A dynamic monitoring system for elderly heart patients

Abstract

The present invention provides a dynamic monitoring system for elderly heart patients, comprising a multimodal sensor module, a data processing module, an analysis module and an alarm module. The multimodal sensor module collects parameters to generate multimodal physiological signals; the data processing module generates pre-processed signals; the analysis module performs real-time analysis of the pre-processed signals based on a preset elderly electrocardiogram characteristic model to generate graded alarm instructions; the alarm module triggers corresponding local reminders, remote notifications or emergency linkage responses according to the graded alarm instructions, wherein the emergency linkage response automatically sends the patient's location and key physiological data to a designated emergency center through the emergency communication module; wherein the adaptive adjustment of the multimodal sensor module includes dynamically optimizing the signal gain according to the skin contact status. The present invention provides a dynamic monitoring system for elderly heart patients, which can solve the problems of poor real-time performance, high false alarm rate and delayed emergency response in dynamic monitoring of elderly heart disease patients.

Description

Technical Field

[0001] The present invention relates to the field of medical electronic technology, to the field of wearable health monitoring equipment, and in particular to a dynamic monitoring system for elderly heart patients. Background Art

[0002] Elderly heart patients face significant limitations with traditional monitoring methods due to declining physiological functions and complex comorbidities. While Holter devices can record 24-hour ECG data, they suffer from data transmission delays and the inability to provide real-time warnings. Furthermore, their bulk makes them uncomfortable to wear, impacting elderly patients' compliance.

[0003] While existing wearable devices (such as smartwatches) can monitor some physiological parameters, single sensors are susceptible to motion interference and have a high false alarm rate. This is particularly true in the elderly, where signal quality deteriorates further due to loose skin and frequent body movements. Remote monitoring systems rely on wireless transmission and manual operation, requiring a high level of technical proficiency from the elderly. Most systems only provide data storage rather than intelligent analysis, making them unable to identify subtle abnormalities such as ST-segment deviations. While artificial intelligence-assisted diagnosis has been applied to arrhythmia classification, general models are not optimized for the ECG characteristics of elderly patients (such as conduction block and atrial fibrillation waveform variation), resulting in insufficient specificity.

[0004] During the emergency response phase, the existing system lacks an automatic docking mechanism with medical institutions. After an abnormal event occurs, manual contact with the emergency center is required, delaying the golden rescue time. In addition, the power consumption and comfort of multi-parameter monitoring equipment are in sharp conflict. Traditional equipment often uses high-frequency sampling to ensure accuracy, which increases battery consumption, while reducing the sampling rate may miss sudden abnormalities. To address the above issues, it is urgent to develop a monitoring system specifically for elderly heart patients. Through sensor fusion, adaptive algorithms and intelligent response mechanisms, it can achieve integrated closed-loop management of high-precision monitoring and emergency linkage while ensuring wearing comfort. Summary of the Invention

[0005] In view of the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a dynamic monitoring system for elderly heart patients, which is used to solve the problems of poor real-time performance, high false alarm rate and delayed emergency response in dynamic monitoring of elderly heart patients. The present invention uses a multimodal sensor to fuse electrocardiogram, blood oxygen and respiratory signals, dynamically optimizes signal gain in combination with skin contact status, and eliminates motion artifacts; adopts a lightweight model that collaborates with edge computing and the cloud to analyze physiological data in real time, identify arrhythmia events and divide risk levels; constructs a graded alarm mechanism, triggers local reminders, remote notifications or emergency linkage according to the risk level, automatically retrieves patient medical records and enables redundant communication links to transmit key data during emergency response, realizes closed-loop management of the entire process from abnormal detection to emergency decision-making, and improves monitoring accuracy and timeliness of treatment.

[0006] The present invention provides a dynamic monitoring system for elderly heart patients, comprising:

[0007] The multimodal sensor module collects ECG, blood oxygen, and respiratory signals from elderly patients in real time, and generates multimodal physiological signals after adaptively adjusting signal acquisition parameters to reduce environmental interference. Adaptive adjustment includes dynamically adjusting the signal acquisition frequency by monitoring impedance changes at the sensor-skin interface. When contact impedance is detected to be outside a preset range, it automatically switches to a backup electrode array and compensates for signal attenuation. It also performs closed-loop adjustment of the photoelectric sensor's transmission power based on ambient temperature fluctuations.

