A blood oxygen monitoring system integrating sleep rhythm modeling

Abstract

This invention relates to the field of medical diagnostic technology and discloses a blood oxygen monitoring system that integrates sleep rhythm modeling. The system includes identifying sleep stage transition times using photoplethysmography (PPG) signals, detecting the blood oxygen saturation trough within the transition window and calculating the response lag time, and combining body movement signals to perform a reliability-weighted assessment of the lag time. This invention uses physiological rhythm transitions as a natural probe of the respiratory system's compensatory capacity, achieving dual diagnosis of sleep stage and respiratory function through a single signal. Simultaneously, it transforms body movement interference into an observation window for respiratory stress recovery capacity, significantly improving diagnostic reliability in home monitoring scenarios. The system completes core analysis at edge nodes and outputs a respiratory function assessment report that can be directly used for clinical decision-making.

Description

Technical Field

[0001] This invention relates to a blood oxygen monitoring system that integrates sleep rhythm modeling, belonging to the field of medical diagnostic technology. Background Technology

[0002] In the field of home-based sleep apnea diagnosis, current mainstream technologies rely on multi-sensor fusion strategies to collect blood oxygen saturation and sleep stage data, and use cloud algorithms to perform static statistics on the separated parameters to calculate indicators such as the average blood oxygen value throughout the night and the proportion of REM sleep.

[0003] However, this model has three systemic limitations: 1. The multi-sensor architecture leads to high equipment costs, and signal cross-interference forces the system to add complex filtering modules, which fundamentally conflicts with the need for simple deployment in home scenarios; 2. When users turn over at night and cause physical disturbances, the existing solution directly discards the data for that period, which not only causes the loss of stress response information of respiratory compensation ability, but also reduces the coverage of effective data due to frequent interruptions; 3. The analytical paradigm of independently processing sleep stage events and blood oxygen fluctuations completely ignores the dynamic coupling relationship between the two at physiological transition nodes, such as the sudden drop in pharyngeal muscle tone during the transition from deep sleep to REM sleep, making it difficult to detect mild to moderate respiratory compensation function degeneration.

[0004] Although recent edge computing optimization solutions have emerged, attempting to complete some signal processing locally, they still cannot break through the mindset of single-dimensional data statistics. For example, while some improved devices reduce cloud dependence, they can only output discrete event alarms and cannot capture the progressive functional decline trend across sleep cycles. This collective neglect of the value of physiological temporal correlation makes it difficult for existing technologies to meet the core diagnostic needs of low cost, all-day coverage, and predictive capabilities in home scenarios. Therefore, how to simultaneously analyze sleep rhythm transition events and blood oxygen response mechanisms through single-channel physiological signals, and transform body movement interference into compensatory capacity assessment parameters in high-interference environments, while simultaneously achieving early warning of respiratory function decline trends, has become the technical problem to be solved by this invention. Summary of the Invention

[0005] This invention provides a blood oxygen monitoring system that integrates sleep rhythm modeling. Its main purpose is to solve the problem of how to synchronously achieve sleep stage identification, dynamic assessment of respiratory compensation capacity, and early warning of functional decline trend through a single PPG signal.

[0006] To achieve the above objectives, the present invention provides a blood oxygen monitoring system that integrates sleep rhythm modeling, the system comprising:

[0007] The signal sensing module is configured to acquire the photoplethysmography (PPG) signal and body movement signal of the monitored subject.

[0008] The rhythm transition recognition module is configured to: identify the start time of the transition from non-rapid eye movement (NREM) to rapid eye movement (REM) sleep state based on the amplitude envelope shape of the photoplethysmography (PPG) signal; wherein, the rhythm transition recognition module identifies the start time by calculating the first-order difference or short-time energy change of the PPG signal amplitude envelope.

[0009] The response lag analysis module is configured to: determine the time when the blood oxygen saturation trough occurs within a predetermined time window after the start time, in conjunction with the photoplethysmography signal, and calculate the response lag duration.

[0010] The risk assessment decision module is configured to: assign a confidence weight to each calculated response lag duration based on body motion signals; accumulate the weighted response lag duration sequence for the entire night; and output the respiratory system functional risk assessment result based on the clustering pattern of response lag duration events above the risk threshold in the weighted response lag duration sequence; the assessment of the risk assessment decision module reveals the respiratory system's instantaneous compensatory capacity to physiological disturbances by analyzing the dynamic response of a single photoplethysmography pulse wave signal at specific physiological rhythm transition points.

[0011] Preferably, the risk assessment decision module is further configured to: when the acceleration variance of the body motion signal exceeds an acceleration variance threshold within a predetermined duration, mark the corresponding response lag duration measurement as unreliable and set the credibility weight to zero to discard the measurement value.

[0012] Preferably, the risk assessment decision module is further configured to: after identifying a period of body motion interference indicated by body motion signals, determine a physiological disturbance index based on the acceleration variance of the body motion signals and the signal-to-noise ratio of the photoplethysmography pulse wave signals within that period; generate a compensation coefficient based on the physiological disturbance index by consulting a preset mapping table; and use the compensation coefficient to calibrate the first response lag duration measured immediately after the period of body motion interference.

[0013] Preferred physiological disturbance index ( The calculation method for ) is as follows: ,in, This represents the variance of acceleration of the body motion signal during the period of body motion interference. This represents the signal-to-noise ratio of the photoplethysmography (PPG) signal during the period of body motion interference.

