Case management system for Internet hospital

Abstract

The invention relates to a case management system for an internet hospital. The case management system comprises an acquisition processing module, an intelligent analysis module, an augmented reality module and a database construction module. Case data and physiological data of a patient and medical case data of an internet hospital and an external database are collected, and the natural language technology is used for processing and forming multi-mode illness state data. Based on a pre-established intelligent analysis model, diagnosis suggestions and treatment schemes are provided for the multi-mode illness state data. And generating medical display of augmented reality according to the multi-mode illness state data, the diagnosis suggestions and the treatment schemes. And fusing the multi-modal health data, the medical case data and the medical display by utilizing a knowledge graph technology to generate a medical database.

Description

Technical Field

[0001] This invention belongs to the field of medical record management, and in particular relates to a medical record management system for internet hospitals. Background Technology

[0002] With the rapid development of internet technology and the advancement of medical informatization, internet hospitals, as a new model of medical service, have gradually gained widespread attention. Internet hospitals utilize internet technology to transfer medical services from offline to online, providing patients with more convenient and efficient medical care.

[0003] However, internet hospitals still face some challenges in case management, such as the collection, processing, and analysis of case data, as well as issues related to the sharing and privacy protection of medical data. Traditional case management systems typically use paper or electronic medical records to record and manage case data, which suffers from incomplete data collection, low processing efficiency, and weak analytical capabilities. Furthermore, due to the heterogeneous and closed nature of medical data, the sharing and utilization of medical data also face numerous limitations. Summary of the Invention

[0004] This application provides a medical record management system for internet hospitals that can solve the above-mentioned technical problems, as follows:

[0005] In a first aspect, this application provides a medical record management system for internet hospitals, characterized in that the system includes a data acquisition and processing module, an intelligent analysis module, an augmented reality module, and a database construction module:

[0006] The data acquisition and processing module is used for:

[0007] Collect patient medical and physiological data;

[0008] Collect medical case data from internet hospitals and external databases;

[0009] Natural language processing technology is used to process physiological and case data, and match them to form multimodal disease data;

[0010] The intelligent analysis module is used for:

[0011] Based on multimodal disease data, corresponding diagnostic suggestions and treatment plan documents are generated using pre-established intelligent analysis models;

[0012] The augmented reality module is used to generate augmented reality medical displays based on multimodal disease data, diagnostic suggestions, and treatment plans. The medical displays include disease display and treatment process display.

[0013] The database building module is used to generate a medical database by leveraging knowledge graph technology and integrating multimodal health data, medical case data, and medical visualizations.

[0014] In one embodiment, the acquisition and processing module further includes a physiological calculation unit, used for:

[0015] Heart rate, blood pressure, and blood glucose data from physiological data can be processed using the following formula:

[0016] Calculate the dynamic heart rate stress index using the following formula:

[0017]

[0018] Wherein, DHRI represents the dynamic heart rate stress index, SDNN represents the standard deviation of the PR interval, and RMSSD represents the root mean square of the difference between adjacent PR intervals. Indicates individualized dynamic baseline heart rate. This represents the moving average heart rate during that period. This represents the standard deviation of heart rate;

[0019] Calculate the blood pressure trend index using the following formula:

[0020]

[0021] Among them, BPTI represents the blood pressure trend index, SBP represents systolic blood pressure, and DBP represents diastolic blood pressure. This indicates the time interval since the last medication or intervention. Indicates the drug's metabolic half-life. This represents the weighting coefficient of the derivative of heart rate variability. The time derivative representing heart rate variability;

[0022] Calculate blood glucose characteristics using the following formula:

[0023]

[0024] in, Represents the basal metabolic rate constant. This represents a function of carbohydrate intake that varies over time. , where represents apparent volume of distribution, BW represents body weight, FFM represents lean body mass, and AGE represents age. Indicates the intestinal absorption rate. Indicates the sympathetic nerve activation coefficient. This indicates the emergency blood glucose level.

[0025] In one embodiment, the intelligent analysis module further includes a disease monitoring unit for:

[0026] A hybrid LSTM-Transformer model is used to obtain time-series feature data of multimodal health data;

[0027] Construct a health baseline model based on Bayesian optimization, and generate a health status benchmark based on medical case data;

[0028] Based on health status benchmarks and time-series feature data, the patient's condition status is assessed in real time. If the time-series feature data exceeds the health status benchmark, a condition alert is generated.

[0029] In one embodiment, the augmented reality module further includes an AR surgical unit, which is used for:

[0030] AR surgical images are generated based on patient medical records and real-time physiological data.

[0031] By using gesture tracking technology, the control information of surgical instruments is obtained, and corresponding operation image effects are generated based on the control information of surgical instruments and mapped onto AR surgical images;

[0032] Based on the control information of surgical instruments, a deep learning model is used to identify the operation intention, and a medical case is matched according to the operation intention to generate AR operation prompts.

[0033] In one embodiment, the AR surgical module is also used for:

[0034] By utilizing a cross-modal feature pyramid network, hierarchical feature extraction is performed on patient case data and real-time physiological data to generate composite AR surgical images that include multi-level anatomical structures and multi-level physiological states.

[0035] Based on the operational intent and the composite AR surgical image, ray tracing technology is used to dynamically adjust the rendering resolution of different layers of the composite AR surgical image.

