System and method for evaluating myocardial load of OSA based on fusion of multi-modal physiological signals

Abstract

This invention relates to the field of respiratory detection technology and discloses a system and method for assessing myocardial load in obstructive pulmonary disease (OSA) based on multimodal physiological signal fusion. The system comprises: a signal acquisition module that acquires blood oxygen saturation, electrocardiogram (ECG), and chest surface electrocardiogram (CTECG) signals; a window positioning and trimming module that determines candidate windows based on blood oxygen saturation and ECG signals, extracts the trunk vibration component of the CTECG signal to determine trigger points, and extracts an effective excitatory sub-window; a feature extraction module that extracts myocardial mechanical load, ECG ischemic components, and timestamps within the sub-window, and calculates the electromechanical delay decoupling time difference; and an assessment and judgment module that calculates a stress index based on the aforementioned components and time difference, and outputs results with phenotypic classification. This invention utilizes multimodal physiological signal fusion to extract the effective excitatory window and extract electromechanical delay features, achieving objective assessment and phenotypic refinement of OSA myocardial load.

Description

Technical Field

[0001] This invention relates to the field of respiratory detection technology, specifically to an OSA myocardial load assessment system and method based on multimodal physiological signal fusion. Background Technology

[0002] Obstructive sleep apnea (OSA) is a respiratory disorder characterized by upper airway collapse during sleep, which causes intermittent hypoxia and dramatic fluctuations in intrathoracic negative pressure. Myocardial load generally refers to the mechanical workload and metabolic oxygen demand borne by the heart to maintain normal ejection function; under pathological conditions such as hypoxia, stress, and impaired pumping, myocardial load increases significantly and exacerbates myocardial oxygen imbalance. Therefore, objectively assessing myocardial load during OSA diagnosis (i.e., detecting whether myocardial injury is present) is a fundamental condition for developing individualized cardiovascular protective intervention plans.

[0003] Currently, multimodal physiological signal fusion is gradually becoming a common method for sleep monitoring. In current clinical practice, OSA assessment does not necessarily require indicators of myocardial damage. Traditional assessment methods typically collect data such as blood oxygen saturation, respiratory airflow, and baseline electrocardiogram simultaneously, identifying respiratory abnormalities by setting amplitude thresholds for independent signals from different channels. The analysis results are highly dependent on statistically analyzing the frequency of events occurring within a unit of time throughout the night, thereby grading the severity of the condition.

[0004] However, focusing solely on the frequency of respiratory events often overlooks the severity of cardiovascular involvement in different patients at equivalent frequencies. Current multimodal physiological signal fusion applications often remain at the surface superposition of multiple signals, failing to utilize physiological trigger points to capture the effective stress window of the body in response to hypoxia. This makes the analysis process susceptible to interference from irrelevant movements during sleep. Furthermore, conventional protocols lack simultaneous analysis of cardiac electrical conduction and physical work, making it difficult to reveal the substantial myocardial load and compression caused by abnormal respiratory events, and even more difficult to quantify the time delay between electrocardiographic characteristics and myocardial mechanical load. This makes it difficult for current systems to effectively detect hidden myocardial damage while assessing OSA, thus limiting the accuracy and comprehensiveness of related OSA myocardial load assessments.

[0005] Therefore, this invention proposes an OSA myocardial load assessment system and method based on multimodal physiological signal fusion to address the shortcomings of existing technologies. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an OSA myocardial load assessment system and method based on multimodal physiological signal fusion. This solves the problems of existing monitoring methods in accurately capturing the effective excitation window and simultaneously quantifying the electrical and mechanical delay characteristics to output objective phenotypic classifications when processing obstructive sleep apnea data.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an OSA myocardial load assessment system based on multimodal physiological signal fusion, comprising: The signal acquisition module is used to acquire blood oxygen saturation signals, single-lead electrocardiogram signals, and chest surface electrocardiogram signals. The window positioning and clipping module is used to determine obstructive sleep apnea events and define candidate time windows based on the blood oxygen saturation signal and the single-lead electrocardiogram signal, extract the trunk muscle contraction and vibration components of the chest surface electrocardiogram signal to determine the micro-arousal trigger point, and truncate the candidate time window to obtain the effective exciton window. The feature extraction module is used to extract the myocardial mechanical load component and the first timestamp of the chest surface electrocardiogram signal within the effective exciton window, extract the myocardial ischemic component and the second timestamp of the single-lead electrocardiogram signal, and calculate the electromechanical delay decoupling time difference based on the first timestamp and the second timestamp. The assessment and judgment module is used to calculate the myocardial mechanical ischemia stress index based on the myocardial mechanical load component, the electrocardiographic ischemia component, and the electromechanical delay decoupling time difference, and output the myocardial load assessment result of obstructive sleep apnea with myocardial vulnerability phenotype classification based on the myocardial mechanical ischemia stress index.

[0008] Preferably, the signal acquisition module assigns globally unique timestamps to the blood oxygen saturation signal, the single-lead electrocardiogram signal, and the chest surface electrocardiogram signal and performs timestamp alignment processing. The signal acquisition module separates the ultra-low frequency signal from the chest surface electrocardiogram signal, extracts the orthogonal projection vector of the ultra-low frequency signal in the three-dimensional spatial coordinate system to construct the body position vector, and calculates the envelope amplitude of the ultra-low frequency signal fluctuation during the respiratory cycle to construct the pleural impedance surrogate parameter. The signal acquisition module updates the ST segment isoelectric reference voltage of the single-lead electrocardiogram signal and the ventricular ejection work reference surface of the chest surface electrocardiogram signal in real time, based on the preset dynamic calibration matrix and using the body position vector and the pleural impedance proxy parameter as input variables.

[0009] Preferably, the window positioning and cropping module calculates the first time derivative of the blood oxygen saturation signal to identify the continuous decline phase of blood oxygen, extracts the RR interval sequence of the single-lead electrocardiogram signal, and calculates the low-frequency heart rate variability power of the RR interval sequence in the low-frequency band. When it is determined that the blood oxygen is continuously decreasing and the corresponding low-frequency heart rate variability power exceeds the baseline threshold, the window positioning and clipping module confirms that an obstructive sleep apnea event has occurred, records the start time of the blood oxygen decrease and the end time of the blood oxygen recovery to a stable baseline, and defines the time from the start time to the end time as the candidate time window.

