A warning system and method for multiple diseases co-infected in the elderly

Abstract

This invention relates to the field of healthcare informatics and discloses an early warning system and method for multiple comorbidities in the elderly. The system includes: a data alignment module for resampling asynchronous physiological signal streams based on a system clock to generate a synchronous physiological dataset; an association matrix generation module for calculating the cross-correlation coefficients between physiological parameters to generate a real-time association matrix; a baseline management module for storing physiological baseline maps that record the health characteristics and compensatory boundaries of the elderly; an offset determination module for calculating the topological offset representation value between the real-time association matrix and the physiological baseline map; and an early warning module for outputting an early warning signal based on the rate of change of the offset representation value. This invention utilizes multidimensional parametric topological mapping to capture functional decoupling characteristics, effectively identifying the pathological deterioration trend masked by multiple concurrent diseases, and outputting a risk warning before physiological compensation collapses.

Description

Technical Field

[0001] This invention relates to an early warning system and method for multiple comorbidities in the elderly, belonging to the field of healthcare informatics technology. Background Technology

[0002] Currently, continuous monitoring of physiological indicators constitutes the core means of early warning. The mainstream approach adopts a multi-level threshold monitoring method based on a single physiological parameter, selecting the safe range of indicators such as heart rate, blood oxygen saturation, and respiratory rate as the judgment benchmark. When the real-time values ​​collected by the sensor cross the preset boundary, an early warning signal is generated. For elderly people with multiple comorbidities, the body's internal compensatory mechanisms cause different pathological states to interfere with the same physiological parameter in a directional conflict. In order to maintain the sensitivity of the early warning, the existing approach lowers the trigger threshold. The technical cost is that it ignores the steady-state distribution of parameters due to compensatory antagonism. This causes the true pathological evolution to be missed under the cover of mutual cancellation of indicators. The evolution trajectory of physiological parameters on the time axis maintains coupling correlation through specific pathological constraints. Before the physiological function suffers substantial failure, the covariance correlation characteristics between multidimensional data streams are distorted before the single physical amplitude.

[0003] Industry experts have attempted to improve accuracy by increasing monitoring dimensions or using logical rule tables, but this linear accumulation scheme increases the consumption of computing resources. Due to the lack of spatiotemporal alignment of multi-source heterogeneous physiological flows in an asynchronous sampling environment, multi-dimensional data analysis faces the problem of insufficient computational confidence and cannot handle baseline drift caused by the static nature of monitoring benchmarks. For example, Chinese invention patent application CN121287082A discloses a physiological monitoring and early warning system that integrates multi-source data. It uses medical semantic representation and statistical feature stacking to improve early warning coverage. The underlying judgment logic highly depends on medical guidelines to define prior rules and discrete medical events. The compensatory increase in heart rate due to multiple diseases masks the downward trend of absolute blood pressure. The rule-guided semantic analysis architecture treats multiple physiological flows as orthogonal independent components, making it difficult to penetrate the compensatory masking layer between physiological parameters and capture the coherent decoupling characteristics between deep physiological systems. As a result, during the compensatory window period before the substantial collapse of physiological function, the existing system is unable to output early warning based on the distortion of underlying structural features, and the static monitoring benchmark cannot adapt to the nonlinear evolution caused by the natural decline of function.

[0004] Therefore, how to select a data processing architecture with spatiotemporal alignment capabilities and calculate the evolution trend of coherent features among multidimensional physiological flows in order to establish an adaptive monitoring mechanism that can offset physiological function drift has become the technical problem to be solved by this invention. Summary of the Invention

[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: an early warning system for multiple comorbidities in the elderly, the system comprising: The data alignment module is used to access multiple physiological signal streams with asynchronous sampling frequencies and obtain the global system clock as a unified time reference. It performs resampling alignment on the multiple physiological signal streams, including interpolation operations, to generate a synchronous physiological dataset that is tightly aligned in the time domain step. The correlation matrix generation module is used to calculate the normalized cross-correlation coefficient between any two physiological parameters in the synchronous physiological dataset within a sliding window of a preset length, and generate a real-time correlation matrix that characterizes the coupling and compensatory state of the multidimensional physiological system of the elderly subjects. The baseline management module is used to manage the pre-established physiological baseline map. The physiological baseline map records the inherent distribution characteristics of various physiological parameters of the elderly subjects in a non-disease-causing healthy state, as well as the collaborative compensatory evolution boundary between various physiological systems, in a matrix topology structure. The offset determination module is used to perform Euclidean space topological mapping between the real-time correlation matrix and the physiological baseline map. By calculating the arithmetic square root of the sum of squares of the differences between the corresponding elements in the real-time correlation matrix and the physiological baseline map, the offset characterization value representing the degree of decoupling of physiological function is obtained. The early warning and forecasting module is used to identify the trajectory of accelerated deterioration of physiological function caused by the concurrent masking of multiple diseases based on the rate of change of the offset characterization value within the current time step, and output an early warning signal indicating that the elderly subjects are in the period of compensatory collapse when the rate of change of the numerical value continues to exceed the evolution threshold for 10 seconds.

[0006] Preferably, the system further includes a baseline adaptive correction module, which is used to correct the physiological baseline map in the baseline management module using a forgetting operator with decay weights; the baseline adaptive correction module adjusts the node weights of the physiological baseline map at a step frequency of 24 hours by acquiring historical physiological data sequences and calculating the long-term drift components of correlation, in order to offset the baseline offset error caused by the natural decline of human physiological functions.