[0036] In one embodiment, a cloud computing module is also included, which is used for:

[0037] Obtain task information and determine whether the task type is a diagnostic task or a query task;

[0038] If the task type is a diagnostic task, an intelligent analysis model is used to process the task information and generate diagnostic suggestions and treatment plans.

[0039] If the task type is a query task, relevant information is matched using the medical database, and feedback information is generated based on the relevant information.

[0040] In one embodiment, the cloud computing module is also used for:

[0041] Extract features from diagnostic task information and classify them;

[0042] Based on the features and classification results, a pre-defined diagnostic rule base is invoked for matching analysis;

[0043] Based on the matching analysis results, an intelligent analysis model is used to generate preliminary diagnostic suggestions and a treatment plan framework;

[0044] Based on the initial diagnostic recommendations and treatment plan framework, optimizations and adjustments were made in conjunction with historical data and expert knowledge base;

[0045] Based on the optimized results, a final diagnostic recommendation and treatment plan document is generated.

[0046] In one embodiment, the cloud computing module is also used for:

[0047] Extract key features from the query task;

[0048] Based on key features, match the preset rule base in the medical database;

[0049] Data retrieval algorithms are used to obtain relevant data from medical databases;

[0050] Based on relevant data, an information matching model is used to generate preliminary feedback information;

[0051] By combining historical data from the medical database, the initial feedback information is optimized to generate the final feedback information.

[0052] In one embodiment, the cloud computing module is also used for:

[0053] Obtain the patient's geographic location information;

[0054] Based on the patient's geographic location information, obtain information on the nearest available storage site;

[0055] Based on the information stored at the site, send the patient's relevant medical data.

[0056] Secondly, this application also provides a method for managing medical records in an internet hospital, characterized by comprising the following steps:

[0057] Collect patient medical and physiological data;

[0058] Collect medical case data from internet hospitals and external databases;

[0059] Natural language processing technology is used to process physiological and case data, and match them to form multimodal disease data;

[0060] Based on multimodal disease data, corresponding diagnostic suggestions and treatment plan documents are generated using pre-established intelligent analysis models;

[0061] Augmented reality medical demonstrations are generated based on multimodal disease data, diagnostic suggestions, and treatment plan documents. These demonstrations include both disease presentations and treatment process presentations.

[0062] By leveraging knowledge graph technology, a medical database is generated by integrating multimodal health data, medical case data, and medical visualizations.

[0063] This application discloses a case management system for internet hospitals. Through multimodal data acquisition and fusion technology, it addresses the problems of single data sources and low processing efficiency in traditional internet hospital case management. It integrates patient physiological data, textual medical records, and external medical cases using natural language processing technology to form comprehensive multimodal disease data. Based on an intelligent analysis model, it automatically generates diagnostic suggestions and treatment plans, improving the efficiency and accuracy of medical decision-making. It constructs a structured medical database using knowledge graph technology, breaking down heterogeneous data silos and achieving secure data sharing. Augmented reality technology transforms complex medical information into intuitive and visual disease displays and treatment process simulations, assisting doctors in quickly understanding case characteristics, optimizing diagnostic and treatment collaboration efficiency, and achieving intelligent upgrades to the entire case management process and efficient utilization of medical resources. Attached Figure Description

[0064] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0065] Figure 1 This is a structural diagram of a medical record management system for an internet hospital according to the present invention;

[0066] Figure 2 This is a flowchart of a case management method for an internet hospital according to the present invention. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0068] In one embodiment, such as Figure 1As shown, a case management system for internet hospitals is provided. This embodiment uses system deployment in a smart healthcare scenario as an example for illustration. It can be understood that this system can be applied not only to internet hospitals, but also to regional medical data centers, remote consultation platforms, and other medical information scenarios. The system architecture supports the collaborative application of cross-institutional medical resources. Through the distributed collaboration of cloud servers and edge computing devices, real-time data flow interaction can be achieved between patient physiological monitoring devices, electronic medical record systems, and AR visualization terminals. In this embodiment, the system includes a data acquisition and processing module 101, an intelligent analysis module 102, an augmented reality module 103, and a database construction module 104.

[0069] The data acquisition and processing module 101 is used for:

[0070] Collect patient medical and physiological data;

[0071] Collect medical case data from internet hospitals and external databases;

[0072] Natural language processing technology is used to process physiological and case data, and match them to form multimodal disease data.

[0073] This can be achieved through various channels, including electronic medical record systems, doctor's diagnostic records, and examination reports. This can involve extracting structured or semi-structured text data, such as medical history, medication records, and allergy information. It can also involve collecting dynamic physiological signals from patients, such as heart rate, blood pressure, blood oxygen saturation, and blood glucose levels, using wearable devices like smart bracelets, ECG monitors, or medical instruments like blood pressure monitors and blood glucose meters. Alternatively, it can be obtained by accessing electronic medical records from partner internet hospitals, remote consultation records, authoritative medical databases such as drug instructions, disease guidelines, clinical research literature, and open medical datasets such as disease case libraries and image databases. Furthermore, it can utilize natural language processing (NLP) technology to perform semantic analysis on unstructured information in the case text, such as extracting disease names, symptoms, and examination indicators, establishing logical relationships between entities, identifying emotional tendencies in the patient's complaints, and aligning the parsed text features, such as abnormally high blood glucose levels, with real-time physiological data, such as current blood glucose levels and fluctuation curves, to create a unified patient health profile. The aligned data is then formatted using standardized protocols, and time-series analysis is used to capture dynamic trends in physiological data, such as blood glucose fluctuation patterns. By combining static case characteristics such as age and past medical history, multimodal disease data of the patient's current condition can be generated.