[0010] Preferably, the window positioning and cropping module applies a bandpass digital filter to the chest surface electrocardiogram signal within the candidate time window to separate the trunk muscle group contraction vibration component, and uses a sliding time window to calculate the short-time root mean square value of the trunk muscle group contraction vibration component to generate a motion energy envelope. The window positioning and cropping module performs a reverse time axis scan from the end time to the start time within the candidate time window, and determines the moment when the motion energy envelope first crosses the dynamic threshold of signal quality as the micro-awakening trigger point. The window positioning and cropping module uses the start time and the micro-awakening trigger point as boundaries to obtain the effective exciton window.

[0011] Preferably, the feature extraction module locates the aortic valve opening feature point and the aortic valve closing feature point for each cardiac cycle within the effective exciton window, calculates the ratio of the pre-ejection period to the left ventricular ejection time, and performs a differential square operation on the ratio and the synchronous ventricular ejection work reference surface to extract the myocardial mechanical load component. The feature extraction module traverses all cardiac cycles within the effective exciton window and records the absolute timestamp of the specific cardiac cycle corresponding to when the myocardial mechanical load component reaches its maximum value as the first timestamp.

[0012] Preferably, the feature extraction module locates the ST segment interval of each cardiac cycle of the single-lead ECG signal within the effective exciton window, calculates the absolute amplitude integral of the voltage deviation of the real signal waveform from the ST segment isoelectric reference voltage, and multiplies the absolute amplitude integral by the rate of decrease of the synchronous blood oxygen saturation signal to extract the ECG ischemic component. The feature extraction module finds the cardiac cycle corresponding to the maximum extreme value of the ischemic component of the electrocardiogram, and records the absolute timestamp of the cardiac cycle in which the maximum extreme value occurs as the second timestamp; The feature extraction module calculates the absolute difference between the first timestamp and the second timestamp on the time axis to generate the force-electric delay decoupling time difference.

[0013] Preferably, the evaluation and judgment module extracts the maximum extreme value of the myocardial mechanical load component and the maximum extreme value of the electrocardiographic ischemia component within the effective exciton window, respectively. The assessment and judgment module uses static reference values ​​of the patient in a conscious, lying position to convert the maximum extreme value of the myocardial mechanical load component into a normalized mechanical extreme value through baseline division, and converts the maximum extreme value of the electrocardiographic ischemic component into a normalized electrical extreme value.

[0014] Preferably, the evaluation and judgment module sums the square of the normalized mechanical extreme value and the square of the normalized electrical extreme value and then performs a square root operation to obtain the comprehensive absolute overload. The evaluation and judgment module multiplies the preset time penalty weight coefficient by the force-electric delay decoupling time difference and uses the result as the exponent of the natural exponential function to obtain the penalty coefficient. The assessment and judgment module calculates the myocardial mechanical ischemic stress index by multiplying the comprehensive absolute overload by the penalty coefficient.

[0015] Preferably, the evaluation and judgment module has a built-in safety judgment threshold and a decoupling critical threshold; When the myocardial mechanical ischemic stress index is determined to be less than the safety threshold, it is determined to be a low-risk compensatory phenotype. When the myocardial mechanical ischemic stress index is greater than or equal to the safety judgment threshold and the electromechanical delay decoupling time difference is greater than the decoupling critical threshold, it is judged as a high-risk phenotype of electromechanical decompensation. When the myocardial mechanical ischemic stress index is greater than or equal to the safety judgment threshold and the mechanical-electrical delay decoupling time difference is less than or equal to the decoupling critical threshold, if the ratio of the normalized mechanical extreme value to the normalized electrical extreme value is greater than or equal to a preset ratio, it is determined to be a mechanical overload dominant phenotype; if the ratio of the normalized electrical extreme value to the normalized mechanical extreme value is greater than or equal to the preset ratio, it is determined to be an ischemic vulnerability dominant phenotype. If the ratio of the normalized mechanical extreme value to the normalized electrical extreme value is within the range of the reciprocal of the preset ratio and the preset ratio, it is determined to be a phenotype of mixed mechanical and electrical impairment.

[0016] This invention also provides a method for assessing myocardial load in OSA based on multimodal physiological signal fusion, comprising the following steps: Acquire blood oxygen saturation signals, single-lead electrocardiogram signals, and chest surface electrocardiogram signals; Based on the blood oxygen saturation signal and the single-lead electrocardiogram signal, obstructive sleep apnea events are determined and candidate time windows are defined. The trunk muscle contraction and vibration components of the chest surface electrocardiogram signal are extracted to determine the micro-arousal trigger point. The candidate time window is truncated to obtain the effective exciton window. Within the effective exciton window, the myocardial mechanical load component and the first timestamp of the chest surface electrocardiogram signal are extracted, and the electrocardiographic ischemic component and the second timestamp of the single-lead electrocardiogram signal are extracted. The electromechanical delay decoupling time difference is calculated based on the first timestamp and the second timestamp. The myocardial mechanical ischemia stress index is calculated based on the myocardial mechanical load component, the electrocardiographic ischemia component, and the time difference between the electromechanical delay and decoupling. Based on the myocardial mechanical ischemia stress index, the myocardial load assessment result of obstructive sleep apnea with myocardial vulnerability phenotype classification is output.

[0017] This invention provides a system and method for assessing myocardial load in obstructive sleep apnea (OSA) based on multimodal physiological signal fusion. It offers the following advantages: 1. This invention, by jointly acquiring blood oxygen saturation, single-lead electrocardiogram, and chest surface electrocardiogram signals, identifies OSA abnormal events and defines candidate time windows based on data fusion of multimodal physiological signals. It extracts the contraction and vibration components of trunk muscle groups to determine micro-arousal trigger points and extracts effective exciton windows, eliminating irrelevant motion interference during sleep monitoring and preserving the body's true physiological stress range in the face of hypoxia and breath-holding. This not only achieves accurate capture of OSA events but also provides an accurate data foundation for subsequent objective assessment of OSA myocardial load.

[0018] 2. This invention simultaneously extracts the myocardial mechanical load component reflecting the physical work pressure of the heart and the electrocardiographic ischemia component reflecting electrophysiological damage within the effective exciton window, and calculates the electro-electrical delay decoupling time difference using the extreme timestamps of the two. This feature processing process based on multimodal physiological signal fusion fills the gap in traditional OSA assessment, which only focuses on respiratory rate and ignores hidden cardiac damage. It deeply quantifies the absolute difference between mechanical and electrical characteristics on the time axis under hypoxic conditions, avoiding the deficiency that a single-dimensional parameter cannot reflect all pathological states, and improves the comprehensiveness and pathological depth of the overall OSA myocardial load assessment.

