Sleep breathing disorder assessment method based on limited lead signals

Abstract

The invention discloses a sleep dyspnea assessment method based on limited lead signals, and belongs to the technical field of sleep dyspnea monitoring. The method aims at solving the problems that in the prior art, an oxygen deduction index (ODI) algorithm lacks a unified standard for baseline definition and event boundaries, the accuracy and consistency of evaluation results are affected, and a traditional lowest blood oxygen saturation (LSpO) index is prone to being interfered by artifacts. According to the method, an adaptive wave crest baseline is combined with a four-node time sequence constraint structure to detect an oxygen reduction event, an ODI {deviation, duration} index is calculated, and the duration and the descending depth of the event are quantified; a quantile interval average value index is introduced to replace traditional LSpO so as to stably evaluate the severity of the hypoxemia; by analyzing the difference value of the heart rate variability indexes before and after sleep, the regulation degree of sleep on the autonomic nervous system is quantified. The system is cooperatively realized by the wearable device and the data processing device, and is especially suitable for household and large-scale screening scenes. The key word in the figure 1 is sleep breathing disorder; obstructive sleep apnea; an oxygen reduction index; a wave crest baseline is self-adapted; detecting a four-node event; heart rate variability; a wearable device; and oxyhemoglobin saturation.

Description

Technical Field

[0001] This invention relates to the field of sleep-disordered breathing (SDB) monitoring technology, and in particular to a method for assessing sleep-disordered breathing based on limited-lead physiological signals (including blood oxygen saturation SpO2, heart rate interval and exercise acceleration signals). Background Technology

[0002] Diagnostic challenges and equipment limitations in sleep-disordered breathing (SDB) Sleep apnea (SDB) has a high prevalence, but its clinical diagnosis still primarily relies on polysomnography (PSG). Due to the complexity and high cost of the equipment, as well as limited patient compliance, it is unsuitable for large-scale population screening. While home sleep apnea monitoring (HSAT) devices can improve diagnostic accessibility to some extent, commonly used Type III devices rely on nasal airflow sensors. These devices suffer from discomfort upon wearing, susceptibility to mouth breathing, and signal loss in the home environment, limiting their stability and data reliability. Type IV devices based on single or few leads have gained attention due to their ease of use, typically reflecting respiratory events through changes in blood oxygen saturation. However, a unified standard for identifying hypoxic events has not yet been established, leading to difficulties in comparing indicators and insufficient accuracy.

[0003] Technical challenges of existing Oxygen Depletion Index (ODI) algorithms ODI has been studied as an alternative indicator for many years, but the ambiguity of its definition limits its widespread application as a reliable diagnostic tool. (1) There are various ways to define the baseline and there is a lack of a unified standard. Existing ODI algorithms employ various baseline definitions. For instance, some schemes use a fixed baseline, such as the pre-sleep average as the benchmark (see patent document CN113116344B); while other studies propose dynamic local average baseline strategies, such as Sharma et al.'s approach of using the average blood oxygen saturation of the first 120 seconds as the benchmark and counting the number of events with a blood oxygen drop ≥4% per hour (reference "Sleep Overnight Monitoring for Apnea in Patients Hospitalized with Heart Failure", Journal of Clinical Sleep Medicine, 2017). This diversity in baseline definitions leads to poor comparability of results between different algorithms, affecting the consistency and reliability of ODI in clinical practice. This is one of the core challenges commonly faced by ODI-based algorithms. Some literature suggests that existing technologies (such as patent document ES2302446A1) have shifted to alternative paths such as frequency domain analysis to circumvent the problems caused by baseline definition. (2) Incomplete event boundary definition affects the accuracy of judgment. Current ODI algorithms lack comprehensive rules for determining oxygen depletion events. Most schemes, when defining events, typically only specify a threshold for the magnitude of the oxygen decrease (e.g., a decrease of ≥3% or 4% from baseline), but generally fail to simultaneously define a criterion for duration or a clear termination condition. For example, the scheme described in patent document CN119380998B counts oxygen depletion events based on the "total number of events below the threshold." This definition primarily covers the magnitude of the oxygen decrease but does not address the minimum duration an event should last or the rules for determining event termination. This incomplete event boundary definition can easily lead to the inappropriate merging of adjacent events or the over-segmentation of the same event when processing continuous oxygenation data, thus affecting the accuracy of event identification. (3) The information dimension is relatively simple, and the internal structural features of the event are not well described. Current technologies typically treat ODI as a simple frequency statistical indicator, lacking structured deconstruction and quantitative analysis of the dynamic characteristics within individual oxygen depletion events. For example, most existing ODI algorithms do not systematically extract and utilize key morphological parameters such as event depth (i.e., the maximum magnitude of blood oxygen decrease) and event width (i.e., the duration from the onset of decline to recovery to baseline). This makes the information dimension carried by the ODI indicator relatively singular, failing to fully tap the rich pathophysiological information contained in the blood oxygen saturation signal waveform, thus limiting its potential as a precise assessment tool to some extent.