[0074] Intelligent analysis module 102 is used for:

[0075] Based on multimodal disease data, corresponding diagnostic suggestions and treatment plan documents are generated using a pre-established intelligent analysis model.

[0076] This can involve using a hybrid model to capture the global correlation of multimodal disease data such as blood glucose fluctuations and electrocardiogram trends, combining Bayesian-optimized health baseline modeling to dynamically assess individual risk, achieving cross-modal feature weighted fusion through knowledge graphs and attention mechanisms, generating graded diagnostic suggestions such as a disease list sorted by confidence level through anomaly detection, similar case matching, and causal chain deduction during the inference stage, and dynamically adjusting treatment plans based on evidence-based medicine guidelines, historical case databases, and real-time physiological parameters, ultimately outputting a structured document.

[0077] Augmented Reality Module 103 is used to generate augmented reality medical displays based on multimodal disease data, diagnostic suggestions, and treatment plans. The medical displays include disease display and treatment process display.

[0078] The augmented reality module 103 transforms abstract medical data into intuitive 3D visualization scenes, building an interactive bridge connecting virtual information with real-world medical scenarios. Based on multimodal disease data such as patient physiological indicators, imaging data, pathology reports, and diagnostic suggestions from the intelligent analysis module, such as surgical plans and medication dosages, combined with typical case features from a medical case library, it generates an augmented reality display dynamically overlaid on the real medical environment. This can be achieved by spatially aligning anatomical data from CT (Computed Tomography) or MRI (Magnetic Resonance Imaging) scans with real-time electrocardiogram signals, blood oxygen concentration, and other physiological parameters through a cross-modal feature pyramid network, generating a multi-level visualization model that includes vascular distribution, lesion location, and organ function status. For example, when a doctor wears an augmented reality headset, they can directly observe a 3D lesion model suspended above the real human body at the patient's bedside. Its color coding can intuitively reflect the degree of inflammation, such as red representing acute inflammation areas, dynamic arrows indicating abnormal blood flow directions, and even simulating the rhythmic flashing of a heartbeat.

[0079] Database building module 104 is used to generate a medical database by using knowledge graph technology to integrate multimodal health data, medical case data and medical displays.

[0080] This involves using knowledge graph technology to deeply integrate multimodal medical data with external knowledge bases, constructing a structured knowledge network. Natural language processing (NLP) techniques are used to extract medical entities and establish semantic relationships, such as mapping chest pain in a patient's complaint to the medical concept of angina, and associating coronary artery stenosis features in imaging data with anatomical knowledge. Nodes and edges are defined through graph pattern design, storing standardized data in triples, such as patient A - diagnosis result - acute myocardial infarction. This data is then linked to heterogeneous data such as lesion locations marked in augmented reality surgery and treatment plans from historical cases, ultimately forming a medical knowledge base.

[0081] The aforementioned case management system for internet hospitals integrates four core technology modules—multimodal data acquisition, intelligent analysis, augmented reality visualization, and knowledge graph construction—to form a closed-loop solution. It uses natural language processing to fuse patient physiological data, textual medical records, and external medical cases into multimodal disease data, addressing data heterogeneity issues. Real-time analysis of dynamic physiological data, matched with a medical case database, automatically generates personalized diagnostic suggestions and treatment plans, improving the efficiency and accuracy of medical decision-making. Augmented reality technology transforms complex disease data into intuitive 3D visualizations, assisting doctors in understanding case characteristics and optimizing diagnostic and treatment collaboration efficiency. Relying on knowledge graph technology to construct a structured medical database, it enables secure cross-institutional data sharing and intelligent reasoning, reducing the risk of medical misdiagnosis and providing a traceable dynamic knowledge network for medical research, ultimately improving the overall quality of medical services and resource utilization efficiency of internet hospitals.

[0082] In one embodiment, the acquisition and processing module 101 further includes a physiological calculation unit, used for:

[0083] Heart rate, blood pressure, and blood glucose data from physiological data can be processed using the following formula:

[0084] Calculate the dynamic heart rate stress index using the following formula:

[0085]

[0086] Wherein, DHRI represents the dynamic heart rate stress index, SDNN represents the standard deviation of the PR interval, and RMSSD represents the root mean square of the difference between adjacent PR intervals. Indicates individualized dynamic baseline heart rate. This represents the moving average heart rate during that period. This represents the standard deviation of heart rate;

[0087] Calculate the blood pressure trend index using the following formula:

[0088]

[0089] Among them, BPTI represents the blood pressure trend index, SBP represents systolic blood pressure, and DBP represents diastolic blood pressure. This indicates the time interval since the last medication or intervention. Indicates the drug's metabolic half-life. This represents the weighting coefficient of the derivative of heart rate variability. The time derivative representing heart rate variability;

[0090] Calculate blood glucose characteristics using the following formula:

[0091]

[0092] in, Represents the basal metabolic rate constant. This represents a function of carbohydrate intake that varies over time. , where represents apparent volume of distribution, BW represents body weight, FFM represents lean body mass, and AGE represents age. Indicates the intestinal absorption rate. Indicates the sympathetic nerve activation coefficient. This indicates the emergency blood glucose level.