[0014] Preferably, the risk assessment decision module is further configured to: divide the overnight response lag duration sequence according to the sleep cycle boundary determined by the sleep cycle segmentation algorithm; calculate at least one statistical feature value characterizing the respiratory stability of each sleep cycle; and determine the evolution trend of respiratory function risk by comparing the statistical feature values ​​of adjacent sleep cycles to provide early warning.

[0015] Preferably, the statistical characteristic values ​​are the average or maximum response lag duration within each sleep cycle; the trend warning is generated based on the cross-cycle deterioration index. The judgment that the respiratory stability is sustained or exceeds a predetermined trend threshold is based on the cross-cycle deterioration index, which is calculated as the ratio of the current cycle statistical characteristic value to the previous cycle statistical characteristic value. The trend warning indicates that there is a progressive risk of decline in respiratory stability.

[0016] Preferably, the change in the amplitude envelope shape of the photoplethysmography signal specifically refers to the fact that during the rapid eye movement (REM) phase, the amplitude of the photoplethysmography signal exhibits a shape with higher irregularity and a decrease in average amplitude compared to the non-REM phase.

[0017] Preferably, the decision logic adopted by the risk assessment decision module includes: comparing the response lag time with at least one critical threshold to generate an event risk level; and within a single sleep cycle, if the frequency of occurrence of an event risk level higher than a predetermined risk level exceeds a predetermined frequency threshold, then the respiratory system function risk assessment level is increased.

[0018] Preferably, the rhythm transition recognition module identifies the starting moment of the transition from non-rapid eye movement (NREM) to rapid eye movement (REM) sleep by analyzing the morphological characteristics of the single-channel photoplethysmography (PPG) wave signal from the wrist.

[0019] Compared with the prior art, the beneficial effects of the present invention are:

[0020] 1. By capturing the key physiological node of the transition from non-rapid eye movement (NREM) to rapid eye movement (REM) phase, the system transforms traditional absolute blood oxygen monitoring into a dynamic assessment of the respiratory system's compensatory capacity. When the pharyngeal muscle tone suddenly drops, healthy individuals show a rapid and stable blood oxygen response, while those with potential airway collapse exhibit identifiable delayed characteristics. This mechanism of transforming physiological rhythm phase transitions into endogenous probes allows a single PPG signal to simultaneously carry dual information on sleep staging and respiratory function diagnosis. This avoids the complex deployment of EEG sensors and reveals the instantaneous compensatory characteristics neglected by traditional polysomnography (PSG) systems.

[0021] 2. The system couples acceleration signal characteristics with PPG signal disorder to generate a physiological disturbance index. When a user turns over, triggering a measurement blind spot in traditional methods, this index dynamically calibrates the first valid measurement value through a preset mapping relationship. This design transforms the body movement period from a data waste zone into an observation window for the respiratory system's stress recovery ability. In the high-interference scenario of the home environment, it maintains the reliability of core indicators and significantly improves the utilization rate of effective data. Furthermore, by dividing the compensatory ability indicator sequence according to the sleep cycle, the system establishes a cross-temporal correlation of characteristic values ​​of adjacent cycles, such as the ratio of the lag duration of the deep sleep period to the REM period. The system extracts macro-trend information from micro-events. When a continuous rise in the cross-cycle deterioration index is detected, even if a single measurement does not reach the risk threshold, it can still trigger an early warning of progressive functional decline. This longitudinal comparison mechanism based on the physiological rhythm framework provides a time window for the intervention of chronic respiratory diseases that cannot be achieved by traditional single-point warnings.

[0022] 3. The system completes core operations such as rhythm conversion recognition, response lag analysis, and body motion correlation calibration at the signal sensing node, and only outputs weighted time-series event reports to the terminal. This marginal transformation from raw signals to diagnostic conclusions avoids the impact of cloud transmission delays on physiological event capture and allows the device to maintain full functionality in low-bandwidth environments. The weighted lag duration sequence and cross-cycle trend indicators generated by overnight monitoring directly correspond to the core parameters of compensatory capacity and functional evolution that are of concern to respiratory clinicians. The output report does not require doctors to analyze raw waveforms or complex data, which can support the initial screening and triage of obstructive sleep apnea, significantly reducing the diagnostic threshold of primary healthcare institutions and alleviating the resource burden on sleep centers in tertiary hospitals. Attached Figure Description

[0023] Figure 1 This is a time series diagram of the hysteresis assessment of blood oxygen monitoring response triggered by sleep rhythm transition in this invention;

[0024] Figure 2 This is a comparison chart of dynamic differences in blood oxygen response lag in this invention;

[0025] Figure 3 This is a schematic diagram of the sleep rhythm transition recognition method based on amplitude envelope difference according to the present invention;

[0026] Figure 4 This is a diagram of the respiratory function risk analysis system that integrates rhythm recognition and body movement assessment according to the present invention.