[0004] Limitations of other relevant evaluation indicators In addition, other commonly used indicators also have certain limitations. For example, the lowest pulse oxygen saturation (LSPO2), as a single extreme point, is easily affected by transient motion artifacts and lacks stability; while assessment indicators based on heart rate variability (HRV) can reflect the activity of the autonomic nervous system, they are still affected by a variety of physiological and external factors, and their specificity needs to be improved.

[0005] In summary, existing sleep apnea assessment methods based on limited-lead signals still face unresolved technical challenges in defining the baseline of ODI, delineating event boundaries, in-depth information mining, and the robustness of other auxiliary indicators. Therefore, there is an urgent need in this field to develop a new technical solution that can achieve clear definitions, possess strong robustness, and provide more accurate structured analysis. Summary of the Invention

[0006] I. Technical problems to be solved This invention aims to propose improvements in the stability, repeatability, and structured analysis capabilities of evaluation results under conditions where the number of device leads is limited. Specifically, the technical problems to be solved include: (1) Solve the problem of oxygen reduction event identification bias caused by inconsistent baseline definition and incomplete event boundary determination; (2) Solve the problem of motion artifact interference and insufficient stability of LSpO2, an index for hypoxemia at extreme points; (3) To address the problem that existing HRV indicators are affected by a variety of non-sleep factors and are difficult to specifically reflect sleep-related autonomic nervous system regulation; (4) The problem of insufficient fusion of multimodal physiological information under limited lead conditions limits the overall accuracy of sleep apnea assessment.

[0007] II. Technical Solution To achieve the above-mentioned technical objectives, the present invention proposes the following technical solution: Option 1: A method for assessing sleep apnea based on limited lead signals The method is based solely on the following three synchronously acquired time series: blood oxygen saturation (SpO2) sequence, heart rate interval sequence, and motion signal sequence; the method includes the following steps: (1) The SpO2 sequence is preprocessed based on motion signal sequence to obtain effective signals that can be used for subsequent analysis; the preprocessing includes at least motion artifact correction to reduce interference from optical path changes caused by body movement and subsequent smoothing. (2) Identify the user's effective sleep time period based on the motion signal sequence; (3) During the effective sleep period, the SpO2 sequence is traversed chronologically, alternating between upward and downward trend segments, and blood oxygenation decline events are detected based on adaptive peak baseline, wherein: (a) The adaptive peak baseline is composed of the current effective SpO2 peak value and is continuously updated to the latest local maximum value during the upward trend segment; when SpO2 turns into a downward trend and meets the initiation conditions of the blood oxygenation decline event, the adaptive peak baseline is locked as the effective peak value of this event. (b) When the SpO2 value decreases relative to the adaptive peak baseline from the effective peak value to a preset deviation threshold, and the duration of this decrease reaches a first preset duration, it is determined as a blood oxygenation decrease event. (c) The end of the blood oxygenation decline event is defined by identifying effective troughs in the subsequent downward trend segment and determining the event end point based on recovery conditions; (4) Calculate the oxygen depletion index ODI_{deviation, duration} based on the number of blood oxygen depletion events detected per unit time. (5) Output oxygen reduction index ODI_{deviation, duration} is used as the first assessment parameter for assessing the risk of sleep-disordered breathing; Preferably, in step (3), the detection of blood oxygenation decline events based on adaptive peak baseline is further enhanced by a four-node temporal constraint structure to fully define an oxygenation reduction event. The four-node temporal constraint structure includes: In the upward trend segment, an effective peak node P and an oxygen reduction initiation node S are determined; wherein, node S satisfies the following condition: after node P, the blood oxygen saturation value decreases relative to the adaptive peak baseline to reach the preset deviation threshold and is continuously maintained for at least the first preset duration; In the downward trend segment, an effective trough node V and an oxygen reduction end node E are determined; wherein, node E satisfies the following condition: after node V, the blood oxygen saturation value rises above the recovery threshold and is maintained continuously for at least a second preset duration; A complete oxygen reduction event is defined by the four nodes: P, S, V, and E. More preferably, the duration of oxygen depletion is calculated based on the time difference between node E and node S, and the depth of oxygen depletion is calculated based on the difference in blood oxygen saturation between node P and node V.