[0093] For example, three mathematical models are used to dynamically analyze patients' physiological data, calculate the ratio of heart rate variability (HRV) to baseline heart rate, quantify cardiac autonomic tone, comprehensively assess dynamic changes in blood pressure and the effects of drug intervention, and predict dynamic blood glucose trajectories based on a physiological pharmacokinetic (PBPK) model. By quantifying cardiovascular stress, drug efficacy, and blood glucose metabolism characteristics, precise risk assessment criteria are provided for the intelligent analysis module. Mathematical modeling transforms physiological data into interpretable medical indicators, addressing the problems of low data utilization and delayed risk warnings in traditional case management systems.

[0094] In one embodiment, the intelligent analysis module further includes a disease monitoring unit for:

[0095] A hybrid LSTM-Transformer model is used to obtain time-series feature data of multimodal health data;

[0096] Construct a health baseline model based on Bayesian optimization, and generate a health status benchmark based on medical case data;

[0097] Based on health status benchmarks and time-series feature data, the patient's condition status is assessed in real time. If the time-series feature data exceeds the health status benchmark, a condition alert is generated.

[0098] Specifically, an LSTM-Transformer hybrid model (temporal feature extraction) is used to capture the long-term temporal dependence of physiological data, such as the blood glucose fluctuation pattern over several weeks. The self-attention mechanism of Transformer is used to model the global correlation of multi-dimensional data such as heart rate, blood pressure, and blood oxygen. Based on the output features of dynamic physiological data such as heart rate per minute, systolic blood pressure, blood glucose concentration, and real-time symptom records such as chest pain intensity and medication logs, a probability distribution model is constructed based on medical case databases, such as data of similar patient groups and guideline recommended thresholds. This model describes the normal fluctuation range of physiological indicators in a healthy state. Bayesian inference can be used to continuously optimize the parameters of the baseline model, such as mean and variance, by combining individual data of newly admitted patients, such as age, medical history, and real-time monitoring results. This enables dynamic calibration of personalized health benchmarks. The current temporal features, such as real-time blood glucose values ​​and ECG ST segment (reflecting the potential changes of ventricular myocardium) offset, are compared with the personalized health baseline. If an indicator exceeds the baseline confidence interval, such as systolic blood pressure > 140 mmHg for 30 minutes, an alarm is triggered.

[0099] In one embodiment, the augmented reality module 103 further includes an AR surgical unit, which is used for:

[0100] AR surgical images are generated based on patient medical records and real-time physiological data.

[0101] By using gesture tracking technology, the control information of surgical instruments is obtained, and corresponding operation image effects are generated based on the control information of surgical instruments and mapped onto AR surgical images;

[0102] Based on the control information of surgical instruments, a deep learning model is used to identify the operation intention, and a medical case is matched according to the operation intention to generate AR operation prompts.

[0103] For example, augmented reality technology can be used to overlay three-dimensional models of the patient's organs, blood vessels, and lesions onto the surgeon's field of vision by integrating patient medical data such as CT or MRI images, pathology reports, and real-time physiological data such as heart rate, blood pressure, and blood oxygen saturation. In cardiac surgery, a model of a beating heart can be displayed, with red areas marking ischemic sites; in tumor resection, the tumor boundary and its position relative to surrounding nerves and blood vessels can be highlighted. The model's state can be adjusted based on real-time physiological data such as blood flow velocity and blood pressure fluctuations to simulate the immediate effects after surgical intervention.

[0104] The system recognizes doctors' hand gestures, such as grasping, moving, and cutting, using optical cameras or sensors. It synchronously maps real instruments in the doctor's hands, such as scalpels and forceps, to virtual surgical tools. For example, if the doctor holds an AR (Augmented Reality) handle, the system automatically matches a virtual ultrasonic scalpel and displays its movement trajectory. As the gesture zooms in, the magnified details of blood vessels in the AR image are updated synchronously. The operation results are rendered in the AR interface, such as hemodynamic changes at the cutting site and tissue deformation effects. Deep learning models analyze the gesture trajectory and instrument movement patterns to determine the doctor's intentions, such as recognizing the sequence of actions to separate tissue along blood vessels, distinguishing the differences between exploring lymph nodes and avoiding the ureter, and retrieving similar surgical cases from a medical case database based on the intention, such as minimally invasive resection plans for similar tumors. Key operational steps and risk avoidance strategies are extracted, and intelligent prompts are overlaid on the AR image, such as generating arrows indicating that the cutting force should be reduced to avoid nerve damage, and generating pop-up windows indicating that the current operation has deviated from the safe area and suggesting angle adjustments based on case studies.

[0105] In one embodiment, the AR surgical module is also used for:

[0106] By utilizing a cross-modal feature pyramid network, hierarchical feature extraction is performed on patient case data and real-time physiological data to generate composite AR surgical images that include multi-level anatomical structures and multi-level physiological states.