[0019] 3. This invention combines the myocardial mechanical load component, the electrocardiographic ischemic component, and the decoupling time difference between the mechanical and electrical delays to calculate a myocardial mechanical ischemic stress index that comprehensively reflects oxygen consumption imbalance and work overload. Based on a multimodal physiological signal fusion-based judgment logic, the stress index outputs results with myocardial vulnerability phenotype classifications, directly distinguishing between low-risk compensated, high-risk mechanical and electrical decompensation, and mechanical overload-dominated types. This successfully reveals substantial myocardial damage while diagnosing OSA, making the final OSA myocardial load assessment results more specific and detailed, and providing a direct basis for developing individualized cardiovascular intervention plans in clinical practice. Attached Figure Description

[0020] Figure 1 This is a system architecture diagram of the OSA myocardial load assessment system based on multimodal physiological signal fusion according to the present invention; Figure 2 This is a flowchart of the OSA myocardial load assessment method based on multimodal physiological signal fusion according to the present invention; Figure 3 This is a flowchart of the classification and determination process for myocardial vulnerability phenotypes according to the present invention; Figure 4 This is a schematic diagram of the cross-modal characteristic evolution and electromechanical delay decoupling of the present invention.

[0021] Among them, 100 is the signal acquisition module; 200 is the window positioning and cropping module; 300 is the feature extraction module; and 400 is the evaluation and judgment module. Detailed Implementation

[0022] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see the appendix Figure 1 This invention provides an OSA myocardial load assessment system based on multimodal physiological signal fusion, comprising: a signal acquisition module 100, a window positioning and cropping module 200, a feature extraction module 300, and an assessment and judgment module 400.

[0024] The signal acquisition module 100 is used to acquire blood oxygen saturation signals, single-lead electrocardiogram signals, and chest surface electrocardiogram signals. The window positioning and clipping module 200 is used to determine obstructive sleep apnea events based on blood oxygen saturation signals and single-lead electrocardiogram signals and to define candidate time windows. It extracts the trunk muscle contraction and vibration components of the chest surface electrocardiogram signal to determine the micro-arousal trigger point and truncates the candidate time window to obtain the effective exciton window. The feature extraction module 300 is used to extract the myocardial mechanical load component and the first timestamp of the chest surface electrocardiogram signal within the effective exciton window, extract the myocardial ischemia component and the second timestamp of the single-lead electrocardiogram signal, and calculate the electromechanical delay decoupling time difference based on the first timestamp and the second timestamp. The assessment and judgment module 400 is used to calculate the myocardial mechanical ischemia stress index based on the myocardial mechanical load component, the electrocardiographic ischemia component and the time difference of decoupling between the electromechanical delay and the myocardial mechanical ischemia stress index, and output the myocardial load assessment result of obstructive sleep apnea with myocardial vulnerability phenotype classification based on the myocardial mechanical ischemia stress index.

[0025] See attached document Figure 2 This invention provides a method for assessing myocardial load in OSA based on multimodal physiological signal fusion, comprising the following steps: S10 acquires blood oxygen saturation signals, single-lead electrocardiogram signals, and chest surface echocardiogram signals; S20, based on blood oxygen saturation signal and single-lead electrocardiogram signal, determine obstructive sleep apnea event and define candidate time windows, extract the trunk muscle contraction vibration component of chest surface electrocardiogram signal to determine micro-arousal trigger point, and truncate the candidate time window to obtain effective exciton window; S30, extract the myocardial mechanical load component and the first timestamp of the chest surface electrocardiogram signal within the effective exciton window, extract the electrocardiographic ischemic component and the second timestamp of the single-lead electrocardiogram signal, and calculate the electromechanical delay decoupling time difference based on the first timestamp and the second timestamp; S40 calculates the myocardial mechanical ischemia stress index based on the myocardial mechanical load component, the electrocardiographic ischemic component, and the time difference between the decoupling of the electromechanical delay. Based on the myocardial mechanical ischemia stress index, it outputs the myocardial load assessment results of obstructive sleep apnea with myocardial vulnerability phenotype classification.

[0026] The OSA myocardial load assessment method and system based on multimodal physiological signal fusion of the present invention belong to the same inventive concept. Each logical module in the assessment system is configured to execute the corresponding steps in the assessment method. Specifically, the signal acquisition module 100 in the assessment system is used to execute step S10 to acquire multimodal physiological signals; the window positioning and pruning module 200 is used to execute step S20 to determine events and truncate to acquire the effective exciton window; the feature extraction module 300 is used to execute step S30 to extract the myocardial mechanical load component and the electrocardiographic ischemia component and calculate the electromechanical delay decoupling time difference; and the assessment judgment module 400 is used to execute step S40 to calculate the stress index and output the phenotypic assessment result. The assessment system relies on the sequential interactive operation of each module to achieve a computational closed loop from physiological data acquisition and feature decoupling to obstructive sleep apnea phenotype recognition.

[0027] To further clarify the implementation of each technical aspect of the present invention, the following will provide a detailed description of the implementation of each functional module involved above and its internal processing flow.

[0028] See attached document Figure 1 In this embodiment, the signal acquisition module 100 in the evaluation system is used to acquire blood oxygen saturation signals, single-lead electrocardiogram (ECG) signals, and chest surface electrocardiogram (CSE) signals. Specifically, the blood oxygen saturation signal reflects the dynamic oxygen-carrying status of peripheral blood during airway obstruction; the single-lead ECG signal characterizes the electrophysiological conduction activity during heartbeat; and the chest surface CSE signal captures the high-frequency, weak physical vibrations of the chest wall caused by ventricular ejection. As a preferred embodiment, the signal acquisition module 100 uses a transmissive photoplethysmometer (which can be worn on the patient's fingertip) to acquire the blood oxygen saturation signal, uses a surface dry electrode (which can be attached to the patient's left anterior chest area) to acquire the single-lead ECG signal, and uses a broadband piezoelectric accelerometer (which can be fixed to the lower part of the patient's sternum) to acquire the chest surface CSE signal. For the selection of the light-emitting and receiving circuits, the impedance matching amplification circuit of the body surface dry electrode, and the wideband piezoelectric accelerometer inside the transmissive photoplethysmography sensor, those skilled in the art can configure them according to conventional medical device design manuals. Their hardware principles are all well-known existing technologies in this field and will not be elaborated here.