[0008] The data processing module performs fusion processing on multimodal physiological signals to eliminate motion artifacts and extract characteristic waveforms to generate preprocessed signals;

[0009] The analysis module analyzes the pre-processed signal in real time based on the preset elderly ECG characteristic model, identifies abnormal ECG events and their risk levels, and generates graded alarm instructions;

[0010] The alarm module triggers corresponding local reminders, remote notifications or emergency linkage responses based on graded alarm instructions. The emergency linkage response automatically sends the patient's location and key physiological data to the designated emergency center through the emergency communication module, and simultaneously retrieves the patient's historical medical records to assist in emergency decision-making. The alarm module implements a three-level judgment logic: the first level initiates local vibration reminders for occasional premature beats and records event characteristics; the second level sends encrypted early warning messages to the preset guardian terminal for sustained tachycardia events; the third level immediately activates the full-duplex transmission channel of the emergency communication module when a life-threatening arrhythmia pattern is detected; the emergency communication module is equipped with a redundant communication link switching mechanism. When the transmission quality of the main wireless channel drops below the threshold, the backup frequency band is automatically activated to establish a parallel data transmission path, and the patient's location information is processed using a differential encryption algorithm;

[0011] Among them, the adaptive adjustment of the multimodal sensor module includes dynamically optimizing the signal gain according to the skin contact status, analyzing the risk level judgment of the module by combining the patient's individualized threshold and the deviation of real-time physiological parameters, and establishing a dynamic baseline based on the patient's historical monitoring data. When the deviation of real-time physiological parameters exceeds the baseline fluctuation range, the online fine-tuning of the model parameters is automatically triggered.

[0012] In one embodiment of the present invention, the fusion processing of the data processing module includes a dynamic filtering strategy based on motion state recognition, constructing a motion interference model through body motion signals collected by inertial sensors, using a time-frequency domain joint analysis method to separate the myoelectric noise component in the electrocardiogram signal, and incorporating the phase relationship between the respiratory signal and the blood oxygen signal into the motion artifact correction algorithm to generate a multimodal waveform data set with time alignment characteristics. After standardization, the data set is bound to the patient identity identifier and stored in a cache queue for the analysis module to call in priority order.

[0013] In one embodiment of the present invention, an elderly ECG feature model is obtained through training using a machine learning method. Its input layer receives preprocessed signal segments segmented by time windows, the middle layer uses an attention mechanism to weightedly fuse multimodal feature vectors, and the output layer generates a probability distribution based on the electrophysiological characteristics of common arrhythmia types in elderly patients. The establishment of individualized thresholds constructs a dynamic baseline based on the patient's historical monitoring data. When the deviation of real-time physiological parameters exceeds the baseline fluctuation range, the online fine-tuning of the model parameters is automatically triggered to reduce the misjudgment rate.

[0014] In one embodiment of the present invention, the analysis module has a built-in timing correlation analysis unit for detecting the rhythm coupling relationship between the electrocardiogram signal and the respiratory signal. When it is found that the respiratory sinus arrhythmia characteristics disappear and are accompanied by a step-down decrease in blood oxygen saturation, the risk level judgment weight is automatically increased, the update cycle of the individualized threshold is adaptively adjusted according to the patient's activity intensity, and the baseline calculation window is extended during the resting stage at night to improve the stability of the judgment.

[0015] In one embodiment of the present invention, a dynamic priority scheduler is provided between the data processing module and the analysis module. When abnormal cross-parameter correlation features are detected in multimodal physiological signals, the processing priority of the corresponding data stream is temporarily increased, and additional computing resources are allocated for multi-dimensional feature cross-validation. The generation process of the preprocessed signal includes timestamp calibration of the characteristic waveform to ensure that the data collected by different sensors remain strictly synchronized in the time domain.

[0016] In one embodiment of the present invention, the alarm module is integrated with an intelligent learning unit, which establishes a personalized feedback model by recording the patient's response behavior to alarms at all levels. When it is detected that a high-risk alarm has been ignored multiple times, the alarm intensity is automatically upgraded and a backup contact call chain is activated. The local reminder device adopts a gradual stimulation strategy, with low-frequency vibration prompts in the initial stage. If no response is received, it switches to a composite sound and light mode, and at the same time sends an attention signal to the data processing module to trigger higher-precision physiological parameter sampling.