[0027] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in further detail below. Obviously, the described embodiments are only a part of the invention and not all of it. This invention provides a blood oxygen monitoring system integrating sleep rhythm modeling. Its overall operation begins with sensor data acquisition, proceeds through real-time signal analysis and dynamic evaluation, and finally outputs a structured risk report. Logically, the system mainly consists of a signal sensing module, a rhythm transition recognition module, a response hysteresis analysis module, and a risk assessment decision module. In a typical application scenario, the user wears a wrist-worn device integrating a single photoplethysmography (PPG) sensor and a single triaxial accelerometer during sleep at night. The signal sensing module is responsible for acquiring the raw data stream from these two sensors. Specifically, the PPG signal is acquired at a frequency of approximately 50Hz, a sampling rate sufficient to capture the fine morphology of the pulse wave, while the body movement signal is acquired at a frequency of approximately 25Hz to accurately record the user's posture changes and limb movements during sleep. The acquired raw data... The signal first undergoes preprocessing via a digital bandpass filter to effectively remove baseline drift caused by breathing and other activities, as well as high-frequency noise introduced by the circuit itself. This provides a high-quality data foundation for subsequent feature extraction and analysis. The processed photoplethysmography (PPG) signal is then fed into a rhythm transition recognition module. The core task of this module is to act as a sentinel for physiological states, accurately identifying the start time of the transition from non-rapid eye movement (NREM) to rapid eye movement (REM) sleep. Given that significant changes occur in the regulation of the autonomic nervous system during the transition from NREM to REM, leading to alterations in peripheral vascular tone, the PPG signal amplitude envelope exhibits a unique morphological characteristic. Compared to NREM, the REM signal amplitude shows higher irregularity and a decreasing average amplitude trend. To accurately capture the start time of this transition, this module is crucial. The rhythm transition recognition module first extracts the amplitude envelope of the photoplethysmography (PPG) signal in real time using Hilbert transform or peak detection and envelope fitting techniques. Then, the module calculates the first-order difference of this amplitude envelope sequence. When this difference value exhibits a sudden spike exceeding a preset negative threshold, or when the energy value shows a significant step drop after calculating the short-time energy of the signal, the system marks this moment as the start of the transition. This recognition process is based entirely on the morphological characteristics of the single-channel photoplethysmography (PPG) signal from the wrist, without the need to introduce additional EEG or eye-tracking sensors.

[0029] Once the rhythm transition recognition module identifies the start time The response lag analysis module is activated, its function being to quantify the dynamic response of the respiratory system to this endogenous physiological challenge. Physiologically, during the transition from non-rapid eye movement (NREM) to rapid eye movement (REM) sleep, the tension of the throat muscles naturally relaxes. For individuals with healthy respiratory function, their respiratory system can quickly compensate, maintaining stable blood oxygen saturation or only experiencing minor, transient fluctuations. However, for individuals at risk of airway collapse, this decrease in muscle tone can induce or exacerbate respiratory resistance, leading to a significant and delayed decrease in blood oxygen saturation. To quantify this process, the response lag analysis module activates at the initial moment... Immediately afterwards, an observation window of 90 to 150 seconds is opened. Within this window, the system continuously calculates the instantaneous blood oxygen saturation value using the red and infrared light absorption ratio in the photoplethysmography (PPG) signal, and searches for the trough value in this blood oxygen saturation sequence in real time. Once the trough value is determined, the time point at which it occurs is recorded. Finally, the system calculates the response lag time. This duration directly reflects the time delay between the transition from a physiological state to a significant stress response in the respiratory system, and has become a core indicator for assessing the compensatory capacity of the respiratory system.

[0030] Meanwhile, the risk assessment decision module processes body motion signals from the accelerometer in parallel and fuses the processing results with the output of the response hysteresis analysis module to ensure the reliability of the assessment results in a high-interference home environment. This module executes a sophisticated, context-based decision logic: First, to address the severe contamination of photoplethysmography (PPG) signals caused by violent body movements such as turning over, this module sets an acceleration variance threshold. If, within the time window corresponding to the calculation of a response hysteresis duration, the acceleration variance of the body motion signal exceeds this threshold for a predetermined duration, the measurement is considered to have suffered irreversible interference, and the measured response hysteresis duration value is then... Data marked as unreliable and with its corresponding credibility weight set to zero are discarded in subsequent cumulative analysis. Furthermore, the system not only effectively removes contaminated data but also transforms periods of bodily movement disturbance into opportunities to observe the respiratory system's stress recovery capacity. Specifically, after identifying a period of bodily movement disturbance indicated by bodily movement signals, the module will analyze the acceleration variance of the bodily movement signals within that period. Signal-to-noise ratio of photoplethysmography (PPG) signal Calculate a physiological disturbance index This index comprehensively quantifies the intensity of body movement interference and its actual impact on physiological signals. Subsequently, the system calculates... The system generates a compensation coefficient by consulting a mapping table pre-set in the device's memory. This compensation coefficient is then used to calibrate the first response lag time measured immediately after the current body movement disturbance period. This transforms body movement disturbance from a purely negative factor into an effective information window for assessing the speed of respiratory stability recovery.

[0031] After obtaining the overnight response lag duration sequence, weighted by confidence and calibrated by body movement, the risk assessment decision-making module enters a higher-dimensional analysis phase to reveal the clustering patterns and evolutionary trends of risks. The module first divides the overnight response lag duration sequence into several subsequences corresponding to sleep cycles, based on the sleep cycle segmentation algorithm and the sleep cycle boundaries. For each sleep cycle, the module calculates at least one statistical characteristic value that characterizes respiratory stability within that cycle, such as the average or maximum value of all valid response lag durations within the cycle. Furthermore, to provide early warning, the module determines the evolutionary trend of respiratory function risk by comparing the statistical characteristic values ​​of adjacent sleep cycles. This is achieved by calculating a cross-cycle deterioration index. This is achieved by defining the index as the ratio of the current period's statistical characteristic value to the previous period's statistical characteristic value. When the respiratory rate consistently or significantly exceeds a predetermined trend threshold, the system generates a trend warning, indicating a potential progressive decline in the user's respiratory stability. Furthermore, within a single sleep cycle, the module compares the response lag time with at least one critical threshold, generating event risk levels such as low, medium, and high risk. If, within a single sleep cycle, the frequency of events with risk levels exceeding a predetermined risk level exceeds a predetermined frequency threshold, the system automatically upgrades the overall assessment level of the user's current respiratory function risk. Finally, the system integrates all analysis results and directly generates a respiratory function assessment report on the edge computing node—the user's wrist-worn device—requiring no secondary interpretation by professionals. This report includes the final risk assessment level, potential trend warning information, and key quantitative indicators such as the overnight weighted average response lag time. This provides users or primary healthcare institutions with direct reference data for clinical decision-making, significantly reducing the threshold and complexity of professional-grade sleep respiratory health monitoring in home settings.