[0029] In order to establish the spatiotemporal basis for joint computation of cross-modal data and eliminate common-mode interference introduced by changes in sleep position, the signal acquisition module 100 performs the following specific processing steps.

[0030] S101 performs timestamp alignment processing on the multimodal signals. The microprocessor inside the signal acquisition module 100 is configured to execute a series of pre-processing logics. The microprocessor front end integrates a high-precision analog-to-digital converter, and the signal acquisition module 100 is equipped with an independent hardware real-time clock unit. While the analog-to-digital converter digitizes the three analog signals, the hardware real-time clock unit uses a hardware interrupt mechanism to assign a globally unique timestamp to each discrete sampling point. Through this timestamp marking, the blood oxygen saturation signal, single-lead ECG signal, and chest surface echocardiogram signal at different sampling rates achieve strict phase alignment on a unified time axis.

[0031] S102, separates the ultra-low frequency signal from the chest surface electrocardiogram signal. Changes in body position and respiratory fluctuations during sleep are extremely low-frequency macroscopic physical movements, with frequencies significantly lower than the high-frequency micro-vibrations generated by the opening and closing of heart valves. Based on this physiological characteristic, the signal acquisition module 100 receives the time-aligned chest surface electrocardiogram signal. A low-pass filter is applied to separate the ultra-low frequency signal. Here, the cutoff frequency of the low-pass filter is set to 0.1Hz. This setting is based on the fact that the respiratory rate of a normal adult during sleep is typically between 0.2Hz and 0.3Hz. Therefore, extracting extremely low frequency components below 0.1Hz can filter out high-frequency heartbeats and normal breathing interference, thereby accurately separating the static gravity projection baseline caused by changes in body position. The calculation process is described as follows: ; In the formula, This represents the impulse response of a low-pass filter with a cutoff frequency of 0.1 Hz. Represents the convolution operator; Represents a time variable.

[0032] S103, constructing a body position vector reflecting the absolute tilt posture. From the perspective of three-dimensional physical space, the static projection components of gravitational acceleration on the three orthogonal sensing axes of the broadband piezoelectric accelerometer directly reflect the absolute tilt posture of the human body on the sleeping bed. Therefore, the signal acquisition module 100 extracts ultra-low frequency signals. Orthogonal projection vectors in a three-dimensional coordinate system are used to construct position vectors. Its mathematical expression is: ; In the formula, , , They represent ultra-low frequency signals respectively Steady-state acceleration components on the X, Y, and Z axes; This represents the matrix transpose operation.

[0033] S104 extracts surrogate parameters of thoracic impedance characterizing thoracic volume deformation. Human respiratory activity causes rhythmic undulation of the thoracic cavity, resulting in periodic micro-variations in the contact stress between the sensor and the chest wall, which manifests as an ultra-low frequency signal. Low-frequency modulation in amplitude. To quantify this physical deformation of the thoracic cavity volume caused by respiration, the signal acquisition module 100 uses Hilbert transform to calculate the envelope amplitude of the ultra-low frequency signal fluctuation during the respiratory cycle, thereby constructing a pleural impedance surrogate parameter. : ; In the formula, This represents the Hilbert transform operator.

[0034] S105, Dynamic Anchoring of Reference Voltage and Reference Surface. Changes in sleep position and thoracic volume not only alter the anatomical projection angle of the heart relative to the surface electrodes but also cause static drift in the impedance of the cardiac conduction medium, leading to irregular baseline wandering in single-lead ECG signals and non-pathological variations in the amplitude of chest surface electrocardiogram waveforms. To address this issue, the signal acquisition module 100 has a pre-configured dynamic calibration matrix. The dynamic calibration matrix is ​​a pre-configured set of linear regression model parameters used to establish a quantitative cancellation relationship between physical deformation of the chest and the physiological signal baseline. This dynamic calibration matrix is ​​obtained through offline clinical data fitting. Specifically, during the research and development phase, synchronous signals from multiple subjects were pre-collected in preset fixed lying positions (e.g., supine, left lateral, right lateral) and at different respiratory depths. Using the body position deflection angle and chest expansion volume as inputs, and the signal baseline offset as the target label, a least-squares method was used for multivariate linear fitting to solve for stable mapping weights.

[0035] In actual operation, the signal acquisition module 100 uses the body position vector acquired in real time. With pleural impedance surrogate parameters As the input variable, the ST segment isoelectric reference voltage of the ECG signal in the current physical attitude list leads is calculated in real time via a linear mapping operation performed using a dynamic calibration matrix. and the ventricular ejection work reference plane of the chest surface electrocardiogram signal The specific mapping relationship is as follows: ; In the formula, This represents a 2×4 multidimensional mapping weight matrix; This represents a bias constant vector with dimension 2×1. In one embodiment, when the system is applied to an adult group with standard body types, the preset bias constant vector... The reference voltage and reference plane were initialized to the initial calibration values ​​when the subject was in a standard supine position and at the end of expiration. This dynamic calibration matrix mechanism based on underlying physical parameters eliminates the common-mode interference caused by body position changes on electrophysiological and mechanical vibration signals at the source.

[0036] See attached document Figure 1 In this embodiment, the window positioning and cropping module 200 performs the following processing steps.

[0037] S201, based on multimodal physiological parameters, jointly determines obstructive sleep apnea events. When a patient experiences airway obstruction, the decrease in alveolar ventilation leads to a corresponding decrease in arterial blood oxygen concentration. Simultaneously, increased sympathetic tone in the autonomic nervous system causes an abnormal enhancement of the low-frequency component in heart rate variability indicators. Combining the above physiological mechanisms, the window positioning and trimming module 200 calculates the first-order time derivative of the received blood oxygen saturation signal, identifies time intervals where the derivative value is continuously less than zero, and thus extracts the phase of continuous decline in blood oxygen. Simultaneously, the window positioning and trimming module 200 uses a peak detection algorithm to extract the R-wave peak point of the single-lead ECG signal and generates an RR interval sequence. For the detection of the R-wave peak of the ECG signal, those skilled in the art can use the standard Pan-Tompkins algorithm or similar threshold difference methods, which are well-known techniques in the field and will not be elaborated here.

