A ward inspection data analysis method and intelligent ward inspection system

Abstract

The present invention relates to the field of data processing, and in particular to a ward inspection data analysis method and an intelligent ward inspection system. The method comprises the following steps: obtaining real-time patient status data and ward inspection monitoring video; obtaining individual physiological characteristic data based on real-time patient status data; performing multi-period analysis on individual physiological characteristic data to generate physiological characteristic change trend data; performing sliding time window division on physiological characteristic change trend data to obtain a short-term change trend sequence; performing feature mutation identification on the short-term change trend sequence to obtain change trend feature points; performing dynamic evolution analysis on physiological characteristic change trend data based on the change trend feature points to generate physiological characteristic evolution trend laws; performing trend prediction based on physiological characteristic evolution trend laws to obtain individual physiological characteristic prediction data; the present invention realizes efficient and accurate ward inspection data analysis.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a ward inspection data analysis method and an intelligent ward inspection system. Background Art

[0002] With the continuous improvement of the intelligence of medical equipment and the level of medical informatization, the analysis and monitoring of ward inspection data has become increasingly important for improving medical quality, patient safety and medical efficiency. However, in the traditional ward inspection process, medical staff need to regularly inspect the ward and record the patient's physiological indicators, equipment operation status and other information. However, due to the limitations of manual inspections, potential problems or abnormal conditions cannot be discovered in time, and there are problems of inaccurate analysis and low efficiency. In order to meet the needs of modern medical environments for ward inspection data analysis, an intelligent method is needed to automatically analyze and process ward inspection data. Summary of the invention

[0003] The present invention provides a ward inspection data analysis method and an intelligent ward inspection system to solve at least one of the above technical problems.

[0004] To achieve the above object, the present invention provides a ward inspection data analysis method, comprising the following steps:

[0005] Step S1: Acquire real-time patient status data and ward patrol monitoring video; obtain individual physiological characteristic data based on the real-time patient status data; perform multi-period analysis on the individual physiological characteristic data to generate physiological characteristic change trend data;

[0006] Step S2: Divide the physiological characteristic change trend data into sliding time windows to obtain a short-term change trend sequence; perform feature mutation identification on the short-term change trend sequence to obtain a change trend feature point;

[0007] Step S3: performing dynamic evolution analysis on the physiological characteristic change trend data based on the change trend feature points, thereby generating the physiological characteristic evolution trend law; performing trend prediction based on the physiological characteristic evolution trend law, thereby obtaining individual physiological characteristic prediction data;

[0008] Step S4: performing object detection on the ward patrol monitoring video to obtain ward object data; performing three-dimensional spatial structure analysis on the ward patrol monitoring video to obtain three-dimensional spatial structure data; performing three-dimensional point cloud modeling based on the three-dimensional spatial structure data and the ward object data to construct a holographic ward model;

[0009] Step S5: performing abnormality detection on the individual physiological characteristic prediction data to generate abnormal physiological characteristic values; performing decision analysis on the abnormal physiological characteristic values ​​to obtain an abnormal characteristic decision strategy;

[0010] Step S6: Based on the abnormal physiological characteristic data, the holographic ward space model is segmented into bounding boxes to obtain bed bounding box data; based on the bed bounding box data, early warning processing is performed to generate an early warning signal; the abnormal characteristic decision strategy and the early warning signal are uploaded to the intelligent ward patrol system to perform ward patrol analysis operations.

[0011] The present invention obtains real-time patient status data and monitoring videos to understand the patient's physiological condition and behavioral activities in real time. Based on the real-time patient status data, the present invention extracts individual physiological characteristic data, including heart rate, blood pressure, etc., to quantify and analyze the patient's physiological state. Multi-period analysis reveals the changing trend of the patient's physiological characteristics, which is helpful to understand the patient's health status and disease changes. Through sliding time window division and feature mutation identification, the physiological characteristic change trend data is divided into short-term change trend sequences, and important change trend feature points are identified to help capture the mutation or important change of physiological characteristics, which is helpful to timely discover potential abnormal conditions or disease deterioration. Through the dynamic evolution analysis of the change trend feature points, the evolution law and trend of the physiological characteristics are revealed, which is helpful to understand the patient's physiological state development, predict the patient's future physiological characteristic change trend, and provide individual physiological characteristic prediction data to make corresponding treatment adjustments or intervention measures. Through object detection and three-dimensional spatial structure of ward patrol monitoring videos Analyze and obtain the object data and three-dimensional spatial structure information in the ward. Based on these data, three-dimensional point cloud modeling is performed to build a holographic ward model. Such a model provides comprehensive and three-dimensional ward environment information, which helps to better understand the patient's behavior and the layout of the ward. By performing abnormal detection on the individual physiological characteristic prediction data, abnormal physiological characteristic values ​​are identified, and potential abnormal conditions or changes in the condition are identified, which helps to timely discover the patient's abnormal state and take corresponding measures. Decision analysis is performed on the abnormal physiological characteristic values, and appropriate abnormal characteristic decision strategies are formulated according to specific circumstances, such as triggering alarms and notifying medical staff. By performing bounding box segmentation on the abnormal physiological characteristic data, the bed bounding box data, that is, the specific location of the patient, is accurately located and extracted. Early warning processing is performed based on the bed bounding box data, and corresponding early warning signals are generated according to the abnormal characteristic decision strategy to remind monitoring personnel or medical staff to pay attention to specific beds or patients, realize intelligent ward patrol analysis, and improve the efficiency of monitoring and response.

[0012] Preferably, step S1 comprises the following steps:

[0013] Step S11: Acquire real-time patient status data and ward patrol monitoring video;

[0014] Step S12: performing individual physiological characteristic analysis on the real-time patient status data to obtain individual physiological characteristic data;

[0015] Step S13: performing multi-period analysis on the individual physiological characteristic data to generate time series physiological characteristic data;

[0016] Step S14: performing time series change trend analysis on the time series physiological characteristic data to generate physiological characteristic change trend data.

[0017] The present invention provides detailed information about the patient's physiological state and behavioral activities by acquiring real-time patient status data and ward patrol monitoring videos. The real-time patient status data includes physiological indicators such as heart rate, blood pressure, and body temperature, while the monitoring video provides the patient's activity trajectory and behavioral pattern in the ward. The acquisition of these data helps to fully understand the patient's health status and behavioral habits, and provides a basis for subsequent analysis and decision-making. By performing individual physiological characteristic analysis on the real-time patient status data, the patient's individual physiological characteristic data, such as heart rate, blood pressure, respiratory rate, etc., are extracted from the real-time patient status data. These physiological characteristic data are quantitative representations of the patient's physiological state, which help to understand the patient's health status and physiological functions. The analysis of individual physiological characteristics provides a basis for subsequent time series analysis and trend analysis. By performing multi-period analysis on individual physiological characteristic data, we can obtain the changes in the patient's physiological characteristics in different time periods, which helps to understand the dynamic changes of the patient's physiological characteristics on different time scales, such as hours, days, weeks, etc. By generating time-series physiological characteristic data, we can reveal the periodic changes, trend changes or sudden changes in physiological characteristics, providing a basis for subsequent trend analysis and anomaly detection. By performing time-series change trend analysis on time-series physiological characteristic data, we can reveal the changing trends and patterns of the patient's physiological characteristics, which helps to understand the long-term, medium-term and short-term changes in the patient's physiological characteristics. The physiological characteristic change trend data is generated to observe and analyze whether the patient's physiological state is stable, improving or worsening, which is of great significance for timely detecting changes in patients and adjusting treatment plans.

[0018] Preferably, the specific steps of step S2 are:

[0019] Step S21: dividing the physiological characteristic change trend data into sliding time windows to obtain block short-term change trend data;

[0020] Step S22: performing sequence connection on the block short-term change trend data to obtain a short-term change trend sequence;

[0021] Step S23: performing convolution sampling on the short-term change trend sequence to obtain the change trend global feature data;

[0022] Step S24: performing feature mutation identification on the global feature data of the change trend, thereby obtaining the feature points of the change trend.