[0008] Option 2: A method for assessing the severity of hypoxemia in sleep-disordered breathing. The method provides a robust assessment of hypoxemia by calculating the quantile average of blood oxygen saturation sequences within an effective sleep period. To achieve this objective, the method includes the following steps: First, the raw SpO2 sequences are preprocessed based on motion signal sequences and effective sleep periods are identified to obtain high-quality analytical data; then, the following core steps are performed: (1) Data preparation: Obtain the SpO2 sequence during the effective sleep period; (2) Index Calculation: Sort the SpO2 data in ascending order of value, take the top q% of data points to form the interval, and calculate the arithmetic mean of all data points in the interval as the AvgMinq%SpO2 index; wherein, the percentage q ranges from 0.5% to 2.5%; (3) Output results: Output the AvgMinq%SpO2 index as an assessment parameter for evaluating the severity of hypoxemia.

[0009] Option 3: A method for assessing changes in autonomic nervous system regulation during sleep. The method quantifies the regulatory effect of sleep on the autonomic nervous system by analyzing changes in heart rate variability (HRV) before and after sleep. The method includes the following steps: (1) Data preparation and time window setting: Based on the effective sleep time period identified through motion signal sequence, a first duration window (representing the "pre-sleep" state) is set at the beginning of the time period, and a second duration window (representing the "post-sleep" state) is set at the end of the time period; the duration of the first and second duration windows is equal, and the duration ranges from 0.5 to 2 hours; Obtain the heartbeat interval sequence of the user within the above two time windows; (2) HRV index calculation: Based on the heart rate interval sequences within the first and second time windows, respectively, at least one identical HRV index is calculated using time-domain analysis, frequency-domain analysis, and / or nonlinear analysis methods. Preferably, the time-domain analysis indicators include, but are not limited to: the standard deviation of the difference between adjacent normal heartbeat intervals (SDSD), the percentage of adjacent normal heartbeat intervals with a difference greater than 20 milliseconds (pNN20), and the median absolute deviation of heart rate (HRMAD). Preferably, the frequency domain analysis metrics include, but are not limited to: low-frequency power (LF), very low-frequency power (VLF), and total power (TP). More preferably, existing heart rate variability analysis toolkits (such as HeartPy) can be used to perform the calculation of the above indicators; (3) Calculation and evaluation of adjustment changes: The difference between the HRV value in the second duration window (after sleep) and the HRV value in the first duration window (before sleep) is calculated to obtain ΔHRV; Preferably, the ΔHRV is comprehensively interpreted based on the physiological significance of different HRV indicators to reflect the degree of improvement or load change of the autonomic nervous system's regulatory state during sleep.

[0010] Option 4: A comprehensive risk assessment method for sleep-disordered breathing By integrating the Oxygen Depletion Index (ODI_{deviation, duration}) and the AvgMinq%SpO2 index, a comprehensive risk assessment system is constructed, including the following steps: Based on a preset first set of judgment thresholds, the severity of the subject's sleep apnea disorder is determined. The severity classification includes at least four levels: "none," "mild," "moderate," and "severe." The severity of the subject's hypoxemia is determined based on a preset second threshold group. This severity grading also includes at least four levels: "none," "mild," "moderate," and "severe." Based on preset rules, a final comprehensive risk assessment report is generated; the report includes at least: A clear classification of the severity of sleep-disordered breathing; Clear grading of the severity of hypoxemia; The overall risk level is derived from the above two factors; Clinical action recommendations corresponding to the overall risk level.