[0038] After acquiring the RR interval sequence, the window positioning and cropping module 200 calculates the power spectral density integral of the sequence in the low-frequency band using short-time Fourier transform to obtain the low-frequency heart rate variability power. In actual operation, the low-frequency band is typically set to 0.04Hz to 0.15Hz to correspond to the characteristic frequency band of sympathetic nerve activity. As a preferred method, the baseline threshold for determining when low-frequency heart rate variability power exceeds the limit is calculated by multiplying the average low-frequency heart rate variability power during the patient's quiet breathing state ten minutes before falling asleep by a set factor, which is set to a value between 1.5 and 2.0. When a continuous decline in blood oxygenation is detected and the corresponding low-frequency heart rate variability power... When the baseline threshold is exceeded, the assessment system confirms an obstructive sleep apnea event and dynamically accumulates the frequency of apnea throughout the night, thus constructing a basic clinical assessment framework for obstructive sleep apnea. At this time, the window positioning and clipping module 200 records the start time of the decrease in blood oxygen saturation and the end time of the recovery to a stable baseline, defining this start-end time interval as the [missing information]. Candidate time window for the secondary event In this embodiment, The sequence number representing an obstructive sleep apnea event. Defined as the first The time stamp in which the blood oxygen saturation signal began to decline continuously during this event. Defined as the time stamp when the blood oxygen saturation signal recovers and remains at a normal level.

[0039] S202, separating the contractile vibration components of the trunk muscle groups. At the end of an obstructive sleep apnea event, extreme apnea and hypoxia often trigger micro-arousals in the cerebral cortex, accompanied by unconscious struggling and violent breathing movements. This macroscopic trunk movement produces mechanical vibration artifacts, thus masking the weak ejection pulsation characteristics of the heart. To quantify this interference, the window positioning and clipping module uses 200 pairs of candidate time windows. Internal chest surface electrocardiogram signal A bandpass digital filter is applied. The passband frequency range of this filter is set to 10Hz to 20Hz, which coincides with the dominant frequency component of the myophonogram generated during the tetanic contraction of human skeletal muscles. This allows the window positioning and clipping module 200 to isolate the pure trunk muscle contraction vibration component from the complex mixed chest wall vibration. .

[0040] S203 generates a motion energy envelope and locates micro-arousal trigger points. To quantitatively assess the transient activity intensity of trunk muscles, the window positioning and clipping module 200 uses a sliding time window to calculate the contractile vibration components of trunk muscle groups. The short-time root mean square value is used to generate the kinetic energy envelope characterizing the intensity of muscle work. Its specific mathematical expression is: ; In the formula, This represents the physical time length of the sliding time window, typically set to 0.5 seconds to balance envelope smoothness with the temporal resolution of abrupt responses. It is the dummy variable for integration during the time integration process; The infinitesimal element represents the integral dummy variable.

[0041] In acquiring the energy envelope of motion Subsequently, the window positioning and cropping module 200 acquires the dynamic threshold for signal quality. Used to determine the onset of motion artifacts. As a preferred parameter setting method, this signal quality dynamic threshold... A fixed multiple of the average value of the motion energy envelope obtained by continuously measuring the patient's energy for one minute while the patient is awake, calm, and lying down is taken. This fixed multiple is typically set to 3 to 5. The window positioning and clipping module 200 is used within the candidate time window. Internally, from the end time Towards the starting point Perform a reverse timeline scan. When the motion energy envelope is detected... The amplitude exceeds the dynamic threshold of the signal quality. At the first moment, the window positioning and clipping module 200 identifies and marks it as a micro-awakening trigger point. Micro-awakening trigger point Defined as candidate time window The time boundary at which significant bodily movement disturbances first appeared within the timeframe.

[0042] S204, adaptive truncation of the time window and acquisition of the effective exciton window. Micro-awakening trigger point. Physically, this signifies the patient's transition from a state of pure, extreme negative pressure in the thoracic cavity to a mixed state accompanied by physical struggle. To ensure that subsequent cross-modal feature extraction targets only the compensatory response of the myocardium under ischemic stress, the window positioning and clipping module 200 uses micro-awakening trigger points. Using time boundaries, candidate time windows A truncation operation is performed. After discarding the tail data containing motion interference, the remaining clean time segment at the front is established as the effective exciton window. Its interval is strictly defined as: ; As a mechanism to ensure the completeness of algorithmic logic, if candidate time windows are detected during the reverse timeline scanning process... The inner kinetic energy envelope It never exceeded the dynamic threshold for signal quality. This indicates that the blocking event was not accompanied by significant micro-awakening trunk movement. In this case, the window positioning and clipping module 200 does not perform a truncation operation, but directly displays the complete candidate time window. Defined as an effective exciton window Through the coordinated operation of the above steps, the assessment system effectively eliminates the interference of motion artifacts, removes irrelevant movement interference during sleep monitoring, and preserves the body's true physiological stress range when facing hypoxia and breath-holding.

[0043] Accurately identifying and capturing the effective exciton window of airway obstruction not only provides a direct assessment basis for determining the independent existence of obstructive sleep apnea events, but also lays a crucial and accurate data foundation for subsequent in-depth quantification of the pathogenic severity of individual apnea events and objective assessment of OSA myocardial load. Airway obstruction causes a sharp increase in intrathoracic negative pressure, leading to a rapid rise in ventricular afterload, which in turn induces myocardial physical compensation. Simultaneously, decreased blood oxygen saturation leads to insufficient myocardial oxygen supply, causing ischemic changes during the cardiac repolarization period. It should be noted that the myocardial mechanical load component is used to quantify the degree of overload of the heart in overcoming the increased afterload under extreme apnea; the cardiac ischemia component is used to characterize the severity of abnormal electrical repolarization of myocardial cells under hypoxic conditions; and the mechano-electrical delay decoupling time difference is used to measure the degree of abnormal time misalignment between cardiac electrophysiological conduction activity and mechanical contractile work under extreme stress. (See Appendix) Figure 1In this embodiment, the feature extraction module 300 is configured to perform the following specific processing steps to quantify the independent stress states of the above-mentioned mechanical and electrophysiological processes and their degree of asynchrony.

[0044] S301 extracts the myocardial mechanical load component and first timestamp of the chest surface seismogram signal within the effective exciton window. Feature extraction module 300 pairs the effective exciton window. Envelope detection and peak finding are performed on the chest surface seismogram signal to locate the aortic valve opening feature points and the aortic valve closing feature points within each cardiac cycle. For the envelope detection and location of aortic valve feature points in the seismogram waveform, those skilled in the art can use the Shannon energy envelope method or wavelet transform multi-scale peak finding algorithm, which are well known techniques in the field and will not be elaborated here.