[0017] In one embodiment of the present invention, a data storage module is provided between the data processing module and the analysis module. The data storage module generates a dynamic storage strategy based on the characteristic waveform energy distribution and abnormal event markers of the preprocessed signal, and stores the stable baseline segment in the multimodal physiological signal in a compressed encoding manner, while retaining the original sampling rate and adding a timestamp index for the signal segment containing the characteristic waveform mutation. The data storage module dynamically adjusts the storage period according to the risk level feedback from the analysis module, and permanently stores the data related to high-risk events and establishes a logical association with the emergency linkage response. When the alarm module triggers the emergency communication module, the multimodal data waveform of the set time period before the incident is automatically extracted from the data storage module, and an auxiliary diagnosis report containing time series evolution characteristics is generated and synchronously transmitted to the emergency center. The execution process of the storage strategy realizes the seamless connection between the real-time data stream and the historical data through the ring buffer, ensuring that the complete physiological parameter change trajectory can be reconstructed when the abnormal event is traced back.

[0018] The present invention provides a dynamic monitoring system for elderly heart patients. The system uses multimodal sensors to fuse electrocardiogram, blood oxygen and respiratory signals, dynamically optimizes signal gain in combination with skin contact status, and eliminates motion artifacts. It adopts a lightweight model that collaborates with edge computing and the cloud to analyze physiological data in real time, identify arrhythmia events and divide risk levels. It builds a graded alarm mechanism to trigger local reminders, remote notifications or emergency linkage according to the risk level. During emergency response, it automatically retrieves patient medical records and enables redundant communication links to transmit key data, realizing closed-loop management of the entire process from abnormality detection to emergency decision-making, thereby improving monitoring accuracy and treatment timeliness. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 This is a system architecture diagram of a dynamic monitoring system for elderly heart patients. DETAILED DESCRIPTION

[0021] The following describes the embodiments of the present invention through specific examples. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. The details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other unless they conflict.

[0022] It should be noted that the illustrations provided in the following embodiments are merely schematic illustrations of the basic concept of the present invention. Therefore, the illustrations only show components related to the present invention and are not drawn according to the number, shape, and size of components in actual implementation. In actual implementation, the type, quantity, and proportion of each component may be changed arbitrarily, and the component layout may also be more complex.

[0023] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it will be apparent to those skilled in the art that the embodiments of the present invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring the embodiments of the present invention.

[0024] See Figure 1 , shown is a dynamic monitoring system for elderly heart patients according to the present invention, comprising a multimodal sensor module, a data processing module, an analysis module, and an alarm module. The multimodal sensor module collects the elderly patient's electrocardiogram (ECG) signal, blood oxygenation signal, and respiratory signal in real time, and generates a multimodal physiological signal after adaptively adjusting the signal acquisition parameters to reduce environmental interference. The data processing module fuses the multimodal physiological signals, eliminates motion artifacts, extracts characteristic waveforms, and generates a preprocessed signal. The analysis module analyzes the preprocessed signal in real time based on a preset elderly ECG characteristic model, identifies abnormal ECG events and their risk levels, and generates graded alarm instructions. The alarm module triggers corresponding local reminders, remote notifications, or emergency linkage responses based on the graded alarm instructions. The emergency linkage response automatically sends the patient's location and key physiological data to a designated emergency center via the emergency communication module, and simultaneously retrieves the patient's historical medical records to assist in emergency decision-making. The adaptive adjustment of the multimodal sensor module includes dynamically optimizing the signal gain based on the skin contact state, and the risk level determination of the analysis module combines the patient's individualized threshold with the deviation of real-time physiological parameters.

