Chronic disease patient physical examination data and complication nursing monitoring system based on Internet of Things

Abstract

The invention relates to the technical field of medical health monitoring, and particularly discloses a chronic disease patient physical examination data and complication nursing monitoring system based on the Internet of Things, multi-dimensional physical examination data such as vital signs and blood biochemical indexes of a patient are collected in real time through a multi-source sensing equipment network, and an individualized data stream is constructed; extracting parameter collaborative change characteristics by adopting multi-time scale analysis, and constructing a dynamic physiological characteristic spectrum; identifying the abnormal offset by calculating the deviation degree of the feature vector distribution and the reference distribution; based on intelligent matching of the spatial distribution features of the feature vectors and a clinical case library, generating structured early warning information including risk assessment and personalized nursing schemes; and finally, visual display and preventive nursing intervention of the complication risk evolution process are realized.

Description

Technical Field

[0001] This invention relates to the field of medical and health monitoring technology, specifically to an Internet of Things-based system for monitoring the physical examination data and complication care of patients with chronic diseases. Background Technology

[0002] With the deepening application of IoT technology in the healthcare field, daily monitoring of chronic disease patients has gradually shifted from traditional single hospital follow-ups to real-time home monitoring. Currently, monitoring systems based on wearable devices and smart terminals can continuously collect various physiological parameters such as blood pressure, blood sugar, and heart rate, and store and perform preliminary analysis through cloud platforms. These systems typically issue warnings for abnormal indicators by setting fixed threshold ranges, thus improving the efficiency and convenience of chronic disease management to some extent. Existing technologies mainly focus on the continuity of data collection, the stability of transmission, and simple rule-based warnings, providing basic technical support for the health management of chronic disease patients.

[0003] The existing technology has the following shortcomings: Existing systems rely on general thresholds and common abnormal patterns derived from large populations for judgment. When individual patients exhibit physiological changes that differ significantly from common symptoms, such as a diabetic patient experiencing a "painless myocardial infarction" without typical chest pain, their physiological parameters (such as heart rate and heart rate variability) may only show slight deviations that do not exceed general thresholds. Because the system cannot understand that such deviations constitute significant abnormalities relative to the individual patient's baseline, this leads to serious underreporting. This "blindness" to individual-specific abnormal patterns prevents the system from providing effective early warnings at critical moments, potentially delaying interventions in life-threatening situations and posing a significant healthcare safety risk. Summary of the Invention

[0004] The purpose of this invention is to provide an Internet of Things-based system for monitoring the physical examination data and complication care of patients with chronic diseases, in order to solve the problems mentioned above.

[0005] The objective of this invention can be achieved through the following technical solutions: An IoT-based monitoring system for physical examination data and complication care of patients with chronic diseases includes: The data acquisition module is used to collect patients' physical examination data in real time and construct an individualized physical examination data stream from the physical examination data; the physical examination data includes vital signs data, blood biochemical indicators, physical activity data and electrocardiogram waveform data; The feature map construction module, based on the physical examination data stream, extracts feature vectors representing physiological state by calculating the change patterns between various physical examination data items within different time windows, and constructs a dynamic physiological feature map. The anomaly detection module identifies anomalous shifts in multi-parameter collaborative patterns by calculating the degree of deviation between the feature vector distribution in the graph and the baseline distribution. The early warning generation module, based on the identified abnormal offsets, analyzes the spatial distribution characteristics of feature vectors and matches them with the clinical case database to generate early warning information that includes risk assessment and care plans. The nursing intervention module combines early warning information with the change trajectory of feature vectors to generate a monitoring report showing the evolution of complication risk. It also adjusts the abnormal judgment boundary of feature vectors based on the implementation effect of nursing plan to achieve preventive nursing intervention for complications.

[0006] As a further aspect of the present invention: the calculation of the variation patterns among various physical examination data items within different time windows specifically includes: The physical examination data stream is divided into three time scales: minute, hour, and day. Within each time scale window, the correlation strength between each physical examination data item is calculated, and the interaction between parameters is quantified by analyzing the synchronicity and lag of parameter changes. The coupling relationship between short-term fluctuations and long-term trends is analyzed. The change patterns of parameters found at different time scales are integrated to form a complete description of the characteristics of parameter synergistic change.

[0007] As a further aspect of the present invention: the extraction of feature vectors representing physiological states specifically includes: Based on the analysis of the variation patterns among parameters, parameter combinations with synergistic variation characteristics are selected. Feature indicators are extracted from three dimensions: time domain, frequency domain, and correlation domain. Time domain features include the amplitude and velocity of parameter changes, frequency domain features include the energy distribution characteristics of the main fluctuation cycle, and correlation domain features include the topological characteristics of the coupling strength between parameters. The feature indicators of the three dimensions are normalized and fused to generate a feature vector that can comprehensively characterize the physiological state.

[0008] As a further aspect of the present invention: the construction of the dynamic physiological feature map specifically includes: A topological connection relationship is constructed by calculating the similarity distance between feature vectors; the dynamic evolution path of feature vectors in time series is analyzed, and the movement trajectory and aggregation state of feature vectors in feature space are recorded; a state transition map of feature vectors is established, and a state transition relationship is constructed by tracking the positional changes of feature vectors at adjacent time points; based on the spatiotemporal distribution characteristics of feature vectors, a dynamic physiological feature map is formed that directly represents the changes in physiological state through the density distribution and evolution trajectory of feature vectors.