[0045] After locating the feature points, the feature extraction module 300 calculates the time interval between the peak of the adjacent ECG R wave and the corresponding aortic valve opening feature point to obtain the transient pre-ejection period. ; Calculate the time interval between the aortic valve opening feature point and the aortic valve closing feature point to obtain the left ventricular ejection time. ,in Indicates the effective exciton window The first One cardiac cycle. Weakened myocardial contractility or increased afterload can lead to a prolonged pre-ejection period and a shortened ejection time. The feature extraction module 300 calculates the ratio of these two ratios and compares it with the ventricular ejection work reference plane dynamically anchored by the signal acquisition module 100. Perform difference square operation to extract the myocardial mechanical load component. The calculation formula is as follows: ; In the formula, For the first The absolute timestamps corresponding to each cardiac cycle. This calculation process eliminates baseline drift caused by body position, retaining only the true pathological mechanical overload caused by airway obstruction. The feature extraction module iterates through the effective exciton window 300 times. Identify the myocardial mechanical load components throughout all cardiac cycles. The specific cardiac cycle corresponding to the maximum extreme value is recorded as the first timestamp. First time stamp This marks the moment when myocardial mechanical overload reaches its limit during airway obstruction.

[0046] S302 extracts the ischemic component of the single-lead ECG signal and the second timestamp within the effective exciton window. Abnormal action potential repolarization of cardiomyocytes under hypoxic conditions manifests as ST segment depression or elevation on a macroscopic ECG. Feature extraction module 300 extracts the signal within the effective exciton window. For each cardiac cycle, locate the ST segment interval of the single-lead ECG signal and calculate the deviation of the actual signal waveform voltage within this interval from the ST segment isoelectric reference voltage. The absolute amplitude integral is calculated. To incorporate the weighted influence of blood oxygenation changes on the degree of ischemia, the feature extraction module 300 combines this integral value with the rate of decrease of the synchronized blood oxygen saturation signal to calculate the ischemic component of the cardiac electrocardiogram. As a preferred method, its mathematical expression is: ; In the formula, This represents a single-lead ECG signal within the effective exciton window; and They represent the first The start and end times of the ST segment in each cardiac cycle; For time integrals and differential elements; Indicates a synchronized blood oxygen saturation signal; Indicates the first Find the maximum absolute value of the rate of descent within the time range of a cardiac cycle.

[0047] After calculating the ischemic component of the cardiac electrical activity in each cardiac cycle, the feature extraction module 300 extracts the signal within the effective exciton window. Internal search for ischemic components in the cardiac circuit The cardiac cycle corresponding to the maximum extreme value is recorded as the second timestamp, and the absolute timestamp of the cardiac cycle in which the maximum offset occurs is recorded. The corresponding second timestamp This marks the extreme moment when hypoxia damage is most severe at the electrophysiological level.

[0048] S303 calculates the decoupling time difference between electromechanical delay based on the first and second timestamps. In a healthy state, cardiac conduction and myocardial mechanical contraction maintain a high degree of electromechanical coupling synchronization. When a patient experiences severe obstructive sleep apnea, the interaction of intense intrathoracic negative pressure and acute hypoxia causes a significant time misalignment between the occurrence of cardiac electrical abnormalities and the occurrence of mechanical decompensation. The degree of this pathological time misalignment directly reflects the microscopic vulnerability at the myocardial cell level. Feature extraction module 300 calculates the first timestamp. With the second timestamp The absolute difference on the time axis generates the force-electric delay decoupling time difference. The calculation logic is as follows: ; Force-electric delay decoupling time difference As an independent cross-modal temporal dimension feature, it quantifies the degree of physical delay in decompensated asynchrony of the myocardium under extreme physical breath-holding stress. This feature processing based on multimodal physiological signal fusion simultaneously extracts mechanical features reflecting the physical work pressure of the heart and electrical features reflecting electrophysiological damage, and deeply quantifies the absolute difference between the two on the time axis. This fills the gap in traditional OSA assessment, which only focuses on respiratory rate and ignores hidden cardiac damage, and greatly improves the comprehensiveness and pathological depth of overall OSA myocardial load assessment.

[0049] In traditional clinical practice, the assessment of obstructive sleep apnea often relies heavily on the apnea-hypopnea index, focusing only on the frequency of airway obstruction while ignoring the heterogeneous damage to cardiovascular function caused by equivalent frequencies of obstruction in different patients. This invention delves into the physiological mechanisms behind obstructive events, suggesting that airway obstruction, as a primary respiratory abnormality, poses its most threatening secondary threat by causing acute hypoxia and mechanical overload of myocardial tissue. To achieve a more prognostic and in-depth assessment of myocardial load in OSA, refer to the appendix... Figure 1 and Figure 3 In this embodiment, the evaluation and judgment module 400 performs the following specific processing steps to transform isolated cross-modal features into comprehensive pathological evaluation indicators.

[0050] S401 performs individualized baseline normalization on cross-modal features. Due to individual differences in cardiac geometry and baseline electrophysiological conduction among patients, the absolute values ​​output by the feature extraction module 300 need to be standardized to eliminate dimensional discrepancies with individual baselines. The evaluation and judgment module 400 extracts the maximum extreme values ​​of the myocardial mechanical load components within the effective exciton window. and the maximum extreme value of the ischemic component of the electrocardiogram. In this embodiment, the maximum extreme value refers to the highest global peak value in the sequence of characteristic values ​​corresponding to each cardiac cycle within the entire effective exciton window time axis, which objectively represents the most severe moment of physical or electrical overload caused by a single airway obstruction event.

[0051] After obtaining the maximum extreme value, the evaluation and judgment module 400 introduces the static reference value when the patient is in a conscious and calm state, and calculates the normalized mechanical extreme value. With normalized electrical extreme values : ; ; In the formula, This represents the baseline value of myocardial mechanical load across multiple consecutive cardiac cycles while the patient is awake and lying flat before falling asleep. This represents the background noise tolerance value of ST segment fluctuations in a patient's awake, supine state before falling asleep. Through the above baseline division operation, the assessment and judgment module 400 maps the absolute physical quantity into a relative multiple parameter characterizing the severity of stress.