[0107] Based on the operational intent and the composite AR surgical image, ray tracing technology is used to dynamically adjust the rendering resolution of different layers of the composite AR surgical image.

[0108] Specifically, the system inputs patient case data such as preoperative CT or MRI images, pathological sections, and real-time physiological data such as hemodynamic parameters and electrophysiological signals. A multi-branch network extracts spatial information such as 3D vascular topology, organ contours, and tumor boundaries; and dynamic features such as blood flow velocity fields, tissue metabolic activity, and inflammatory response areas. A multi-scale feature pyramid is constructed, with low-level networks focusing on local details such as microcalcifications in the vessel wall, and high-level networks capturing global relationships such as the spatial relationship between the lesion and surrounding nerve bundles. This allows for the fusion of anatomical structures and physiological features to generate AR images containing the following levels: bottom layer, high-resolution anatomical structures such as millimeter-level vascular branches; middle layer, visualization of dynamic physiological parameters such as oxyhemoglobin concentration distribution; and top layer, surgical safety warning areas such as flashing high-risk bleeding zones. Gesture tracking devices such as optical cameras and IMU (Inertial Measurement Unit) sensors capture the doctor's actions in real time, such as instrument movement direction and zoom of the focused area. Combined with a semantic segmentation model, the system identifies the intention behind the operation, such as probing the tumor edge or avoiding the optic nerve. The rendering strategy is dynamically adjusted based on the user's intent. When a doctor zooms in to observe a certain anatomical structure, ray-traced specular rendering is enabled for that area, such as displaying the refraction effect of blood flow in blood vessels in real time. For areas far from the user's field of view, the texture resolution is reduced, and only key anatomical markers are retained, such as the outline of the heart, while simplifying the details of the myocardial texture.

[0109] Physiological data such as blood pressure fluctuations and myocardial contractility are mapped to the physical properties of virtual tissues. Dynamic effects are simulated through ray tracing. When blood pressure rises, the light transmittance of the blood vessel wall is adjusted in real time according to the pressure change. During tumor resection, tissue deformation and bleeding diffusion paths are simulated.

[0110] In one embodiment, a cloud computing module 105 is also included, which is used for:

[0111] Obtain task information and determine whether the task type is a diagnostic task or a query task;

[0112] If the task type is a diagnostic task, an intelligent analysis model is used to process the task information and generate diagnostic suggestions and treatment plans.

[0113] If the task type is a query task, relevant information is matched using the medical database, and feedback information is generated based on the relevant information.

[0114] For example, the system receives task requests from terminal devices such as doctor workstations and mobile apps, and parses the task objectives, such as generating diagnostic suggestions or querying medication guidelines for a specific disease. If the task involves patient condition analysis, such as determining whether a patient has a myocardial infarction based on electrocardiogram data and symptoms, the data is forwarded to the intelligent analysis model. If the task is information retrieval, such as finding a list of the best antihypertensive drugs for hypertensive patients, the medical database is directly accessed. Patient multimodal data is input into a pre-trained intelligent analysis model to generate diagnostic and treatment suggestions. Natural language processing is used to parse the query semantics, such as dietary restrictions for diabetic patients, and relevant data is matched in the medical database. This can be done by directly extracting recommended items from guidelines, such as daily carbohydrate intake ≤200g, or by retrieving relevant research abstracts from a literature database, such as the impact of low-carbohydrate diets on blood sugar control. The search results are then integrated to generate feedback information.

[0115] In one embodiment, the cloud computing module 105 is further configured to:

[0116] Extract features from diagnostic task information and classify them;

[0117] Based on the features and classification results, a pre-defined diagnostic rule base is invoked for matching analysis;

[0118] Based on the matching analysis results, an intelligent analysis model is used to generate preliminary diagnostic suggestions and a treatment plan framework;

[0119] Based on the initial diagnostic recommendations and treatment plan framework, optimizations and adjustments were made in conjunction with historical data and expert knowledge base;

[0120] Based on the optimized results, a final diagnostic recommendation and treatment plan document is generated.

[0121] Specifically, the system receives multimodal patient data and uses pre-trained classification algorithms such as random forests and CNNs (Convolutional Neural Networks) to label features. For example, if the primary diagnosis is acute myocardial infarction, the complication is cardiogenic shock, and the comorbidity is type 2 diabetes, a rule base is established based on clinical guidelines, evidence-based medicine, and expert experience. For instance, if a patient meets the criteria of chest pain + ST-segment elevation + elevated troponin, the STEMI (ST-segment elevation myocardial infarction) diagnostic pathway is triggered. If the patient also has diabetes, the risk of microvascular complications should be assessed first. The extracted features are logically matched with the conditions in the rule base to filter out diagnostic pathways that meet the conditions. If the patient's features completely match the STEMI rules, the corresponding diagnostic suggestion is triggered. The classification results are input into an intelligent analysis model, combined with the disease association network in the knowledge graph, to generate a diagnostic conclusion. For example, if the primary diagnosis is acute ST-segment elevation myocardial infarction, the differential diagnosis excludes aortic dissection, and emergency PCI (percutaneous coronary intervention) + dual antiplatelet therapy is recommended. By reviewing the patient's medical history and combining it with the current examination results, diagnostic weights are adjusted. If the patient has previously had a stent implanted, the risk of in-stent restenosis needs to be additionally assessed. Expert-annotated cases are incorporated through federated learning to optimize treatment priorities; for example, the original recommendation to get out of bed 24 hours post-surgery is adjusted to 48 hours of absolute bed rest. The optimized results are then converted into standardized document output.