[0023] The present invention divides the data into multiple blocks by performing sliding time window division on the physiological characteristic change trend data, each block represents the physiological characteristic change trend in a short period of time, which helps to subdivide the change trend in a long period of time into smaller time periods to better capture the local changes in the patient's physiological characteristics. The block short-term change trend data provides more detailed and accurate change information on the time scale, providing a basis for subsequent sequence connection and feature analysis. By performing sequence connection on the block short-term change trend data, the short-term change trend data of each block are connected together in chronological order to form a complete short-term change trend sequence, which helps to integrate the change trend data that is continuous but scattered in time into a continuous sequence, better representing the overall changes of the patient. The short-term change trend sequence provides an analysis of the patient's physiological state in different time periods. Overall grasp provides a basis for subsequent global feature extraction and feature mutation identification. By performing convolution sampling on short-term change trend sequences, global feature information in the sequence is extracted. The convolution operation captures local patterns and change trends in the sequence, while the sampling operation reduces the dimension of the data and extracts higher-level feature representations. These change trend global feature data better describe the overall change trend of the patient and provide a basis for subsequent feature mutation identification. By performing feature mutation identification on the change trend global feature data, special points or abnormal points in the change trend, namely, change trend feature points, are detected. These feature points represent important changes in physiological characteristics or turning points in disease status. By identifying change trend feature points, important reference information is provided to help medical staff promptly detect changes in the patient's physiological state or disease progression, and then adjust the treatment plan or take necessary intervention measures.

[0024] Preferably, the specific steps of step S3 are:

[0025] Step S31: performing dynamic evolution analysis on the physiological characteristic change trend data based on the change trend feature points to generate a trend dynamic evolution trajectory;

[0026] Step S32: performing cross-modal deep structure data analysis on the trend dynamic evolution trajectory, thereby generating a physiological characteristic evolution trend law;

[0027] Step S33: performing trend prediction on the time series physiological characteristic data based on the physiological characteristic evolution trend law to obtain individual physiological characteristic prediction data.

[0028] The present invention performs dynamic evolution analysis on physiological characteristic change trend data based on change trend feature points to reveal the evolution trajectory of the patient's physiological characteristics, help understand the change patterns and trends of physiological characteristics in different time periods, including periodic changes, trend changes and sudden changes, generate trend dynamic evolution trajectories to provide an overall grasp of the changes in the patient's physiological state, and provide a basis for subsequent trend analysis and prediction. By performing cross-modal deep structure data analysis on the trend dynamic evolution trajectory, the association and laws between different modal data are mined to help discover the potential laws and patterns of physiological characteristic evolution trends, such as the mutual influence, correlation and time series relationship between different physiological characteristics, and generate physiological characteristic evolution trend laws to provide more comprehensive and integrated analysis results, providing a basis for subsequent trend prediction and individual characteristic prediction. By performing trend prediction on time series physiological characteristic data based on physiological characteristic evolution trend laws, individual physiological characteristic changes in future time periods are inferred to help predict the patient's physiological state and disease progression trend, and provide a basis for medical staff to provide early warning and intervention. Individual physiological characteristic prediction data helps medical staff to adjust treatment plans, optimize monitoring strategies or take other necessary measures in a timely manner to provide more accurate and personalized medical care.

[0029] Preferably, the specific steps of step S4 are:

[0030] Step S41: performing object detection on the ward patrol monitoring video to obtain ward object data; the ward object data includes hospital beds and medical equipment;

[0031] Step S42: performing a three-dimensional spatial structure analysis on the ward patrol monitoring video to obtain three-dimensional spatial structure data;

[0032] Step S43: Performing three-dimensional point cloud modeling on the three-dimensional spatial structure data and the ward object data to construct a three-dimensional spatial model;

[0033] Step S44: dynamically render and reconstruct the three-dimensional space model based on the individual physiological characteristic data to construct a holographic ward model.

[0034] The present invention detects objects in the ward patrol monitoring video, identifies objects such as beds and medical equipment in the video, and extracts relevant object data. It helps to understand the layout and equipment distribution in the ward, and provides a basis for subsequent ward analysis and decision-making. The ward object data provides information about the environment and resources in the ward, which helps medical staff to make reasonable resource allocation and work arrangements. By performing three-dimensional spatial structure analysis on the ward patrol monitoring video, the three-dimensional spatial structure information in the ward is inferred, including the position, size and relative relationship of the object. It helps to understand the layout and spatial distribution in the ward, and provides a basis for subsequent three-dimensional point cloud modeling. The three-dimensional spatial structure data provides spatial perception of the internal environment of the ward, which helps medical staff to carry out spatial planning and resource management. By performing three-dimensional point cloud modeling on the three-dimensional spatial structure data and the ward object data, the objects and spatial structures of the ward are modeled in the form of three-dimensional point clouds. It helps to build an accurate three-dimensional spatial model of the ward, including beds, medical equipment and other room layouts. The three-dimensional spatial model provides an intuitive presentation of the internal environment of the ward, and provides a basis for the subsequent construction of a holographic ward model. By dynamically rendering and reconstructing the three-dimensional space model based on individual physiological characteristics data, the data of individual physiological characteristics are integrated with the three-dimensional space model to build a holographic ward model. The patient's physiological characteristics are combined with the ward environment to achieve a visual presentation of the patient's condition. The holographic ward model provides medical staff with a comprehensive visualization tool to help them understand the patient's condition and environment more intuitively and make corresponding decisions and interventions.

[0035] Preferably, the specific steps of step S43 are:

[0036] Step S431: generating three-dimensional point cloud data based on the three-dimensional spatial structure data;

[0037] Step S432: using the ward object data to mark the position of the three-dimensional point cloud data to obtain the object point cloud position data;

[0038] Step S433: performing multi-object point cloud segmentation based on the object point cloud position data to construct a three-dimensional mesh model of multiple objects;

[0039] Step S434: performing three-dimensional fusion on the three-dimensional mesh models of multiple objects to construct a three-dimensional point cloud model;

[0040] Step S435: Perform point cloud space correction on the three-dimensional point cloud model to construct a three-dimensional space model.

[0041] The present invention generates three-dimensional point cloud data based on three-dimensional spatial structure data, represents objects and spatial structures in the ward in the form of point clouds, converts the three-dimensional information of the ward into a series of discrete points, and provides a basis for subsequent object segmentation and model construction. By using the ward object data to mark the position of the three-dimensional point cloud data, the object to which each point in the point cloud belongs and its position information are determined, and different objects in the three-dimensional point cloud are helped to be identified and segmented, providing a basis for subsequent object segmentation and model construction. The object point cloud position data provides information on the specific position and spatial relationship of each object in the ward. By performing multi-object point cloud segmentation based on the object point cloud position data, the point cloud data is segmented into Multiple separate object parts help separate the points of different objects in the 3D point cloud data, and build a separate 3D mesh model for each object. Multi-object point cloud segmentation provides independent representation of different objects in the ward, providing a basis for subsequent model construction and analysis. By 3D fusion of multiple object 3D mesh models, the segmented object models are fused into a complete 3D point cloud model, and the mesh models of each object are merged into an overall point cloud model, retaining the spatial position and shape information of each object. By performing point cloud space correction on the 3D point cloud model, the point cloud data is further processed and optimized to build an accurate and complete 3D spatial model. Correct the noise, missing or overlapping problems in the point cloud data to improve the quality and accuracy of the model.

[0042] Preferably, the step S5 performs abnormal detection on the individual physiological characteristic prediction data to generate abnormal physiological characteristic values; specifically, the step S5 includes:

[0043] Standardizing the individual physiological characteristic prediction data to obtain standard physiological characteristic prediction data;

[0044] Calculate the confidence of the standard physiological feature prediction data to obtain the feature prediction confidence;

[0045] Predictively optimizing the standard physiological feature prediction data based on the feature prediction confidence level to generate optimized physiological feature prediction data;

[0046] Based on the preset physiological parameter range threshold, the abnormal deviation range of the optimized physiological characteristic prediction data is identified, and when the abnormal physiological characteristic value exceeds the preset physiological parameter range threshold, an abnormal physiological characteristic value is generated.

