A real-time monitoring and judging system for sleep period arrhythmia

Abstract

The present application relates to the technical field of physiological signal diagnosis, and discloses a real-time monitoring and judging system for sleep arrhythmia, comprising a sensor for synchronously collecting physiological signals and motion signals and a processor, wherein the motion signals are used for first re-gating to eliminate motion pollution, then a signal quality index is calculated for a resting data segment by a signal quality calculation module, and a reference signal captured in a motion-resting conversion event is used to dynamically set a diagnosis quality threshold to perform second re-gating by a dynamic self-calibration module; finally, an arrhythmia judging module performs arrhythmia judgment only on the data segment passing the double gating, the system avoids double interference of motion artifacts and low-quality signals in resting state by using serial double gating, and the unique dynamic self-calibration mechanism also translates motion interference into individualized calibration beacons, thereby solving the problem that a static threshold fails due to individual differences and ensuring the reliability of diagnosis.

Description

Technical Field

[0001] This invention relates to a real-time monitoring and judgment system for sleep-related arrhythmias, belonging to the field of physiological signal diagnosis technology. Background Technology

[0002] Currently, long-term, continuous, and effective monitoring of paroxysmal arrhythmias during sleep is a key technological requirement for preventing serious cardiovascular events. Existing technologies mainly fall into two categories. One category consists of medical-grade diagnostic devices, such as multi-lead Holter monitoring. While these devices provide rich diagnostic information, the complex lead wires and electrode attachment methods severely disrupt user sleep, leading to low user compliance and making them unsuitable for routine home monitoring. The other category comprises consumer-grade wearable devices based on photoplethysmography (PPG) or simple single-lead ECG. These devices are comfortable to wear and meet the compliance requirements for long-term monitoring, but they have inherent limitations in diagnostic accuracy. The main reason for this is that in the uncontrolled home sleep environment, users' unconscious turning over and limb movements produce strong motion artifacts. These motion artifacts are highly similar in signal morphology and temporal rhythm to various key pathological features, such as the irregular rhythm of atrial fibrillation or the morphological distortion of premature beats, resulting in severe pathological artifact morphology confusion.

[0003] To address this challenge, some technical solutions attempt to integrate multiple sensors, such as EEG and radar, to obtain richer physiological information. However, such methods often increase system complexity and fail to fundamentally resolve the confusion between motion interference and pathological features. For example, Chinese invention patent application CN120130950A discloses a sleep monitoring method and system based on the combination of EEG and millimeter-wave radar. This solution attempts to achieve comprehensive diagnosis by fusing EEG signals for sleep staging and millimeter-wave radar signals for respiration and heart rate. However, this multimodal fusion approach, on the one hand... This significantly increases the complexity and deployment difficulty of the system, deviating from the original intention of providing convenience for consumer-grade monitoring. On the other hand, its core millimeter-wave radar faces similar challenges to wearable devices when acquiring signals such as heart rate and respiration. That is, motion artifacts such as the user turning over and limb movement can still seriously contaminate the physiological signals, resulting in a lack of reliable signal quality assurance for subsequent judgments of arrhythmias and respiratory events. This method does not provide an effective mechanism to actively identify and remove these motion-contaminated data. Therefore, it still has limitations in solving the core problem of confusion between pathological features and artifact morphology.

[0004] Therefore, the technical problem to be solved by this invention is how to provide a brand-new monitoring and judgment system to avoid the problems of false alarms and missed diagnoses caused by the confusion of pathological artifacts, and to achieve reliable diagnosis and judgment while ensuring user comfort. Summary of the Invention

[0005] This invention provides a real-time monitoring and judgment system for sleep arrhythmias. Its main purpose is to solve the problem that existing technologies in sleep monitoring are prone to false alarms or missed diagnoses due to the confusing morphology of pathological artifacts, and it is difficult to balance wearing comfort and diagnostic reliability.

[0006] To achieve the above objectives, the present invention provides a real-time monitoring and judgment system for sleep-related arrhythmias, the system comprising methods for synchronously acquiring the user's first physiological signal. Motion signals that characterize user body movements The sensor, and a processor connected to the sensor; the processor is configured to perform the following module functions:

[0007] The motion intensity analysis module is used to analyze motion signals. Generate an exercise intensity index (MII) in real time;

[0008] The first gating module compares the MII with a preset diagnostic invalidity threshold (DIT) to determine the first physiological signal. Data segments are marked as either resting valid state or motion contamination-invalid state;

[0009] The signal quality calculation module is used to calculate the signal quality of signals marked as being in a resting valid state. The data segment calculates its signal quality index (SQI) in real time.

[0010] The dynamic self-calibration module is a dynamic self-calibration procedure for setting the diagnostic quality threshold (DQT). The dynamic self-calibration procedure includes: (a) real-time monitoring of the MII to identify a motion-resting transition event where the MII value drops from above the DIT to below the DIT; and (b) automatically extracting the immediately following motion-resting transition event that is marked as a valid resting state. (c) Call the signal quality calculation module to calculate a reference signal quality index (SQI) for the reference signal. Base (d) According to SQI Base Configure DQT;

[0011] The second gating module is used to calculate the effective values ​​of each resting state. The SQI of the data segment is compared with the DQT set by the dynamic self-calibration module; if the SQI is lower than the DQT, then the corresponding... The data segment was remarked as a silent contamination-invalid state;

[0012] The arrhythmia judgment module is subject to an execution constraint that limits the arrhythmia judgment module to perform arrhythmia diagnosis operations only on data segments that are marked as being in a resting valid state and whose SQI is greater than or equal to DQT.