[0011] Option 5: A sleep-disorder breathing assessment system This invention provides a sleep apnea assessment system for performing any of the aforementioned sleep apnea assessment methods based on limited-lead signals. The system achieves convenient and accurate home or clinical sleep monitoring and risk assessment through integrated hardware acquisition and software analysis. The system is characterized by comprising two main parts: an integrated wearable data acquisition device and a data processing device. (1) Integrated wearable data acquisition device The device is an integrated hardware unit used to synchronously collect the user's core physiological signals, and its components include: Oxygen saturation sensing module: used to acquire photoplethysmography (PPG) signals and generate oxygen saturation (SpO2) sequences and heart rate interval sequences based on these signals; Motion sensing module: used to synchronously collect the user's motion signal sequence for signal preprocessing and effective sleep period identification; Communication module: used to wirelessly transmit multiple collected physiological sequence data to external data processing equipment; The blood oxygen saturation sensing module and the motion sensing module share the same high-precision real-time clock to ensure that the collected SpO2 sequence, heart rate interval sequence and motion signal sequence are synchronized on the time axis. (2) Data processing equipment The device is a computing unit that receives and processes physiological data, and its components include: Communication interface: configured to receive the physiological sequence data from the wearable data acquisition device via wireless communication; Processing module: configured to perform operations on received data, the operations including at least one of the methods described in any of the aforementioned schemes one to four, such as oxygen depletion event detection, oxygen depletion index calculation, hypoxemia severity assessment, autonomic nervous system regulation change assessment, and comprehensive risk assessment. Output module: configured to output the sleep apnea assessment results generated by the processing module; The specific form of the data processing device includes, but is not limited to: smartphones, tablets, personal computers, cloud servers, or dedicated medical diagnostic terminals.

[0012] III. Beneficial Effects Compared with the prior art, the technical solution provided by the present invention can bring one or more of the following beneficial effects: 1. The accuracy and robustness of the oxygen depletion event detection method have been significantly improved, with outstanding clinical validation results. This invention, by introducing adaptive peak baseline technology and a four-node temporal constraint structure, fundamentally solves the shortcomings of existing methods, such as the lack of adaptability of fixed baselines and fuzzy event boundaries, effectively improving the accuracy and reliability of oxygen depletion event detection; To objectively verify the technical effects of this invention, the inventors, based on clinical research data including 499 adults, systematically compared the ODI_{bias, duration} calculation paradigm proposed in this invention (such as ODI4_8, ODI4_9, etc.) with existing ODI algorithms (ODI3_pobm, ODI4_pobm) that are publicly disclosed in the prior art, have clear literature support (Jeremy Levy et al., NPJ Digital Medicine, 2021), and have been open-sourced (POBM toolkit). Clinical validation results show that, within the parameter range defined in the claims (bias threshold 3% to 5%, first preset duration 3 to 15 seconds, second preset duration 1 to 8 seconds), the ODI parameters generated by this invention have significantly better correlation (r) and consistency (R²) with the gold standard AHI than those of the prior art. Representative example ODI4_8: r=0.955, R²=0.913, RMSE=6.43; Representative example ODI4_9: r=0.956, R²=0.914, RMSE=6.483; Compared to existing technologies (ODI3_pobm: r=0.852, R²=0.727; ODI4_pobm: r=0.875, R²=0.766), the present invention demonstrates significant advantages in its ODI calculation paradigm within the parameter range defined in the claims: r values ​​are all >0.95, R² values ​​are all >0.90, and RMSE values ​​are all <7.5. In particular, ODI4_8 achieves the lowest root mean square error (RMSE=6.43) and the smallest system bias (average difference = -0.004) while maintaining high correlation (r=0.955), fully demonstrating the comprehensive technical advantages of the present invention within the scope defined in the claims. By setting different combinations of deviation thresholds (3%, 4%, 5%) and durations (3-15 seconds, 1-8 seconds), this invention provides several high-performance ODI calculation paradigms, all with r-values ​​> 0.95 and R² > 0.90, offering flexible and reliable assessment options for different clinical rigor or application scenarios. This result clearly demonstrates that, within the parameter range defined in the claims, this invention can consistently provide superior oxygen depletion event detection performance compared to existing technologies, effectively solving problems such as large differences in identification results and the inability to accurately quantify event duration due to improper parameter selection in existing technologies.