[0025] Figure 1As shown in Figure 1, the system consists of four core modules: a multimodal sensor module, a data processing module, an analysis module, and an alarm module. Each module is tightly coupled through signal flow and functional logic to form a complete monitoring closed loop. The multimodal sensor module serves as the data acquisition front end, integrating flexible electrodes, photoelectric sensors, and impedance sensors for synchronously capturing ECG, blood oxygenation, and respiratory signals, respectively. The module's core innovation lies in its adaptive adjustment mechanism, which dynamically optimizes signal gain based on the skin contact state of elderly patients. For example, when changes in contact impedance due to skin laxity or sweat are detected, the module automatically adjusts the electrode's input impedance matching parameters to reduce signal distortion. For blood oxygenation signal acquisition, the module incorporates a built-in closed-loop regulation system that compensates for photoelectric sensor transmission power by real-time monitoring of ambient temperature fluctuations, ensuring signal stability under varying wearing conditions. Respiratory signals are acquired using the chest impedance method, combined with a motion state recognition algorithm to filter out noise interference caused by limb movement. All sensor data undergo preliminary processing to generate multimodal physiological signals, which are transmitted to the data processing module via a low-power wireless transmission protocol. The data processing module performs signal fusion and preprocessing, with its core technologies focusing on motion artifact removal and feature waveform extraction. After receiving multimodal physiological signals, the module first performs baseline drift correction on the ECG signal and uses a wavelet transform-based filtering algorithm to separate high-frequency myoelectric noise from low-frequency respiratory interference. To address the time-domain correlation between the blood oxygenation signal and the respiratory signal, the module uses phase synchronization technology to align their sampling time axes, eliminating data misalignment caused by sensor response delays. Motion artifacts are eliminated using a dynamic filtering strategy. Built-in inertial sensors (such as accelerometers) capture patient motion data in real time, construct a three-dimensional motion model, and calculate its coupling relationship with physiological signals. Finally, an adaptive filter is used to reversely eliminate motion interference components. Key feature waveforms are extracted from the processed signals, such as the QRS complex morphology and ST segment slope of the ECG signal and the pulse waveform period of the blood oxygenation signal. These features are standardized and encoded to generate a preprocessed signal, which is transmitted to the analysis module for in-depth analysis.

[0026] Furthermore, the core of the analysis module lies in a pre-configured geriatric ECG feature model, trained and optimized using machine learning methods specifically for the ECG characteristics of elderly patients. The model's input layer receives preprocessed signal segments segmented by time windows and employs a convolutional neural network to extract spatiotemporal features of the multimodal signal, such as the rhythmic coupling between the ECG and respiratory signals and the correlation between the rate of oxygen saturation desaturation and ST-segment deviation. An attention mechanism is incorporated into the middle layer to dynamically adjust the weights of different feature vectors based on common pathological features in elderly patients (such as atrial fibrillation waveform variability and conduction block patterns). The output layer uses a patient-specific threshold to determine risk level. This threshold is not a fixed value but a dynamic baseline constructed based on the patient's historical monitoring data. For example, the heart rate variability coefficient over the past 72 hours is used as a reference. When real-time heart rate deviation is detected beyond the baseline fluctuation range, the model parameters are automatically fine-tuned online to reduce false positives. For identified abnormal events, the module categorizes them into three levels of risk based on preset classification rules: Level 1 corresponds to minor abnormalities such as occasional premature beats, Level 2 targets moderate symptoms such as persistent tachycardia, and Level 3 covers critical conditions such as ST-segment elevation myocardial infarction. The risk assessment results are converted into graded alarm instructions and transmitted to the alarm module via encrypted data packets. The alarm module transforms data analysis into actual intervention, and its response mechanism strictly corresponds to the risk level. For Level 1 risk, the module activates the local reminder device, using a gradually increasing stimulation strategy. Initially, low-frequency vibrations alert the patient. If no response is received, the system switches to a composite audio and visual alarm. Simultaneously, a simple notification is sent to the family's terminal via short-range wireless communication. When Level 2 risk is triggered, in addition to increasing the intensity of the local alarm, the module also sends an encrypted early warning message to the preset guardian via the cellular network. The message contains the type of abnormal event, the time of occurrence, and brief waveform characteristics. When a Level 3 high-risk event is detected, the module immediately activates the emergency communication module's full-duplex transmission channel. This channel utilizes a redundant communication design. The primary channel prioritizes 5G networks for high-priority data packets, while the backup channel utilizes LoRa technology to ensure the transmission of essential information. During the emergency response, the module automatically retrieves key information from the patient's electronic medical record (including recent medication history, allergy history, and previous diagnoses), combines it with real-time physiological data, and encapsulates it into an emergency decision-making support report. This report is then converted to an HL7-compliant format via medical data middleware and directly sent to the emergency center's dispatch system. To ensure reliable data transmission, patient location information is processed using a differential encryption algorithm, enabling precise positioning while ensuring privacy and security.