[0032] Furthermore, the preset mapping table is embedded within the device as a set of piecewise functions, which will calculate the physiological disturbance indicators. The value is directly mapped to a compensation coefficient for response lag time. In a specific implementation, this functional relationship is expressed as: when... When the value is lower than the preset resting baseline value of 1.5, the compensation coefficient is 1.0, while when... When the value is in the range of 1.5 to 7.0, the compensation coefficient increases accordingly. The value increases along the formula Coefficient = 1.0 + 0.09 * ( -1.5) linearly increasing, and when When the value exceeds 7.0, the compensation coefficient is set to an upper limit of 1.495. This functional relationship is obtained by performing piecewise regression analysis on over 1000 valid body movement events and their subsequent blood oxygen response data collected under simultaneous monitoring of a polysomnography system from at least 50 subjects with different physical characteristics. A one-time individualized baseline calibration procedure is performed upon initial use, guiding the user to continuously collect physiological signals for three minutes while lying still without significant body movement. The system uses the second-by-second variance data of the triaxial acceleration signal within these three minutes, taking the 95th percentile of its probability distribution, and sets it as the acceleration variance threshold for the user's subsequent judgment of severe body movement interference. Simultaneously, the system calculates the mean μ and standard deviation σ of the first-order difference sequence of the photoplethysmography pulse wave signal amplitude envelope during this period, and sets the individualized rhythm transition recognition threshold to μ minus 3.5 times σ. The cross-cycle deterioration index... The trend threshold is uniformly set to 1.05. This value represents that the rate of change of the core statistical feature value of adjacent sleep cycles exceeds the upper limit of the normal physiological fluctuation range observed in large physiological databases. All of these are extended implementation methods known to those skilled in the art.

[0033] Example 1: In a specific application scenario, the operation of the technical solution of the present invention is as follows: A user, who did not show any clear signs of respiratory disease in a routine physical examination, but suffered from daytime sleepiness and occasional nighttime awakenings due to shortness of breath, and whose overnight average blood oxygen saturation readings of a traditional home pulse oximeter remained within the normal range, failed to provide effective diagnostic clues. The user wore a wrist device integrating the system of the present invention for seven consecutive days of home sleep monitoring in order to reveal the true dynamic response characteristics of their respiratory function under physiological rhythm transitions. On the first night of monitoring, after the system was started, the signal sensing module began to synchronously collect the user's photoplethysmography (PPG) wave signal and body movement signal, and sent the data stream to the subsequent processing unit. After entering the first sleep cycle, when the user's sleep state transitioned from non-rapid eye movement (NREM) to rapid eye movement (REM) sleep, the rhythm transition recognition module analyzed the negative abrupt change in the first-order difference of the PPG wave signal amplitude envelope to determine the start time of the transition. Mark it down, and the response hysteresis analysis module will then... A 120-second observation window was then opened. Within this window, the instantaneous blood oxygen saturation sequence was continuously calculated by analyzing the photoplethysmography (PPG) signal. A blood oxygen trough was captured at the 58th second, and its occurrence time was recorded as [the value is missing from the original text]. The system calculates the initial response lag time based on this. Its value is 58 seconds. Here, the precise time marker of the rhythm transition recognition module provides an unbiased timing starting point for the response lag analysis module, making... The measurement accurately reflects the respiratory system's compensatory delay triggered by changes in endogenous pharyngeal muscle tone, rather than random blood oxygen fluctuations. This transforms an isolated physiological signal into a diagnostic carrier carrying dual information on sleep stage and respiratory function. During the middle of the night, when the user experiences a relatively vigorous turning-over movement, the system faces a technical dilemma in maintaining data integrity and reliability under high interference. Traditional solutions would discard this data segment directly if the acceleration variance of the body movement signal exceeded a threshold, creating an information blind spot. The system of this invention treats this period of body movement interference as an observation window for the respiratory system's stress recovery ability. The module first bases the data on the acceleration variance of the body movement signal during this period... Signal-to-noise ratio of photoplethysmography (PPG) signal Calculate a physiological disturbance index The value of this indicator is generated by consulting a mapping table preset in the device's memory to produce a compensation coefficient. This mapping table was established during the product development phase through regression analysis of reference polysomnography data under different intensities of body movement disturbances, ensuring the physiological significance and accuracy of the compensation coefficient. For the first response lag time measured immediately after the end of the body movement disturbance period, the system automatically applies the compensation coefficient for calibration. In this way, the system not only avoids data loss but also transforms an disturbance event into a quantitative assessment of the time required for the user's respiratory control system to recover stability after being subjected to external physical disturbances, resolving the inherent contradiction between data loss and measurement inaccuracy. At the end of the seven-day monitoring cycle, the system does not perform a full calibration on all data. Instead of using static statistics, the analysis framework is restructured to shift the focus from isolated respiratory event risks to assessing the long-term evolutionary trend of respiratory function stability. The risk assessment decision module first divides the weighted response lag duration sequence for each night into independent sleep cycle subsequences based on a sleep cycle segmentation algorithm, and calculates the maximum response lag duration within each cycle as the statistical characteristic value for that cycle. Subsequently, the module calculates the cross-cycle deterioration index between adjacent sleep cycles. This refers to the ratio of the current period's statistical characteristic value to the previous period's statistical characteristic value. Monitoring data shows that this user... During the second half of the night's sleep cycle, the value showed a pattern of being consistently higher than the preset trend threshold for three consecutive nights. Although the duration of its single response lag never reached the critical threshold of severe risk, this longitudinal comparison based on the physiological rhythm framework revealed a hidden risk of progressive decline in its respiratory compensation ability during the deepening of sleep. Ultimately, in the respiratory function assessment report generated by the system at the edge node, in addition to the usual quantitative indicators, a special warning was output, indicating that there was a potential progressive decline trend in its respiratory stability. This provided a decisive time window for early intervention before the user showed obvious clinical signs.