[0007] Preferably, the data alignment module is further subdivided into the following processing units: an asynchronous buffer unit, used for non-blocking buffering of physiological signal streams with different sampling frequencies; and a time-stamp normalization unit, used to obtain the system clock and interpolate the data frames in the asynchronous buffer unit based on the system clock, so that each physiological signal stream is aligned on the same time axis.

[0008] Preferably, the correlation matrix generation module is further subdivided into the following processing units: a window control unit, used to extract synchronous physiological dataset segments with a fixed duration of 5 to 30 minutes; and a normalization operation unit, used to perform covariance operation on the synchronous physiological dataset segments and eliminate dimensions to generate a real-time correlation matrix characterizing the degree of linear correlation between parameters.

[0009] Preferably, the physiological baseline map managed by the baseline management module includes coupled nodes consisting of circulatory system parameters, respiratory system parameters, and metabolic system parameters; the logical weights between the coupled nodes are used to characterize the collaborative compensation boundaries between the physiological systems of the elderly subjects in a non-pathological state, and serve as the global constraint baseline for the offset determination module to perform topological mapping.

[0010] Preferably, the offset determination module further includes a conflict discrimination logic unit, which is used to identify data coupling anomalies caused by the concurrent occurrence of multiple diseases by calculating the overall structural similarity of the real-time correlation matrix relative to the physiological baseline map when a single physiological parameter experiences masking fluctuations, and to avoid the risk of missed reports by maintaining the output continuity of the offset characterization value after the anomaly is identified.

[0011] Preferably, the early warning and forecasting module is further subdivided into the following processing units: a trend extraction unit, used to perform first-order derivative calculation on the offset characterization value to obtain the trend value representing the rate of pathological deterioration; and a threshold comparison unit, used to dynamically compare the trend value with the stored evolution threshold to determine whether the offset characterization value is in an accelerated deviation state.

[0012] Preferably, the early warning and forecasting module further includes a trend prediction unit, which is used to predict the numerical evolution range of the offset characterization value in the next 4 to 12 hours based on the historical sequence of the offset characterization value through a state space mapping model, and feeds back the numerical evolution range to the early warning and forecasting module to adjust the sensitivity of the evolution threshold.

[0013] Preferably, the synchronized physiological dataset includes at least heart rate, blood oxygen saturation, respiratory rate, and blood pressure; the system is deployed on the monitoring center server and is used to receive and parse data packets reported by wireless physiological sensors through a network interface to realize continuous physiological function monitoring and pathological trend identification for elderly people with multiple comorbidities.

[0014] An early warning method for an early warning system for multiple comorbidities in the elderly, the method comprising the following steps: Step S1101: Acquire multiple physiological signal streams with asynchronous sampling frequencies, and use the global system clock as a unified time reference to perform resampling alignment on the multiple physiological signal streams, including interpolation value operation, to generate a synchronous physiological dataset that is tightly aligned in the time domain step size. Step S1102: Calculate the normalized cross-correlation coefficient between any two physiological parameters in the synchronous physiological dataset within a sliding window of a preset length, and generate a real-time correlation matrix that characterizes the coupling compensation state of the multidimensional physiological system of the elderly subjects. Step S1103: Retrieve the physiological baseline map of the pre-recorded physiological parameters of the elderly subjects in a healthy state without disease, and obtain the offset characterization value that characterizes the degree of decoupling of physiological function by calculating the arithmetic square root of the sum of the squares of the differences between the real-time correlation matrix and the corresponding elements in the physiological baseline map. Step S1104: Calculate the rate of change of the offset characterization value within the current time step, and when the rate of change of the value is found to exceed the evolution threshold for 10 seconds, output a warning signal indicating that the elderly subject is in the period of compensatory collapse risk.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. In the early warning of multiple diseases in the elderly, the spatiotemporal alignment unit receives multi-source physiological data streams with sampling frequencies of varying magnitudes. It generates a real-time state matrix with a unified time reference through cubic spline interpolation. This non-uniform temporal interpolation processing based on confidence weights eliminates data matrix distortion caused by asynchronous sampling from different monitoring devices and network transmission jitter, reconstructing the originally fragmented discrete physiological parameters into a high-fidelity temporal data sequence. This processing mechanism solves the data alignment bottleneck faced by medical information systems when accessing heterogeneous monitoring terminals, and supports subsequent underlying data analysis of the interactive evolution trend of multiple diseases.

[0016] 2. The coherence tensor generation unit calculates the dynamic correlation coefficients between data streams of each dimension in the real-time state matrix and generates a real-time coherence tensor that characterizes the coupling strength of physiological parameters. This mechanism breaks the mathematical assumption in existing technologies that physiological parameters are mutually orthogonal and operate independently, and instead focuses on the entangled state between multiple data streams. When the target object experiences systemic pathological deterioration, the system captures the abnormal decoupling characteristics of data streams caused by complications by monitoring the collapse of the inherent covariance relationship between various physiological indicators. This dimensional leap from one-dimensional amplitude monitoring to high-dimensional feature extraction enables the system to accurately identify the microscopic signals of disease deterioration during the pathological compensation period before each independent physiological indicator crosses the absolute threshold.