[0013] Preferably, the processor is also configured to perform the following function: calculate the total duration of all data segments marked as being in a resting valid state and whose SQI is greater than or equal to DQT. And count the total monitoring time. ;based on and Calculate an Effective Diagnostic Coverage (EDC) index; and output the arrhythmia diagnosis results obtained by the arrhythmia judgment module together with the Effective Diagnostic Coverage (EDC) index.

[0014] Preferred, first physiological signal The photoplethysmography (PPG) signal is a motion signal. This is the ACC signal from the triaxial accelerometer.

[0015] Preferably, the Motion Intensity Index (MII) is based on motion signals. The variance is calculated within a preset time window.

[0016] Preferably, the Motion Intensity Index (MII) is based on motion signals. It is calculated using the absolute value of the integral within a preset time window.

[0017] Preferably, the signal quality calculation module, when calculating the signal quality index (SQI), includes: analyzing the resting effective signal quality index (SQI). Statistical characteristics of the data segments to quantify their morphological reliability; these statistical characteristics include... Kurtosis of the data segment is used to characterize the sharpness of the pulse waveform; statistical characteristics also include signal purity, which is defined as the ratio of the peak energy of the main pulse wave to the baseline noise energy.

[0018] Preferably, the dynamic self-calibration module sets the diagnostic quality threshold (DQT) based on the reference signal quality index (SQI). Base Dynamically set, it follows the rule: DQT = k × SQI Base ,in, The preset scaling factor is 0.3 to 0.8.

[0019] Preferably, the processor is further configured to: continuously monitor the effective diagnostic coverage (EDC) within a rolling time window; determine whether the EDC is below a preset paradigm failure threshold; if the EDC is below the paradigm failure threshold, suspend the execution constraints of the arrhythmia judgment module and instead execute a downgraded diagnostic algorithm, which applies the first physiological signal... Perform long-term average heart rate analysis.

[0020] Preferably, the processor is also configured to perform motion pathogenesis analysis: collect all events marked as motion contamination-invalid states to form a motion event sequence; for a single motion event in the sequence, backtrack and extract events marked as resting valid and with a SQI greater than or equal to DQT within a preset time window prior to its occurrence. Signal; Analysis Physiological precursor characteristics of signals; etiological classification of motion events based on physiological precursor characteristics.

[0021] Preferably, the physiological precursor features include heart rate or heart rate variability; and the etiological classification includes classifying the motor event as motor activity triggered by autonomic arousal or spontaneous movement.

[0022] Compared with the prior art, the beneficial effects of the present invention are:

[0023] 1. In the real-time monitoring and judgment of arrhythmias during sleep, a dual-gating mechanism is used to establish a physical resting condition benchmark based on motion signals. This actively eliminates all data segments whose physiological signal morphology is distorted due to limb movement, avoiding the technical challenge of confusing pathological features with motion artifacts in the diagnostic field. Furthermore, within the data segments confirmed to be at rest, a quality check based on the morphological characteristics of the physiological signals themselves is introduced to identify and eliminate low signal-to-noise ratio data caused by non-motor factors such as low perfusion or poor contact. This serial, orthogonal recognition process, which first judges body motion interference and then judges signal quality, ensures that the data used for arrhythmia judgment has dual validity guarantees at both the physical and signal levels before entering the diagnostic algorithm. This avoids the dual risks of false alarms from motion artifacts and false alarms from low-quality resting data that existing technologies cannot simultaneously address.

[0024] 2. During the arrhythmia assessment process, a novel indicator for evaluating the reliability of the monitoring is calculated in parallel using the resting-state valid and motion-contaminated invalid status markers generated by the gating mechanism. This indicator, namely the effective diagnostic coverage rate, enables the system to output not only a diagnostic conclusion but also a quantitative assessment of the reliability of that conclusion. This approach, from the perspective of the diagnostic system architecture, solves the technical bottleneck that the validity of negative conclusions cannot be verified in long-term monitoring. For the first time, the monitoring report has the ability to self-assess its diagnostic quality, providing an objective decision-making basis for clinical judgment of whether a negative result is a true negative or invalid monitoring.

[0025] 3. This system further constructs a dynamic self-calibrating closed-loop logic, transforming the motion signal analysis mechanism from a contamination screening tool into an optimal sample capture triggering tool. Specifically, at the window immediately following a strenuous exercise session, it actively anchors the physiological signal that is in the optimal perfusion state at this time and uses it as an individualized benchmark calibration signal. The system reuses the signal quality index calculation function to quantify the quality of this benchmark signal and uses it as an anchor point to dynamically reset the diagnostic quality threshold. This operating mechanism, which translates the maximum interference source (strenuous exercise) into the optimal calibration source, enables the secondary quality gating scale to automatically adapt to the physiological baseline differences of different individuals, such as elderly people with low perfusion or the same individual with similar body temperature changes during different sleep stages. This solves the robustness problem that static thresholds inevitably fail due to individual differences in real diagnostic scenarios. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the real-time monitoring and judgment based on dual gating of the present invention;

[0027] Figure 2 This is a schematic diagram illustrating the changes in signal quality during the dual-gating process of the present invention;

[0028] Figure 3 The internal interaction logic diagram for setting the dynamic diagnostic threshold of this invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below; it should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the scope of protection of the invention.