[0047] The present invention standardizes individual physiological characteristic prediction data and converts the data into standardized data with the same scale and distribution, which helps to eliminate the dimensional differences between different characteristics and makes the data easier to compare and analyze. The standardized data provides a more accurate and consistent feature representation, providing a basis for subsequent calculations and processing. By calculating the confidence of standard physiological characteristic prediction data, the reliability of each feature prediction value is evaluated, which helps to determine the accuracy and credibility of feature predictions and provides a basis for subsequent prediction optimization and abnormality identification. The feature prediction confidence is used to judge the reliability of feature values, thereby improving the trust in physiological characteristic data. The physiological characteristic prediction data is predicted and optimized, and the characteristic prediction is weighted or adjusted according to the confidence level to improve the accuracy and reliability of the characteristic prediction, making the physiological characteristic prediction data more precise and reliable. The optimized physiological characteristic prediction data provides a more accurate and reliable estimate of the individual physiological state. By identifying the abnormal deviation range based on the preset physiological parameter range threshold, it is determined whether the optimized physiological characteristic prediction data is within the normal range. When the physiological characteristic value exceeds the preset range threshold, it is marked as an abnormal physiological characteristic value, which helps to timely discover and identify the individual's abnormal physiological state and take corresponding treatment measures. The generation of abnormal physiological characteristic values ​​provides early warning and monitoring of the individual's health status.

[0048] Preferably, the abnormal physiological characteristic values ​​are analyzed in step S5 to obtain an abnormal characteristic decision strategy; specifically ...

[0049] Performing frequency statistics on abnormal physiological characteristic values ​​to obtain abnormal physiological characteristic frequency data;

[0050] Performing abnormal correlation analysis on abnormal physiological characteristic values ​​based on abnormal physiological characteristic frequency data to obtain abnormal physiological characteristic correlation data;

[0051] Conduct risk quantitative analysis on abnormal physiological characteristic correlation data to obtain risk trend data;

[0052] Based on the risk trend data, decision analysis is performed on abnormal physiological characteristic values ​​to obtain abnormal characteristic decision strategies.

[0053] The present invention performs frequency statistics on abnormal physiological characteristic values ​​to understand the frequency of occurrence of abnormal physiological characteristics in the ward, which is helpful to evaluate the prevalence and severity of abnormal physiological phenomena and provide data support for subsequent abnormal association analysis and risk quantitative analysis. The abnormal physiological characteristic frequency data provides an overall understanding of abnormal conditions in the ward. By performing abnormal association analysis based on the abnormal physiological characteristic frequency data, the association and mutual influence between different physiological characteristics are identified, which helps to discover the potential relationship and pattern between abnormal physiological characteristics and provide a basis for subsequent risk analysis and decision-making. The abnormal physiological characteristic association data provides a correlation analysis of abnormal conditions in the ward. By performing risk quantitative analysis on the abnormal physiological characteristic association data, the risk degree of abnormal physiological characteristics to the health status of patients is quantified, which helps to evaluate the importance and urgency of abnormal physiological characteristics and provide decision-making reference for medical staff. The risk trend data provides risk assessment and prediction of abnormal conditions in the ward. By performing decision analysis on abnormal physiological characteristic values ​​based on risk trend data, a decision strategy for abnormal physiological characteristics is formulated to help medical staff identify and deal with high-risk abnormal conditions and take corresponding intervention measures in a timely manner. The abnormal characteristic decision strategy provides decision guidance for abnormal conditions in the ward.

[0054] Preferably, the specific steps of step S6 are:

[0055] Step S61: performing bounding box segmentation on the holographic ward space model based on the abnormal physiological feature data to obtain bed bounding box data;

[0056] Step S62: generating bed position data based on the bed boundary box data;

[0057] Step S63: Performing early warning processing based on the bed position data to generate an early warning signal;

[0058] Step S64: Upload the abnormal feature decision strategy and warning signal to the intelligent ward inspection system to perform ward inspection analysis operations.

[0059] The present invention extracts the bed area from the entire ward space by performing bounding box segmentation of the holographic ward space model on the abnormal physiological characteristic data, which helps to focus on the bed area and reduce the analysis and processing of other areas. The bed bounding box data provides an accurate description of the bed position, providing a basis for subsequent bed position data generation and early warning processing. By generating bed position data based on the bed bounding box data, the position and layout of the bed in the ward space are accurately determined, which helps medical staff understand the distribution of the beds and provides a positioning basis for subsequent early warning processing and patrol analysis. The bed position data provides detailed information on the layout and distribution of the beds. Early warning processing of bed location data, timely detection and identification of existing abnormal situations, helps to discover potential risks and problems in advance, and provides early warning signals for medical staff so that they can take appropriate measures. The early warning signals provide warnings and reminders of abnormal situations to help ensure the safety and health of patients. By uploading abnormal feature decision-making strategies and early warning signals to the smart ward patrol system, automated ward patrol analysis operations are realized, which helps to improve analysis efficiency and accuracy and reduce the workload of medical staff. The smart ward patrol system automatically performs patrol analysis tasks based on the uploaded abnormal feature decision-making strategies and early warning signals, and promptly feeds back results and suggestions to medical staff.

[0060] In this specification, an intelligent ward inspection system is provided, which is used to execute the ward inspection data analysis method as described above, including:

[0061] Physiological characteristic module, obtains real-time patient status data and ward patrol monitoring video; obtains individual physiological characteristic data based on real-time patient status data; performs multi-period analysis on individual physiological characteristic data to generate physiological characteristic change trend data;

[0062] The change trend module divides the physiological characteristic change trend data into sliding time windows to obtain a short-term change trend sequence; it identifies characteristic mutations of the short-term change trend sequence to obtain the change trend feature points;

[0063] The trend prediction module dynamically analyzes the physiological characteristic change trend data based on the change trend feature points, thereby generating the physiological characteristic evolution trend law; and performs trend prediction based on the physiological characteristic evolution trend law to obtain individual physiological characteristic prediction data;

[0064] The three-dimensional space module performs object detection on the ward patrol monitoring video to obtain the ward object data; performs three-dimensional space structure analysis on the ward patrol monitoring video to obtain three-dimensional space structure data; performs three-dimensional point cloud modeling based on the three-dimensional space structure data and ward object data to construct a holographic ward model;

[0065] The anomaly detection module performs anomaly detection on the individual physiological characteristic prediction data to generate abnormal physiological characteristic values; performs decision analysis on the abnormal physiological characteristic values ​​to obtain abnormal characteristic decision strategies;

[0066] The patrol analysis module performs bounding box segmentation on the holographic ward space model based on abnormal physiological feature data to obtain bed bounding box data; performs early warning processing based on the bed bounding box data to generate early warning signals; and uploads the abnormal feature decision strategy and early warning signals to the intelligent ward patrol system to execute ward patrol analysis operations.

[0067] The present invention obtains real-time patient status data and ward patrol monitoring video through the physiological characteristic module, monitors the patient's physiological characteristics, including heart rate, respiratory rate, body temperature, etc., which helps to timely understand the patient's health status. By performing multi-period analysis on individual physiological characteristic data, the change trend data of the physiological characteristics is obtained, which helps to identify whether the patient's physiological state has changed, so as to perform further processing and prediction. The change trend module divides the physiological characteristic change trend data into short-term change trend sequences through sliding time windows, which helps to capture local changes and trends of physiological characteristics. The short-term change trend sequence is identified by feature mutations to determine the feature points of the change trend. These feature points represent significant changes in physiological characteristics, which are of great significance for abnormality detection and trend prediction. The trend prediction module performs dynamic evolution analysis based on the change trend feature points to reveal the evolution trend law of the physiological characteristics, which helps to understand the changes in the patient's physiological state and provide a basis for subsequent trend prediction. Based on the evolution trend law of the physiological characteristics, trend prediction is performed to predict the individual's physiological characteristic data, which helps to discover potential health problems in advance and take corresponding intervention measures. The three-dimensional space module performs object detection and The three-dimensional spatial structure analysis builds a holographic ward model, which provides a three-dimensional representation of objects in the ward, which helps to understand the spatial layout and environment of the ward. Based on the abnormal physiological feature data, the bounding box segmentation of the holographic ward space model is carried out, and the bounding box data of the bed is identified, which helps to associate the abnormal features with the specific bed and perform corresponding early warning processing. The anomaly detection module performs anomaly detection on the individual physiological feature prediction data and identifies abnormal physiological feature values, which helps to timely discover the patient's health problems and abnormal conditions. Decision analysis is performed on the abnormal physiological feature values ​​to generate corresponding abnormal feature decision strategies. It helps to guide medical staff to take appropriate measures and interventions. The patrol analysis module performs bounding box segmentation on the holographic ward space model based on abnormal physiological feature data, extracts the bounding box data of the bed, helps to locate and associate abnormal features with specific beds, performs early warning processing based on the bed bounding box data, generates early warning signals related to abnormal features, helps to notify medical staff in time and take corresponding treatment measures, uploads abnormal feature decision-making strategies and early warning signals to the intelligent ward patrol system, and executes ward patrol analysis operations, which helps to provide comprehensive ward patrol and monitoring functions to ensure the safety and health of patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 A schematic diagram of the steps of a ward rounds data analysis method of the present invention;