[0009] As a further aspect of the present invention: the degree of deviation between the feature vector distribution and the baseline distribution in the calculated spectrum specifically includes: The study compares and analyzes the spatial density distribution, temporal evolution distribution, and correlation structure distribution of feature vectors. By calculating the density change rate of feature vectors in different regions of the feature space, it identifies abnormal clustering or dispersion of distribution density. At the same time, it analyzes the continuity of feature vector evolution trajectory to detect trajectory interruption or abrupt change. The deviation indicators of the three dimensions are weighted and integrated to generate a comprehensive deviation evaluation result.

[0010] As a further aspect of the present invention: the identification of abnormal offsets in multi-parameter cooperative modes specifically includes: By analyzing the distance relationship and topological structure between feature vectors, feature representations of multi-parameter cooperative patterns are extracted; the evolution trajectory of cooperative patterns in feature space is tracked in real time; when the evolution trajectory of cooperative patterns is detected to exceed the normal fluctuation range, a graded early warning signal is generated based on the offset direction and amplitude; and the start time and evolution trend of abnormal offsets are identified through offset pattern analysis.

[0011] As a further aspect of the present invention: the matching of the spatial distribution features of the feature vectors with the clinical case database specifically includes: A quantitative description system for spatial distribution characteristics is established by calculating the clustering density, distribution boundary, and topological structure of feature vectors in the feature space to quantify spatial distribution characteristics. Based on the quantitative results of spatial distribution characteristics, multi-dimensional similarity matching is performed in a clinical case database. The credibility of the current matching is evaluated by analyzing the clinical validation results and historical warning accuracy of the matched cases. Based on the matching results and credibility evaluation, a case matching report with confidence index is generated.

[0012] As a further aspect of the present invention: the generation of early warning information including risk assessment and care plan specifically includes: Risk levels are determined based on the degree of deviation of feature vectors and case matching results; customized nursing plans are generated based on effective interventions in matched clinical cases and combined with individual patient characteristics; risk levels, nursing plans, matched case information, and confidence indicators are integrated into structured early warning information; and the processing priority of early warning information is determined according to risk levels and timeliness requirements.

[0013] As a further aspect of the present invention: the generation of the monitoring report demonstrating the evolution of complication risk specifically includes: The system integrates and displays the movement trajectory of feature vectors in the feature space with time axis information; constructs a multi-dimensional risk evolution map, and displays the dynamic development process of complication risk through density changes, direction shifts, and velocity fluctuations of feature vector trajectories; marks key time points and spatial locations where risk levels change significantly in the feature vector trajectories; and predicts the evolution trend of complication risk in the future based on the historical movement patterns of feature vectors through trajectory trend prediction analysis.

[0014] As a further aspect of the present invention: the preventive nursing intervention for complications specifically includes: Preventive nursing measures are initiated when the feature vector approaches but does not reach the abnormality judgment boundary; the intensity and content of the nursing plan are adjusted according to the real-time feedback of the feature vector; at the same time, a closed-loop monitoring system for intervention effect is established to continuously track the response of the feature vector after nursing intervention; and a preventive nursing knowledge base is established to accumulate successful preventive nursing cases.

[0015] The beneficial effects of this invention are: (1) This invention constructs a dynamic physiological feature map and analyzes the distribution patterns of feature vectors from multiple dimensions, including spatial density, temporal evolution, and correlation structure. This enables the identification of subtle collaborative pattern anomalies that are difficult to detect using traditional methods. Preventive care can be triggered when the feature vectors approach but do not reach the abnormal boundary, advancing the intervention time from the "already diseased" stage to the "potentially diseased" stage. This early warning mechanism based on multi-parameter collaborative analysis improves the early detection capability of chronic disease complications such as diabetic cardiovascular disease and hypertension target organ damage, thus gaining a valuable time window for clinical intervention.

[0016] (2) The abnormal judgment boundary is dynamically adjusted based on the feature vector change trajectory, so that the early warning standard is adaptively updated with changes in the patient's physiological state, overcoming the limitations of the traditional fixed threshold. At the same time, by building a preventive nursing knowledge base, intervention programs are continuously accumulated and optimized, enabling the system to have continuous learning capabilities. This personalized and adaptive management approach avoids over-medicalization and provides precise nursing care based on the patient's real-time status, improving the scientificity and effectiveness of chronic disease management. Attached Figure Description

[0017] The invention will now be further described with reference to the accompanying drawings.