[0030] This invention provides a real-time monitoring and judgment system for sleep-related arrhythmias, constructing a serial dual-gating mechanism. This mechanism utilizes physical motion signals to eliminate motion artifacts. Within the confirmed physically resting data, a second screening is performed using the signal's own quality indicators to eliminate low-quality signals caused by non-motion factors. Finally, the arrhythmia judgment module is only allowed to execute on high-confidence data segments that pass this dual-gating, thus avoiding false alarms and missed diagnoses at the source. The implementation of this system typically relies on an integrated first physiological signal... Sensors and motion signals Wearable devices with sensors, such as a wristwatch or ring, have an internal processor responsible for executing various functional modules. The preferred sensor is a photoplethysmography (PPG) sensor, used to acquire pulse wave signals that reflect the cardiac cycle; The preferred sensor is a triaxial accelerometer (ACC) sensor, used to acquire acceleration signals characterizing the user's body motion; the first stage of system operation is the synchronous acquisition of data and motion gating; the processor acquires PPG signals in parallel through the sensor interface with strictly synchronized timestamps. With ACC signal This strict temporal synchronization forms the physical basis for distinguishing between motion contamination and resting contamination; the motion intensity analysis module is configured to analyze the ACC signal in real time. Generate a sports intensity index A computationally simple implementation involves the module calculating the sum of variances of the three-axis ACC signals within a preset rolling time window (e.g., 1 second), or calculating the absolute value of its integral. The resulting scalar value quantifies the intensity of the user's body movement at the current moment, denoted as . .

[0031] The first-level gating module will calculate in real time. With a preset diagnostic invalidity threshold Compare; this It wasn't set arbitrarily; it's a key parameter determined through offline calibration experiments during the equipment development phase. These experiments calibrated... Value and The correlation between signal (PPG) morphological distortion, It is set to such a critical value: when Above this value, the waveform characteristics of the PPG signal become unusable for reliable pathological diagnosis due to the introduction of motion artifacts; based on this comparison result, the first gating module will select each... The data segment is logically marked as one of two states: if Below Marked as a valid resting state, if Greater than or equal to If the signal is marked as motion contamination - invalid, the second stage of system operation is a secondary quality gating for resting valid data. In an uncontrolled sleep environment, resting does not equate to reliable signal; non-motor factors such as low perfusion in the extremities or loose device wear can also lead to poor signal quality. Therefore, the signal quality calculation module only applies to data already marked as resting valid. The data segment calculates its signal quality index in real time. ; The calculation process aims to quantify the morphological reliability of a signal; a specific calculation procedure includes analysis. Statistical characteristics of the data segment: First, calculate the kurtosis of the data segment. This index is used to characterize the sharpness of the pulse waveform. A high kurtosis value usually corresponds to a clear and steep pulse wave peak, while a low kurtosis value corresponds to a flat waveform and a signal-to-noise ratio. Second, calculate the signal purity. This index can be defined as the ratio of the pulse wave's main peak energy (usually in the 0.5-3Hz frequency band) to the baseline noise energy, usually in a higher frequency band, such as 5-10Hz. A high ratio means that the main signal is prominent and the baseline is clean.

[0032] The third stage of system operation is the dynamic self-calibration of the gating threshold. To address the challenge of static quality thresholds failing due to individual physiological differences, such as low perfusion in the elderly, or baseline differences in the same individual at different time phases, such as body temperature changes, this system introduces a dynamic self-calibration module. The core idea of ​​this module is to translate motion disturbances into calibration beacons; this module monitors in real time... To identify once Value from higher than The state of motion, falling back to The following exercise-resting transition events; the physiological mechanism is that, immediately after strenuous exercise, although the limbs are at rest, the peripheral perfusion level is often at a relative peak in the user's current state; therefore, upon capturing an exercise-resting transition event, the dynamic self-calibration module automatically extracts the subsequent events marked as valid resting states. The data segment is selected and designated as the baseline signal, representing the optimal signal sample obtainable by the user at the current physiological baseline. Subsequently, the module calls the signal quality calculation module. The algorithm calculates the reference signal quality index (SQI) for this reference signal. Base ; and based on this SQI Base Dynamically set diagnostic quality thresholds This setting follows a proportional rule: DQT = k × SQI Base ;in, This is a preset scaling factor, whose value ranges from 0.3 to 0.8. For example, if... Setting it to 0.5 means that the signal quality of subsequent resting valid data must reach more than 50% of the user's own best sample quality to be considered diagnostically valuable; the fourth stage of system operation is to perform dual gating and constraint diagnosis; the second gating module will convert the resting valid data calculated by the signal quality calculation module into signal quality values. data segment The dynamic self-calibration module dynamically sets the parameters. Compare; if Below If the data segment is in a resting state, its signal quality is substandard, and therefore it is remarked as a resting contaminated-invalid state. Finally, the arrhythmia detection module (such as the atrial fibrillation detection algorithm, premature beat identification algorithm, etc.) is subject to a strict execution constraint. This constraint limits the module to only processing data segments marked as resting valid states, and whose... Greater than or equal to dynamic The data segments are used to perform arrhythmia diagnosis calculations; all data segments marked as motion contamination-invalid or resting contamination-invalid are actively removed and do not participate in the final diagnosis.