[0017] 3. The topology deviation determination unit maps the real-time coherence tensor to the benchmark constraint map stored in memory. It quantifies the degree of deviation of the pathological state by calculating the Euclidean space topological distance between the two. This path, which transforms complex nonlinear pathological conflicts into a calculation path of a single topological distance scalar, solves the alarm conflicts and redundancy caused by the traditional single-variable multi-level threshold monitoring logic when facing fluctuations in data of multiple co-occurring diseases. This scheme extracts the first derivative of the topological distance as the early warning criterion, thereby achieving a time lead for capturing deterioration trends. From the processing architecture perspective, it avoids the risk of systemic underreporting caused by the cross-covering of multiple diseases and improves the depth of medical monitoring data flow. Attached Figure Description

[0018] Figure 1 This is a flowchart of the early warning processing of multidimensional parameter topology association in this invention; Figure 2 This is a diagram illustrating the data interaction and modular logic operation of the early warning system of this invention.

[0019] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0021] An early warning system for multiple comorbidities in the elderly, the system comprising: The data alignment module is used to access multiple physiological signal streams with asynchronous sampling frequencies and obtain the global system clock as a unified time reference. It performs resampling alignment on the multiple physiological signal streams, including interpolation operations, to generate a synchronous physiological dataset that is tightly aligned in the time domain step. The correlation matrix generation module is used to calculate the normalized cross-correlation coefficient between any two physiological parameters in the synchronous physiological dataset within a sliding window of a preset length, and generate a real-time correlation matrix that characterizes the coupling and compensatory state of the multidimensional physiological system of the elderly subjects. The baseline management module is used to manage the pre-established physiological baseline map. The physiological baseline map records the inherent distribution characteristics of various physiological parameters of the elderly subjects in a non-disease-causing healthy state, as well as the collaborative compensatory evolution boundary between various physiological systems, in a matrix topology structure. The offset determination module is used to perform Euclidean space topological mapping between the real-time correlation matrix and the physiological baseline map. By calculating the arithmetic square root of the sum of squares of the differences between the corresponding elements in the real-time correlation matrix and the physiological baseline map, the offset characterization value representing the degree of decoupling of physiological function is obtained. The early warning and forecasting module is used to identify the trajectory of accelerated deterioration of physiological function caused by the concurrent masking of multiple diseases based on the rate of change of the offset characterization value within the current time step, and output an early warning signal indicating that the elderly subjects are in the period of compensatory collapse when the rate of change of the numerical value continues to exceed the evolution threshold for 10 seconds.

[0022] Preferably, the system further includes a baseline adaptive correction module, which is used to correct the physiological baseline map in the baseline management module using a forgetting operator with decay weights; the baseline adaptive correction module adjusts the node weights of the physiological baseline map at a step frequency of 24 hours by acquiring historical physiological data sequences and calculating the long-term drift components of correlation, in order to offset the baseline offset error caused by the natural decline of human physiological functions.

[0023] Preferably, the data alignment module is further subdivided into the following processing units: an asynchronous buffer unit, used for non-blocking buffering of physiological signal streams with different sampling frequencies; and a time-stamp normalization unit, used to obtain the system clock and interpolate the data frames in the asynchronous buffer unit based on the system clock, so that each physiological signal stream is aligned on the same time axis.

[0024] Preferably, the correlation matrix generation module is further subdivided into the following processing units: a window control unit, used to extract synchronous physiological dataset segments with a fixed duration of 5 to 30 minutes; and a normalization operation unit, used to perform covariance operation on the synchronous physiological dataset segments and eliminate dimensions to generate a real-time correlation matrix characterizing the degree of linear correlation between parameters.

[0025] Preferably, the physiological baseline map managed by the baseline management module includes coupled nodes consisting of circulatory system parameters, respiratory system parameters, and metabolic system parameters; the logical weights between the coupled nodes are used to characterize the collaborative compensation boundaries between the physiological systems of the elderly subjects in a non-pathological state, and serve as the global constraint baseline for the offset determination module to perform topological mapping.

[0026] Preferably, the offset determination module further includes a conflict discrimination logic unit, which is used to identify data coupling anomalies caused by the concurrent occurrence of multiple diseases by calculating the overall structural similarity of the real-time correlation matrix relative to the physiological baseline map when a single physiological parameter experiences masking fluctuations, and to avoid the risk of missed reports by maintaining the output continuity of the offset characterization value after the anomaly is identified.

[0027] Preferably, the early warning and forecasting module is further subdivided into the following processing units: a trend extraction unit, used to perform first-order derivative calculation on the offset characterization value to obtain the trend value representing the rate of pathological deterioration; and a threshold comparison unit, used to dynamically compare the trend value with the stored evolution threshold to determine whether the offset characterization value is in an accelerated deviation state.

[0028] Preferably, the early warning and forecasting module further includes a trend prediction unit, which is used to predict the numerical evolution range of the offset characterization value in the next 4 to 12 hours based on the historical sequence of the offset characterization value through a state space mapping model, and feeds back the numerical evolution range to the early warning and forecasting module to adjust the sensitivity of the evolution threshold.

[0029] Preferably, the synchronized physiological dataset includes at least heart rate, blood oxygen saturation, respiratory rate, and blood pressure; the system is deployed on the monitoring center server and is used to receive and parse data packets reported by wireless physiological sensors through a network interface to realize continuous physiological function monitoring and pathological trend identification for elderly people with multiple comorbidities.