[0069] Figure 2 Detailed implementation flow chart of step S1;

[0070] Figure 3Detailed implementation flow chart of step S2;

[0071] Figure 4 Detailed implementation flow chart of step S3. DETAILED DESCRIPTION

[0072] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0073] The present application example provides a ward inspection data analysis method and an intelligent ward inspection system. The execution subjects of the ward inspection data analysis method and the intelligent ward inspection system include but are not limited to: mechanical equipment, data processing platform, cloud server node, network upload device, etc. equipped with the system can be regarded as the general computing node of the present application, and the data processing platform includes but is not limited to: at least one of an audio image management system, an information management system, and a cloud data management system.

[0074] See also Figures 1 to 4 The present invention provides a ward inspection data analysis method, comprising the following steps:

[0075] Step S1: Acquire real-time patient status data and ward patrol monitoring video; obtain individual physiological characteristic data based on the real-time patient status data; perform multi-period analysis on the individual physiological characteristic data to generate physiological characteristic change trend data;

[0076] Step S2: Divide the physiological characteristic change trend data into sliding time windows to obtain a short-term change trend sequence; perform feature mutation identification on the short-term change trend sequence to obtain a change trend feature point;

[0077] Step S3: performing dynamic evolution analysis on the physiological characteristic change trend data based on the change trend feature points, thereby generating the physiological characteristic evolution trend law; performing trend prediction based on the physiological characteristic evolution trend law, thereby obtaining individual physiological characteristic prediction data;

[0078] Step S4: performing object detection on the ward patrol monitoring video to obtain ward object data; performing three-dimensional spatial structure analysis on the ward patrol monitoring video to obtain three-dimensional spatial structure data; performing three-dimensional point cloud modeling based on the three-dimensional spatial structure data and the ward object data to construct a holographic ward model;

[0079] Step S5: performing abnormality detection on the individual physiological characteristic prediction data to generate abnormal physiological characteristic values; performing decision analysis on the abnormal physiological characteristic values ​​to obtain an abnormal characteristic decision strategy;

[0080] Step S6: Based on the abnormal physiological characteristic data, the holographic ward space model is segmented into bounding boxes to obtain bed bounding box data; based on the bed bounding box data, early warning processing is performed to generate an early warning signal; the abnormal characteristic decision strategy and the early warning signal are uploaded to the intelligent ward patrol system to perform ward patrol analysis operations.

[0081] The present invention obtains real-time patient status data and monitoring videos to understand the patient's physiological condition and behavioral activities in real time. Based on the real-time patient status data, the present invention extracts individual physiological characteristic data, including heart rate, blood pressure, etc., to quantify and analyze the patient's physiological state. Multi-period analysis reveals the changing trend of the patient's physiological characteristics, which is helpful to understand the patient's health status and disease changes. Through sliding time window division and feature mutation identification, the physiological characteristic change trend data is divided into short-term change trend sequences, and important change trend feature points are identified to help capture the mutation or important change of physiological characteristics, which is helpful to timely discover potential abnormal conditions or disease deterioration. Through the dynamic evolution analysis of the change trend feature points, the evolution law and trend of the physiological characteristics are revealed, which is helpful to understand the patient's physiological state development, predict the patient's future physiological characteristic change trend, and provide individual physiological characteristic prediction data to make corresponding treatment adjustments or intervention measures. Through object detection and three-dimensional spatial structure of ward patrol monitoring videos Analyze and obtain the object data and three-dimensional spatial structure information in the ward. Based on these data, three-dimensional point cloud modeling is performed to build a holographic ward model. Such a model provides comprehensive and three-dimensional ward environment information, which helps to better understand the patient's behavior and the layout of the ward. By performing abnormal detection on the individual physiological characteristic prediction data, abnormal physiological characteristic values ​​are identified, and potential abnormal conditions or changes in the condition are identified, which helps to timely discover the patient's abnormal state and take corresponding measures. Decision analysis is performed on the abnormal physiological characteristic values, and appropriate abnormal characteristic decision strategies are formulated according to specific circumstances, such as triggering alarms and notifying medical staff. By performing bounding box segmentation on the abnormal physiological characteristic data, the bed bounding box data, that is, the specific location of the patient, is accurately located and extracted. Early warning processing is performed based on the bed bounding box data, and corresponding early warning signals are generated according to the abnormal characteristic decision strategy to remind monitoring personnel or medical staff to pay attention to specific beds or patients, realize intelligent ward patrol analysis, and improve the efficiency of monitoring and response.

[0082] In the embodiment of the present invention, refer to Figure 1 , is a schematic flow chart of the steps of a ward rounds data analysis method of the present invention. In this example, the steps of the ward rounds data analysis method include:

[0083] Step S1: Acquire real-time patient status data and ward patrol monitoring video; obtain individual physiological characteristic data based on the real-time patient status data; perform multi-period analysis on the individual physiological characteristic data to generate physiological characteristic change trend data;

[0084] In this embodiment, by installing physiological monitoring equipment, such as a heart rate monitor, a blood pressure monitor, a respiratory rate sensor, etc., the patient's physiological parameters are monitored in real time and the data is transmitted to the system. A surveillance camera is installed inside the ward to obtain a real-time video stream of the ward. A data acquisition node is set to receive data from the sensor device and transmit it to a central data storage server. The data transmission is performed by wired or wireless means to ensure the real-time and integrity of the data. The video surveillance system is used to obtain the real-time video stream in the ward, which is achieved through a server or network video transmission protocol connected to the surveillance camera. The received real-time patient status data is preprocessed, including data cleaning, denoising, missing value processing, etc., to ensure the quality and accuracy of the data. Individual physiological characteristics are extracted from the preprocessed data, and the average heart rate, blood pressure change amplitude, respiratory rate fluctuation degree and other characteristics are calculated. The individual physiological characteristic data are divided into time periods, and the physiological characteristic data in each time period are analyzed. The statistical indicators such as the mean, standard deviation, and change rate of the characteristics are calculated to obtain the change trend of the physiological characteristics. The generated physiological characteristic change trend data is stored in a database or data warehouse for subsequent analysis and use.

[0085] Step S2: Divide the physiological characteristic change trend data into sliding time windows to obtain a short-term change trend sequence; perform feature mutation identification on the short-term change trend sequence to obtain a change trend feature point;

[0086] In this embodiment, the physiological characteristic change trend data is divided into a plurality of continuous time windows. The size of the time window is set according to the demand. The window size is selected as 30 minutes or 1 hour. The physiological characteristic data in each time window is analyzed to extract the short-term change trend sequence. Various methods are used, such as calculating the average value, maximum value, minimum value or standard deviation in each window as the representation of the short-term change trend sequence. A threshold is set. When a data point in the short-term change trend sequence exceeds or falls below the threshold, it is marked as a characteristic mutation point. The selection of the threshold is based on domain knowledge or determined according to the statistical characteristics of the data. The change rate between adjacent data points in the short-term change trend sequence is calculated. When the change rate exceeds the predetermined threshold, the current data point is marked as a characteristic mutation point. Machine learning algorithms such as support vector machine (SVM) and random forest (Random Forest) are used. orest) or neural network, etc., and train the model to identify the characteristic mutation points in the short-term change trend sequence. The training data includes the marked characteristic mutation points and the data points of normal change. According to the recognition result of the characteristic mutation, the change trend characteristic points are extracted from the short-term change trend sequence. The characteristic point is the mutation point itself, and also other information related to the mutation point, such as the position, amplitude or duration of the mutation point.