[0033] To address the bottleneck of the inability to verify the validity of negative results in long-term monitoring, this system also calculates an effective diagnostic coverage rate in parallel. The processor calculates the total duration of all data segments that passed through dual gating (i.e., those used for diagnostics). And count the total monitoring time. ;Calculation based on both index( The final diagnosis of arrhythmia will be related to this. The system outputs all indicators simultaneously, providing a quantitative assessment of the reliability of the diagnostic conclusions. Furthermore, it includes a robust safeguard mechanism to handle extreme conditions, such as a user experiencing continuous agitation throughout the night due to a specific illness. Extremely low levels caused the main diagnostic module to fail; the processor continuously monitors within a rolling time window. If judged If the failure rate falls below a preset paradigm failure threshold of 10%, the processor suspends the execution constraints of the arrhythmia judgment module and instead executes a degradation diagnostic algorithm. This degradation algorithm does not perform fine morphological or rhythm analysis on the signal, but rather performs all... The system performs a more robust long-term average heart rate analysis on contaminated signals, calculating the average heart rate over a 5-minute window to determine the presence of persistent severe bradycardia or tachycardia. Simultaneously, the system reuses contaminated information removed by the first gating mechanism. The processor collects all events marked as motion contaminated or invalid, forming a motion event sequence. For each motion event in the sequence, the processor backtracks and extracts events that occurred in a high-confidence state (resting valid and...) prior to their occurrence. )of The signal; analyzing the physiological characteristics of the precursor signal can reveal instantaneous changes in heart rate or heart rate variability (HRV); based on the physiological precursor characteristics, the etiological classification of the exercise event can be performed; if the HRV is stable before exercise, it can be classified as spontaneous exercise (possibly associated with PLMD, etc.); if there is a sharp drop in HRV or a sharp rise in heart rate before exercise, it can be classified as exercise triggered by autonomic nervous system arousal (possibly associated with arousal caused by sleep apnea), thus opening up two diagnostic dimensions of heart rhythm and sleep movement in a single monitoring.

[0034] Example 1: In a specific sleep monitoring application, the monitoring subject is an elderly user whose vital signs include poor circulation in the extremities, accompanied by periodic micro-movements. This scenario presents a dual technical challenge: limb movements are prone to morphological artifacts, while poor circulation leads to an already low baseline signal-to-noise ratio for physiological signals at rest. When the user's periodic micro-movements occur, the ACC signal collected synchronously... Fluctuations occur, and the exercise intensity analysis module calculates the exercise intensity index in real time. Subsequently, it instantaneously exceeded the preset diagnostic invalidity threshold. The first-level gating module immediately executes the first-level gating, and sends the PPG signal corresponding to this timestamp. The data segment is marked as motion contamination-invalid, and is therefore prevented from entering the backend arrhythmia detection module, thus avoiding the risk of falsely reporting such motion artifacts as arrhythmias.

[0035] During the resting period of the limbs, the user Below , The data segment was initially marked as resting valid; however, due to poor end-cycle performance, its PPG signal amplitude was low, resulting in a lower signal quality index (based on signal kurtosis and purity). It remains at a low baseline level; if a static baseline based on statistical data of a healthy population is used... Then almost all of the user's valid resting data will be lost. Below this static Relabeled as resting contamination-ineffective, ultimately leading to reduced effective diagnostic coverage ( When the value approaches zero, the monitoring system fails; the dynamic self-calibration module used in this system is activated under this condition; this module detects a single... Higher than The movement, and the fall back to Following the following exercise-resting transition event, the subsequent resting effective period, where the physiological perfusion level is briefly at the user's individual peak, will be automatically converted to active physiological perfusion. The data segment is identified as the reference signal; this module calls the aforementioned signal quality calculation module to calculate the reference signal and obtain an SQI that reflects the optimal signal state of the individual user. Base Furthermore, this module is based on In this embodiment, the rules are as follows: The value is 0.4, which is used to set a dynamic and individualized value for this module. ;this Values ​​lower than the aforementioned static However, it is still higher than the user's noise baseline; this is individualized. Under this benchmark, the user subsequently generated [data / information] during the resting period. Low value but reliable morphology All data segments can meet the requirements. Greater than or equal to the dynamic The conditions are met, thus passing through a series of dual gating; the arrhythmia judgment module ultimately executes on these confirmed valid data and outputs a result with high accuracy. The diagnostic results of the indicators; simultaneously, the motion contamination-invalid event sequences eliminated by the first gating are used in parallel for time series pattern analysis, with the processor tracing back their precursors. The signal analysis showed that the heart rate and HRV were stable, so it was classified as spontaneous movement and output together with the periodic analysis results, indicating signs related to sleep movement.