[0030] An early warning method for an early warning system for multiple comorbidities in the elderly, the method comprising the following steps: Step S1101: Acquire multiple physiological signal streams with asynchronous sampling frequencies, and use the global system clock as a unified time reference to perform resampling alignment on the multiple physiological signal streams, including interpolation value operation, to generate a synchronous physiological dataset that is tightly aligned in the time domain step size. Step S1102: Calculate the normalized cross-correlation coefficient between any two physiological parameters in the synchronous physiological dataset within a sliding window of a preset length, and generate a real-time correlation matrix that characterizes the coupling compensation state of the multidimensional physiological system of the elderly subjects. Step S1103: Retrieve the physiological baseline map of the pre-recorded physiological parameters of the elderly subjects in a healthy state without disease, and obtain the offset characterization value that characterizes the degree of decoupling of physiological function by calculating the arithmetic square root of the sum of the squares of the differences between the real-time correlation matrix and the corresponding elements in the physiological baseline map. Step S1104: Calculate the rate of change of the offset characterization value within the current time step, and when the rate of change of the value is found to exceed the evolution threshold for 10 seconds, output a warning signal indicating that the elderly subject is in the period of compensatory collapse risk.

[0031] Example 1: In a cloud-based medical information processing hub scenario where a continuous multi-source physiological data stream from elderly patients tagged with hypertension and chronic obstructive pulmonary disease is continuously received within a monitoring center, the input data to the system originates from wireless physiological sensors from different manufacturers. The sampling frequencies of heart rate, blood oxygen saturation, respiratory rate, and blood pressure data reported by these sensors are distributed in a discrete range from 0.1Hz to 10Hz. Network transmission delay causes disordered timestamp drift. Due to the multiple disease concurrence mechanisms of the target elderly subjects, there is a high degree of cross-compensation effect between the cardiovascular and respiratory systems. When early heart failure occurs, the physiological response of compensatory heart rate increase masks the downward trend of absolute blood pressure, causing the traditional monitoring logic that relies on fixed threshold comparison of independent physiological parameters to fail. After receiving data packets, the communication interface uses a hash mapping algorithm to strip the patient's name and identification code contained in the data packet, converts the original data into a desensitized data stream carrying only device identification and timestamps, and routes the desensitized multi-channel physiological signal streams with asynchronous sampling frequencies to the data alignment module.

[0032] The data alignment module uses the global system clock as a unified time reference and interpolates the data frames stored in the asynchronous buffer unit according to the system clock. It resamples and aligns multiple discrete data streams with sampling frequencies differing by orders of magnitude to the same time axis, generating a synchronized physiological dataset. The correlation matrix generation module extracts segments of the synchronized physiological dataset with fixed durations of 5 to 30 minutes within a preset sliding window, calculates the covariance of these segments to eliminate dimensions, and calculates the normalized cross-correlation coefficient between any two physiological parameters. This covariance calculation preprocessing includes Z-score standardization preprocessing for each physiological parameter. Subtracting the mean within each window and dividing by the standard deviation transforms the covariance calculated for the standardized data into a dimensionless form, ensuring mathematical equivalence. This operation converts independently fluctuating heart rate and blood oxygen sequences into real-time correlation matrices characterizing the coupled compensatory state of a multidimensional physiological system. This resolves the physical conflict between single-data amplitude monitoring and multiple pathological compensatory mechanisms, transforming the observed parameters of disease data evolution from univariate amplitude overshooting to structural distortions in multivariate covariance relationships. During system initialization, the benchmark management module acquires 72 consecutive hours of physiological records of the subject's health status and calculates the Pearson correlation coefficient after aligning the physiological parameter sequences using a 1-minute step size. The calculation results are arranged into a symmetric covariance initial matrix. The main diagonal elements are assigned a value of 1, and the off-diagonal elements are extracted as node logical weights to generate the subject's inherent characteristic physiological baseline map.

[0033] The offset determination module reads the real-time correlation matrix and retrieves the pre-established physiological baseline map from the baseline management module. The physiological baseline map records the inherent distribution characteristics of various physiological parameters and the collaborative compensation evolution boundary of the elderly subjects in a non-pathological, healthy state using a matrix topology structure. The offset determination module calculates the arithmetic square root of the sum of squares of the differences between the elements at corresponding positions in the real-time correlation matrix and the physiological baseline map, and outputs a scalar offset characterization value that quantifies the degree of physiological function decoupling. The trend extraction unit in the early warning and forecasting module calculates the first derivative of the offset characterization value within the current time step to obtain the numerical change rate. The threshold comparison unit dynamically compares the numerical change rate with the stored evolution threshold. When the numerical change rate continuously exceeds the evolution threshold for 10 seconds, an early warning signal indicating that the elderly subjects are in a period of compensatory collapse risk is generated. Simultaneously with the generation of the early warning, the trend prediction unit in the early warning and forecasting module is activated, using the offset characterization value and its first derivative as discrete state variables, and constructing a Kalman filter state transition equation with the variance of the historical offset sequence as the system noise matrix. The process utilizes the current evolution threshold as the observation boundary condition, and solves the probability distribution of state variables within the next 4 to 12 hours through multi-step iterative deduction, thereby outputting numerical evolution interval feedback to adjust the threshold sensitivity. The early warning and forecasting module extracts physiological data sequences containing compensatory collapse events from the historical medical record database, calculates the second derivative of the offset characterization value, extracts the numerical change rate of the pathological deterioration inflection point and constructs a benchmark numerical set; calculates the probability density distribution of the set, and selects the 90th percentile value as the evolution threshold; when the numerical change rate continuously exceeds the evolution threshold for 10 seconds, an early warning signal is output; the 10-second time window spans the transient sympathetic reflex cycle induced by body position change, cuts off non-pathological fluctuation data caused by deep breathing or posture change, and generates an early warning signal based on the stable objective state of continuous pathological evolution. This processing link replaces the linear superposition judgment of independent thresholds with the convergence quantization of the topological distance of multidimensional data streams. When the absolute amplitude of all individual physiological parameters is within the normal physiological fluctuation range, a digital risk instruction characterizing the trajectory of complication deterioration is output.