[0027] In one embodiment of the present invention, the adaptive adjustment mechanism of the multimodal sensor module focuses on addressing the signal acquisition challenges caused by the physiological characteristics of elderly patients. The impedance monitoring system at the sensor-skin interface samples the contact impedance at a millisecond frequency. When the impedance value exceeds a preset safety range (e.g., due to a sudden rise in impedance caused by electrode detachment), the module automatically switches to a backup electrode array. This array utilizes a distributed layout and selects the optimal signal source by scanning the contact quality of each backup electrode. A digital signal reconstruction algorithm is then applied to compensate for signal attenuation caused by electrode switching. For blood oxygen monitoring using a photoelectric sensor, the module incorporates a built-in temperature-light intensity compensation curve that dynamically adjusts the LED emission power based on the real-time ambient temperature. For example, in low-temperature environments, infrared light intensity is appropriately increased to penetrate thickened epidermal tissue, while in hot and sweaty environments, green light intensity is reduced to reduce epidermal reflection interference. The compensation parameters generated during the signal gain optimization process are fed back to the data processing module via a dedicated data bus to correct for baseline drift in the multimodal physiological signals. For example, when the blood oxygen sensor detects ambient light interference, this parameter triggers the adjustment of the center frequency of the band-stop filter in the data processing module, forming a combined feedforward-feedback control loop. The data processing module expands on the fusion processing algorithm and proposes a dynamic filtering strategy based on motion state recognition. This strategy captures patient motion data in real time using a six-axis inertial sensor (consisting of a three-axis accelerometer and a three-axis gyroscope). A Kalman filter is used to fuse the multi-dimensional motion signals to construct feature vectors representing motion amplitude, frequency, and direction. These vectors are input into a pretrained motion interference classification model to identify typical elderly patient activity patterns, such as walking, turning over, and trembling, and generate corresponding motion artifact templates. During the joint time-frequency domain analysis phase, the module performs a short-time Fourier transform on the ECG signal to extract the energy distribution characteristics of each frequency band. This is then convolved with the motion artifact template to isolate the pure ECG component. For respiratory signal processing, the algorithm performs a correlation analysis between the baseline respiratory waveform measured by impedance spectroscopy and the volume change signal caused by chest and abdominal movement to eliminate spurious respiratory fluctuations caused by postural changes. The resulting preprocessed signal not only contains time-aligned multimodal waveform data but also includes metadata tags (such as motion state encoding and signal quality score) to prioritize data segments during high-risk periods in subsequent analysis modules. This fusion processing mechanism specifically optimizes the ability to eliminate small-amplitude high-frequency tremor interference common in elderly patients, improving signal fidelity by approximately 40% compared to traditional filtering methods.

[0028] like Figure 1Figure 2 illustrates the construction and dynamic optimization mechanism of an ECG feature model for the elderly. This model is trained using a supervised learning method. The training dataset specifically collects multimodal physiological signals from elderly patients with heart disease, covering common arrhythmias such as atrial fibrillation, premature ventricular contractions, and ST-segment abnormalities. The input layer receives preprocessed signal segments segmented into time windows. The length of each time window is adaptively adjusted based on ECG signal characteristics. For example, when increased RR interval variability is detected, the window is automatically extended to capture the complete arrhythmia cycle. The middle layer employs a multi-head attention mechanism to assign differentiated weights to the feature vectors of different physiological signals. For example, it prioritizes features such as absent P waves or widened QRS complexes, which are common in elderly patients. The output layer generates probability distributions for each abnormality type based on a softmax function and combines them with individualized thresholds for risk stratification. The individualized thresholds are established by constructing a dynamic baseline based on historical monitoring data. Specifically, the standard deviation of parameters such as heart rate variability and blood oxygen fluctuation range is calculated over a sliding time window (e.g., 72 hours) to serve as a reference for the current threshold. When the deviation of real-time physiological parameters exceeds a preset multiple of the baseline fluctuation range, the system triggers online fine-tuning of the model parameters. This fine-tuning process employs a transfer learning strategy, making only minor adjustments to the weights of the fully connected layers to preserve the generalization capabilities of the pre-trained model. Furthermore, the model incorporates an adversarial training mechanism, augmenting training samples with noisy data simulating changes in skin impedance in the elderly, improving classification robustness under low signal-to-noise ratio conditions.