[0087] Step S3: performing dynamic evolution analysis on the physiological characteristic change trend data based on the change trend feature points, thereby generating the physiological characteristic evolution trend law; performing trend prediction based on the physiological characteristic evolution trend law, thereby obtaining individual physiological characteristic prediction data;

[0088] In this embodiment, the extracted change trend feature points are sorted in chronological order to ensure the correctness of the evolution analysis. By analyzing the relationship between the change trend feature points and the distribution pattern of the feature points, the evolution law of the physiological characteristics is extracted, including periodic patterns, trend growth / decline, stage changes, etc. According to the acquired real-time patient status data and physiological characteristic change trend data, an input data set of the prediction model is constructed, and historical data and current data are considered as input. According to the characteristics of the data and the prediction requirements, a suitable prediction model is selected. Commonly used prediction models include time series models (such as ARI MA, LSTM, etc.), regression models (such as linear regression, support vector regression, etc.) or machine learning models (such as random forest, neural network, etc.). The prediction model is trained using historical data. According to the selected model, the model parameters are adjusted and iterative training is performed to obtain the best prediction performance. The trained model is used to predict the individual physiological characteristics in the future. According to the type of prediction model and the input data, the physiological characteristic prediction data for a period of time in the future is obtained.

[0089] Step S4: performing object detection on the ward patrol monitoring video to obtain ward object data; performing three-dimensional spatial structure analysis on the ward patrol monitoring video to obtain three-dimensional spatial structure data; performing three-dimensional point cloud modeling based on the three-dimensional spatial structure data and the ward object data to construct a holographic ward model;

[0090] In this embodiment, video data is obtained from the ward patrol monitoring camera, and object detection is performed on the ward patrol monitoring video. For each frame of video, the object is detected and its position and category information is obtained. The monitoring camera is calibrated to obtain the internal and external parameters of the camera, including the focal length, distortion parameters, rotation matrix and translation vector of the camera, etc. Feature points are extracted from the frames in the video, and feature point matching algorithms (such as SIF T, SUMO) are used to obtain the feature points. RF, ORB, etc.) are used to match feature points to obtain the correspondence between adjacent frames, and the correspondence between camera calibration parameters and feature points is used to perform three-dimensional spatial structure analysis on the video using a three-dimensional reconstruction algorithm (such as structured light, multi-view geometry, etc.) to obtain the three-dimensional coordinate information of objects in the ward. The ward object data obtained in the object detection step and the three-dimensional coordinate information obtained in the three-dimensional spatial structure analysis step are integrated to establish an association between the object and its corresponding three-dimensional coordinates. Based on the integrated data, a three-dimensional point cloud modeling algorithm (such as a grid-based method, a geometric feature-based method, etc.) is used to model the three-dimensional space of the ward to generate a holographic ward model, which contains objects in the ward such as walls, furniture, and equipment, and retains their geometric shapes and relative position information.

[0091] Step S5: performing abnormality detection on the individual physiological characteristic prediction data to generate abnormal physiological characteristic values; performing decision analysis on the abnormal physiological characteristic values ​​to obtain an abnormal characteristic decision strategy;

[0092] In this embodiment, feature extraction is performed on the collected physiological characteristic prediction data, and the original data is converted into a representative feature vector using statistical features, frequency domain features, time domain features and other methods. The individual physiological characteristic prediction data is subjected to abnormality detection based on a preset threshold value to determine whether each feature vector is within a normal range. For feature vectors determined to be abnormal, corresponding abnormal physiological characteristic values ​​are generated, and the generated abnormal physiological characteristic values ​​are integrated into an abnormal feature vector, in which each feature represents an abnormal physiological characteristic. An abnormal characteristic decision strategy is formulated, and corresponding treatment measures are taken according to the severity and type of the abnormal characteristics. These measures include issuing an alarm, notifying medical staff, adjusting treatment plans, etc. The abnormal decision strategy is optimized and adjusted according to actual conditions and feedback information to improve judgment accuracy and treatment effects.

[0093] Step S6: Based on the abnormal physiological characteristic data, the holographic ward space model is segmented into bounding boxes to obtain bed bounding box data; based on the bed bounding box data, early warning processing is performed to generate an early warning signal; the abnormal characteristic decision strategy and the early warning signal are uploaded to the intelligent ward patrol system to perform ward patrol analysis operations.

[0094] In this embodiment, in the holographic ward space model, the bounding box of each bed is marked, and the bounding box data of each bed, including information such as position, size and direction, is extracted from the holographic ward space model through computer vision technology, such as target detection or image segmentation algorithm. According to the rules and measures formulated by the abnormal feature decision strategy, corresponding early warning processing plans are prepared. These plans include issuing alarms, sending notifications, adjusting monitoring equipment parameters, etc., and the abnormal features are associated with the corresponding bed bounding boxes to ensure that each abnormal feature corresponds to the corresponding bed. According to the severity of the abnormal features and the bed bounding box data, a corresponding early warning signal is generated. The early warning signal is a signal in the form of digital, sound, light or other forms, which is used to remind medical staff or trigger subsequent processing measures. The formulated abnormal feature decision strategy is uploaded to the intelligent ward patrol system. In this way, the system analyzes and makes decisions according to the strategy, and uploads the generated early warning signal to the intelligent ward patrol system. The system triggers the corresponding patrol analysis operation according to the early warning signal.

[0095] In this embodiment, refer to Figure 2 , is a flowchart of detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include:

[0096] Step S11: Acquire real-time patient status data and ward patrol monitoring video;

[0097] Step S12: performing individual physiological characteristic analysis on the real-time patient status data to obtain individual physiological characteristic data;

[0098] Step S13: performing multi-period analysis on the individual physiological characteristic data to generate time series physiological characteristic data;

[0099] Step S14: performing time series change trend analysis on the time series physiological characteristic data to generate physiological characteristic change trend data.

[0100] In this embodiment, sensors and monitoring equipment are set to obtain real-time patient status data and ward patrol monitoring videos. The sensors include heart rate monitors, blood pressure monitors, respiratory monitors, etc., which are used to collect the patient's physiological data. The monitoring equipment is a camera or other visualization equipment, which is used to obtain real-time video of the ward. The real-time patient status data is preprocessed, including noise removal, data interpolation or filling missing values, etc., to ensure the quality and integrity of the data. According to the knowledge and algorithms in the medical field, individual physiological characteristics are extracted from the preprocessed data, including indicators such as heart rate, blood pressure, and respiratory rate. Different feature extraction methods are selected according to specific needs and disease characteristics. Individual physiological characteristic data Divide into multiple time periods, such as by hour, by day or by disease stage, etc. According to specific needs and disease characteristics, choose the appropriate time period division method, integrate and count the individual physiological characteristic data in each time period, generate summary data within the time period, including statistical indicators such as the average value, maximum value, minimum value, standard deviation within the time period, calculate the time series physiological characteristic data to obtain the time series change trend of the physiological characteristics, use statistical analysis or machine learning methods, such as moving average, exponential smoothing, regression analysis, etc., to visualize the time series change trend results of the physiological characteristics, use line charts, trend charts, heat maps, etc., so that medical staff can intuitively understand the changing trend of physiological characteristics.

[0101] In this embodiment, refer to Figure 3 , is a flowchart of detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include:

[0102] Step S21: dividing the physiological characteristic change trend data into sliding time windows to obtain block short-term change trend data;

[0103] Step S22: performing sequence connection on the block short-term change trend data to obtain a short-term change trend sequence;

[0104] Step S23: performing convolution sampling on the short-term change trend sequence to obtain the change trend global feature data;

[0105] Step S24: performing feature mutation identification on the global feature data of the change trend, thereby obtaining the feature points of the change trend.

[0106] In this embodiment, the size and step length of the sliding time window are determined. The size of the time window determines the time span of each block, and the step length determines the degree of overlap between adjacent time windows. The physiological characteristic change trend data is divided into blocks according to the sliding time window. The data in each time window is a continuous section or a discrete time point. It is determined according to the characteristics and needs of the data. The size and shape of the convolution kernel are determined. The size of the convolution kernel determines the sampling range of the short-term change trend sequence, and the shape is selected according to specific needs. The short-term change trend sequence is convolved. The global features in the change trend sequence are obtained by sliding the convolution kernel. The convolution operation uses a common convolutional neural network algorithm or other applicable convolution methods. According to a specific algorithm or rule, the change trend global feature data is analyzed to identify feature mutation points. These feature mutation points represent significant changes or abnormal events in the physiological state. The feature mutation recognition algorithm is applied to detect the change trend global feature data and mark potential feature mutation points. These feature mutation points are extreme points of the data, trend change points or other points with significant changes.