[0013] 2. Provides multi-dimensional, structured information on oxygen reduction events. The technical solution of this invention provides the ability to extract multi-dimensional and structured oxygen reduction event information; Compared to the limitations of the traditional oxygen depletion index, which only reflects the frequency of events, this invention establishes a detection model containing four feature nodes: P, S, V, and E, to achieve precise quantification of key parameters for each oxygen depletion event: including the duration of the event based on the time difference between the S and E nodes, and the depth of decrease based on the difference in blood oxygen saturation between the P and V nodes. Therefore, this invention expands the hypoxia index from a single event count indicator into a multi-dimensional structured indicator that can simultaneously assess the severity of hypoxia, thereby providing more comprehensive and accurate information support for clinical diagnosis.

[0014] 3. The diagnostic performance and robustness of the indicators for assessing the severity of hypoxemia have been significantly improved. This invention innovatively introduces the AvgMinq%SpO2 index (specifically, AvgMin1%SpO2), effectively addressing the inherent limitations of the traditional lowest oxygen saturation (LSpO2) as an extreme point indicator. Traditional LSpO2 is highly susceptible to transient motion artifacts or signal noise, leading to misjudgments of the severity of hypoxemia. To objectively verify the technical effects of this invention, the inventors conducted a systematic comparison of AvgMin 1%SpO2 with traditional LSpO2 and AvgMin 0.5%SpO2 based on a clinical study involving 499 adults. The clinical results showed that AvgMin 1%SpO2 exhibited significant advantages in diagnosing moderate to severe OSA: The diagnostic accuracy reached 84.33% (optimal cutoff value ≤ 85.56%). Sensitivity: 87.24%, Specificity: 80.86%; Compared to the comparative indicators, AvgMin1%SpO2 showed significantly better diagnostic accuracy than traditional LSpO2 (78.76%) and AvgMin0.5%SpO2 (83.97%). This result demonstrates that the 1% quantile is the clinically validated optimal parameter selection in this invention: its performance surpasses that of traditional LSpO2, confirming the effectiveness of smoothing noise through integrated interval information; simultaneously, its performance is also superior to the 0.5% quantile, indicating that the 1% interval better balances the representativeness and stability of the data—effectively avoiding the impact of a single extreme event while fully covering the most severe hypoxic load at night. The AvgMinq%SpO2 index proposed in this invention, especially the clinically validated and optimized AvgMin1%SpO2, is a hypoxemia severity assessment method that has been clinically validated on a large scale and is significantly superior to traditional and similar improved indexes in terms of diagnostic accuracy and robustness, effectively improving the clinical practical value of the overall sleep apnea assessment scheme.

[0015] 4. Achieved specific quantitative assessment of autonomic nervous system regulation function during sleep. By calculating the ΔHRV of specific time windows before and after sleep, this invention provides an innovative and quantifiable method to evaluate the regulatory and restorative effects of sleep on the autonomic nervous system. This metric, combined with reliable effective sleep duration determined by the same device, ensures consistency in data analysis benchmarks and provides a valuable additional dimension for comprehensive sleep health assessment.

[0016] 5. The system architecture combines high performance, high availability, and multi-dimensional evaluation capabilities. The integrated system provided by this invention, based on limited physiological signals that can be synchronously collected by a single wearable device, achieves SDB assessment results highly consistent with the gold standard. Its advantages are specifically reflected in: High assessment accuracy: Through the innovative adaptive peak baseline method and four-node event structure, oxygen reduction events are accurately defined, significantly improving the accuracy of indicators such as ODI; The system exhibits excellent robustness: it performs artifact correction and effective sleep recognition through motion signals, and combines robust metrics such as AvgMinq%SpO2 to effectively resist interference and reduce misjudgments. The assessment has multiple dimensions: it provides parameters such as event frequency, hypoxia load, event morphology, and autonomic nervous system regulation (ΔHRV), thus constructing a more comprehensive assessment system; Excellent user experience and high accessibility: Relying on a single wearable device, it enables comfortable home-based monitoring, greatly improving user compliance and making it very suitable for large-scale screening and long-term health management. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the architecture of a sleep-disorder breathing assessment system (hardware and data flow) provided in an embodiment of the present invention; Figure 2 This is a flowchart illustrating the overall process of the sleep-disordered breathing assessment method provided by the present invention. Figure 3 This is a schematic diagram of the adaptive peak baseline method and four-node oxygen reduction event detection in one embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the calculation of the AvgMinq%SpO2 index in one embodiment of the present invention; Figure 5 This is a flowchart of a method for assessing the degree of regulation of the autonomic nervous system in one embodiment of the present invention. Detailed Implementation

[0018] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. These embodiments are implemented based on the technical solution of the present invention, but the scope of protection of the present invention is not limited to the following embodiments. Embodiment: Comprehensive assessment of sleep apnea based on a wristwatch device. This embodiment uses a wristwatch-style pulse oximeter integrating a photoplethysmography (PPG) sensor and a triaxial accelerometer as the data acquisition device, and an external computing device (such as a personal computer or cloud server) as the data processing device. The system architecture diagram is shown below. Figure 1 As shown.