[0029] like Figure 1As shown in the figure, the three-level judgment logic and execution process of the graded alarm instructions. The first-level alarm is for occasional physiological abnormalities, such as a single atrial premature beat or a brief drop in blood oxygen saturation. The system activates the local vibration reminder device, which adopts a progressive stimulation strategy: in the initial stage, it uses low-frequency intermittent vibration to remind. If the patient does not confirm the receipt by pressing the physical button within the set time, the vibration frequency is gradually increased and a soft light effect is superimposed to avoid scaring the elderly patient. The second-level alarm responds to persistent abnormal events, such as sinus tachycardia accompanied by accelerated respiratory rate for ten consecutive minutes. At this time, the system sends a warning message to the preset guardian terminal through an encrypted communication protocol. The message body is encapsulated in JSON format and contains the abnormality type code, occurrence timestamp, current physiological parameter summary and recommended treatment measures. The encryption process uses the national secret SM4 algorithm to ensure the security of data transmission. A Level 3 alert is activated immediately upon the detection of life-threatening symptoms (such as sustained ST-segment elevation exceeding the threshold or ventricular fibrillation waveforms). The system performs three tasks in parallel: First, the emergency communication module uses redundant channels to simultaneously transmit the patient's geographic location, real-time vital signs, and high-resolution physiological waveforms from the last 15 minutes. Second, the medical data middleware automatically retrieves the patient's electronic medical record from the hospital information system, extracting key information such as recent medication history, allergy history, previous surgical records, and recent examination reports, and converting it into a structured document that meets the emergency center's data standards. Finally, the system activates the voice guidance module, using clear and concise voice commands to guide the patient or on-site personnel in implementing initial first aid measures (such as administering nitroglycerin or maintaining a specific position). Key physiological data selection follows clinical emergency guidelines, prioritizing abnormal ECG segments, continuous blood pressure trend charts, blood oxygen saturation curves, and current activity status (using inertial sensor data to determine whether the patient has fallen or is still).

[0030] Furthermore, the transmission reliability and data security of the emergency communication module are enhanced. The redundant communication link switching mechanism is based on real-time channel quality assessment. The primary channel prioritizes the 5G network's eMBB service type to ensure data transmission speeds under high bandwidth requirements. The backup channel is configured as NB-IoT and LoRa dual-mode standby to ensure that basic data transmission can be maintained even when 5G signal coverage is insufficient. The channel switching decision algorithm comprehensively evaluates three indicators: signal strength, bit error rate, and network latency. When any indicator exceeds the preset threshold, the switch is automatically triggered. The switching process uses seamless technology, and data packet sequence number synchronization and cache retransmission mechanisms are used to prevent information loss. Patient location information is processed using differential privacy protection technology. The specific method is: random noise that conforms to the Laplace distribution is added to the original GPS coordinates. This ensures that the location information obtained by the emergency center is valid within a 100-meter accuracy range, but cannot be traced back to a specific street or house number. The medical data middleware features a built-in protocol conversion engine, supporting bidirectional conversion across multiple medical data standards, including HL7, FHIR, and DICOM. When retrieving historical medical records, it automatically matches key fields required for emergency response. For example, it can integrate dispersed test results, imaging reports, and medication records into a unified timeline for diagnosis and treatment summaries. Furthermore, the middleware implements data desensitization, masking sensitive information such as patient ID numbers and contact information while retaining diagnostic conclusions and treatment records for emergency reference.