[0107] In this embodiment, reference Figure 4 The above is a schematic flow chart of the detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include:

[0108] Step S31: performing dynamic evolution analysis on the physiological characteristic change trend data based on the change trend feature points to generate a trend dynamic evolution trajectory;

[0109] Step S32: performing cross-modal deep structure data analysis on the trend dynamic evolution trajectory, thereby generating a physiological characteristic evolution trend law;

[0110] Step S33: performing trend prediction on the time series physiological characteristic data based on the physiological characteristic evolution trend law to obtain individual physiological characteristic prediction data.

[0111] In this embodiment, the values ​​of these feature points are extracted from the physiological characteristic change trend data. The feature points are data points with significant changes, such as peak values, valley values, mutation points, etc. The extracted feature points are connected in chronological order to form a trend dynamic evolution trajectory. The connection adopts linear interpolation or other suitable interpolation methods to fill the data between the feature points to obtain a smooth evolution trajectory. The trend dynamic evolution trajectory is associated with other related deep-level structural data for analysis. These deep-level structural data are anatomical structures, histological characteristics, biochemical indicators, etc. of physiological characteristics. Cross-modal data analysis is achieved through data fusion, feature extraction, pattern recognition and other methods. According to the results of cross-modal data analysis, the evolution trend laws of physiological characteristics are extracted. These laws It is a form of mathematical model, statistical law, trend line fitting, etc., which is used to describe the temporal change law of physiological characteristics. According to the evolution trend law of physiological characteristics, a suitable trend prediction model is selected. Commonly used prediction models include linear regression, ARIMA model, exponential smoothing model, neural network model, etc. The time series physiological characteristic data is preprocessed, including noise removal, smoothing, missing value processing, etc. The selected trend prediction model is applied to perform trend prediction on the preprocessed physiological characteristic data. The prediction is a single point prediction or a time series prediction. The time span and accuracy of the prediction are determined according to the needs. According to the results of the prediction model, the individual physiological characteristic prediction data is generated. The predicted data is the value at a specific time point in the future, and it is also the trend change in the future.

[0112] In this embodiment, step S4 includes the following steps:

[0113] Step S41: performing object detection on the ward patrol monitoring video to obtain ward object data; the ward object data includes hospital beds and medical equipment;

[0114] Step S42: performing a three-dimensional spatial structure analysis on the ward patrol monitoring video to obtain three-dimensional spatial structure data;

[0115] Step S43: Performing three-dimensional point cloud modeling on the three-dimensional spatial structure data and the ward object data to construct a three-dimensional spatial model;

[0116] Step S44: dynamically render and reconstruct the three-dimensional space model based on the individual physiological characteristic data to construct a holographic ward model.

[0117] In this embodiment, an object detection algorithm (such as a target detection algorithm based on deep learning) is applied to perform object detection on the preprocessed ward patrol monitoring video. This step aims to identify objects in the ward, such as beds, medical equipment, etc., and obtain their position and bounding box information. Ward object data, including the category, position, size, etc. of the object, are extracted from the object detection results. These data are organized into a format that can be used for subsequent analysis and processing. The camera used for ward patrol monitoring is calibrated to obtain the camera's intrinsic and extrinsic parameters, which describe the camera's viewing angle, distortion, and other information. The ward patrol monitoring video is processed, such as removing distortion and image stabilization, to improve the subsequent three-dimensional reconstruction quality. The camera calibration parameters and the position information of the moving object are used to perform three-dimensional reconstruction and structural analysis to obtain the three-dimensional spatial structure data of the ward. Stereoscopic vision technology, structured light scanning, and other methods are used to achieve three-dimensional reconstruction. The obtained three-dimensional spatial structure data is converted into point cloud data representation. The point cloud data is composed of a large number of three-dimensional points, each of which contains The position, color and other information of the ward object are obtained and merged with the point cloud data, the position and shape information of the object are added to the point cloud data, and the point cloud data is used for 3D modeling. The point cloud data is converted into a 3D space model by using point cloud registration, point cloud segmentation, surface reconstruction and other technologies. Software tools (such as PointCloudLibrary, MeshLab, etc.) are used for point cloud processing and modeling operations, and the individual's physiological characteristic data is associated with the 3D space model. According to the changes in the physiological characteristic data, the model is dynamically rendered and reconstructed. According to the individual's heart rate data, the beating rate and amplitude of the patient's heart model are adjusted; according to the body temperature data, the patient's skin color is changed; according to the blood pressure data, the expansion degree of the patient's blood vessel model is adjusted, etc. The 3D space model reconstructed by dynamic rendering is combined with the individual's physiological characteristic data to generate a holographic ward model, which presents the 3D structure of objects in the ward, and displays dynamic effects according to the individual's physiological characteristic data, providing more intuitive and comprehensive ward information.

[0118] In this embodiment, the specific steps of step S43 are:

[0119] Step S431: generating three-dimensional point cloud data based on the three-dimensional spatial structure data;

[0120] Step S432: using the ward object data to mark the position of the three-dimensional point cloud data to obtain the object point cloud position data;

[0121] Step S433: performing multi-object point cloud segmentation based on the object point cloud position data to construct a three-dimensional mesh model of multiple objects;

[0122] Step S434: performing three-dimensional fusion on the three-dimensional mesh models of multiple objects to construct a three-dimensional point cloud model;

[0123] Step S435: Perform point cloud space correction on the three-dimensional point cloud model to construct a three-dimensional space model.

[0124] In this embodiment, the obtained three-dimensional spatial structure data is converted into a three-dimensional point cloud data representation, and the position information of each point is used as the coordinate of the point cloud data to form a point cloud data set. According to the converted three-dimensional spatial structure data, the corresponding three-dimensional point cloud data is generated, and each point cloud contains the position and other attributes, such as color, normal, etc. The position information in the ward object data is matched with the three-dimensional point cloud data, and the points corresponding to the objects such as beds and medical equipment in the point cloud data are marked using the position information in the ward object data. The object point cloud position data, including the position and attribute information of the object, are extracted from the point cloud data. These data will be used for subsequent multi-object point cloud segmentation and three-dimensional fusion. The object point cloud position data is used to perform multi-object point cloud segmentation, and the point cloud data is segmented into point cloud subsets of multiple objects. A clustering algorithm (such as DBS CAN, Mean Shift, etc.) or a segmentation algorithm (such as a semantic segmentation algorithm based on deep learning) is used to implement point cloud segmentation. For each segmented object point cloud subset, a three-dimensional mesh model is generated, and a surface reconstruction algorithm (such as Poisson reconstruction, Marching Cloud Reconstruction, etc.) is used to perform point cloud segmentation. ubes, etc.) convert point cloud data into a 3D mesh model, fuse the 3D mesh models of multiple objects to generate an overall 3D mesh model, use a mesh fusion algorithm (such as normal consistency, voxelization, etc.) to achieve mesh fusion, convert the fused 3D mesh model into a 3D point cloud model, and achieve this by converting vertices on the mesh model into point cloud data, perform spatial correction on the 3D point cloud model to eliminate existing noise, holes or incomplete parts, use a point cloud filtering algorithm (such as Gaussian filtering, statistical filtering, etc.) to correct and optimize the point cloud model, convert the corrected 3D point cloud model into a 3D space model, use software tools (such as Mesh Lab, Unit, etc.) to load and visualize point cloud data, and generate a 3D space model that can be used and displayed.

[0125] In this embodiment, the abnormality detection is performed on the individual physiological characteristic prediction data in step S5 to generate abnormal physiological characteristic values; specifically:

[0126] Standardizing the individual physiological characteristic prediction data to obtain standard physiological characteristic prediction data;

[0127] Calculate the confidence of the standard physiological feature prediction data to obtain the feature prediction confidence;

[0128] Predictively optimizing the standard physiological feature prediction data based on the feature prediction confidence level to generate optimized physiological feature prediction data;

[0129] Based on the preset physiological parameter range threshold, the abnormal deviation range of the optimized physiological characteristic prediction data is identified, and when the abnormal physiological characteristic value exceeds the preset physiological parameter range threshold, an abnormal physiological characteristic value is generated.