[0018] Figure 1 This is a flowchart of the system of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 As shown, this invention is an Internet of Things-based monitoring system for physical examination data and complication care of patients with chronic diseases, comprising: The data acquisition module is used to collect patients' physical examination data in real time and construct an individualized physical examination data stream from the physical examination data; the physical examination data includes vital signs data, blood biochemical indicators, physical activity data and electrocardiogram waveform data; The feature map construction module, based on the physical examination data stream, extracts feature vectors representing physiological state by calculating the change patterns between various physical examination data items within different time windows, and constructs a dynamic physiological feature map. The anomaly detection module identifies anomalous shifts in multi-parameter collaborative patterns by calculating the degree of deviation between the feature vector distribution in the graph and the baseline distribution. The early warning generation module, based on the identified abnormal offsets, analyzes the spatial distribution characteristics of feature vectors and matches them with the clinical case database to generate early warning information that includes risk assessment and care plans. The nursing intervention module combines early warning information with the change trajectory of feature vectors to generate a monitoring report showing the evolution of complication risk. It also adjusts the abnormal judgment boundary of feature vectors based on the implementation effect of nursing plan to achieve preventive nursing intervention for complications.

[0021] In the data acquisition module, a data acquisition network is formed by deploying various sensing devices on different parts of the patient's body and in the environment to simultaneously collect the patient's vital signs, blood biochemical indicators, physical activity data, and electrocardiogram (ECG) waveform data. Specifically, a smart ECG patch is worn on the patient's torso to collect ECG waveforms; a multi-functional smart bracelet is worn on the wrist to collect blood oxygen saturation and body surface temperature; a smart blood pressure monitor and blood glucose meter are placed in the home environment to collect blood pressure and blood glucose values; and a smart scale is laid on the indoor floor to collect weight and body fat data. These devices continuously acquire analog physiological signals at a preset sampling frequency through built-in sensor elements, which are then converted into digital signals by the device's internal analog-to-digital converter.

[0022] A unified time synchronization mechanism is established to ensure time consistency of multi-source data. A master clock device is set up in the system, and all other acquisition devices synchronize with the master clock device via a wireless communication protocol. The master clock device broadcasts a time calibration signal to all devices in the network every five minutes, and each acquisition device adjusts its local clock upon receiving the signal. The start time, end time, and device identifier are recorded in each data packet. When the data is transmitted to the data processing center, the data collected by different devices are aligned to a unified timeline based on the timestamp information.

[0023] The system performs quality assessment and enhancement processing on the collected raw data. It calculates the signal-to-noise ratio, data integrity, and acquisition stability indicators for each data source in real time. When a data source's quality indicator is detected to be below a preset threshold, the system automatically triggers a data re-acquisition mechanism or switches to a backup acquisition device. For data loss caused by temporary signal interruptions, a time-series forecasting method is used to fill in the missing data and generate reasonable estimates of the missing intervals.

[0024] The enhanced data is standardized according to a unified format. The system defines a standard data structure and transmission protocol to convert heterogeneous data collected from different devices into unified JSON format data packets. Each data packet contains a patient identifier, collection timestamp, data type, data value, and quality identifier fields. The processed data is arranged in chronological order to form a continuous and complete personalized physical examination data stream, providing a standardized data foundation for subsequent feature extraction and analysis.

[0025] In the feature map construction module, the collected physical examination data stream undergoes preprocessing and multi-time-scale segmentation. Data preprocessing includes smoothing and filtering the raw data, and using a moving average method to eliminate instantaneous interference signals, ensuring data continuity and stability. Subsequently, the processed data stream is simultaneously divided into three time scales: minute-level windows (1 minute), hour-level windows (1 hour), and day-level windows (24 hours). Data within each time window is used to create independent datasets. The minute-level windows record the instantaneous fluctuation characteristics of the parameters, the hour-level windows reflect the periodic changes in the parameters, and the day-level windows display the long-term trends of the parameters. During the segmentation process, the integrity and continuity of the data at each time scale are ensured to avoid data loss or gaps.

[0026] At the minute-level timescale, the synchronicity of parameter changes is analyzed. The instantaneous response relationship between parameters is quantified by calculating the consistency ratio of different parameter change directions within a 1-minute window. Specifically, for each pair of parameters, the proportion of times they change in the same direction at the same time point is counted out of the total number of observations; a higher proportion indicates stronger synchronicity between parameters. At the hour-level timescale, the lag relationship of parameter changes is analyzed. Time series cross-correlation analysis is used to calculate the time difference between the peak occurrences of different parameter changes, and the optimal lag time is determined by finding the maximum value of the cross-correlation coefficient. At the daily timescale, the long-term trend coupling relationship of parameters is analyzed. Trends are extracted from parameter sequences within a 24-hour window, the consistency of trend lines for different parameters is compared, and the similarity index between trend lines is calculated. Finally, the analysis results from the three timescales are integrated to establish a complete description of the coordinated parameter changes.

[0027] Based on the description of parameter co-variation characteristics, feature indicators are extracted from three dimensions: time domain, frequency domain, and correlation domain. Time domain feature extraction includes calculations of both the magnitude and rate of change. The magnitude of change is characterized by calculating the range and standard deviation of the parameter within a fixed time window; the range reflects the overall range of parameter change, and the standard deviation characterizes the intensity of parameter fluctuation. The rate of change is measured using the first-order difference sequence of parameter values, calculating the mean and variance of the difference sequence to reflect the speed and stability of parameter change. Frequency domain feature extraction first performs a Fourier transform on the parameter sequence to identify the main fluctuation periods and calculates the energy distribution ratio of each frequency band, paying particular attention to frequency bands related to human physiological rhythms, such as the frequency range corresponding to physiological activities like breathing and heartbeat. Correlation domain feature extraction constructs a parameter correlation network, calculating topological indicators such as degree centrality, compactness centrality, and eigenvector centrality for each parameter node to quantify the importance and influence of the parameter in the co-variation network.