[0034] Example 2: When the cloud-based medical information processing hub faces the objective task of verifying the early warning efficacy of a multi-disease compensatory mechanism, the data input to the system is extracted from a publicly available multi-parameter physiological dataset for intensive care. This dataset contains heart rate, blood oxygen saturation, and blood pressure waveforms of elderly patients with comorbid hypertension and chronic obstructive pulmonary disease. To reproduce the asynchronous transmission environment of the edge network, the communication interface superimposes Gaussian white noise with a signal-to-noise ratio of 20dB onto the original data stream to simulate patient motion artifacts. Simultaneously, a random packet loss mechanism is introduced, causing the sampling frequency of multiple physiological signals to jump between 0.1Hz and 10Hz. The multi-disease concurrency mechanism in the target elderly subjects leads to a cross-compensatory effect between the cardiovascular and respiratory systems. When early heart failure occurs, the physiological response of increased compensatory heart rate masks the downward trend of absolute blood pressure, causing the traditional monitoring logic relying on fixed threshold comparisons of independent physiological parameters to fail. After receiving the data packets, the communication interface uses a hash mapping algorithm to extract the patient's name from the data packets. Along with the identification code, the raw data is converted into a desensitized data stream carrying only device identification and timestamps. The communication interface routes the desensitized multi-channel physiological signal streams with asynchronous sampling frequencies to the data alignment module. The fixed duration of the sliding window in the correlation matrix generation module is used to balance the real-time performance of data acquisition and the stability of the statistical feature matrix. The set decision logic rule is to dynamically determine the time boundary based on the discreteness of the input buffer signal. When the variance of the multi-source signal is distributed in the preset low-frequency range, the window duration is shortened; when the variance crosses the mutation threshold, the window duration is extended. Based on this logic and combined with the human microcirculation constant, 15 minutes is calculated as the preferred sliding window duration under typical working conditions. Exceeding the upper limit of 30 minutes causes the matrix update to lag behind the acute heart failure attack cycle, while falling below the lower limit of 5 minutes causes high-frequency local noise to mask the normalized cross-correlation coefficient, causing discrete data deviation. This calculation logic establishes an effective working window of 5 minutes to 30 minutes to satisfy the topological mapping operation.

[0035] The experiment established a control group relying on a single physiological parameter's absolute amplitude exceeding the alarm limit, and an experimental group using a multi-dimensional topological mapping path. A disease progression gradient model was constructed, encompassing three deterioration stages: mild compensation, extreme compensation, and compensation collapse. The data alignment module acquired the global system clock and calculated interpolated values ​​for noisy asynchronous data, generating a synchronized physiological dataset. The correlation matrix generation module calculated the normalized cross-correlation coefficients between heart rate and blood pressure parameters within sliding windows of 5 min, 15 min, and 30 min. During the extreme compensation gradient stage, raw data containing Gaussian white noise showed that the heart rate scalar monitored in the control group remained stable at 88 bpm, and the systolic blood pressure scalar remained within the range of 125 mmHg. At this point, the control group could not trigger an alarm. In the experimental group, within a 15-minute sliding window, the correlation matrix generation module output key intermediate parameters, including the normalized cross-correlation coefficients of heart rate and systolic blood pressure. The cross-correlation coefficient changed from 0.76 at the healthy baseline to -0.41. This change indicates the decoupling characteristics of the cardiovascular sympathetic nervous system altering heart rate reserve to maintain blood pressure balance. The offset determination module extracts the real-time correlation matrix and calculates the square root of the sum of squares of the differences between the real-time correlation matrix and the corresponding elements in the physiological baseline map. It outputs the offset characterization value that quantifies the degree of physiological function decoupling. This intermediate value climbed from the normal baseline of 0.82 to the abrupt value of 3.54 during the extreme compensation stage. The system calculates the structural deviation of the multidimensional data stream collaborative structure before a single image value shows a pathological drop. Comparing different parameter boundaries, the offset characterization value under the 5-minute sliding window oscillates violently between 1.15 and 3.82 due to noise interference, while the offset characterization value under the 30-minute sliding window reaches the abrupt value after a delay of 18.5 minutes, confirming that the 15-minute setting constitutes the optimal working window.