[0052] S402 calculates the myocardial mechanical ischemic stress index. In the pathological process of airway obstruction, mechanical overload and electrical ischemia are often not simply linearly superimposed, but rather exhibit a mutually deteriorating orthogonal characteristic. When the heart simultaneously faces increased ejection resistance and oxygen supply cutoff, myocardial cells will accelerate exhaustion. If this is accompanied by a time misalignment between electrophysiological conduction and physical contraction, it indicates that microscopic decompensation has occurred within the myocardium. Based on the above physiological logic, the evaluation and judgment module 400 constructs a multi-dimensional fusion calculation model. As a preferred approach, the evaluation and judgment module 400 is based on normalized mechanical extreme values. Normalized electrical extreme values and the time difference of force-electric delay decoupling Calculate the myocardial mechanical ischemic stress index The specific mathematical expression is: ; In the formula, It is a natural exponential function; This represents the preset time penalty weighting coefficient. In practical applications, those skilled in the art can set this using large-sample clinical statistical data. The value of is typically between 1.5 and 3.0, used to mitigate the deteriorating effect of nonlinear amplification time decoupling on the excitation exponent. The first part of the formula quantifies the combined absolute overload of the mechanical and electrical dimensions using the square of the Euclidean spatial distance formed by the sum of the squares of the two components. The second part uses exponential operation logic to apply a weighted penalty to the asynchronous decompensation phenomenon.

[0053] S403 performs myocardial vulnerability phenotype classification based on the rule engine. The assessment module 400 has a built-in phenotype mapping rule matrix for analyzing the myocardial mechanical ischemic stress index. The composition ratio of its internal components. As a specific implementation, the phenotypic mapping rule matrix is ​​a decision mapping table containing multi-dimensional conditional branches. The input parameters of this mapping table include the myocardial mechanical ischemic stress index. Absolute amplitude, normalized mechanical extreme value With normalized electrical extreme values The ratio relationship, and the time difference of force-electric delay decoupling. The span value; its output corresponds to the specific clinical phenotype label.

[0054] To execute the decision mapping table, the evaluation and decision module 400 has a pre-set safety decision threshold. and decoupling critical threshold Safety judgment threshold The upper limit of the index for spontaneous micro-arousals during normal sleep in healthy adults is typically set at 2.0 to 2.5; decoupling critical threshold. The maximum electrical conduction delay time that a healthy myocardium can tolerate is typically set to 150 to 200 milliseconds.

[0055] When assessing the myocardial ischemic stress index Less than the safety threshold When the assessment and judgment module 400 determines that the obstructive event is within the heart's tolerable range, it marks it as a low-risk compensatory phenotype, that is, a relatively safe state in which the heart can maintain normal blood pumping function through its own regulation.

[0056] When assessing the myocardial ischemic stress index Greater than or equal to the safety judgment threshold At that time, the evaluation and judgment module 400 further checks the force-electric synchronization status. It only needs to determine the force-electric delay decoupling time difference. Greater than the decoupling critical threshold The assessment and judgment module 400 will forcibly trigger a high-risk alarm and classify it as a high-risk phenotype of electrokinetic decompensation, indicating that the patient has a great risk of nocturnal cardiogenic adverse events during sleep apnea, that is, a critical state in which cardiac electrical signal conduction and muscle contraction are severely disconnected and the patient loses the ability to self-regulate.

[0057] If the decoupling crisis threshold is not exceeded, the evaluation and judgment module 400 compares... and The proportion of contribution. If Divide by A ratio greater than or equal to 1.5 indicates that the patient's airway obstruction was primarily due to resistance-induced obstruction of blood ejection caused by negative thoracic pressure. The assessment module 400 categorizes this as a mechanical overload-dominant phenotype, meaning the heart is under excessive workload to overcome the enormous resistance caused by airway collapse. Divide by A ratio greater than or equal to 1.5 indicates that the patient is sensitive to ischemia and prone to sudden cardiac repolarization abnormalities during airway obstruction. The assessment module 400 classifies this as an ischemic vulnerability-dominant phenotype, meaning that myocardial cells are highly susceptible to cardiac conduction disturbances due to hypoxia. As a guarantee of the completeness of the classification logic, if the ratio of the above two values ​​is within the range of 1 / 1.5 to 1.5, it indicates that the myocardium is subjected to the dual pressure of high workload and electrical hypoxia to an equal degree. The assessment module 400 classifies this as a mixed mechanical and electrical involvement phenotype, meaning that the heart is simultaneously facing the combined impact of resistance overload and ischemic hypoxia.

[0058] S404 outputs structured obstructive sleep apnea myocardial load assessment results. As the final data output of the entire assessment system, the assessment module 400 not only outputs routine respiratory event statistics, but also performs multi-dimensional clinical qualitative analysis of all airway obstruction events that occurred during the overnight monitoring period. The assessment module 400 summarizes the above calculation and judgment logic to generate a structured report, outputting obstructive sleep apnea myocardial load assessment results with myocardial vulnerability phenotype classification. This assessment result includes the frequency characteristics of the patient's sleep apnea and clearly identifies whether each apnea event induced secondary damage to myocardial tissue and the specific vulnerability phenotype classification. This assessment scheme, which deconstructs the physical obstruction of the airway and myocardial and vascular stress damage from the surface to the core, successfully reveals substantial myocardial damage while diagnosing OSA, making the final assessment results more specific and detailed. This provides clinicians with intuitive, accurate, and direct evidence and data support for developing individualized airway intervention or cardiovascular protection strategies.

[0059] Specific application examples: The following provides a specific application example of an OSA myocardial load assessment system and method based on multimodal physiological signal fusion. This example demonstrates the calculation process of each logical module in the assessment system configured to perform the corresponding steps in the assessment method, highlighting the practical application of multimodal physiological signal fusion in OSA myocardial load assessment.

[0060] Taking the nighttime sleep monitoring process of a 55-year-old male subject as an example, the operation flow of the OSA myocardial load assessment system based on multimodal physiological signal fusion is as follows: Signal Acquisition and Event Triggering: The signal acquisition module 100 acquires the subject's blood oxygen saturation signal, single-lead electrocardiogram signal, and chest surface echocardiogram signal, and performs timestamp alignment processing on the multimodal signals.

[0061] At 02:30, the window positioning and clipping module 200 determines obstructive sleep apnea events and defines candidate time windows based on blood oxygen saturation signals and single-lead electrocardiogram signals. Specifically, the window positioning and clipping module 200 detected a continuous drop in blood oxygen saturation signal from 96% to 84%, and simultaneously, the extracted low-frequency heart rate variability power exceeded the baseline threshold, determining that an obstructive sleep apnea event had occurred. It recorded the start and recovery times of the blood oxygen drop and defined candidate time windows. .