[0028] After normalizing the feature indicators extracted from the three dimensions, they are fused according to predetermined weights to generate a comprehensive feature vector. The normalization process uses a min-max normalization method to transform each feature indicator to the range of 0 to 1. During feature fusion, different weights are assigned based on the importance of each dimension: 0.4 for time-domain features, 0.3 for frequency-domain features, and 0.3 for correlation-domain features. The final feature vector is obtained through weighted summation. When constructing a dynamic physiological feature map based on the feature vectors, the cosine similarity between feature vectors is first calculated. Topological connections are established based on the similarity magnitude, and feature vectors with a similarity greater than 0.8 are grouped into the same cluster. Then, the positional changes of feature vectors over time are analyzed, recording their movement trajectories and clustering states. By tracking the positional changes of feature vectors at adjacent time points, a state transition probability matrix is ​​constructed to record the transition patterns between states. Finally, based on the spatiotemporal distribution density and evolution trajectory of the feature vectors, a dynamic physiological feature map is formed.

[0029] After the dynamic physiological feature atlas is established, the system regularly updates and optimizes it with the continuous input of new data. The update cycle is set to 24 hours, with the atlas recalculated at a fixed time each day. During the update process, new feature vectors are first added to the feature space, and the distribution density and clustering of the feature vectors are recalculated. Then, based on the data from the most recent 7 days, the state transition probability matrix is ​​adjusted to ensure that the atlas reflects the latest physiological state changes. Simultaneously, the system monitors the stability of the atlas. When a significant change in feature distribution is detected, the atlas reconstruction mechanism is automatically triggered to recalculate the topological connectivity and state transition paths of the feature vectors, ensuring the accuracy and timeliness of the atlas.

[0030] In the anomaly identification module, historical feature vector data of patients in a stable, healthy state are collected, with a time span of no less than 30 days. Distribution characteristics of this historical data are analyzed, comparing and contrasting it from three dimensions: spatial density distribution, temporal evolution distribution, and correlation structure distribution. The spatial density distribution benchmark is obtained by calculating the distribution density of feature vectors in each region of the feature space, dividing the feature space into several uniform grid cells, and statistically analyzing the frequency of feature vector occurrence in each cell. The temporal evolution distribution benchmark is established by analyzing the changing patterns of feature vectors over time, calculating the average amplitude and direction distribution of feature vector position changes at adjacent time points. The correlation structure distribution benchmark is established by analyzing the connection relationships between feature vectors, calculating the average distance and connection strength between different feature vectors.

[0031] In spatial density distribution deviation analysis, the feature space is first divided into 100 uniform grid cells. The distribution density of the current feature vector in each grid cell is calculated, i.e., the number of feature vectors per unit area. The current density distribution is compared with the baseline density distribution, and the density change rate of each grid cell is calculated. The density change rate is obtained by the ratio of the current density value to the baseline density value. When the ratio exceeds 1.5 or is less than 0.5, an abnormal density change is considered to have occurred in that area. Simultaneously, abnormal clustering of feature vectors in space is identified. When the density value of a certain area is significantly higher than the surrounding areas and exceeds twice the baseline density value, it is marked as an abnormal clustering area. Similarly, when the density value of a certain area is significantly less than half of the baseline density value, it is marked as an abnormal dispersion area. The identification of these abnormal density areas provides a spatial dimension basis for subsequent abnormal offset judgments.

[0032] In the temporal evolution distribution deviation analysis, the focus is on the continuity of the eigenvector evolution trajectory. First, a temporal trajectory sequence of the eigenvectors is established, recording the position coordinates of the eigenvector in the feature space at each time point. Then, the movement distance and direction between adjacent time points are calculated. When the movement distance at three consecutive time points exceeds three times the baseline movement distance, a trajectory abrupt change is considered to have occurred. Simultaneously, the continuity of the trajectory is analyzed; when an eigenvector experiences a positional jump at a certain time point, and the jump distance exceeds five times the baseline movement distance, it is marked as a trajectory interruption. Furthermore, a trajectory smoothness index is calculated, evaluated by the degree of change in trajectory curvature; when the curvature change exceeds twice the baseline value, abnormal fluctuations in the trajectory are considered. These abnormal changes in temporal evolution characteristics reflect unstable changes in physiological states.

[0033] In the analysis of deviations in the distribution of association structures, changes in the connectivity relationships between feature vectors are analyzed. First, an association network of feature vectors is established, with feature vectors as nodes and the similarity between feature vectors as edge weights. The topological characteristics of the current association network are calculated, including average path length, clustering coefficient, and degree distribution. These topological characteristics are compared with a baseline network, and their relative rates of change are calculated. When the average path length increases by more than 20% of the baseline value, network connectivity is considered to have decreased; when the clustering coefficient decreases by more than 15% of the baseline value, the degree of network clustering is considered to have decreased. Simultaneously, changes in the position of feature vectors within the association network are analyzed. When a feature vector's centrality index changes significantly within the network, such as a change in degree centrality exceeding 25% of the baseline value, the association relationship of that feature vector is considered abnormal. These changes in association structures reflect alterations in multi-parameter collaborative patterns.