[0019] I. Data Acquisition and Preprocessing Signal acquisition: The wristwatch-style pulse oximeter simultaneously acquires raw photoplethysmography (PPG) signals and three-dimensional acceleration signals at a sampling frequency of 200 Hz; Based on the PPG signal, the device calculates and records the blood oxygen saturation (SpO2) sequence in real time at a fixed frequency of 1 Hz; simultaneously, by detecting the pulse wave peak in the PPG waveform, it records the occurrence time of each heartbeat in an event-driven manner. Given that the sampling frequency of the PPG signal is 200 Hz, the timestamp resolution of the occurrence time can reach 5 milliseconds. Based on this, a heartbeat interval sequence is generated; The vector amplitude of the triaxial accelerometer was calculated and recorded as a motion signal sequence at a frequency of 1 Hz; All sequences share the device's internal high-precision clock, ensuring a time synchronization error of less than ±100 ms; SpO2 pretreatment: Motion artifact correction: An adaptive filtering algorithm based on acceleration signals is employed. When the amplitude of the motion signal exceeds a set quiet threshold, the SpO2 signal is deemed to be severely interfered with during that period, and is marked accordingly. In subsequent analysis, it is given a lower weight or subjected to interpolation correction. Smoothing: The effective SpO2 sequence after labeling is smoothed using a moving average filter with a window length of 5 seconds to suppress high-frequency noise.

[0020] II. Identification of Effective Sleep Periods Based on the preprocessed motion signal sequence, an activity threshold method is used to identify effective sleep periods. Specifically, the average activity level within each 5-minute window is calculated, and periods where the activity level remains below a preset threshold are initially identified as "potential sleep periods." Then, combined with time information (such as the user-set bedtime), logical judgments are made to ultimately determine the start and end points of the "effective sleep period."

[0021] III. Oxygen Loss Event Detection and ODI Calculation During the effective sleep period, the following algorithm is executed: Dynamic traversal and trend judgment: The algorithm iterates through the SpO2 sequence in chronological order, switching between two states: "upward trend" and "downward trend". Effective peak confirmation and oxygen reduction initiation (nodes P and S): In an upward trend, record the maximum value encountered as a candidate peak P; When the SpO2 value is detected to have decreased by 4% from point P (a preset deviation threshold), timing begins. If this decreasing state (SpO2 value ≤ P value - 4%) continues for at least 8 seconds (a first preset duration), then P is confirmed as a valid peak value, and the starting point of timing is marked as the oxygen reduction initiation point S. The algorithm then switches to a decreasing trend. Effective trough confirmation and end of oxygen reduction (nodes V and E): In a downtrend, record the minimum point encountered as a candidate valley value V; When the SpO2 value is detected to rise from point V to 4% (relative recovery), or to 97% (absolute threshold), timing begins. If this recovery state is maintained for at least 3 seconds (second preset duration), V is confirmed as a valid valley value, and the starting point of timing is marked as the oxygen reduction end point E. The algorithm then reverses its upward trend and begins searching for the next peak. Event definition and ODI calculation: From a complete PSVE sequence (such as Figure 3 (As shown) Define an oxygen depletion event. Count the total number of oxygen depletion events detected throughout the entire effective sleep period; The Oxygen Depletion Index (ODI) is calculated as follows: ODI_{4,8} = (Total number of oxygen depletion events / Total effective sleep duration (hours)), with the unit being "events / hour". (Note: The formula can also be equivalently expressed as (Total number of events / Total effective sleep seconds) × 3600).

[0022] IV. Calculation of AvgMin1%SpO2 index All SpO2 data points within the effective sleep period are sorted in ascending order of value. The lowest 1% quantile is determined, which is the set of data points that fall within the top 1% after sorting. A schematic diagram of this calculation is shown below. Figure 4 As shown. Calculate the arithmetic mean of all data points within this interval, which is the AvgMin1%SpO2 index.