[0034] Example 2: To objectively verify the effectiveness and engineering feasibility of the technical solution of this invention in a simulated real-world high-interference home environment, particularly to examine the system's core function of converting body movement interference into a respiratory stress recovery observation window, and the accuracy of early warning of functional decline based on the cross-cycle deterioration index, this example was designed and implemented. In a typical application scenario, specifically for a test subject who frequently tosses and turns during sleep at night, traditional monitoring equipment would discard a large amount of data due to body movement, failing to form a continuous and complete assessment of respiratory stability, potentially missing crucial data. The present invention aims to verify how the system of the present invention can not only maintain the continuity of monitoring in this challenging scenario, but also extract deeper physiological information by taking advantage of the challenge itself. The test platform was built in a standard sleep laboratory. The test subject was a volunteer who met the characteristics of mild obstructive sleep apnea syndrome after initial screening and had frequent body movement events in the sleep report. The polysomnography (PSG) system was used as the gold standard, but the system of the present invention operated independently, using only its wrist device, which integrates a single photoplethysmography pulse wave sensor and a single triaxial accelerometer.

[0035] First, the acceleration variance threshold used to identify unreliable measurements was designed to accurately distinguish between benign, minor posture adjustments and violent bodily movements sufficient to contaminate the photoplethysmography (PPG) signal quality. The core technical trade-off in this design was to avoid misjudging valid data as interference and discarding it, while simultaneously preventing the inclusion of severely interfered data from reducing the overall accuracy of the assessment. To this end, a deterministic calibration procedure was implemented before the formal experiment: test subjects were required to wear the device and, as instructed, sequentially perform three standard movements: lying still, wrist micro-movements, and turning over in bed, simultaneously recording acceleration signals. By calculating the variance of the acceleration signal under each movement, and using the decimal place of the variance distribution for the turning-over-bed movement as the initial threshold, an objective and individualized benchmark was established for the selection of this parameter. Second, for the core component of the risk assessment decision module, namely the physiological disturbance index… The pre-defined mapping table between the compensation coefficients and the experimental parameters is not based on empirical inference, but rather originates from a pioneering study. The essence of this study was to establish a quantitative model between the intensity of physiological perturbations and the calibration requirements for respiratory response delays. In this study, multiple test subjects were recruited and, under PSG monitoring, actively induced body movements through external physical stimuli, while simultaneously recording the acceleration variance measured by the system of this invention. Photoplethysmography (PPG) signal-to-noise ratio The blood oxygen recovery time after apnea or hypoventilation events triggered by this perturbation, precisely captured by PSG, was established through regression analysis. The mathematical relationship between the actual blood oxygenation recovery delay and the actual blood oxygenation recovery delay was used to generate this mapping table, ensuring the physiological basis for subsequent compensation calibration. During the experiment, when the system was running stably in the first half of the night, it accurately identified the start time of several transitions from non-rapid eye movement (NREM) to rapid eye movement (REM) phases and calculated the response lag duration. The baseline value remained stable within a certain range, indicating the test subject's basic respiratory compensation ability without significant disturbance. At 2:16 AM, the accelerometer detected a violent body movement lasting longer than a preset value, with its acceleration variance significantly exceeding the preset acceleration variance threshold. At this point, conventional algorithms deployed on traditional devices would directly mark this period and subsequent time windows as invalid and discard them, causing a breakpoint in the assessment. However, the risk assessment decision module of this invention was triggered by this body movement event and entered its unique processing flow. The module did not discard this event information but immediately processed the data based on the acceleration variance during the period of body movement disturbance. Signal-to-noise ratio of photoplethysmography (PPG) signal A physiological disturbance index was calculated. Value, based on this The value is consulted in the preset mapping table to obtain a specific compensation coefficient, and this coefficient is used to calibrate the first response lag duration measured immediately after the motion disturbance period. The embedded table below presents representative data points before and after this key milestone, as well as during another sleep cycle. See Table 1:

[0036] Table 1. Experimental Data Monitoring and Calibration Table.

[0037]

[0038] As can be seen from the above data, at point B after the body motion disturbance, the uncalibrated The apparent improvement compared to the baseline value at point A is physiologically illogical, as physical stress typically increases the burden on the respiratory system. This invention, however, introduces a compensatory mechanism to... The calibration time was 21.0 seconds. This value is not only higher than the baseline, but also more accurately reflects the instantaneous negative impact of this physiological disturbance on respiratory compensation. In other words, the system will weight and calibrate all data from the entire night. The sequence is divided into sleep cycles, and the statistical characteristic value of each cycle is calculated, which in this case is the average. By calculating the cross-cycle deterioration index between adjacent cycles That is, the average of the current period Compared with the previous period average The ratio was found from the second to the fourth sleep cycle. The values ​​were 1.04, 1.08, and 1.11, respectively, showing a unidirectional increasing trend that was continuously higher than the preset trend threshold of 1.05, which caused the system to generate a trend warning in the final report.