[0034] In the comprehensive deviation assessment, the deviation indicators of the three dimensions are weighted and integrated. The weight of spatial density distribution deviation is set to 0.4, the weight of temporal evolution distribution deviation is set to 0.35, and the weight of association structure distribution deviation is set to 0.25. The deviation value of each dimension is calculated by the weighted average of all abnormal indicators under that dimension. Spatial density distribution deviation is jointly determined by the density change rate anomaly indicator and the number of abnormal regions indicator, where the density change rate anomaly indicator has a weight of 0.6 and the number of abnormal regions indicator has a weight of 0.4. Temporal evolution distribution deviation consists of trajectory mutation indicator, trajectory interruption indicator, and trajectory smoothness indicator, with weights of 0.4, 0.4, and 0.2, respectively. Association structure distribution deviation consists of network connectivity indicator, clustering degree indicator, and centrality change indicator, with weights of 0.4, 0.3, and 0.3, respectively. The final comprehensive deviation is obtained by weighted summation of the deviations of the three dimensions. When the comprehensive deviation exceeds 0.7, an abnormal offset warning is triggered.

[0035] In the process of identifying abnormal deviations, the starting time of the abnormal deviation is first determined. By retrospectively analyzing the change trajectory of the feature vector, the time point when the overall deviation first exceeds 0.5 is found as the starting time of the abnormal deviation. Then, the evolution trend of the abnormal deviation is analyzed, and the slope of the change in the overall deviation over the past 6 hours is calculated. When the slope is positive and greater than 0.05, the abnormal deviation is considered to be aggravating; when the slope is negative, the abnormal deviation is considered to be mitigating. Based on the magnitude and trend of the overall deviation, the warning level is divided into three levels: a yellow warning is issued when the overall deviation is between 0.7 and 0.8; an orange warning is issued when it is between 0.8 and 0.9; and a red warning is issued when it exceeds 0.9. At the same time, the characteristic patterns of the abnormal deviation are recorded, including the main dimensions of occurrence, duration, and trend of change, to provide a basis for subsequent nursing interventions.

[0036] In the early warning generation module, the spatial distribution characteristics of feature vectors are quantitatively described. The feature space is divided into several uniform grid cells, the size of which is adaptively adjusted according to the distribution density of feature vectors. The number of feature vectors in each grid cell is calculated to obtain the spatial clustering density distribution. The degree of spatial clustering is quantified by statistically analyzing the ratio of the number of feature vectors in each grid cell to the total number of feature vectors. Simultaneously, the distribution boundary of feature vectors is calculated by finding the maximum and minimum values ​​of feature vectors in each dimension of the feature space to determine the boundary range of the distribution area. For the quantification of topology, the connectivity relationships between feature vectors are calculated, the average distance between each feature vector and its nearest neighbor is statistically analyzed, and the number of clusters formed by the feature vectors and the size distribution of each cluster are also determined. These quantitative indicators together constitute a complete descriptive system for the spatial distribution characteristics.

[0037] When performing multi-dimensional similarity matching in a clinical case database, a matching index system is first established. The matching index includes three dimensions: spatial clustering density similarity, distribution boundary similarity, and topological structure similarity. Spatial clustering density similarity is obtained by comparing the density distribution curves of the current feature vector with those of case feature vectors in the database, calculating the ratio of the overlapping area of ​​the two density distribution curves. Distribution boundary similarity is obtained by comparing the boundary positions and shapes of the distribution regions, calculating the ratio of the overlapping volume of the two distribution regions to the total volume. Topological structure similarity is obtained by comparing the connectivity relationships between feature vectors, calculating the similarity between two feature vector sets in terms of nearest neighbor distance distribution and clustering structure. The similarity scores of the three dimensions are then weighted and averaged according to predetermined weights of 0.4, 0.3, and 0.3 to obtain a comprehensive similarity score.

[0038] When assessing the reliability of matching results, multiple evaluation factors are considered. First, the clinical validation results of the matched cases are analyzed, and the historical accuracy rate of the warnings for each case is statistically analyzed, i.e., the degree of consistency between the case's characteristic pattern and the actual occurrence of complications. The accuracy rate of the matched cases in the past 30 warnings is calculated as a baseline reliability indicator. The time relevance of the matching is also considered, prioritizing recently validated cases; for cases that have not been validated for more than 90 days, their reliability weight is reduced. Furthermore, the completeness of the matching is analyzed; when the similarity scores across the three dimensions are relatively balanced, the matching result is considered more reliable. Finally, considering the number of matched cases, when multiple cases are successfully matched simultaneously, the weighted average of the reliability scores of each case is taken as the final reliability assessment result.

[0039] Risk levels are determined based on the degree of deviation of the eigenvector and the case matching results. Risk levels are divided into three levels: low risk, medium risk, and high risk. The determination of the risk level is based on two main factors: the eigenvector deviation score and the case matching reliability score. The eigenvector deviation score is obtained from the previously calculated overall deviation, and the case matching reliability score is obtained through a matching reliability assessment. A low risk level is defined as a eigenvector deviation score below 0.6 and a case matching reliability score below 0.7; a medium risk level is defined as a eigenvector deviation score between 0.6 and 0.8 or a case matching reliability score between 0.7 and 0.9; and a high risk level is defined as a eigenvector deviation score exceeding 0.8 or a case matching reliability score exceeding 0.9. If the two indicators belong to different risk levels, the higher risk level is taken as the final result.