[0130] In this embodiment, a standardized method is used to convert individual physiological characteristic prediction data into standard physiological characteristic prediction data, feature prediction data for which confidence calculation is required is extracted from the standard physiological characteristic prediction data, and confidence calculation is performed on the feature prediction data using an appropriate algorithm or method, including calculating the confidence score of the feature prediction based on statistical analysis, machine learning models, or other reliable methods. For feature prediction data with lower confidence, prediction optimization is performed, including using interpolation, regression, rule-based methods, or other appropriate methods to fill or correct these data. For feature prediction data with higher confidence, its original value is kept unchanged, and the optimized feature prediction data is merged with the unchanged feature prediction data to generate optimized physiological characteristic prediction data. According to medical knowledge and the advice of field experts, a preset physiological parameter range threshold is set. For blood pressure, a minimum and maximum value of systolic blood pressure are set. The optimized physiological characteristic prediction data is compared with the preset physiological parameter range threshold to identify whether there is an abnormal deviation. If a feature prediction value exceeds the set range threshold, it is marked as an abnormal physiological characteristic value. According to the mark of the abnormal deviation, the abnormal feature value is extracted from the optimized physiological characteristic prediction data and used as the abnormal physiological characteristic value.

[0131] In this embodiment, the abnormal physiological characteristic values ​​are analyzed in step S5 to obtain an abnormal characteristic decision strategy; specifically, the abnormal characteristic decision strategy is:

[0132] Performing frequency statistics on abnormal physiological characteristic values ​​to obtain abnormal physiological characteristic frequency data;

[0133] Performing abnormal correlation analysis on abnormal physiological characteristic values ​​based on abnormal physiological characteristic frequency data to obtain abnormal physiological characteristic correlation data;

[0134] Conduct risk quantitative analysis on abnormal physiological characteristic correlation data to obtain risk trend data;

[0135] Based on the risk trend data, decision analysis is performed on abnormal physiological characteristic values ​​to obtain abnormal characteristic decision strategies.

[0136] In this embodiment, frequency statistics are performed on abnormal physiological characteristic values ​​to obtain abnormal physiological characteristic frequency data, the number of times or frequency of each abnormal characteristic value appears is counted to understand the occurrence of abnormalities, abnormal association analysis is performed based on the abnormal physiological characteristic frequency data to obtain abnormal physiological characteristic association data, by analyzing the correlation between abnormal characteristic values, the association pattern or law between them is found, and methods such as association rule analysis, collaborative filtering, and correlation coefficient calculation are used to achieve this, and risk quantitative analysis is performed on the abnormal physiological characteristic association data to obtain risk trend data, by quantifying the degree of correlation and risk level between abnormal characteristics, the importance and potential degree of danger of abnormal characteristics are evaluated, and risk assessment methods such as probability models, statistical analysis, or machine learning algorithms are used to achieve this, and decision analysis is performed on abnormal physiological characteristic values ​​based on risk trend data to obtain abnormal characteristic decision strategies, and according to the risk trend data, appropriate decision strategies are formulated to deal with the occurrence of abnormal characteristics, and the decision strategies include taking further examinations, adjusting treatment plans, and reminding patients or medical staff to ensure timely intervention and management of abnormal situations.

[0137] In this embodiment, the specific steps of step S6 are:

[0138] Step S61: performing bounding box segmentation on the holographic ward space model based on the abnormal physiological feature data to obtain bed bounding box data;

[0139] Step S62: generating bed position data based on the bed boundary box data;

[0140] Step S63: Performing early warning processing based on the bed position data to generate an early warning signal;

[0141] Step S64: Upload the abnormal feature decision strategy and warning signal to the intelligent ward inspection system to perform ward inspection analysis operations.

[0142] In this embodiment, the holographic ward space model is segmented by bounding boxes based on the abnormal feature data. The position and range of the bed are determined by drawing the bounding box in the model. The extracted bed position information is sorted and converted into bed position data, including the coordinates, direction or other relevant parameters of the bed. The patient's bed position data is locked based on the abnormal physiological feature data, and an early warning signal is generated and uploaded to the system. The abnormal feature decision strategy and the early warning signal are uploaded to the intelligent ward patrol system. The intelligent ward patrol system performs ward patrol analysis operations based on the uploaded abnormal feature decision strategy and early warning signal. This includes patrolling objects in the ward, monitoring the condition of patients on the bed, identifying safety risks, etc.

[0143] In this embodiment, an intelligent ward inspection system is provided, which is used to execute the ward inspection data analysis method as described above, including:

[0144] Physiological characteristic module, obtains real-time patient status data and ward patrol monitoring video; obtains individual physiological characteristic data based on real-time patient status data; performs multi-period analysis on individual physiological characteristic data to generate physiological characteristic change trend data;

[0145] The change trend module divides the physiological characteristic change trend data into sliding time windows to obtain a short-term change trend sequence; it identifies characteristic mutations of the short-term change trend sequence to obtain the change trend feature points;

[0146] The trend prediction module dynamically analyzes the physiological characteristic change trend data based on the change trend feature points, thereby generating the physiological characteristic evolution trend law; and performs trend prediction based on the physiological characteristic evolution trend law to obtain individual physiological characteristic prediction data;

[0147] The three-dimensional space module performs object detection on the ward patrol monitoring video to obtain the ward object data; performs three-dimensional space structure analysis on the ward patrol monitoring video to obtain three-dimensional space structure data; performs three-dimensional point cloud modeling based on the three-dimensional space structure data and ward object data to construct a holographic ward model;

[0148] The anomaly detection module performs anomaly detection on the individual physiological characteristic prediction data to generate abnormal physiological characteristic values; performs decision analysis on the abnormal physiological characteristic values ​​to obtain abnormal characteristic decision strategies;

[0149] The patrol analysis module performs bounding box segmentation on the holographic ward space model based on abnormal physiological feature data to obtain bed bounding box data; performs early warning processing based on the bed bounding box data to generate early warning signals; and uploads the abnormal feature decision strategy and early warning signals to the intelligent ward patrol system to execute ward patrol analysis operations.

[0150] The present invention obtains real-time patient status data and ward patrol monitoring video through the physiological characteristic module, monitors the patient's physiological characteristics, including heart rate, respiratory rate, body temperature, etc., which helps to timely understand the patient's health status. By performing multi-period analysis on individual physiological characteristic data, the change trend data of the physiological characteristics is obtained, which helps to identify whether the patient's physiological state has changed, so as to perform further processing and prediction. The change trend module divides the physiological characteristic change trend data into short-term change trend sequences through sliding time windows, which helps to capture local changes and trends of physiological characteristics. The short-term change trend sequence is identified by feature mutations to determine the feature points of the change trend. These feature points represent significant changes in physiological characteristics, which are of great significance for abnormality detection and trend prediction. The trend prediction module performs dynamic evolution analysis based on the change trend feature points to reveal the evolution trend law of the physiological characteristics, which helps to understand the changes in the patient's physiological state and provide a basis for subsequent trend prediction. Based on the evolution trend law of the physiological characteristics, trend prediction is performed to predict the individual's physiological characteristic data, which helps to discover potential health problems in advance and take corresponding intervention measures. The three-dimensional space module performs object detection and The three-dimensional spatial structure analysis builds a holographic ward model, which provides a three-dimensional representation of objects in the ward, which helps to understand the spatial layout and environment of the ward. Based on the abnormal physiological feature data, the bounding box segmentation of the holographic ward space model is carried out, and the bounding box data of the bed is identified, which helps to associate the abnormal features with the specific bed and perform corresponding early warning processing. The anomaly detection module performs anomaly detection on the individual physiological feature prediction data and identifies abnormal physiological feature values, which helps to timely discover the patient's health problems and abnormal conditions. Decision analysis is performed on the abnormal physiological feature values ​​to generate corresponding abnormal feature decision strategies. It helps to guide medical staff to take appropriate measures and interventions. The patrol analysis module performs bounding box segmentation on the holographic ward space model based on abnormal physiological feature data, extracts the bounding box data of the bed, helps to locate and associate abnormal features with specific beds, performs early warning processing based on the bed bounding box data, generates early warning signals related to abnormal features, helps to notify medical staff in time and take corresponding treatment measures, uploads abnormal feature decision-making strategies and early warning signals to the intelligent ward patrol system, and executes ward patrol analysis operations, which helps to provide comprehensive ward patrol and monitoring functions to ensure the safety and health of patients.