[0036] Example 2: To objectively verify the technical effectiveness of the tandem dual-gating mechanism in avoiding false alarms and improving diagnostic indicators, this example constructs a comparative experiment. This experiment includes gold standard data for reference. The experimental data comes from a synchronously collected sleep physiological database containing 50 subjects. During the subjects' sleep, this database simultaneously records the 12-lead ECG signal used to provide the gold standard reference and the first physiological signal used as algorithm input. That is, wrist-worn PPG signals and motion signals The data included wrist-worn ACC signals; ECG data were annotated frame by frame by a professional cardiologist, clarifying the exact start and end times and types of all arrhythmic events; the sampling rate of both PPG and ACC signals was 100Hz; the experiment selected data segments with a total duration of 100 hours, including typical sleep states (quiet sleep, accompanied by vigorous movements such as turning over, accompanied by limb micro-movements) and typical signal quality (high signal-to-noise ratio, low signal-to-noise ratio due to low perfusion).

[0037] The experiment set up three analysis groups, all using the same back-end arrhythmia detection algorithm, i.e., the rhythm irregularity algorithm used to detect atrial fibrillation, but the front-end data preprocessing methods were different: Control group 1 (filtering only): a conventional bandpass filter (0.5-8Hz) was used for PPG signals. The data is processed and then all data is sent to the backend algorithm; Control group 2 (single motion gating method): only the ACC signal is used. calculate and only Below The resting effective data is sent to the backend algorithm; the sample group of this invention (dual gating method): adopts the complete method of this invention, that is, based on the control group 2, the resting effective data is processed twice. Verification, and using the dynamic self-calibration procedure set. Gating is implemented, and only data that passes the double gating is sent to the backend algorithm. Table 1 shows the processing procedure and judgment results for three typical challenging data segments. and All of these are threshold values ​​calibrated or dynamically set according to the method of this invention.

[0038] Table 1: Examples of Processing and Judgment for Typical Challenging Data Segments

[0039]

[0040] Referring to Table 1, control group 1 could not distinguish motion artifacts in A-001 and A-003, nor resting-state low-quality signals in A-002 and A-004, resulting in a large number of false positives (FPs). Control group 2 successfully eliminated A-001 and A-003 through motion gating, avoiding false alarms from motion artifacts, but still could not process the resting-state low-quality signals of A-002 and A-004, also producing false alarms. The sample group of this invention, through cascaded dual gating, not only eliminated motion artifacts (A-001 and A-003), but also eliminated resting-state low-quality signals. Numbers (A-002, A-004) were used only for high-confidence data such as B-001 (true positive, TP) and C-001 (true negative, TN), thus eliminating false alarms in these segments. Statistical analysis was performed on all 100 hours of data to evaluate the overall diagnostic performance of each method; the results are shown in Table 2. The gold standard included 125 atrial fibrillation events (used to calculate TP and false negative FN) and 890 non-atrial fibrillation interference events (including motion and low-quality signals, used to calculate FP and TN). Sensitivity was defined as... Specificity is defined as follows: Positive predictive value (PPV) is defined as follows: .

[0041] Table 2: Statistical Comparison of Diagnostic Performance of Three Methods on a 100-Hour Dataset

[0042]

[0043] Experimental data show that the control group 1 method has extremely low specificity (78.1%) and positive predictive value (38.1%), generating 195 false positive events, which will cause serious false alarm problems for users in application. Control group 2, by removing motion data, reduced the number of false positives from 195 to 47 and improved the specificity to 94.7%, but it is still affected by low-quality resting state signals. The present invention sample group, based on control group 2, further removed low-quality resting state signals, further reducing the number of false positives from 47 to 4, achieving a specificity of 99.5% and a positive predictive value of 96.6%. The sensitivity (91.2%) and effective diagnostic coverage (79.1%) of the present invention sample group are slightly lower than those of control group 2, indicating that the system actively abandons a small number of real event data that occur during motion or low-quality periods in order to improve the reliability of the diagnosis. The experimental data objectively confirm the role of the present invention's dual gating mechanism in avoiding the dual interference of motion artifacts and low-quality resting state signals during sleep and improving the reliability of diagnostic conclusions.

[0044] Example 3: This example uses a sleep physiology database identical to that in Example 2, containing a 100-hour gold standard ECG reference, to evaluate the actual effect of a technical solution that does not employ the dual gating of this invention, but instead uses the background technique of first filtering and denoising, followed by diagnostic analysis, i.e., control group 1 in Example 2; control group 1 only uses PPG signals. However, without collecting or using the ACC signal. In control group 1, the 100-hour PPG signal collected... After being processed by an adaptive filtering algorithm commonly used in the art, which aims to filter out baseline drift and high-frequency noise and attempt to attenuate body motion artifacts, all filtered data are fed into the back-end arrhythmia judgment algorithm that is exactly the same as the sample group of the present invention in Example 2.