[0040] Based on effective interventions from matched clinical cases, customized nursing care plans are generated, taking into account individual patient characteristics. First, historically proven effective interventions are extracted from successfully matched cases, including medication regimens, exercise recommendations, dietary control, and monitoring frequency. Then, adjustments are made according to the patient's individual characteristics, considering factors such as age, weight, underlying diseases, and history of drug allergies. For elderly patients, exercise intensity and medication dosage are appropriately reduced; for patients with impaired renal function, drugs metabolized by the kidneys are avoided; and for patients with specific food allergies, dietary recommendations are adjusted. The patient's current living environment and medical conditions are also considered to ensure the feasibility of the generated nursing care plan. Finally, the interventions are prioritized according to implementation time and importance to form a complete nursing care plan.

[0041] Risk levels, nursing protocols, matched case information, and confidence indices are integrated into structured early warning information. This information uses a unified data structure, including a basic information section, a risk assessment section, a nursing recommendation section, and a reference section. The basic information section records patient identification information, the warning generation time, and a unique warning identifier. The risk assessment section includes the risk level, risk description, and expected development trend. The nursing recommendation section details the implementation content, timeline, and precautions for each nursing intervention. The reference section provides basic information about the matched cases, historical verification results, and the confidence index for this match. The processing priority of early warning information is determined based on risk level and timeliness requirements: high-risk warnings are pushed out immediately, medium-risk warnings are processed within 2 hours, and low-risk warnings are processed within 24 hours. A tracking mechanism for early warning information is also established to record the processing status and subsequent effectiveness evaluation results.

[0042] In the nursing intervention module, when generating monitoring reports, the movement trajectory of feature vectors in the feature space is first integrated and displayed with time axis information. A spatiotemporal coordinate system is established, with the horizontal axis representing the time series and the vertical axis representing the position coordinates in the feature space. For each feature vector, its three-dimensional coordinate position in the feature space at different time points is recorded, forming a continuous movement trajectory. The feature vector positions at adjacent time points are connected by curves to form a complete movement path. During display, different colors are used to distinguish trajectories for different time periods; recent trajectories are displayed in warm colors, and historical trajectories are displayed in cool colors. At the same time, time markers are marked on the trajectory lines, with a time tag set every 6 hours, clearly showing the movement process of the feature vectors over time. To enhance the visualization effect, density rendering is added to areas with drastic trajectory changes, and the color depth reflects the length of time the feature vector stays in that area.

[0043] When constructing a multi-dimensional risk evolution map, the dynamic development process of complication risk is displayed from three dimensions. In the density change dimension, the number of feature vectors distributed per unit volume in the feature space is calculated, and the density distribution is displayed as a heatmap. Higher density indicates more frequent occurrence of feature vectors in that area, potentially corresponding to a specific physiological state. In the directional offset dimension, the overall directional trend of the feature vector movement trajectory is analyzed, and the angle between the trajectory direction and the direction of the normal healthy area is calculated; a larger angle indicates a more severe offset. In the velocity fluctuation dimension, the movement velocity of the feature vector in the feature space is calculated, including instantaneous velocity and average velocity; abnormal acceleration or deceleration of velocity may indicate changes in health status. Information from these three dimensions is presented in a layered manner within the same map, and users can select to display one or more dimensions as needed.

[0044] When marking key time points in the feature vector trajectory, the criteria for a significant change in risk level are first determined. A significant change is considered to have occurred when the overall risk score of the feature vector changes by more than 0.3 within 2 hours, or when the risk level upgrades at 3 consecutive time points. Conspicuous markers are placed at these time points, recording the specific time and location coordinates. Simultaneously, the state characteristics of the feature vector before and after this time point are recorded, including the direction of movement, speed, and surrounding density. Based on the historical movement patterns of the feature vector, a time series prediction method is used to analyze the trajectory's development trend. By calculating the movement direction and speed of the trajectory over the past 24 hours, combined with periodic change patterns, the possible movement path and final destination area of ​​the feature vector within the next 12 hours are predicted. The prediction results are displayed on the map as a dashed trajectory, with the prediction confidence interval marked.

[0045] Preventive nursing measures are initiated when the feature vector approaches but does not reach the abnormality threshold. A warning threshold is set at 80% of the abnormality threshold. When the distance between the feature vector and the abnormality threshold is less than this threshold, the system automatically triggers the preventive nursing mechanism. First, the movement trend of the feature vector is assessed. If the feature vector is moving towards the abnormal area, primary prevention measures are immediately initiated; if the feature vector hovers near the warning threshold, secondary prevention measures are initiated. Preventive measures include increasing monitoring frequency, adjusting lifestyle habits, and providing preventive treatment. The system dynamically adjusts the intensity of preventive measures based on the distance and movement speed of the feature vector from the abnormal area; the closer the distance and the faster the movement, the stronger the preventive measures. Simultaneously, the initiation time, measures, and initial state of each preventive nursing intervention are recorded to provide basic data for effectiveness evaluation.