[0039] Example 3: This example combines Figures 1 to 4 This paper describes the implementation of a blood oxygen monitoring system that integrates sleep rhythm modeling. Figure 1 As shown, the process begins with the rhythm transition trigger module detecting the transition start time t0 as the initial event. Subsequently, the observation window control module sets the observation window [t0, t0+150s], and determines the window length to be 90~150 seconds through statistical optimization based on physiological modeling. During this window, the blood oxygen calculation module starts the blood oxygen monitoring process. In the iterative calculation loop within the window time, the system obtains the red light / infrared light ratio and calculates the instantaneous blood oxygenation rate. Value, and at the same time The sequence is transmitted to the trough detection module, which updates the lowest value record in real time. After a trough is detected, the trough time tv is recorded, and the physiological process is annotated: sudden drop in pharyngeal muscle tone → decrease in ventilation → decrease in blood oxygenation, clearly explaining the physiological mechanism of the decrease in blood oxygenation. When the observation window ends, the system sends a window end signal to the trough detection module, which transmits the trough time tv to the lag calculation module. This module calculates the response lag time from the transition point to the occurrence of the blood oxygen trough using the formula lag time = tv - t0, and then calculates the output value of the lag time. Finally, the output result module performs a functional interpretation based on this lag time value: 4-6 seconds for healthy individuals and 15-20 seconds for at-risk individuals. This value can reflect an individual's compensatory ability.

[0040] like Figure 2 As shown in the figure, the horizontal axis represents time (seconds), and the vertical axis represents arterial blood oxygen saturation (%), depicting the temporal evolution of blood oxygen response within the converted window period. Healthy individuals are represented by solid lines, and at-risk individuals by dashed lines. After the (transition moment), the healthy individual's The curve remained basically stable, with only a slight dip followed by a rapid rebound, while the risk of individuals... The curve shows a significant delayed decline, and only gradually recovers after reaching its lowest point at tv (the trough moment), as clearly marked in the graph. =58 seconds, indicating that the at-risk individual from The response lag time between the transition time (transition time) and the tv (valley time) is used to quantify the time difference of the compensatory response of the respiratory system to the physiological stress caused by the transition of sleep rhythm.

[0041] like Figure 3 As shown, the original PPG signal in the upper part exhibits high amplitude regularity during the NREM stage, while transitioning to a low amplitude irregularity during the REM stage. The marked area transition represents the transition process from the non-rapid eye movement (NREM) stage to the rapid eye movement (REM) stage. In the amplitude envelope diagram in the middle, in A significant decrease occurs at this point, marking a morphological abrupt change in the amplitude envelope. The first-order difference curve of the bottom amplitude envelope further reveals the numerical change of this abrupt change. A mutation point appears at a corresponding location. If the mutation point is below the threshold, the discrimination condition of the rhythm conversion recognition module is met, and the accurate identification of the start time of the conversion is completed.

[0042] like Figure 4 As shown, the top layer contains sensor inputs, including a photoplethysmography (PPG) sensor and a triaxial accelerometer, used to acquire photoelectric and body motion signals during sleep. Below this is the signal sensing module, responsible for acquiring PPG signals (≈50Hz) and body motion signals (≈25Hz), and transmitting the signals to the subsequent processing module. The rhythm transition recognition module identifies the NREM→REM transition time based on the acquired signals. By analyzing the PPG amplitude envelope shape and calculating first-order difference / short-time energy mutation, it accurately determines the starting point of rhythm change. After identifying the transition point, the system simultaneously starts the response hysteresis analysis module and the body motion intervention module. The system includes a disturbance processing module, where the response lag analysis module identifies the trough of blood oxygen saturation and calculates the response lag duration within a 90-150 second window. The body motion disturbance processing module calculates the acceleration variance and determines the signal reliability of the acceleration signal, generating a physiological disturbance index (PDI). The system then proceeds to the risk assessment decision module, where it assigns a reliability weight to the response lag duration, divides the data according to sleep cycles, calculates the cross-cycle deterioration index (CDI), identifies risk events, and assesses patterns. Finally, the system outputs a respiratory function risk assessment report, which includes risk level, trend warning, and a summary of key indicators.