[0045] When processing the typical challenging data segments shown in Table 1 of Example 2, the performance of the control group 1 was as follows: For segments A-001 and A-003 (violent turning movements), although the adaptive filtering algorithm filtered out some noise, the motion artifacts were highly similar to pathological features such as atrial fibrillation in morphology and rhythm. The algorithm could not effectively distinguish them from the real pathological signals, resulting in the filtered signals still carrying strong artifact features. Therefore, the backend algorithm judged both segments as abnormal and generated two false positives (FPs). For segments A-002 and A-004 (resting but low perfusion), the signal itself had an extremely low signal-to-noise ratio. The filtering algorithm could not recover the effective pulse wave morphology from the noise. When analyzing these low-quality signals, the backend algorithm also misjudged the irregular noise morphology as pathological features and generated two more false positives (FPs). The diagnostic performance of the control group 1 and the sample group of the present invention (i.e., the sample group of the present invention in Example 2) on the entire 100-hour dataset was directly compared. The statistical results are shown in Table 3.

[0046] Table 3: Comparison of diagnostic performance between control group 1 and the sample group of the present invention

[0047]

[0048] Experimental results showed that, while control group 1, which used a filtering-then-diagnostic approach, maintained high sensitivity (96.0%), it resulted in 195 false positives, with a specificity of only 78.1% and a positive predictive value as low as 38.1%, meaning that over 60% of the detected abnormalities were false alarms. In contrast, the sample group of this invention, by introducing a filtering-then-diagnostic approach... The first level of gating of the signal and based on The second layer of gating actively eliminated data segments with pathological artifact morphological confusion, significantly reducing false positives from 195 to 4, thereby increasing specificity to 99.5% and positive predictive value to 96.6%. Data confirms that without the synchronous motion signal-based gating mechanism of this invention, relying solely on... The filtering of the signal itself cannot solve the problem of false alarms caused by motion artifacts and low-quality resting state signals in sleep monitoring.

[0049] Example 4: This example combines Figures 1 to 3 This describes a real-time monitoring and judgment system for sleep-related arrhythmias, such as... Figure 1As shown, the system first synchronously acquires signals to obtain the first physiological signal and motion signal, and calculates the motion intensity index to quantify the intensity of the user's physical activity. This index undergoes a first-level gating screening; data exceeding the diagnostic invalidity threshold is marked as motion-contaminated invalid data and discarded, while data below the threshold enters the signal quality index calculation stage to quantify the reliability of the signal morphology. Simultaneously, the motion intensity index is used as input to monitor motion-resting transition events in parallel. Based on this, the system automatically extracts the baseline signal, calculates the baseline signal quality index, and dynamically sets the diagnostic quality threshold. Subsequently, the signal quality index and the dynamic diagnostic quality threshold converge at the second-level gating. The system determines whether the former is greater than or equal to the latter. If it is below the threshold, it is remarked as resting-state contamination invalid data and discarded. If the conditions are met, it enters the arrhythmia judgment stage, which is constrained to diagnose only data that passes the dual gating. Finally, the system outputs the arrhythmia diagnosis conclusion and simultaneously calculates the effective diagnostic coverage rate in parallel. After statistically analyzing the total effective data duration and total monitoring duration, it outputs the EDC index.

[0050] like Figure 2 As shown, the horizontal axis of the graph represents processing, and the vertical axis represents signal quality / intensity percentage. The graph contains three curves: one representing physiological signal strength, and the other representing the signal quality / intensity percentage. (Dashed line) represents motion signals. (Dotted lines) and (solid lines) representing valid data segments, during the signal acquisition phase, and All are at a high level; after calculating the exercise intensity index, they are at the first level of gating. The intensity dropped sharply, and The decline is only slight, at which point valid data segments begin to appear, with their initial quality being lower than [previous level]. With the subsequent calculation of signal quality index, dynamic self-calibration, and the execution of the second gating, The curve and the effective data segment curve descended smoothly in sync, and the quality of the effective data segment was consistently lower than that of the curve. The original quality is ultimately used to determine the arrhythmia.

[0051] like Figure 3 As shown, this process involves a processor, a motion monitoring module, a reference signal extraction module, a quality calculation module, and a threshold setting module. When the process starts, the processor instructs the motion monitoring module to monitor MII changes in real time. When the motion monitoring module detects motion and discovers a motion-to-rest transition event, it triggers the reference signal extraction module. After confirming that the MII has dropped from above the DIT to below the DIT, the reference signal extraction module extracts the valid PPG data segment after the transition and identifies it as the reference signal. This reference signal is then transmitted to the quality calculation module, which calculates its reference signal quality index (SQI). Base SQI BaseThis is then passed to the threshold setting module, which performs dynamic setting of DQT=k×SQI. Base After the calculation, the threshold setting module updates the diagnostic quality threshold (DQT) to the processor, and the processor uses this new DQT for gating in subsequent data segments.