[0046] The intensity and content of the nursing care plan are adjusted based on real-time feedback from the feature vector. A feedback evaluation mechanism is established to assess the effectiveness of the nursing care plan every 4 hours. Evaluation indicators include whether the movement direction of the feature vector deviates from the abnormal area, whether the movement speed slows down, and whether the distance from the abnormal area increases. If the feature vector continues to move towards the abnormal area without slowing down, the intensity of the nursing care plan is increased, such as increasing the medication dosage or extending the monitoring time. If the feature vector begins to return to the healthy area, the intensity of nursing care is appropriately reduced to avoid over-intervention. Simultaneously, the content of the nursing care plan is adjusted according to the specific location of the feature vector. When the feature vector is in a specific area, special nursing measures targeting the physiological characteristics of that area are activated. All adjustment records are meticulously saved, forming a complete history of nursing care plan adjustments.

[0047] Establish a closed-loop monitoring system for intervention effects to continuously track the response of eigenvectors after nursing intervention. The monitoring cycle is set at 8 hours, and an effectiveness evaluation report is generated at the end of each cycle. The evaluation content includes quantitative indicators such as changes in the location of eigenvectors, the degree of improvement in their movement trajectory, and changes in risk level. The effectiveness score of the nursing intervention is calculated by comparing the state of the eigenvectors before and after the intervention. When the effectiveness score is below 0.6, the current nursing plan is considered ineffective, and a new intervention strategy needs to be developed. Simultaneously, the response time of the eigenvectors to the nursing intervention is monitored, recording the time interval from the implementation of nursing care to the onset of a response in the eigenvectors, to assess the timeliness of the nursing intervention. An early warning mechanism is established; if the eigenvectors do not show significant improvement within 24 hours after the nursing intervention, an alert for escalation of nursing care is issued to ensure the effectiveness of the intervention.

[0048] Establish a preventative nursing knowledge base to accumulate successful preventative nursing cases. Each case record includes basic patient information, initial state of the feature vector, implemented nursing interventions, the response process of the feature vector, and final effect evaluation. Cases are categorized and stored according to the location and movement trend of the feature vector for easy and rapid matching later. A case effectiveness evaluation system is established in the knowledge base, classifying cases into three levels—excellent, good, and average—based on the degree and speed of improvement of the feature vector. The cases in the knowledge base are regularly updated and maintained, eliminating ineffective cases and adding new successful ones. A case retrieval mechanism is also established to quickly find similar successful cases in the knowledge base based on the current state characteristics of the feature vector, providing a reference for developing nursing plans. The knowledge base undergoes a comprehensive evaluation and optimization quarterly to ensure the accuracy and practicality of its content.

[0049] The working principle of this invention is as follows: A data acquisition module collects multi-source physical examination data in real time, including patients' vital signs, blood biochemical indicators, physical activity data, and electrocardiogram waveforms, and constructs an individualized physical examination data stream. Subsequently, a feature map construction module performs multi-timescale analysis on the data stream, extracting time-domain, frequency-domain, and correlation-domain features and fusing them to generate feature vectors, thereby constructing a dynamic physiological feature map. An anomaly identification module identifies abnormal deviations in multi-parameter collaborative patterns by calculating the deviation of the feature vector distribution from the baseline distribution. An early warning generation module analyzes the spatial distribution characteristics of the feature vectors based on the abnormal deviation and matches them with a clinical case database to generate structured early warning information containing risk assessment and customized nursing plans. Finally, a nursing intervention module combines the early warning information with the feature vector change trajectory to generate a monitoring report and dynamically adjusts the abnormal judgment boundary based on the nursing plan's implementation effect, achieving preventative nursing intervention for complications and forming a closed-loop management system from data acquisition to intervention feedback.

[0050] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A system for monitoring physical examination data and complication care of patients with chronic diseases based on the Internet of Things, characterized in that, include: The data acquisition module is used to collect patients' physical examination data in real time and build an individualized physical examination data stream from the physical examination data; The physical examination data includes vital signs data, blood biochemical indicators, physical activity data, and electrocardiogram waveform data; The feature map construction module, based on the physical examination data stream, extracts feature vectors representing physiological state by calculating the change patterns between various physical examination data items within different time windows, and constructs a dynamic physiological feature map. The anomaly detection module identifies anomalous shifts in multi-parameter collaborative patterns by calculating the degree of deviation between the feature vector distribution in the graph and the baseline distribution. The early warning generation module, based on the identified abnormal offsets, analyzes the spatial distribution characteristics of feature vectors and matches them with the clinical case database to generate early warning information that includes risk assessment and care plans. The nursing intervention module combines early warning information with the change trajectory of feature vectors to generate a monitoring report showing the evolution of complication risk. It also adjusts the abnormal judgment boundary of feature vectors based on the implementation effect of the nursing plan, thereby realizing preventive nursing intervention for complications.

2. The IoT-based chronic disease patient physical examination data and complication nursing monitoring system according to claim 1, characterized in that, The calculation of the variation patterns among various physical examination data items within different time windows specifically includes: The physical examination data stream is divided into three time scales: minute, hour, and day. Within each time scale window, the correlation strength between each physical examination data item is calculated, and the interaction between parameters is quantified by analyzing the synchronicity and lag of parameter changes. The coupling relationship between short-term fluctuations and long-term trends is analyzed. The change patterns of parameters found at different time scales are integrated to form a complete description of the characteristics of parameter synergistic change.