[0023] V. Assessment of the degree of regulation of the autonomic nervous system The method flow is as follows Figure 5 As shown, the specific steps include: Data extraction: The heart rate interval sequences were extracted from the first 60 minutes (first predetermined duration window, ranging from 0.5 to 2 hours) and the last 60 minutes (second predetermined duration window, ranging from 0.5 to 2 hours) of the effective sleep period. In this specific embodiment, 60 minutes was selected as the preferred duration to achieve a balance between data sufficiency and time window independence. Indicator Calculation: Based on the heart rate interval sequences of the two time periods mentioned above, calculate at least one heart rate variability (HRV) indicator; the HRV indicator includes, but is not limited to: low-frequency power (LF), very low-frequency power (VLF), total power (TP), standard deviation of the difference between adjacent normal heart rate intervals (SDSD), percentage of adjacent normal heart rate intervals with a difference greater than 20 milliseconds (pNN20), and median absolute deviation of heart rate (HR MAD). Change calculation: Calculate the difference between the same HRV index before and after sleep, i.e., ΔHRV; Assessment Application: The ΔHRV value is used to assess the degree of regulation and recovery of the autonomic nervous system during sleep.

[0024] VI. Results Output and Evaluation The system analyzes the processed data and outputs the following core evaluation parameters and a comprehensive report: Oxygen Depletion Index (ODI_{4,8}): [Example value: 67.1] Events / hour Minimum blood oxygen saturation (AvgMin1%SpO2): [Example value: 78.1] % Autonomic nervous system regulation parameters (post-sleep values ​​- pre-sleep values): ΔLF: [Example value: 1560] ms² ΔTP (Total Power): [Example value: 4073] ms² ΔSDSD (Standard deviation of the difference between adjacent normal heartbeat intervals): [Example value: 17.38] ms ΔpNN20 (percentage of consecutive normal heartbeat intervals greater than 20 milliseconds): [Example value: 0.33] ΔHR MAD (Median Absolute Deviation of Heart Rate): [Example value: 0.21] Based on the combination of the above parameters, the system generates a comprehensive risk assessment report. The report's conclusions are based on a preset risk determination logic, such as: When ODI_{4,8} and AvgMin1%SpO2 indicate severe sleep apnea and severe hypoxia, and autonomic nervous system regulatory parameters (such as significantly increased ΔLF) show a specific pattern, the report may indicate: "Indicates a risk of severe sleep apnea, accompanied by severe nocturnal hypoxia and enhanced sympathetic dominance after sleep."

Claims

1. A method for assessing sleep apnea based on limited-lead signals, characterized in that, The method is based solely on the following three synchronously acquired time series: blood oxygen saturation (SpO2) sequence, heart rate interval sequence, and motion signal sequence; the method includes the following steps: The SpO2 sequence is preprocessed based on motion signal sequence to obtain effective signals that can be used for subsequent analysis. The preprocessing includes at least motion artifact correction to reduce interference from optical path changes caused by body movement and subsequent smoothing. The effective sleep time period of the user is identified based on the motion signal sequence; Within the effective sleep period, the SpO2 sequence is traversed chronologically, alternating between upward and downward trend segments, and blood oxygenation decline events are detected based on an adaptive peak baseline, wherein: (a) The adaptive peak baseline is composed of the current effective SpO2 peak value and is continuously updated to the latest local maximum value during the upward trend segment; when SpO2 turns into a downward trend and meets the initiation conditions of the blood oxygenation decline event, the adaptive peak baseline is locked as the effective peak value of this event; the adaptive peak baseline update is only reactivated when a new upward trend segment is subsequently detected. (b) When the SpO2 value decreases relative to the adaptive peak baseline from the effective peak value to a preset deviation threshold, and the duration of this decrease reaches a first preset duration, it is determined as a blood oxygenation decrease event. (c) The end of the blood oxygenation decline event is defined by identifying effective troughs in the subsequent downward trend segment and determining the event end point based on recovery conditions; The oxygen depletion index (ODI) is calculated based on the number of blood oxygen depletion events detected per unit time. The output oxygen depletion index (ODI)_{deviation, duration} is used as the primary assessment parameter for evaluating the risk of sleep-disordered breathing.