[0052] Example 5: Before deploying the method in wearable devices, it is necessary to preset two core parameters, namely the diagnostic invalidity threshold (…). ) and diagnostic quality threshold ( The dynamic setting ratio coefficient Offline calibration was performed; this calibration process relied on a pre-acquired, high-resolution reference database containing at least 200 hours of synchronized physiological data streams from subjects of different ages and health conditions (including healthy, those with poor peripheral circulation, and those with atrial fibrillation), specifically including wrist-worn PPG signals. (Sampling rate 100Hz), wrist-worn ACC signal (Sampling rate 100Hz), and the synchronous 12-lead ECG signal as the gold standard reference; all The signal segments were manually labeled by professional technicians based on their ECG synchronization signals as follows: State 1: Can be used for diagnosis (clear signal and normal ECG or atrial fibrillation), State 2: Motion contamination (ECG confirms motion artifacts), or State 3: Resting contamination (ECG confirms limb rest but PPG is distorted due to low perfusion or poor contact). The calibration procedure is as follows: First, the processor calculates the sum of the three-axis variances of the ACC within a 1-second window in this embodiment, using the simplified algorithm of the motion intensity analysis module. Value; second step, establish The third step involves analyzing the correlation between the values ​​and the gold standard artificial label (state one combined with state three is considered resting, and state two is considered moving); and then analyzing the correlation between the values ​​and the gold standard artificial label (state one combined with state three is considered resting, and state two is considered moving). The true positive rate (correctly removing state two: motion contamination) and false positive rate (incorrectly removing state one / three: resting contamination) below the threshold; the fourth step is to... A specific operating point on the ROC curve is set to achieve a preset engineering balance. In the calibration of this embodiment, [the following is done]: The maximum true positive rate (excluding motion data) that can be achieved when the false positive rate (false rejection of resting data) is below 5%. The critical value was determined to be 1.35 (dimensionless). The specific calculation procedure, i.e., the implementation of the signal quality calculation module, is as follows: For a given signal that has been determined to be silent and valid by the first gating module... In this embodiment, the window length of the data segment is 5 seconds: Calculate the kurtosis of the 5-second data segment. Perform a Fast Fourier Transform (FFT) on the data segment to obtain its power spectrum; calculate the main peak energy of the signal. That is, the power integral within the frequency band from 0.5Hz to 3Hz; calculate the baseline noise energy. That is, the power integral within the 5Hz to 10Hz frequency band; calculate the signal purity ( ), ;Will and Each part is normalized (mapped to the 0-1 interval), and then weighted and combined to obtain the final result. The value, the synthesis method was determined to be .

[0053] Dynamic scaling factor The calibration procedure is as follows: This calibration aims to address the individual variability challenges encountered in Example 1, and its goal is to find a... The value makes the dynamic threshold DQT = k × SQI Base It can effectively remove State 3: resting contamination data, without mistakenly removing State 1: data that can be used for diagnosis (especially baseline data). (Low perfusion user data); Calibration is as follows: First, from the reference database, simulate the logic of the dynamic self-calibration module, extract the reference signal after all motion-resting transition events, and use the above... The calculation procedure calculates the SQI for each subject. Base The baseline value; the second step, also from the database, is to extract all data segments marked as State 3: resting contamination by the gold standard, and calculate their... Value (denoted as) The third step is to test a series of [tests] in a loop. Values ​​(from 0.30 to 0.80, in steps of 0.05); for each The value is used to calculate the corresponding DQT = k × SQI. Base Then, the statistics are compiled. Can it be successfully removed? (Right now The proportion of ) is recorded as the resting contamination removal rate; the fourth step is to select the minimum value corresponding to a resting contamination removal rate higher than 98%. This measure aims to ensure a high rejection rate while maximizing the retention of valid signals. Through this procedure, the... The value was determined to be 0.45; following the above calibration procedure, the core preset parameters of the method were determined: , ;at the same time The calculation procedures are fixed; these parameters and procedures are pre-installed in all wearable devices shipped from the factory, serving as the basis for the operation of the first-level gating module, signal quality calculation module, and dynamic self-calibration module.

[0054] Example 6: This example illustrates the robustness guarantee mechanism of the method under an extreme condition, namely the activation and execution of the degradation diagnosis algorithm. The scenario is set as follows: a monitored user suffers from severe restless legs syndrome (RLS), causing them to continuously perform high-frequency limb movements during a monitoring period of up to one hour; during this period, the monitoring device executes the method as usual; due to the user's continuous limb movements, the calculated movement intensity index (…) Most of the time, over 95% of the tests were above the diagnostic invalidity threshold set by the first-level gating module. Therefore, the vast majority of All data segments were correctly marked as motion-contaminated - invalid by the first level of gating; effective diagnostic coverage was continuously calculated within a rolling time window (e.g., 30 minutes). The system determines the time frame within this 30-minute window. The figure is only 4.8%, which is lower than the preset paradigm failure threshold of 10%.

[0055] Given The threshold for paradigm failure has been lowered, and the system determines that the gating diagnostic method has failed, triggering a diagnostic mode downgrade. The system proactively suspends the execution constraints of the arrhythmia judgment module, i.e., pauses the execution of all fine-grained arrhythmia judgment algorithms that are highly sensitive to signal quality. Simultaneously, the system activates and executes a downgraded diagnostic algorithm; this downgraded algorithm does not... The signal is gated, but all signals within that hour are gated. The system performs long-term average heart rate analysis on signals, including data marked as motion contamination-invalid. Specifically, the system calculates the median heart rate within a 5-minute sliding window, a statistical method robust to short-term motion artifact spikes. In the latter half of this 1-hour period, the user's long-term average heart rate analysis shows that their 5-minute average heart rate remains consistently below 35 beats per minute. Based on the output of this downgraded diagnostic algorithm, the system generates a warning message for persistent severe bradycardia and correlates it with... The metric (4.8%) is output along with the data; this mechanism ensures that even in extreme cases where the primary diagnostic method fails due to data contamination, the method can still monitor high-risk persistent heart rate abnormalities rather than paroxysmal events.

