature quantities of the monitoring data, and performs predictive analysis on the feature quantities based on the data labels, and outputs the predicted events and the predicted occurrence time of the predicted events.

[0083] (2) Arrange all predicted events in sequence based on the predicted occurrence time to generate a monitoring report.

[0084] Specifically, the extracted features are input into the trained monitoring model, which then outputs predicted events and predicted time points based on probability values. The predicted event point is set as the predicted occurrence time. Here, the predicted event refers to the type of event in which the monitored patient is likely to become infected.

[0085] All predicted events are arranged sequentially based on their predicted occurrence time. Typically, earlier events are listed first, and later events are listed last. Monitoring data and predicted events can be displayed using charts and graphs to generate monitoring reports.

[0086] In one specific embodiment, a medical knowledge base is constructed, and medical methods corresponding to monitoring reports are retrieved from the medical knowledge base based on natural language algorithms. Feedback measures are then generated based on the medical methods.

[0087] Specifically, a medical knowledge base refers to a knowledge base containing infection events and medical diagnostic procedures. Natural Language Processing (NLP) models are a class of computer algorithms and techniques used to understand and process human language. The goal of NLP models is to enable computers to understand, interpret, generate, and manipulate human language. Therefore, using natural language algorithms to filter medical diagnostic procedures from the medical knowledge base constitutes medical methods, and all medical methods are set as feedback measures.

[0088] The above describes the AI-based hypothermia infection monitoring method in the embodiments of this application. The following describes the AI-based hypothermia infection monitoring system in the embodiments of this application. Please refer to [link / reference]. Figure 4 One embodiment of the artificial intelligence-based hypothermia infection monitoring system in this application includes:

[0089] The collection module 201 is used to collect comprehensive medical information of surgical patients, set body temperature ranges, extract the type and time of infection events from the comprehensive medical information based on the body temperature ranges, extract the first sample data from all comprehensive medical information based on the type and time of infection, set the time resolution, and generate the second sample data corresponding to the first sample data based on the time resolution.

[0090] The training module 202 is used to obtain the data type and data label of the second sample data, preprocess the second sample data based on the data type, extract data features, and train and generate a monitoring model based on the data features.

[0091] The prediction module 203 is used to collect monitoring data of the patients to be monitored in real time according to the data tags, input the preprocessed monitoring data into the monitoring model, and output the predicted events and the predicted occurrence time of the predicted events.

[0092] Feedback module 204 is used to summarize all predicted events and their predicted occurrence times, generate a monitoring report, and output feedback measures based on the monitoring report.

[0093] Through the collaborative efforts of the aforementioned components, firstly, a first sample of data is selected by collecting comprehensive medical information. This first sample data is then supplemented and expanded based on temporal resolution to generate a second sample of data, improving the completeness and accuracy of the sample data. Next, a neural network model is trained using the second sample data. Based on the first analysis results, a second analysis result is generated until the model error is less than a first preset value, and a monitoring model is generated. This improves the generalization ability and predictive accuracy of the monitoring model while ensuring the security of the second sample data. Finally, the monitoring model performs predictive analysis on the monitoring data of the patients to be monitored, outputting predicted events and generating monitoring reports and corresponding feedback measures. This improves the accuracy of detecting hypothermia infection in the monitored patients, enhancing the efficiency and reliability of hypothermia infection monitoring.

[0094] This application also uses an automatic encoder to perform validity analysis on vital sign data and generate monitoring data, which can improve the completeness and accuracy of the monitoring data.

[0095] This application also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the artificial intelligence-based low body temperature infection monitoring method.

[0096] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0097] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0098] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for monitoring hypothermia infection based on artificial intelligence, characterized in that, The artificial intelligence-based method for monitoring hypothermia infection includes: Collect comprehensive medical information of surgical patients, set body temperature ranges, extract the type and time of infection events from the comprehensive medical information based on the body temperature ranges, extract first sample data from all the comprehensive medical information based on the type and time of infection, set time resolution, and generate second sample data corresponding to the first sample data based on the time resolution; The process of generating second sample data corresponding to the first sample data based on the time resolution includes: The process involves: acquiring basic information about the surgical patient; obtaining subsample data from all the first sample data that are of the same category and have a time resolution higher than a day; extracting medical data from the subsample data; obtaining specific changes in any of the medical data; marking the time information corresponding to the specific changes; generating an estimated event and an estimated occurrence time corresponding to the estimated event based on the time information; wherein the specific changes include changes in data values, changes in collection frequency, and changes in data items; and adding the estimated event and the estimated occurrence time to the first sample data to generate the second sample data. The process of generating an estimated event and its corresponding estimated occurrence time based on the time information includes: A medical event database is established. Supplementary events associated with the infection event are obtained from the medical event database. Related events are extracted from the supplementary events based on the specific change, and the related events are set as the estimated events. A change timestamp is output based on the change type of the specific change, and a lag time is output based on the change type. The change timestamp and the lag time are combined and set as the estimated occurrence time. Obtain the data type and data label of the second sample data, preprocess the second sample data based on the data type and extract data features, and train and generate a monitoring model based on the data features; Based on the data tags, the monitoring data of the patients to be monitored are collected in real time. After the monitoring data is preprocessed, it is input into the monitoring model and the predicted event and the predicted occurrence time of the predicted event are output. Summarize all predicted events and their predicted occurrence times to generate a monitoring report, and output feedback measures based on the monitoring report.