[0031] like Figure 1 Figure 2 shows the design of a time series correlation analysis unit, which uses time series correlation of multimodal signals to identify potential pathological features. The analysis unit first performs phase synchronization analysis on the RR interval sequence of the ECG signal and the inspiration-expiration cycle of the respiratory signal, calculating the mutual information entropy between the two to assess autonomic nervous system regulation. When respiratory sinus arrhythmia features are significantly weakened or eliminated (as indicated by the correlation coefficient between the RR interval variability and the respiratory cycle falling below a threshold), the system automatically increases the risk level. In particular, if a step-wise decrease in blood oxygen saturation is also present, this combined feature is flagged as a precursor to respiratory failure. An activity-intensity adaptive mechanism is incorporated into the personalized threshold update cycle. Metabolic equivalents (METs) are calculated using inertial sensor data. During low-intensity activity periods (such as sleep), the baseline calculation window is extended to 2 hours, and a moving average algorithm is used to smooth fluctuations in physiological parameters. During high-intensity activity periods (such as walking or stair climbing), the window is shortened to 30 minutes, and the sampling frequency is increased to capture rapidly changing trends. To address the common circadian rhythm differences among elderly patients, the system establishes two independent baseline models for daytime and nighttime use. This smoothes baseline transitions when switching between modes to avoid false alarms caused by sudden threshold changes. Furthermore, the analysis unit integrates a pharmacokinetic model. Upon detecting a patient taking cardiovascular medications such as beta-blockers, the analysis unit automatically adjusts the time-decay curve of the heart rate abnormality threshold to avoid misjudgments due to physiological parameter changes caused by the medication's effects.

[0032] The present invention provides a dynamic monitoring system for elderly heart patients. It fuses electrocardiogram, blood oxygen and respiratory signals through multimodal sensors, dynamically optimizes signal gain in combination with skin contact status, and eliminates motion artifacts. It adopts a lightweight model that collaborates with edge computing and the cloud to analyze physiological data in real time, identify arrhythmia events and classify risk levels. It builds a graded alarm mechanism to trigger local reminders, remote notifications or emergency linkage according to the risk level. When responding to emergency response, it automatically retrieves the patient's medical records and enables redundant communication links to transmit key data, realizing closed-loop management of the entire process from abnormality detection to emergency decision-making, thereby improving monitoring accuracy and treatment timeliness.

[0033] Therefore, the present invention provides a dynamic monitoring system for elderly heart patients, which can solve the problems of poor real-time performance, high false alarm rate and delayed emergency response in dynamic monitoring of elderly heart disease patients.

[0034] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the present invention. Anyone skilled in the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by one of ordinary skill in the art without departing from the spirit and technical principles disclosed herein are intended to be covered by the claims of the present invention.

Claims

1. A dynamic monitoring system for elderly heart patients, characterized in that: include: The multimodal sensor module collects ECG, blood oxygen, and respiratory signals from elderly patients in real time, and generates multimodal physiological signals after adaptively adjusting signal acquisition parameters to reduce environmental interference. This adaptive adjustment includes dynamically adjusting the signal acquisition frequency by monitoring impedance changes at the sensor-skin interface. When the contact impedance is detected to be outside a preset range, it automatically switches to a backup electrode array and compensates for signal attenuation. It also performs closed-loop regulation of the photoelectric sensor's transmission power based on ambient temperature fluctuations. The data processing module performs fusion processing on the multimodal physiological signals to eliminate motion artifacts and extract characteristic waveforms to generate preprocessed signals; The analysis module analyzes the pre-processed signal in real time based on a preset elderly ECG characteristic model, identifies abnormal ECG events and their risk levels, and generates graded alarm instructions; The alarm module triggers corresponding local reminders, remote notifications, or emergency linkage responses based on the graded alarm instructions. The emergency linkage response automatically sends the patient's location and key physiological data to the designated emergency center through the emergency communication module, and simultaneously retrieves the patient's historical medical records to assist in emergency decision-making. The alarm module implements a three-level decision logic: the first level initiates a local vibration reminder for occasional premature beats and records event characteristics; the second level sends an encrypted early warning message to a preset guardian terminal for sustained tachycardia events; and the third level immediately activates the full-duplex transmission channel of the emergency communication module upon detecting a life-threatening arrhythmia pattern; the emergency communication module is configured with a redundant communication link switching mechanism. When the transmission quality of the primary wireless channel drops below a threshold, the backup frequency band is automatically activated to establish a parallel data transmission path, and the patient's location information is processed using a differential encryption algorithm. Among them, the adaptive adjustment of the multimodal sensor module includes dynamically optimizing the signal gain according to the skin contact status, the risk level determination of the analysis module combines the patient's individualized threshold and the real-time physiological parameter deviation, the establishment of the individualized threshold is based on the patient's historical monitoring data to build a dynamic baseline, and when the real-time physiological parameter deviation exceeds the baseline fluctuation range, the online fine-tuning of the model parameters is automatically triggered.