[0056] 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.

[0057] 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. A real-time monitoring and judgment system for sleep-related arrhythmias, characterized in that, The system includes a primary physiological signal for synchronous acquisition of the user. Motion signals that characterize user body movements The sensor, and a processor connected to the sensor; the processor is configured to perform the following module functions: The motion intensity analysis module is used to analyze motion signals. Generate an exercise intensity index (MII) in real time; The first gating module compares the MII with a preset diagnostic invalidity threshold (DIT) to determine the first physiological signal. Data segments are marked as either resting valid state or motion contamination-invalid state; The signal quality calculation module is used to calculate the signal quality of signals marked as being in a resting valid state. The data segment calculates its signal quality index (SQI) in real time. The dynamic self-calibration module is a dynamic self-calibration procedure for setting the diagnostic quality threshold (DQT). The dynamic self-calibration procedure includes: (a) real-time monitoring of the MII to identify a motion-resting transition event where the MII value drops from above the DIT to below the DIT; and (b) automatically extracting the immediately following motion-resting transition event that is marked as a valid resting state. (c) Call the signal quality calculation module to calculate a reference signal quality index (SQI) for the reference signal. Base (d) According to SQI Base Configure DQT; The second gating module is used to calculate the effective values ​​of each resting state. The SQI of the data segment is compared with the DQT set by the dynamic self-calibration module; if the SQI is lower than the DQT, then the corresponding... The data segment was remarked as a silent contamination-invalid state; The arrhythmia judgment module is subject to an execution constraint, which limits the arrhythmia judgment module to perform arrhythmia diagnosis operations only on data segments that are marked as resting valid states and whose SQI is greater than or equal to DQT. The dynamic self-calibration module sets the diagnostic quality threshold (DQT) based on the reference signal quality index (SQI). Base Dynamically set, it follows the rule: DQT = k × SQI Base ,in, The preset scaling factor is 0.3 to 0.

8.

2. The real-time monitoring and judgment system for sleep-related arrhythmias according to claim 1, characterized in that, The processor is also configured to perform the following function: calculate the total duration of all data segments marked as resting valid and whose SQI is greater than or equal to DQT. And count the total monitoring time. ;based on and Calculate an effective diagnostic coverage (EDC) index; It also outputs the arrhythmia diagnosis results obtained by the arrhythmia judgment module, along with the effective diagnostic coverage (EDC) index.

3. The real-time monitoring and judgment system for sleep-related arrhythmias according to claim 1, characterized in that, First physiological signal The photoplethysmography (PPG) signal is a motion signal. This is the ACC signal from the triaxial accelerometer.

4. The real-time monitoring and judgment system for sleep-related arrhythmias according to claim 1, characterized in that, The Motion Intensity Index (MII) is based on motion signals. The variance is calculated within a preset time window.

5. The real-time monitoring and judgment system for sleep-related arrhythmias according to claim 1, characterized in that, The Motion Intensity Index (MII) is based on motion signals. It is calculated using the absolute value of the integral within a preset time window.

6. The real-time monitoring and judgment system for sleep-related arrhythmias according to claim 1, characterized in that, The signal quality calculation module, when calculating the signal quality index (SQI), includes: analyzing the resting effective signal. Statistical characteristics of the data segments to quantify their morphological reliability; these statistical characteristics include... Kurtosis of the data segment is used to characterize the sharpness of the pulse waveform; statistical characteristics also include signal purity, which is defined as the ratio of the peak energy of the main pulse wave to the baseline noise energy.

7. The real-time monitoring and judgment system for sleep-related arrhythmias according to claim 2, characterized in that, The processor is also configured to: continuously monitor the effective diagnostic coverage (EDC) within a rolling time window; determine whether the EDC is below a preset paradigm failure threshold; if the EDC is below the paradigm failure threshold, suspend the execution constraints of the arrhythmia judgment module and instead execute a downgraded diagnostic algorithm, which applies the first physiological signal... Perform long-term average heart rate analysis.

8. The real-time monitoring and judgment system for sleep-related arrhythmias according to claim 1, characterized in that, The processor is also configured to perform motion pathogenesis analysis: collect all events marked as motion contamination-invalid states to form a motion event sequence; for a single motion event in the sequence, backtrack and extract events marked as resting valid and with a SQI greater than or equal to DQT within a preset time window prior to its occurrence. Signal; Analysis Physiological precursor characteristics of the signal; Based on the characteristics of physiological precursors, motion events are classified etiologically.

9. A real-time monitoring and judgment system for sleep-related arrhythmias according to claim 8, characterized in that, Physiological precursor features include heart rate or heart rate variability; and etiological classifications include classifying motor events as autonomically triggered or spontaneous movements.

Images