[0043] Example 4: In a specific application scenario, a new user uses a wrist device integrated with the system of this invention for the first time. After the user puts on the device, the system does not immediately enter a passive background monitoring mode. Instead, it actively guides the user to perform a one-time system initialization and individualized calibration procedure lasting about five minutes. The underlying technical logic of this procedure is that any universally applicable empirical threshold may introduce judgment bias when faced with individual physiological differences. Only a benchmark established based on the user's own physiological data under controllable conditions can ensure the accuracy of subsequent assessments. The procedure first instructs the user to keep their wrist still for one minute while sitting. During this period, the signal sensing module continuously collects photoplethysmography (PPG) signals and triaxial acceleration signals at its standard sampling rate. The system algorithm calculates the second-by-second variance of the acceleration signal for this minute and determines the 98th percentile value of this variance sequence as the user's individualized acceleration variance threshold. This ensures that the threshold used to identify severe body movement interference is set based on the user's own resting noise baseline, rather than a universal, potentially overly sensitive or insensitive fixed value. Next, the procedure enters the second phase, guiding the user through two minutes of rhythmic deep breathing training via device prompts. This involves deep inhalation and exhalation at a frequency of six times per minute. During this controlled breathing regulation process without external physiological load, the system can still identify subtle changes in vascular tone caused by autonomic nervous system fine-tuning, similar to the transition from non-rapid eye movement (NREM) to rapid eye movement (REM) phases. Based on this, the system calculates a series of baseline values ​​for response lag durations. The system then constructs a distribution sequence from all response lag duration values ​​measured within these 120 seconds and calculates... The system uses the mean and standard deviation to establish the mean as the user's individualized health status response benchmark. The sum of this benchmark plus twice the standard deviation is then set as the individualized threshold between low-risk and medium-risk events for that user, thus firmly anchoring the risk grading scale to the user's own physiological characteristics. During continuous monitoring throughout the night, the system's internal sleep cycle segmentation algorithm operates based on a deterministic temporal clustering logic. This algorithm identifies the start times of the transition from non-rapid eye movement (NREM) to rapid eye movement (REM) phases, as marked by the rhythm transition recognition module. As a time sequence, the algorithm iterates through the sequence and calculates any two adjacent starting times. and The time interval between ,like If the sleep cycle falls within the physiologically recognized typical range of 80 to 130 minutes, the system will... arrive The entire time period is identified as a complete sleep cycle. This method of dividing cycles based on physiological rhythm anchors, compared to traditional segmentation methods that rely on external clocks or simple body movement signals, can more accurately correlate respiratory stability analysis with the actual sleep structure. Furthermore, the construction process of the preset mapping relationship table embedded in the risk assessment decision module, used to calibrate the response lag time after body movement disturbances, follows a rigorous offline modeling procedure. During the product development phase, a group of test subjects with different physical characteristics were recruited. Under the synchronous monitoring of a professional polysomnography system, body movements of varying intensities were induced through controlled external physical stimuli. For each body movement event, the system synchronously recorded the physiological disturbance index calculated by the device of this invention. And the respiratory disturbances induced by the physical activity and the subsequent true blood oxygen recovery time, precisely measured by a polysomnography system, were analyzed using a large amount of collected data. Piecewise linear regression analysis was performed on the blood oxygen recovery time and the value, and finally a curve that could accurately reflect the quantitative relationship between the intensity of physiological disturbance and the required compensatory recovery time was fitted. After the curve was discretized, it was solidified into the mapping relationship table in the device memory.

[0044] After completing overnight monitoring, the system executes a hierarchical risk aggregation logic to generate a final assessment report. First, for each defined sleep cycle, the system calculates two core indicators based on all weighted and calibrated response lag duration data within the cycle: one is the risk event frequency, i.e., the number of events where the response lag duration exceeds an individualized critical threshold; the other is the cross-cycle deterioration index. Subsequently, the system calculates a comprehensive risk score for each sleep cycle. This score is derived by weighting the frequency of risk events with a cross-cycle deterioration index. The weighting coefficients used are the optimal values ​​determined through regression analysis in the aforementioned offline modeling procedure. Finally, the system calculates the total risk index for the entire night, which is the weighted average of the risk scores for all sleep cycles. The weight of the second half of the night's sleep cycles is appropriately increased to reflect the physiological characteristic that REM sleep dominates in the second half of the night. This total risk index is ultimately mapped to a three-tiered risk level system consisting of low risk, trend warning, and high risk, and combined with... The continuous changing trend of the data jointly determines the output of the final report. Thus, through the above series of interconnected deterministic procedures, the system of this invention transforms a nighttime monitoring into an individualized, full-time, multi-dimensional in-depth analysis of the user's respiratory system's compensatory capacity. The final assessment report not only reveals the current risk status but also indicates its evolutionary direction within the framework of physiological rhythms. This achieves the established technical goal of transforming body movement disturbances into an effective observation window and providing early warning for chronic respiratory function decline.

[0045] Example 5: In the system's internal parameter configuration, a preset mapping table is statically stored in the non-volatile storage unit of the wrist device. The establishment of this mapping table is based on an offline modeling procedure developed during the research and development phase of this invention. The process revolves around the logical correlation between the Physiological Disturbances Index (PDI) and the dynamic delay characteristics of the respiratory system's compensatory capacity. Specifically, in this procedure, the system first selects multiple test subjects with significant differences in vital signs and, under the synchronous recording of a polysomnography system, applies controllable body movement stimulation to induce real physiological disturbance events. Simultaneously, the system uses the acceleration variance of the acceleration signal and the signal-to-noise ratio of the photoplethysmography pulse wave signal to acquire the acceleration variance in real time. Based on this, the corresponding Physiological Disturbances Index (PDI) is calculated. Subsequently, the system matches the PDI with the blood oxygen recovery time revealed in the polysomnography data, using PDI as the independent variable and blood oxygen lag calibration requirements as the dependent variable, and executes multiple rounds of... Piecewise regression modeling is used, and the regression results are smoothed and discretized to form the mapping structure used for table lookup compensation. The compensation coefficients are called on demand during system operation without relying on external computing resources, and their values ​​are matched and looked up based on real-time PDI calculation results. This enables rapid calibration of the first response lag time under different disturbance intensities. It should also be noted that although the mapping table is built into a general form at the factory, the system also supports local reconstruction and dynamic optimization of it through an individualized calibration process during specific user initialization. Specifically, during the initial use phase, the user completes a set action sequence to induce representative physical events. The system uses this as an anchor point to resample the PDI value of the corresponding time period and automatically adjusts the calibration coefficients of the corresponding segment in the mapping table in combination with the lag performance, so that the overall compensation curve covers common disturbance ranges while better matching the actual physiological response characteristics of the individual.