3. The IoT-based chronic disease patient physical examination data and complication nursing monitoring system according to claim 1, characterized in that, The extraction of feature vectors representing physiological states specifically includes: Based on the analysis of the variation patterns among parameters, parameter combinations with synergistic variation characteristics are selected. Feature indicators are extracted from three dimensions: time domain, frequency domain, and correlation domain. Time domain features include the amplitude and velocity of parameter changes, frequency domain features include the energy distribution characteristics of the main fluctuation cycle, and correlation domain features include the topological characteristics of the coupling strength between parameters. The feature indicators of the three dimensions are normalized and fused to generate a feature vector that can comprehensively characterize the physiological state.

4. The IoT-based chronic disease patient physical examination data and complication nursing monitoring system according to claim 1, characterized in that, The construction of the dynamic physiological feature map specifically includes: A topological connection relationship is constructed by calculating the similarity distance between feature vectors; the dynamic evolution path of feature vectors in time series is analyzed, and the movement trajectory and aggregation state of feature vectors in feature space are recorded; a state transition map of feature vectors is established, and a state transition relationship is constructed by tracking the positional changes of feature vectors at adjacent time points; based on the spatiotemporal distribution characteristics of feature vectors, a dynamic physiological feature map is formed that directly represents the changes in physiological state through the density distribution and evolution trajectory of feature vectors.

5. The IoT-based chronic disease patient physical examination data and complication nursing monitoring system according to claim 1, characterized in that, The degree of deviation between the feature vector distribution and the baseline distribution in the calculated graph specifically includes: The study compares and analyzes the spatial density distribution, temporal evolution distribution, and correlation structure distribution of feature vectors. By calculating the density change rate of feature vectors in different regions of the feature space, it identifies abnormal clustering or dispersion of distribution density. At the same time, it analyzes the continuity of feature vector evolution trajectory to detect trajectory interruption or abrupt change. The deviation indicators of the three dimensions are weighted and integrated to generate a comprehensive deviation evaluation result.

6. The IoT-based chronic disease patient physical examination data and complication nursing monitoring system according to claim 1, characterized in that, The identification of anomalous shifts in multi-parameter collaborative patterns specifically includes: By analyzing the distance relationship and topological structure between feature vectors, feature representations of multi-parameter cooperative patterns are extracted; the evolution trajectory of cooperative patterns in feature space is tracked in real time; when the evolution trajectory of cooperative patterns is detected to exceed the normal fluctuation range, a graded early warning signal is generated based on the offset direction and amplitude; and the start time and evolution trend of abnormal offsets are identified through offset pattern analysis.

7. The IoT-based chronic disease patient physical examination data and complication nursing monitoring system according to claim 1, characterized in that, The matching process, which involves analyzing the spatial distribution characteristics of feature vectors and matching them with a clinical case database, specifically includes: A quantitative description system for spatial distribution characteristics is established by calculating the clustering density, distribution boundary, and topological structure of feature vectors in the feature space to quantify spatial distribution characteristics. Based on the quantitative results of spatial distribution characteristics, multi-dimensional similarity matching is performed in a clinical case database. The credibility of the current matching is evaluated by analyzing the clinical validation results and historical warning accuracy of the matched cases. Based on the matching results and credibility evaluation, a case matching report with confidence index is generated.

8. The IoT-based chronic disease patient physical examination data and complication nursing monitoring system according to claim 1, characterized in that, The generation of early warning information, which includes risk assessment and care plans, specifically includes: Risk levels are determined based on the degree of deviation of feature vectors and case matching results; customized nursing plans are generated based on effective interventions in matched clinical cases and combined with individual patient characteristics; risk levels, nursing plans, matched case information, and confidence indicators are integrated into structured early warning information; and the processing priority of early warning information is determined according to risk levels and timeliness requirements.

9. The IoT-based chronic disease patient physical examination data and complication nursing monitoring system according to claim 1, characterized in that, The generation of the monitoring report, which displays the evolution of complication risk, specifically includes: The system integrates and displays the movement trajectory of feature vectors in the feature space with time axis information; constructs a multi-dimensional risk evolution map, and displays the dynamic development process of complication risk through density changes, direction shifts, and velocity fluctuations of feature vector trajectories; marks key time points and spatial locations where risk levels change significantly in the feature vector trajectories; and predicts the evolution trend of complication risk in the future based on the historical movement patterns of feature vectors through trajectory trend prediction analysis.

10. The IoT-based chronic disease patient physical examination data and complication nursing monitoring system according to claim 1, characterized in that, The preventive nursing interventions for complications specifically include: Preventive nursing measures are initiated when the feature vector approaches but does not reach the abnormality judgment boundary; the intensity and content of the nursing plan are adjusted according to the real-time feedback of the feature vector; at the same time, a closed-loop monitoring system for intervention effect is established to continuously track the response of the feature vector after nursing intervention; and a preventive nursing knowledge base is established to accumulate successful preventive nursing cases.

Images