[0122] In one embodiment, the cloud computing module 105 is further configured to:

[0123] Extract key features from the query task;

[0124] Based on key features, match the preset rule base in the medical database;

[0125] Data retrieval algorithms are used to obtain relevant data from medical databases;

[0126] Based on relevant data, an information matching model is used to generate preliminary feedback information;

[0127] By combining historical data from the medical database, the initial feedback information is optimized to generate the final feedback information.

[0128] For example, when receiving a query request, such as dietary restrictions for diabetic patients, natural language processing is used to identify keywords and, combined with context, determine the query intent, such as whether the restriction refers to food type or lifestyle habits. Structured feature labels are then generated, such as: Disease: Diabetes, Category: Diet, Management Attribute: Restriction. A rule base is constructed based on clinical guidelines, authoritative medical literature, and expert consensus. For example, diabetic patients should follow the "low GI diet" principle; hypertensive patients should limit sodium intake. Rules most relevant to the query features are selected using keyword and semantic similarity algorithms. An efficient search engine quickly locates relevant data in medical databases, including structured data (recommendations from guidelines, such as daily carbohydrate intake ≤200g) and unstructured data (research conclusions from literature, such as the Mediterranean diet reducing the risk of diabetic complications). Data is ranked according to authority (guidelines > expert consensus > general literature) and timeliness. An information extraction model extracts key information from the search results to generate structured feedback. The feedback content is adjusted by incorporating patient history data, such as allergy history and medication records. For example, if a patient is allergic to sulfonamides, the feedback should indicate the need to avoid sulfonylurea hypoglycemic drugs. Expert-annotated cases are introduced through federated learning to optimize recommendation priorities and generate final feedback information.

[0129] In one embodiment, the cloud computing module 105 is further configured to:

[0130] Obtain the patient's geographic location information;

[0131] Based on the patient's geographic location information, obtain information on the nearest available storage site;

[0132] Based on the information stored at the site, send the patient's relevant medical data.

[0133] For example, GPS (Global Positioning System) coordinates can be obtained in real time from wearable devices worn by the patient, such as smartwatches, medical terminals such as monitors, or mobile apps, or the registered address can be extracted from the registration information of the hospital information system. Medical data is distributed and stored across multiple edge computing nodes or regional data centers globally, with the storage capacity, bandwidth, and real-time load status of each site recorded in a cloud directory service. A weighted nearest neighbor search algorithm is used to calculate the optimal site by considering the following factors: the straight-line distance or travel time between the patient's location and the site; the estimated transmission speed based on Ping value (the time required for a data packet to be sent from one computer to another and return) or historical transmission latency data; avoiding sending data to sites that are already full; and prioritizing sites that comply with the data storage regulations of the patient's country or region.

[0134] This application provides a case management system for internet hospitals. By deeply integrating multimodal data acquisition, intelligent analysis, augmented reality, and knowledge graph technologies, it constructs a closed-loop medical management system covering data integration, dynamic decision-making, visual interaction, and secure storage, achieving an intelligent upgrade of case management in internet hospitals. Natural language processing technology breaks down barriers between heterogeneous data such as text, images, and physiological signals. Combined with a dynamic physiological indicator calculation model, fragmented medical information is transformed into unified multimodal disease data, providing a structured foundation for precision diagnosis and treatment. Based on an LSTM-Transformer hybrid model and Bayesian optimization of the health baseline, dynamic assessment of individualized patient risk is achieved. Graded diagnostic suggestions are generated through anomaly detection and causal reasoning. Treatment plans are optimized using a cloud-based rule base and historical data, significantly improving response speed and decision-making rigor in acute and critical care scenarios. The AR surgical unit, through a cross-modal feature pyramid network and ray tracing technology, visualizes anatomical structures, physiological states, and surgical intentions in three dimensions, supporting real-time intraoperative navigation, risk warnings, and the reuse of expert experience, reducing operational complexity and improving surgical accuracy. By building a medical knowledge base based on knowledge graphs and combining edge computing and blockchain technologies to achieve dynamic data storage and secure transmission, and optimizing storage scheduling through a weighted nearest neighbor algorithm to balance data availability and compliance, a reliable foundation is provided for regional medical collaboration and scientific research analysis. From real-time physiological data alerts to intelligent generation of treatment plans, and then to AR surgical simulation and iterative database optimization, a closed-loop management system covering the entire lifecycle of prevention, diagnosis, treatment, and research is formed, effectively reducing the rate of human error, shortening decision-making time, and promoting the efficient allocation of medical resources.

[0135] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0136] Based on the same inventive concept, this application also provides a method for implementing the aforementioned medical record management system for an internet hospital. The solution provided by this method is similar to the implementation described in the above system. Therefore, the specific limitations of one or more embodiments of the medical record management method for an internet hospital provided below can be found in the above-described limitations of the medical record management system for an internet hospital, and will not be repeated here.