2. A dynamic monitoring system for elderly heart patients according to claim 1, characterized in that: The fusion processing of the data processing module includes a dynamic filtering strategy based on motion state recognition, constructing a motion interference model through body motion signals collected by inertial sensors, using a time-frequency domain joint analysis method to separate the myoelectric noise component in the electrocardiogram signal, and incorporating the phase relationship between the respiratory signal and the blood oxygen signal into the motion artifact correction algorithm to generate a multimodal waveform data set with time alignment characteristics. After standardization, this data set is bound to the patient identity and stored in a cache queue for the analysis module to call in priority order.

3. A dynamic monitoring system for elderly heart patients according to claim 1, characterized in that: The elderly ECG feature model is obtained through training using a machine learning method. Its input layer receives preprocessed signal segments segmented by time windows. The middle layer uses an attention mechanism to weightedly fuse multimodal feature vectors. The output layer generates a probability distribution based on the electrophysiological characteristics of common arrhythmia types in elderly patients. The establishment of the individualized threshold builds a dynamic baseline based on the patient's historical monitoring data. When the deviation of real-time physiological parameters exceeds the baseline fluctuation range, the online fine-tuning of the model parameters is automatically triggered to reduce the misjudgment rate.

4. A dynamic monitoring system for elderly heart patients according to claim 1, characterized in that: The analysis module has a built-in timing correlation analysis unit for detecting the rhythm coupling relationship between the electrocardiogram signal and the respiratory signal. When the respiratory sinus arrhythmia feature disappears and is accompanied by a step-down decrease in blood oxygen saturation, the risk level judgment weight is automatically increased. The update cycle of the individualized threshold is adaptively adjusted according to the patient's activity intensity, and the baseline calculation window is extended during the resting stage at night to improve the stability of the judgment.

5. The dynamic monitoring system for elderly heart patients according to claim 1, characterized in that: A dynamic priority scheduler is provided between the data processing module and the analysis module. When abnormal cross-parameter correlation features are detected in multimodal physiological signals, the processing priority of the corresponding data stream is temporarily increased, and additional computing resources are allocated for multi-dimensional feature cross-validation. The generation process of the preprocessed signal includes timestamp calibration of the characteristic waveform to ensure that the data collected by different sensors remain strictly synchronized in the time domain.

6. A dynamic monitoring system for elderly heart patients according to claim 1, characterized in that: The alarm module is integrated with an intelligent learning unit, which establishes a personalized feedback model by recording the patient's response behavior to alarms at all levels. When it monitors that high-risk alarms have been ignored multiple times, the alarm intensity is automatically upgraded and a call chain of backup contacts is activated. The local reminder adopts a gradual stimulation strategy, with low-frequency vibration prompts in the initial stage. If no response is received, it switches to a composite sound and light mode, and simultaneously sends an attention signal to the data processing module to trigger more accurate physiological parameter sampling.

7. The dynamic monitoring system for elderly heart patients according to claim 1, characterized in that: A data storage module is provided between the data processing module and the analysis module. The data storage module generates a dynamic storage strategy based on the characteristic waveform energy distribution and abnormal event markers of the preprocessed signal. The stable baseline segment in the multimodal physiological signal is stored in a compressed encoding manner, while the signal segment containing characteristic waveform mutations retains the original sampling rate and adds a timestamp index. The data storage module dynamically adjusts the storage period according to the risk level feedback from the analysis module. The data related to high-risk events is permanently stored and a logical association with the emergency linkage response is established. When the alarm module triggers the emergency communication module, the multimodal data waveform of the set time period before the incident is automatically extracted from the data storage module, and an auxiliary diagnosis report containing time series evolution characteristics is generated and synchronously transmitted to the emergency center. The execution process of the storage strategy realizes the seamless connection between real-time data stream and historical data through a ring buffer, ensuring that the complete physiological parameter change trajectory can be reconstructed when the abnormal event is traced back.

Images