[0036] The trend extraction unit within the early warning and forecasting module receives the offset characterization value, calculates the first derivative to generate a numerical rate of change, and when the numerical rate of change exceeds the evolution threshold and lasts for 10 seconds, the experimental group outputs an early warning signal. Under a 15-minute optimized sliding window, this triggering node precedes the control group's blood oxygen saturation falling below the 90% absolute threshold by 45.3 minutes. As the noise intensity of the simulated body motion artifact increases gradient to a signal-to-noise ratio of 10 dB, the control group is forced to generate high-density false alarm commands due to high-frequency fluctuations. The experimental group's correlation matrix generation module filters out high-frequency random disturbances by utilizing the statistical smoothing properties of covariance calculation, maintaining the continuity of the offset characterization value output and maintaining an early warning lead of 42.1 minutes during the extreme compensation period. The output quantitative results confirm that the numerical rate of change calculated based on the Euclidean spatial distance between the real-time correlation matrix and the physiological baseline map shows a positively correlated accelerating growth trend with the deterioration of the multidimensional physiological compensation coupling depth. This calculation process solves the monitoring dilemma of the absolute amplitude of variables in the multi-disease concurrent state list not exceeding the limit.

[0037] Example 3: This example combines Figures 1 to 2 This document describes an early warning system and method for treating multiple comorbidities in the elderly, such as... Figure 1 As shown, in step S1101, multiple physiological signal streams with asynchronous sampling frequencies are acquired, and resampling alignment, including interpolation, is performed on the multiple physiological signal streams using the global system clock as a unified time reference, generating a synchronous physiological dataset that is tightly aligned in the time domain step. In step S1102, the normalized cross-correlation coefficient between any two physiological parameters in the synchronous physiological dataset is calculated within a sliding window of a preset length, generating a real-time correlation matrix that characterizes the coupling compensation state of the multidimensional physiological system of the elderly subjects. Then, in step S1103, a physiological baseline map of the inherent distribution characteristics of physiological parameters of the elderly subjects in a non-disease-prone healthy state is retrieved. By calculating the square root of the sum of squares of the differences between the real-time correlation matrix and the corresponding elements in the physiological baseline map, the offset characterization value representing the degree of physiological function decoupling is obtained. Finally, in step S1104, the rate of change of the offset characterization value within the current time step is calculated, and when the rate of change exceeds the evolution threshold for 10 seconds, a warning signal indicating that the elderly subjects are in a period of compensatory collapse risk is output.

[0038] like Figure 2 As shown, the system receives the data to be analyzed and determines whether the timestamps of the data source are consistent. If the determination result is negative, the asynchronous sampling signal stream is routed to the data alignment module to perform interpolation calculation and resample to a unified frequency, thereby converting it into a synchronous physiological dataset. If the determination result is positive, the correlation matrix generation module calculates the normalized cross-correlation coefficient and generates a real-time correlation matrix. At the logical convergence point of confirming whether a real-time correlation matrix has been generated, the system combines the operation of providing physiological baseline maps and defining the boundaries of collaborative compensation evolution performed by the baseline management module to determine whether the offset characterization value after mapping is greater than the evolution threshold. If this determination is negative, the conclusion is that there is no significant functional shift and the system continues to monitor. If this determination is positive, the conclusion is that there is a functional shift and the early warning and forecasting module is triggered. This module identifies the multiple diseases that concurrently mask the accelerated deterioration trajectory based on the continuous rate of change of the numerical value exceeding the evolution threshold, and finally generates an early warning signal indicating the risk period of compensatory collapse.

[0039] Example 4: When the cloud-based medical information processing hub receives continuous physiological data streams from elderly subjects for an extended period, the system faces challenges such as baseline shifts caused by natural physiological decline and masking fluctuations in individual physiological parameters due to the concurrent occurrence of multiple diseases. Fixed graph weights cannot match the aging trajectory of human function, and the local stability of independent parameters masks the true risk of systemic compensatory collapse, leading to false negative results from the early warning module. The desensitized data stream continues to accumulate in the asynchronous buffer unit, and the data alignment module outputs a synchronized physiological dataset based on the system clock. The baseline management module extracts a 24-hour synchronized physiological dataset, calculates the moving average and covariance matrix of each physiological parameter within this period, obtains the long-term drift component of the correlation, and sets a time decay factor. ,in, A calibration constant used to characterize the rate of physiological decline. The calibration constant λ is calculated by extracting the moving average variance sequence of the offline resting heart rate and systolic blood pressure of the elderly subjects for at least 3 months from the baseline management module, and multiplying the absolute value of the linear fitting slope of the variance sequence by the preset basal metabolic equivalent constant. The baseline management module multiplies the initial logical weight of each coupled node in the physiological baseline map by this time decay factor, and compensates the long-term drift component of the correlation to the decayed coupled nodes to generate an updated physiological baseline map. This calibration procedure reconstructs the matrix topology boundary according to the objective time parameters at a step frequency of 24 hours to offset the baseline quantification error caused by the decline of human physiological function.

[0040] The conflict discrimination logic unit within the offset determination module receives the updated physiological baseline map and the real-time correlation matrix. When the absolute amplitude of a single heart rate parameter remains within the normal range of 60 bpm to 100 bpm, while the correlation trend of the blood pressure parameter changes abruptly, the conflict discrimination logic unit extracts the feature values ​​of the real-time correlation matrix and the updated physiological baseline map, calculates the Frobenius norm distance between the two, and transforms this Frobenius norm distance into an overall structural similarity scalar ranging from 0 to 1. When the overall structural similarity scalar is below 0.6, the conflict discrimination logic unit outputs a data coupling abnormality status. The set threshold of 0.6 is extracted from retrospective statistics of monitoring data from tens of thousands of clinical patients with concurrent hypertension and respiratory failure. Data clustering analysis shows that compensatory decoupling... One to two hours before a systemic collapse, the topological similarity of multidimensional parameters irreversibly falls below the lower bound of this feature. After identifying the anomaly, the conflict discrimination logic unit locks the offset representation value of the previous time step, rejects the injection of real-time distorted data, and maintains the continuity of the offset representation value output to the early warning and forecasting module. While locking the main link offset representation value, the system simultaneously starts a bypass high-frequency monitoring thread to specifically extract the second-order infinitesimal change of the real-time correlation matrix feature value during the locked period. Once the bypass thread detects that the cumulative integral of this change exceeds the upper limit of the compensatory fluctuation boundary, it forcibly unlocks the state and outputs the rapid change trend at this time without delay. This underlying computing architecture eliminates univariate misleading, adapts to long-term benchmark evolution, and ensures that the output of the early warning and forecasting module reflects the objective risk instructions of the real pathological evolution.