2. The method according to claim 1, characterized in that, The adaptive peak baseline-based detection of blood oxygenation decline events further employs a four-node temporal constraint structure, including: The blood oxygen saturation sequence is subjected to adaptive peak-baseline traversal in chronological order, and the upward trend segment and the downward trend segment are identified alternately. In the upward trend segment, an effective peak node P and an oxygen reduction initiation node S are determined, wherein the node S satisfies the following condition: after the node P, the blood oxygen saturation value decreases relative to the adaptive peak baseline to a preset deviation threshold and is continuously maintained for at least the first preset duration. In the downward trend segment, an effective trough node V and an oxygen reduction end node E are determined, wherein node E satisfies the following condition: after node V, the blood oxygen saturation value rises above the recovery threshold and is maintained continuously for at least a second preset duration. A complete oxygen reduction event is defined by the four nodes: P, S, V, and E. Specifically, the oxygen reduction index is calculated based on the number of oxygen reduction events per unit time, the duration of oxygen reduction is calculated based on the time difference between node E and node S, and the depth of oxygen reduction is calculated based on the difference in blood oxygen saturation between node P and node V.

3. The method according to claim 1 or 2, characterized in that: The deviation threshold is a decrease in SpO2 of 3% to 5%; The first preset duration is 3 to 15 seconds; The second preset duration is 1 to 8 seconds.

4. The method according to claim 3, characterized in that: The preset deviation threshold is 4%, and the first preset duration is 8 seconds; and / or The recovery threshold is the blood oxygen saturation value of the effective valley node V increasing by 4% or reaching the absolute threshold of 97%, and the second preset duration is 3 seconds.

5. A method for assessing the severity of hypoxemia in sleep-disordered breathing, characterized in that, include: Based on the synchronously acquired motion signal sequence, the SpO2 sequence is obtained within the user's effective sleep period from the collected SpO2 sequence. The SpO2 data within the effective sleep period are sorted in ascending order of value to determine the lowest q quantile interval, and the arithmetic mean of all data points within this interval is calculated as the AvgMinq%SpO2 index. The percentage q corresponding to the lowest quantile interval ranges from 0.5% to 2.5%.

6. A method for assessing the degree of autonomic nervous system regulation during sleep, characterized in that, include: Based on the collected motion signal sequences, the user's effective sleep time period is identified, and the heart rate interval sequence within that time period is obtained; Within the effective sleep period, a first duration window for the start and a second duration window for the end are set respectively; The first duration window and the second duration window have the same duration, and the duration is between 0.5 and 2 hours; Calculate at least one identical heart rate variability (HRV) index based on the heartbeat interval sequence within the first and second time windows, respectively. The difference between the HRV index value of the second duration window and the HRV index value of the first duration window is calculated to obtain ΔHRV; The degree of regulation of the autonomic nervous system during sleep is assessed based on the value of ΔHRV.

7. The method according to claim 6, characterized in that, The HRV metric includes at least one of the following: Low Frequency Power (LF); Very Low Frequency Power (VLF); Total Power (TP); Standard Deviation of Successive Differences (SDSD) of the difference between adjacent normal heartbeat intervals. The percentage of consecutive normal heartbeat intervals with a difference greater than 20 milliseconds (pNN20); Heart Rate Median Absolute Deviation (HR MAD).

8. A comprehensive risk assessment method for sleep-disordered breathing, characterized in that, Combine the oxygen reduction index ODI_{deviation, duration} calculated according to claim 1 with the AvgMinq%SpO2 index calculated according to claim 5; Among them, based on the preset judgment rules, the severity of sleep apnea is determined by the ODI_{deviation, duration}, and the severity of the accompanying hypoxemia is determined by the AvgMinq%SpO2. The final risk assessment report is generated by combining the two severity levels mentioned above.

9. A sleep-disordered breathing assessment system, characterized in that, include: Wearable data acquisition devices are used to simultaneously collect and generate the following three time series: (a) Blood oxygen saturation (SpO2) sequence; (b) Intercardiac sequence; (c) Motion signal sequence; The wearable data acquisition device has a clock module for time synchronization, ensuring that the time synchronization error of the three sequences does not exceed ±100 ms; The communication module is used to send the three sequences to the data processing device; A data processing apparatus, comprising a processor and a memory, the processor being configured to perform the method as described in any one of claims 1 to 8; The output module is used to output the assessment results of sleep apnea disorder. The data processing device can be any one of a smart terminal, a computer system, or a cloud server.

Images