[0151] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0152] The above is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A ward inspection data analysis method, characterized in that: The following steps are involved: Step S1: Acquire real-time patient status data and ward patrol monitoring video; Obtain individual physiological characteristic data based on real-time patient status data; Conduct multi-period analysis on individual physiological characteristic data to generate physiological characteristic change trend data; Step S2: dividing the physiological characteristic change trend data into sliding time windows to obtain a short-term change trend sequence; Identify characteristic mutations of short-term trend sequences to obtain characteristic points of trend changes; Step S3: performing dynamic evolution analysis on the physiological characteristic change trend data based on the change trend feature points, thereby generating a physiological characteristic evolution trend law; Based on the evolution trend of physiological characteristics, trend prediction is performed to obtain individual physiological characteristics prediction data; Step S4: performing object detection on the ward patrol monitoring video to obtain ward object data; Conduct three-dimensional spatial structure analysis on ward patrol monitoring videos to obtain three-dimensional spatial structure data; Based on the three-dimensional spatial structure data and the ward object data, three-dimensional point cloud modeling is performed to construct a holographic ward model; the specific steps of step S4 are: Step S41: performing object detection on the ward patrol monitoring video to obtain ward object data; the ward object data includes hospital beds and medical equipment; Step S42: performing a three-dimensional spatial structure analysis on the ward patrol monitoring video to obtain three-dimensional spatial structure data; Step S43: Perform three-dimensional point cloud modeling on the three-dimensional spatial structure data and the ward object data to construct a three-dimensional spatial model; the specific steps of step S43 are: Step S431: generating three-dimensional point cloud data based on the three-dimensional spatial structure data; Step S432: using the ward object data to mark the position of the three-dimensional point cloud data to obtain the object point cloud position data; Step S433: performing multi-object point cloud segmentation based on the object point cloud position data to construct a three-dimensional mesh model of multiple objects; Step S434: performing three-dimensional fusion on the three-dimensional mesh models of multiple objects to construct a three-dimensional point cloud model; Step S435: performing point cloud space correction on the three-dimensional point cloud model to construct a three-dimensional space model; Step S44: dynamically rendering and reconstructing the three-dimensional space model based on the individual physiological characteristic data to construct a holographic ward model; Step S5: performing abnormality detection on the individual physiological characteristic prediction data to generate abnormal physiological characteristic values; Perform decision analysis on abnormal physiological characteristic values ​​to obtain abnormal characteristic decision strategies; Step S6: Based on the abnormal physiological characteristic data, the holographic ward space model is segmented into bounding boxes to obtain bed bounding box data; based on the bed bounding box data, early warning processing is performed to generate an early warning signal; the abnormal characteristic decision strategy and the early warning signal are uploaded to the intelligent ward patrol system to execute the ward patrol analysis operation.

2. The ward inspection data analysis method according to claim 1, characterized in that: The specific steps of step S1 are: Step S11: Acquire real-time patient status data and ward patrol monitoring video; Step S12: performing individual physiological characteristic analysis on the real-time patient status data to obtain individual physiological characteristic data; Step S13: performing multi-period analysis on the individual physiological characteristic data to generate time series physiological characteristic data; Step S14: performing time series change trend analysis on the time series physiological characteristic data to generate physiological characteristic change trend data.

3. The ward inspection data analysis method according to claim 1, characterized in that: The specific steps of step S2 are: Step S21: dividing the physiological characteristic change trend data into sliding time windows to obtain block short-term change trend data; Step S22: performing sequence connection on the block short-term change trend data to obtain a short-term change trend sequence; Step S23: performing convolution sampling on the short-term change trend sequence to obtain the change trend global feature data; Step S24: performing feature mutation identification on the global feature data of the change trend, thereby obtaining the feature points of the change trend.

4. The ward inspection data analysis method according to claim 1, characterized in that: The specific steps of step S3 are: Step S31: performing dynamic evolution analysis on the physiological characteristic change trend data based on the change trend feature points to generate a trend dynamic evolution trajectory; Step S32: performing cross-modal deep structure data analysis on the trend dynamic evolution trajectory, thereby generating a physiological characteristic evolution trend law; Step S33: performing trend prediction on the time series physiological characteristic data based on the physiological characteristic evolution trend law to obtain individual physiological characteristic prediction data.

5. The ward inspection data analysis method according to claim 1, characterized in that: The step S5 performs abnormal detection on the individual physiological characteristic prediction data to generate abnormal physiological characteristic values; specifically: Standardizing the individual physiological characteristic prediction data to obtain standard physiological characteristic prediction data; Calculate the confidence of the standard physiological feature prediction data to obtain the feature prediction confidence; Predictively optimizing the standard physiological feature prediction data based on the feature prediction confidence level to generate optimized physiological feature prediction data; Based on the preset physiological parameter range threshold, the abnormal deviation range of the optimized physiological characteristic prediction data is identified, and when the abnormal physiological characteristic value exceeds the preset physiological parameter range threshold, an abnormal physiological characteristic value is generated.

6. The ward inspection data analysis method according to claim 1, characterized in that: In step S5, the abnormal physiological characteristic values ​​are analyzed to obtain an abnormal characteristic decision strategy; Specifically: Performing frequency statistics on abnormal physiological characteristic values ​​to obtain abnormal physiological characteristic frequency data; Performing abnormal correlation analysis on abnormal physiological characteristic values ​​based on abnormal physiological characteristic frequency data to obtain abnormal physiological characteristic correlation data; Conduct risk quantitative analysis on abnormal physiological characteristic correlation data to obtain risk trend data; Based on the risk trend data, decision analysis is performed on abnormal physiological characteristic values ​​to obtain abnormal characteristic decision strategies.

7. The ward inspection data analysis method according to claim 1, characterized in that: The specific steps of step S6 are: Step S61: performing bounding box segmentation on the holographic ward space model based on the abnormal physiological feature data to obtain bed bounding box data; Step S62: generating bed position data based on the bed boundary box data; Step S63: Performing early warning processing based on the bed position data to generate an early warning signal; Step S64: Upload the abnormal feature decision strategy and warning signal to the intelligent ward inspection system to perform ward inspection analysis operations.

8. An intelligent ward inspection system, characterized in that: The method for analyzing ward rounds data according to claim 1 comprises: Physiological characteristic module, obtains real-time patient status data and ward patrol monitoring video; obtains individual physiological characteristic data based on real-time patient status data; performs multi-period analysis on individual physiological characteristic data to generate physiological characteristic change trend data; The change trend module divides the physiological characteristic change trend data into sliding time windows to obtain a short-term change trend sequence; it identifies characteristic mutations of the short-term change trend sequence to obtain the change trend feature points; The trend prediction module dynamically analyzes the physiological characteristic change trend data based on the change trend feature points, thereby generating the physiological characteristic evolution trend law; and performs trend prediction based on the physiological characteristic evolution trend law to obtain individual physiological characteristic prediction data; The three-dimensional space module performs object detection on the ward patrol monitoring video to obtain the ward object data; performs three-dimensional space structure analysis on the ward patrol monitoring video to obtain three-dimensional space structure data; performs three-dimensional point cloud modeling based on the three-dimensional space structure data and ward object data to construct a holographic ward model; The anomaly detection module performs anomaly detection on the individual physiological characteristic prediction data to generate abnormal physiological characteristic values; performs decision analysis on the abnormal physiological characteristic values ​​to obtain abnormal characteristic decision strategies; The patrol analysis module performs bounding box segmentation on the holographic ward space model based on abnormal physiological feature data to obtain bed bounding box data; performs early warning processing based on the bed bounding box data to generate early warning signals; and uploads the abnormal feature decision strategy and early warning signals to the intelligent ward patrol system to execute ward patrol analysis operations.

Images