[0137] In one exemplary embodiment, such as Figure 2 As shown, a method for managing medical records in an internet hospital is provided, characterized by the following steps:

[0138] S1, collect patient medical and physiological data;

[0139] S2 collects medical case data from internet hospitals and external databases;

[0140] S3 utilizes natural language processing technology to process physiological and case data, matching them to form multimodal disease data;

[0141] S4, based on multimodal disease data, uses a pre-established intelligent analysis model to generate corresponding diagnostic suggestions and treatment plan documents;

[0142] S5 generates augmented reality medical demonstrations based on multimodal disease data, diagnostic suggestions, and treatment plan documents. The medical demonstrations include disease display and treatment process display.

[0143] S6 utilizes knowledge graph technology to integrate multimodal health data, medical case data, and medical visualizations to generate a medical database.

[0144] In one embodiment, the method further includes:

[0145] Heart rate, blood pressure, and blood glucose data from physiological data can be processed using the following formula:

[0146] Calculate the dynamic heart rate stress index using the following formula:

[0147]

[0148] Wherein, DHRI represents the dynamic heart rate stress index, SDNN represents the standard deviation of the PR interval, and RMSSD represents the root mean square of the difference between adjacent PR intervals. Indicates individualized dynamic baseline heart rate. This represents the moving average heart rate during that period. This represents the standard deviation of heart rate;

[0149] Calculate the blood pressure trend index using the following formula:

[0150]

[0151] Among them, BPTI represents the blood pressure trend index, SBP represents systolic blood pressure, and DBP represents diastolic blood pressure. This indicates the time interval since the last medication or intervention. Indicates the drug's metabolic half-life. This represents the weighting coefficient of the derivative of heart rate variability. The time derivative representing heart rate variability;

[0152] Calculate blood glucose characteristics using the following formula:

[0153]

[0154] in, Represents the basal metabolic rate constant. This represents a function of carbohydrate intake that varies over time. , where represents apparent volume of distribution, BW represents body weight, FFM represents lean body mass, and AGE represents age. Indicates the intestinal absorption rate. Indicates the sympathetic nerve activation coefficient. This indicates the emergency blood glucose level.

[0155] In one embodiment, the method further includes:

[0156] A hybrid LSTM-Transformer model is used to obtain time-series feature data of multimodal health data;

[0157] Construct a health baseline model based on Bayesian optimization, and generate a health status benchmark based on medical case data;

[0158] Based on health status benchmarks and time-series feature data, the patient's condition status is assessed in real time. If the time-series feature data exceeds the health status benchmark, a condition alert is generated.

[0159] In one embodiment, the method further includes:

[0160] AR surgical images are generated based on patient medical records and real-time physiological data.

[0161] By using gesture tracking technology, the control information of surgical instruments is obtained, and corresponding operation image effects are generated based on the control information of surgical instruments and mapped onto AR surgical images;

[0162] Based on the control information of surgical instruments, a deep learning model is used to identify the operation intention, and a medical case is matched according to the operation intention to generate AR operation prompts.

[0163] In one embodiment, the method further includes:

[0164] By utilizing a cross-modal feature pyramid network, hierarchical feature extraction is performed on patient case data and real-time physiological data to generate composite AR surgical images that include multi-level anatomical structures and multi-level physiological states.

[0165] Based on the operational intent and the composite AR surgical image, ray tracing technology is used to dynamically adjust the rendering resolution of different layers of the composite AR surgical image.

[0166] In one embodiment, the method further includes:

[0167] Obtain task information and determine whether the task type is a diagnostic task or a query task;

[0168] If the task type is a diagnostic task, an intelligent analysis model is used to process the task information and generate diagnostic suggestions and treatment plans.

[0169] If the task type is a query task, relevant information is matched using the medical database, and feedback information is generated based on the relevant information.

[0170] In one embodiment, the method further includes:

[0171] Extract features from diagnostic task information and classify them;

[0172] Based on the features and classification results, a pre-defined diagnostic rule base is invoked for matching analysis;

[0173] Based on the matching analysis results, an intelligent analysis model is used to generate preliminary diagnostic suggestions and a treatment plan framework;

[0174] Based on the initial diagnostic recommendations and treatment plan framework, optimizations and adjustments were made in conjunction with historical data and expert knowledge base;

[0175] Based on the optimized results, a final diagnostic recommendation and treatment plan document is generated.

[0176] In one embodiment, the method further includes:

[0177] Extract key features from the query task;

[0178] Based on key features, match the preset rule base in the medical database;

[0179] Data retrieval algorithms are used to obtain relevant data from medical databases;

[0180] Based on relevant data, an information matching model is used to generate preliminary feedback information;

[0181] By combining historical data from the medical database, the initial feedback information is optimized to generate the final feedback information.

[0182] In one embodiment, the method further includes:

[0183] Obtain the patient's geographic location information;

[0184] Based on the patient's geographic location information, obtain information on the nearest available storage site;

[0185] Based on the information stored at the site, send the patient's relevant medical data.

[0186] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of a case management method for an internet hospital as described above.