[0041] Example 5: When the cloud-based medical information processing hub faces the situation of new device access and physiological baseline map initialization, the baseline management module reads the offline electronic medical record steady-state dataset of the elderly subjects within the past 6 months. The data alignment module calculates the interpolation value of the heart rate sequence and systolic blood pressure sequence contained in the offline electronic medical record steady-state dataset with a step size of 1 second, and generates an offline synchronization matrix. The baseline management module calculates the data variance of the offline synchronization matrix, extracts the time period with variance value less than the preset steady-state threshold and marks it as the healthy steady-state baseline window. The baseline management module extracts the multidimensional physiological data in the healthy steady-state baseline window, calculates the initial normalized cross-correlation coefficient, and writes it into the corresponding coupling node to generate the initial physiological baseline map.

[0042] The offset determination module retrieves the initial physiological baseline map. Before receiving the real-time desensitized data stream, the communication interface injects a 10-minute Gaussian white noise test sequence. The correlation matrix generation module generates a test correlation matrix based on the Gaussian white noise test sequence. The conflict discrimination logic unit calculates the Frobenius norm distance between the test correlation matrix and the initial physiological baseline map. The conflict discrimination logic unit sets the reciprocal of the Frobenius norm distance as the environmental compensation weight coefficient and writes the environmental compensation weight coefficient into the evolution threshold calculation loop. The threshold comparison unit multiplies the pre-stored evolution threshold by the environmental compensation weight coefficient to generate a dynamic evolution threshold. The early warning and forecasting module compares the first derivative of the real-time offset characterization value based on the dynamic evolution threshold and outputs an early warning signal. This calibration procedure enables the system to output a quantitative judgment boundary that adapts to environmental changes during the cold start phase.

[0043] Example 6: When the system faces the task of extracting the rate of change of the offset characterization value of discrete time series, the trend extraction unit in the early warning and forecasting module reads the real-time offset characterization value at the current sampling time and the historical offset characterization value before the preset delay period, calculates the scalar difference between the two and divides it by the preset delay period to generate a discrete rate of change of the numerical value; the system sets the preset delay period to 50 times the system resampling step size of 1s; the system allocates a first-in-first-out circular queue with a fixed address length according to the available cache space of the local processor, and pushes the real-time offset characterization values ​​continuously output by the offset determination module into the first-in-first-out circular queue in the order of increasing timestamp; under the synchronous triggering beat of the global system clock, the trend extraction unit reads the first element and the last element of the first-in-first-out circular queue to calculate the backward difference quotient value, and outputs the rate of change of the numerical value to the threshold comparison unit.

[0044] When the system faces the on-site preprocessing situation of connecting new devices and calibrating the abrupt change threshold of the sliding window duration, the baseline management module extracts the initial physiological desensitization data of the elderly subjects in a closed-eye resting state for 2 hours; the data alignment module calculates the interpolation value of the initial physiological desensitization data to generate the calibration synchronization dataset; the correlation matrix generation module calculates the baseline variance sequence of the calibration synchronization dataset with a step period of 1 minute; the baseline management module extracts the maximum and minimum values ​​of the baseline variance sequence and calculates the absolute difference to generate the baseline range scalar; the baseline management module multiplies the baseline range scalar by a tolerance factor of 1.5 to calculate the abrupt change threshold; the system writes the abrupt change threshold into the logic comparison register of the correlation matrix generation module; when processing real-time multi-source signals, the correlation matrix generation module continuously calculates the real-time variance of the input buffer signal, obtains the comparison level signal between the real-time variance and the abrupt change threshold, and dynamically increases or decreases the duration constant of the sliding window within a physical interval of 5 minutes to 30 minutes based on the comparison level signal.

[0045] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0046] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An early warning system for multiple comorbidities in the elderly, characterized in that, The system includes: The data alignment module is used to access multiple physiological signal streams with asynchronous sampling frequencies and obtain the global system clock as a unified time reference. It performs resampling alignment on the multiple physiological signal streams, including interpolation operations, to generate a synchronous physiological dataset that is tightly aligned in the time domain step. The correlation matrix generation module is used to calculate the normalized cross-correlation coefficient between any two physiological parameters in the synchronous physiological dataset within a sliding window of a preset length, and generate a real-time correlation matrix that characterizes the coupling and compensatory state of the multidimensional physiological system of the elderly subjects. The baseline management module is used to manage the pre-established physiological baseline map. The physiological baseline map records the inherent distribution characteristics of various physiological parameters of the elderly subjects in a non-disease-causing healthy state, as well as the collaborative compensatory evolution boundary between various physiological systems, in a matrix topology structure. The offset determination module is used to perform Euclidean space topological mapping between the real-time correlation matrix and the physiological baseline map. By calculating the arithmetic square root of the sum of squares of the differences between the corresponding elements in the real-time correlation matrix and the physiological baseline map, the offset characterization value representing the degree of decoupling of physiological function is obtained. The early warning and forecasting module is used to identify the trajectory of accelerated deterioration of physiological function caused by the concurrent masking of multiple diseases based on the rate of change of the offset characterization value within the current time step, and output an early warning signal indicating that the elderly subjects are in the period of compensatory collapse when the rate of change of the numerical value continues to exceed the evolution threshold for 10 seconds.