[0046] Furthermore, the trend threshold used for the Cross-Cycle Deterioration Index (CDI) is not based on empirical judgment, but rather on a trade-off mechanism guided by engineering logic. The core of this threshold setting lies in coordinating the technical balance between the system's early warning capability and its tolerance for false alarms. If the threshold is set too low, the system will tend to be oversensitive, easily misjudging natural physiological fluctuations as abnormal trends and reducing the specificity of the assessment. Conversely, if the threshold is set too high, it will suppress the response to early deterioration signals and may miss the intervention window. Therefore, the threshold is ultimately set at a slightly increasing range that covers the upper limit of normal compensatory fluctuations within a typical sleep cycle. The specific value is calculated by combining the dynamic stability boundary of the photoplethysmography (PPG) signal used in this system under high interference conditions and the upper limit of the confidence band calculated from the mean rate of change of statistical characteristic values ​​within consecutive sleep cycles. This is calibrated through a harmonic regression model, ensuring that the system has both the sensitivity of judgment and can effectively avoid misjudgment caused by statistically random fluctuations.

[0047] 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.

[0048] 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 blood oxygen monitoring system integrating sleep rhythm modeling, characterized in that, The system includes: The signal sensing module is configured to acquire the photoplethysmography (PPG) signal and body movement signal of the monitored subject. The rhythm transition recognition module is configured to: identify the start time of the transition from non-rapid eye movement (NREM) to rapid eye movement (REM) sleep state based on the amplitude envelope shape of the photoplethysmography (PPG) signal; wherein, the rhythm transition recognition module identifies the start time by calculating the first-order difference or short-time energy change of the PPG signal amplitude envelope. The response lag analysis module is configured to: determine the time when the blood oxygen saturation trough occurs within a predetermined time window after the start time, in conjunction with the photoplethysmography signal, and calculate the response lag duration. The risk assessment decision module is configured to: assign a confidence weight to each calculated response lag duration based on body motion signals; accumulate the weighted response lag duration sequence for the entire night; and output the respiratory system functional risk assessment result based on the clustering pattern of response lag duration events above the risk threshold in the weighted response lag duration sequence; the assessment of the risk assessment decision module reveals the respiratory system's instantaneous compensatory capacity to physiological disturbances by analyzing the dynamic response of a single photoplethysmography pulse wave signal at specific physiological rhythm transition points. The risk assessment decision module is also configured to: after identifying a period of body motion interference indicated by body motion signals, determine a physiological disturbance index based on the acceleration variance of the body motion signals and the signal-to-noise ratio of the photoplethysmography pulse wave signals within that period; generate a compensation coefficient based on the physiological disturbance index by consulting a preset mapping table; and use the compensation coefficient to calibrate the first response lag duration measured immediately after the period of body motion interference. Physiological disturbance index ( The calculation method for ) is as follows: ,in, This represents the variance of acceleration of the body motion signal during the period of body motion interference. This represents the signal-to-noise ratio of the photoplethysmography (PPG) signal during the period of body motion interference. The risk assessment decision module is also configured to: divide the overnight response lag duration sequence according to the sleep cycle boundaries determined by the sleep cycle segmentation algorithm; calculate at least one statistical feature value characterizing the respiratory stability of each sleep cycle; and determine the evolution trend of respiratory function risk by comparing the statistical feature values ​​of adjacent sleep cycles to provide early warning. The statistical characteristic values ​​are the average or maximum response lag duration within each sleep cycle; trend warnings are generated based on a cross-cycle deterioration index. The judgment that the respiratory stability is sustained or exceeds a predetermined trend threshold is based on the cross-cycle deterioration index, which is calculated as the ratio of the current cycle statistical characteristic value to the previous cycle statistical characteristic value. The trend warning indicates that there is a progressive risk of decline in respiratory stability.

2. The blood oxygen monitoring system integrating sleep rhythm modeling according to claim 1, characterized in that, The risk assessment decision module is also configured to: when the acceleration variance of the body motion signal exceeds an acceleration variance threshold within a predetermined duration, mark the corresponding response lag duration measurement as unreliable and set the credibility weight to zero to discard the measurement value.

3. The blood oxygen monitoring system integrating sleep rhythm modeling according to claim 1, characterized in that, The change in the amplitude envelope morphology of the photoplethysmography (PPG) signal specifically refers to the fact that during the rapid eye movement (REM) phase, the amplitude of the PPG signal exhibits a higher degree of irregularity and a decrease in average amplitude compared to the non-REM phase.

4. The blood oxygen monitoring system integrating sleep rhythm modeling according to claim 1, characterized in that, The decision-making logic adopted by the risk assessment decision module includes: comparing the response lag time with at least one critical threshold to generate an event risk level; and within a single sleep cycle, if the frequency of an event with a risk level higher than a predetermined risk level exceeds a predetermined frequency threshold, then the respiratory system function risk assessment level is increased.

5. The blood oxygen monitoring system integrating sleep rhythm modeling according to claim 1, characterized in that, The rhythm transition recognition module identifies the starting moment of the sleep state transitioning from non-rapid eye movement (NREM) to rapid eye movement (REM) by analyzing the morphological characteristics of the single-channel photoplethysmography (PPG) wave signal from the wrist.

Images