[0187] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0188] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0189] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A medical record management system for internet hospitals, characterized in that, The system includes a data acquisition and processing module, an intelligent analysis module, an augmented reality module, and a database construction module. The acquisition and processing module is used for: Collect patient medical and physiological data; Collect medical case data from internet hospitals and external databases; The physiological data and the case data are processed using natural language processing technology to match and form multimodal disease data; The intelligent analysis module is used for: Based on the multimodal disease data, corresponding diagnostic suggestions and treatment plan documents are generated using a pre-established intelligent analysis model; The augmented reality module is used to generate an augmented reality medical display based on the multimodal disease data, diagnostic suggestions, and treatment plans. The medical display includes a disease display and a treatment process display. The database construction module is used to generate a medical database by integrating the multimodal health data, the medical case data, and the medical display using knowledge graph technology.

2. The system according to claim 1, characterized in that, The data acquisition and processing module also includes a physiological calculation unit, used for: The heart rate, blood pressure, and blood glucose data in the physiological data are processed using the following formula: Calculate the dynamic heart rate stress index using the following formula: Wherein, DHRI represents the dynamic heart rate stress index, SDNN represents the standard deviation of the PR interval, and RMSSD represents the root mean square of the difference between adjacent PR intervals. Indicates individualized dynamic baseline heart rate. This represents the moving average heart rate during that period. This represents the standard deviation of heart rate; Calculate the blood pressure trend index using the following formula: Among them, BPTI represents the blood pressure trend index, SBP represents systolic blood pressure, and DBP represents diastolic blood pressure. This indicates the time interval since the last medication or intervention. Indicates the drug's metabolic half-life. This represents the weighting coefficient of the derivative of heart rate variability. The time derivative representing heart rate variability; Calculate blood glucose characteristics using the following formula: in, Represents the basal metabolic rate constant. This represents a function of carbohydrate intake that varies over time. , where represents apparent volume of distribution, BW represents body weight, FFM represents lean body mass, and AGE represents age. Indicates the intestinal absorption rate. Indicates the sympathetic nerve activation coefficient. This indicates the emergency blood glucose level.

3. The system according to claim 2, characterized in that, The intelligent analysis module also includes a disease monitoring unit, used for: The LSTM-Transformer hybrid model is used to obtain the temporal feature data of the multimodal health data; Construct a health baseline model based on Bayesian optimization, and generate a health status benchmark based on the medical case data; Based on the health status benchmark and the time-series feature data, the patient's condition status is judged in real time. If the time-series feature data exceeds the health status benchmark, a condition alarm message is generated.

4. The system according to claim 1, characterized in that, The augmented reality module also includes an AR surgical unit, which is used for: AR surgical images are generated based on patient medical records and real-time physiological data. Using gesture tracking technology, the control information of surgical instruments is obtained, and corresponding operation image effects are generated based on the control information of the surgical instruments and mapped onto the AR surgical image; Based on the control information of the surgical instruments, a deep learning model is used to identify the operation intention, and a medical case is matched according to the operation intention to generate AR operation prompts based on the medical case.

5. The system according to claim 4, characterized in that, The AR surgical module is also used for: By utilizing a cross-modal feature pyramid network, hierarchical feature extraction is performed on patient case data and real-time physiological data to generate composite AR surgical images that include multi-level anatomical structures and multi-level physiological states. Based on the stated operational intent and the composite AR surgical image, ray tracing technology is used to dynamically adjust the rendering resolution of different layers of the composite AR surgical image.

6. The system according to claim 1, characterized in that, It also includes a cloud computing module, which is used for: Obtain task information and determine whether the task type is a diagnostic task or a query task; If the task type is a diagnostic task, the intelligent analysis model is used to process the task information and generate diagnostic suggestions and treatment plans. If the task type is a query task, relevant information is matched using the medical database, and feedback information is generated based on the relevant information.

7. The system according to claim 6, characterized in that, The cloud computing module is also used for: Extract features from diagnostic task information and classify them; Based on the features and classification results, a preset diagnostic rule base is invoked for matching analysis; Based on the matching analysis results, the intelligent analysis model is used to generate preliminary diagnostic suggestions and a treatment plan framework. Based on the preliminary diagnostic recommendations and treatment plan framework, optimizations and adjustments were made in conjunction with historical data and expert knowledge base; Based on the optimized results, a final diagnostic recommendation and treatment plan document is generated.

8. The system according to claim 7, characterized in that, The cloud computing module is also used for: Extract the key features from the query task; Based on the key features, match the preset rule base in the medical database; Data retrieval algorithms are used to obtain relevant data from the medical database; Based on the relevant data, preliminary feedback information is generated using an information matching model; By combining historical data from the medical database, the preliminary feedback information is optimized to generate the final feedback information.

9. The system according to claim 6, characterized in that, The cloud computing module is also used for: Obtain the patient's geographic location information; Based on the patient's geographic location information, obtain information on the nearest available storage site; Based on the information from the storage site, the patient's relevant medical data is sent.

10. A method for managing medical records in an internet hospital, characterized in that, Includes the following steps: Collect patient medical and physiological data; Collect medical case data from internet hospitals and external databases; The physiological data and the case data are processed using natural language processing technology to match and form multimodal disease data; Based on the multimodal disease data, corresponding diagnostic suggestions and treatment plan documents are generated using a pre-established intelligent analysis model; Augmented reality medical demonstrations are generated based on the multimodal disease data, diagnostic suggestions, and treatment plan documents. These demonstrations include a display of the disease condition and a display of the treatment process. A medical database is generated by integrating the multimodal health data, the medical case data, and the medical demonstrations using knowledge graph technology.

Images