2. The early warning system for multiple comorbidities in the elderly according to claim 1, characterized in that, The system also includes a baseline adaptive correction module, which uses a forgetting operator with decay weights to correct the physiological baseline map in the baseline management module; The baseline adaptive correction module acquires historical physiological data sequences and calculates long-term correlation drift components. It then adjusts the node weights of the physiological baseline map at a step frequency of 24 hours to offset the baseline offset error caused by the natural decline of human physiological functions.

3. The early warning system for multiple comorbidities in the elderly according to claim 1, characterized in that, The data alignment module is further subdivided into the following processing units: an asynchronous buffer unit, used for non-blocking buffering of physiological signal streams with different sampling frequencies; and a time-stamp normalization unit, used to obtain the system clock and interpolate the data frames in the asynchronous buffer unit based on the system clock, so that each physiological signal stream is aligned on the same time axis.

4. The early warning system for multiple comorbidities in the elderly according to claim 1, characterized in that, The correlation matrix generation module is further subdivided into the following processing units: a window control unit, used to extract synchronous physiological dataset segments with a fixed duration of 5 to 30 minutes; and a normalization operation unit, used to perform covariance calculation on the synchronous physiological dataset segments and eliminate dimensions to generate a real-time correlation matrix characterizing the degree of linear correlation between parameters.

5. The early warning system for multiple comorbidities in the elderly according to claim 1, characterized in that, The physiological baseline map managed by the baseline management module contains coupled nodes consisting of circulatory system parameters, respiratory system parameters, and metabolic system parameters. The logical weights between the coupled nodes are used to characterize the collaborative compensation boundaries between the physiological systems of the elderly subjects in non-pathological states, and serve as the global constraint baseline for the offset determination module to perform topological mapping.

6. The early warning system for multiple comorbidities in the elderly according to claim 1, characterized in that, The offset determination module also includes a conflict discrimination logic unit, which is used to identify data coupling anomalies caused by the concurrent occurrence of multiple diseases by calculating the overall structural similarity of the real-time correlation matrix relative to the physiological baseline map when a single physiological parameter experiences masking fluctuations. After identifying the anomaly, it avoids the risk of missed reports by maintaining the output continuity of the offset characterization value.

7. The early warning system for multiple comorbidities in the elderly according to claim 1, characterized in that, The early warning and forecasting module is further subdivided into the following processing units: a trend extraction unit, which is used to perform first-order derivative calculation on the offset characterization value to obtain the trend value representing the rate of pathological deterioration; The threshold comparison unit is used to dynamically compare the trend value with the stored evolution threshold to determine whether the offset characterization value is in an accelerated deviation state.

8. The early warning system for multiple comorbidities in the elderly according to claim 1, characterized in that, The early warning and forecasting module also includes a trend prediction unit, which is used to predict the numerical evolution range of the offset characterization value in the next 4 to 12 hours based on the historical sequence of the offset characterization value through a state space mapping model, and feeds back the numerical evolution range to the early warning and forecasting module to adjust the sensitivity of the evolution threshold.

9. The early warning system for multiple comorbidities in the elderly according to claim 1, characterized in that, The synchronized physiological dataset includes at least heart rate, blood oxygen saturation, respiratory rate, and blood pressure. The system is deployed on the monitoring center server and is used to receive and parse data packets reported by wireless physiological sensors through a network interface to achieve continuous physiological function monitoring and pathological trend identification for elderly people with multiple comorbidities.

10. A method for early warning of multiple comorbidities in the elderly, used to implement the early warning system for multiple comorbidities in the elderly as described in claim 1, characterized in that, The method includes the following steps: Step S1101: Acquire multiple physiological signal streams with asynchronous sampling frequencies, and use the global system clock as a unified time reference to perform resampling alignment on the multiple physiological signal streams, including interpolation value operation, to generate a synchronous physiological dataset that is tightly aligned in the time domain step size. Step S1102: Calculate the normalized cross-correlation coefficient between any two physiological parameters in the synchronous physiological dataset within a sliding window of a preset length, and generate a real-time correlation matrix that characterizes the coupling compensation state of the multidimensional physiological system of the elderly subjects. Step S1103: Retrieve the physiological baseline map of the pre-recorded physiological parameters of the elderly subjects in a healthy state without disease, and obtain the offset characterization value that characterizes the degree of decoupling of physiological function by calculating the arithmetic square root of the sum of the squares of the differences between the real-time correlation matrix and the corresponding elements in the physiological baseline map. Step S1104: Calculate the rate of change of the offset characterization value within the current time step, and when the rate of change of the value is found to exceed the evolution threshold for 10 seconds, output a warning signal indicating that the elderly subject is in the period of compensatory collapse risk.

Images