2. The artificial intelligence-based method for monitoring hypothermia infection according to claim 1, characterized in that, The process of training and generating a monitoring model based on the data features includes: Based on the data labels, the data features are labeled with category labels, the category labels corresponding to the infection events are set as target variables, and all the data features are encoded based on the label encoding and then converted into a numerical sample set; A neural network model is constructed by standardizing the numerical sample set and inputting it into the neural network model. The neural network model is trained based on the target variable by setting a training set and a validation set. The hyperparameters of the neural network model are adjusted based on the grid search method, and the adjusted neural network model is set as the monitoring model.

3. The artificial intelligence-based method for monitoring hypothermia infection according to claim 2, characterized in that, The step of training the neural network model based on the target variable by setting training and validation sets includes: The neural network model includes an input layer, an intermediate layer, and an output layer. The input layer and a portion of the first neurons contained in the intermediate layer are designated as a first training unit, and the output layer and the remaining second neurons contained in the intermediate layer are designated as a second training unit. The numerical sample set is divided into the training set, the validation set, and the test set. Training data matching the category label and the label encoding is obtained in the training set. The training data is input into the first training unit to generate a first analysis result. The second training unit receives the first analysis result and generates a second analysis result. The second analysis result is compared with the category label based on the target variable, and the model error is calculated based on the loss function. This step is repeated until the model error is less than the first preset value.

4. The artificial intelligence-based method for monitoring hypothermia infection according to claim 1, characterized in that, The real-time collection of monitoring data of the patient to be monitored based on the data tags includes: Based on the data tags, a sensor device is deployed to collect vital sign data of the patient to be monitored. Based on the operating status of the sensor device, it is determined whether all the vital sign data are valid. If all are valid, the vital sign data is set as the monitoring data. Otherwise, all the vital sign data is divided into a valid dataset and an invalid dataset. The encoder and decoder of the autoencoder are trained using the valid dataset, and the encoder is trained using the invalid dataset. The encoder maps the invalid dataset to the latent space of the valid dataset. The trained encoder encodes and decodes the invalid data in the invalid dataset, recovers the invalid data, and sets it as the monitoring data.

5. The artificial intelligence-based method for monitoring hypothermia infection according to claim 1, characterized in that, The generation of the monitoring report includes: The monitoring model acquires the feature quantities of the monitoring data, performs predictive analysis on the feature quantities based on the data labels, and outputs the predicted event and the predicted occurrence time of the predicted event; Based on the predicted occurrence time, all the predicted events are arranged sequentially to generate the monitoring report.

6. The artificial intelligence-based method for monitoring hypothermia infection according to claim 5, characterized in that, The feedback measures based on the monitoring report include: constructing a medical knowledge base, retrieving the medical method corresponding to the monitoring report from the medical knowledge base using a natural language algorithm, and generating the feedback measures based on the medical method.

7. A hypothermia infection monitoring system based on artificial intelligence, characterized in that, The AI-based hypothermia infection monitoring system includes: A collection module is used to collect comprehensive medical information of surgical patients, set a body temperature range, extract the type and time of infection events from the comprehensive medical information based on the body temperature range, extract first sample data from all the comprehensive medical information based on the type and time of infection, set a time resolution, and generate second sample data corresponding to the first sample data based on the time resolution; wherein, generating second sample data corresponding to the first sample data based on the time resolution includes: obtaining basic information of the surgical patient, obtaining sub-sample data with the same type and time resolution higher than daily units from all the first sample data based on the basic information; extracting medical data from the sub-sample data, obtaining specific changes in any of the medical data, and marking the time information corresponding to the specific changes, based on... The time information generates an estimated event and its corresponding estimated occurrence time, wherein the specific change includes changes in data value, changes in collection frequency, and changes in data items; the estimated event and its corresponding estimated occurrence time are added to the first sample data to generate the second sample data; wherein generating the estimated event and its corresponding estimated occurrence time based on the time information includes: establishing a medical event database, obtaining supplementary events associated with the infection event from the medical event database, extracting related events from the supplementary events based on the specific change, and setting the related events as the estimated event; outputting a change timestamp based on the change type of the specific change, outputting a lag time based on the change type, and setting the change timestamp and the lag time together as the estimated occurrence time; The training module is used to obtain the data type and data label of the second sample data, preprocess the second sample data based on the data type, extract data features, and train and generate a monitoring model based on the data features. The prediction module is used to collect monitoring data of the patients to be monitored in real time according to the data tags, input the monitoring data into the monitoring model after preprocessing the monitoring data, and output the predicted event and the predicted occurrence time of the predicted event. The feedback module is used to summarize all the predicted events and the predicted occurrence time, generate a monitoring report, and output feedback measures based on the monitoring report.

8. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement the artificial intelligence-based method for monitoring hypothermic infections as described in any one of claims 1-6.

Images