[0062] Subsequently, the window positioning and cropping module 200 extracts the trunk muscle contraction vibration components of the chest surface electrocardiogram signal to determine the micro-arousal trigger point. The system generates a motion energy envelope, and the scan reveals that the motion energy envelope first crosses the dynamic threshold of signal quality at 02:30:45. This moment is identified as the micro-awakening trigger point. The window positioning and cropping module 200 uses this micro-awakening trigger point. Using time boundaries, candidate time windows Truncating to obtain the effective exciton window .

[0063] Cross-modal feature extraction: Feature extraction module 300 in effective exciton window Myocardial mechanical load component extracted from chest surface electrocardiogram signal and first timestamp In this embodiment, the myocardial mechanical load component was identified. The absolute timestamp of the specific cardiac cycle corresponding to the point where the global maximum value is reached is recorded as the first timestamp. The specific values ​​are 02:30:40.100.

[0064] Feature extraction module 300 in effective exciton window Internal extraction of the ischemic component of single-lead ECG signals and second timestamp In this embodiment, the ischemic component of the electrocardiogram was identified. The absolute timestamp of the cardiac cycle corresponding to the maximum value is recorded as the second timestamp. The specific values ​​are 02:30:40.220.

[0065] Based on the above results, the feature extraction module 300 uses the first timestamp. With the second timestamp Calculate the time difference between force and electrical delay decoupling The calculation logic is as follows: =∣02:30:40.100-02:30:40.220∣=120 milliseconds (converted to standard unit 0.12 seconds when substituted into subsequent exponent calculations).

[0066] Assessment and Phenotypic Classification: The evaluation and judgment module 400 performs individualized baseline normalization on the cross-modal features. In this embodiment, the normalized mechanical extrema are... It was calculated to be 2.8, the normalized electrical extreme value. It was calculated as 1.1. Please refer to [link / reference]. Figure 4 , Figure 4 There are two marked feature points: the first point is located on the black solid line curve (horizontal coordinate 0.100 seconds, vertical coordinate 2.8), representing the first timestamp obtained in this embodiment. With normalized mechanical extrema The second fixed point is located on the black dashed curve (horizontal coordinate 0.220 seconds, vertical coordinate 1.1), representing the second timestamp obtained. With normalized electrical extreme values . Figure 4 The two vertical dashed lines passing through these two fixed points are used to anchor the instant of the extreme value occurrence on the time axis; while the horizontal line segment with a double-headed arrow between the two vertical dashed lines clearly indicates the force-electric delay decoupling time difference calculated above. The physical span (i.e., a time difference of 0.12 seconds on the horizontal axis). After obtaining the normalized features, the evaluation and judgment module 400 presets a time penalty weight coefficient λ of 2.0, and a safety judgment threshold. The decoupling critical threshold is set to 2.0. It takes 150 milliseconds.

[0067] Evaluation and Judgment Module 400 is based on normalized mechanical extreme values Normalized electrical extreme values Decoupling time difference with force and electrical delay Calculation of myocardial ischemic stress index : ; ; The assessment module 400 is based on the myocardial biomechanics ischemic stress index. Output the myocardial load assessment results for obstructive sleep apnea with myocardial vulnerability phenotype classification. The specific rule engine execution logic is as follows: First, determine the myocardial mechanical ischemic stress index. (3.82) Greater than the safety judgment threshold (2.0). Next, determine the time difference between force and electrical delay decoupling. (120 milliseconds) is less than the decoupling critical threshold (150 milliseconds) No high-risk alarm triggered. The final evaluation and judgment module 400 calculates the normalized mechanical extreme value. With normalized electrical extreme values The ratio of . The ratio is greater than or equal to 1.5. Based on the above results, the assessment and judgment module 400 classifies it as the mechanical overload dominant phenotype and outputs a structured assessment and judgment result of myocardial load in obstructive sleep apnea.

[0068] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An OSA myocardial load assessment system based on multimodal physiological signal fusion, characterized in that, include: The signal acquisition module is used to acquire blood oxygen saturation signals, single-lead electrocardiogram signals, and chest surface electrocardiogram signals. The window positioning and clipping module is used to determine obstructive sleep apnea events and define candidate time windows based on the blood oxygen saturation signal and the single-lead electrocardiogram signal, extract the trunk muscle contraction and vibration components of the chest surface electrocardiogram signal to determine the micro-arousal trigger point, and truncate the candidate time window to obtain the effective exciton window. The feature extraction module is used to extract the myocardial mechanical load component and the first timestamp of the chest surface electrocardiogram signal within the effective exciton window, extract the myocardial ischemic component and the second timestamp of the single-lead electrocardiogram signal, and calculate the electromechanical delay decoupling time difference based on the first timestamp and the second timestamp. The assessment and judgment module is used to calculate the myocardial mechanical ischemia stress index based on the myocardial mechanical load component, the electrocardiographic ischemia component, and the electromechanical delay decoupling time difference, and output the myocardial load assessment result of obstructive sleep apnea with myocardial vulnerability phenotype classification based on the myocardial mechanical ischemia stress index.

2. The OSA myocardial load assessment system based on multimodal physiological signal fusion according to claim 1, characterized in that, The signal acquisition module assigns globally unique timestamps to the blood oxygen saturation signal, the single-lead electrocardiogram signal, and the chest surface electrocardiogram signal and performs timestamp alignment processing. The signal acquisition module separates the ultra-low frequency signal from the chest surface electrocardiogram signal, extracts the orthogonal projection vector of the ultra-low frequency signal in the three-dimensional spatial coordinate system to construct the body position vector, and calculates the envelope amplitude of the ultra-low frequency signal fluctuation during the respiratory cycle to construct the pleural impedance surrogate parameter. The signal acquisition module updates the ST segment isoelectric reference voltage of the single-lead electrocardiogram signal and the ventricular ejection work reference surface of the chest surface electrocardiogram signal in real time, based on the preset dynamic calibration matrix and using the body position vector and the pleural impedance proxy parameter as input variables.

3. The OSA myocardial load assessment system based on multimodal physiological signal fusion according to claim 1, characterized in that, The window positioning and cropping module calculates the first time derivative of the blood oxygen saturation signal to identify the continuous decline phase of blood oxygen, extracts the RR interval sequence of the single-lead electrocardiogram signal and calculates the low-frequency heart rate variability power of the RR interval sequence in the low-frequency band. When it is determined that the blood oxygen is continuously decreasing and the corresponding low-frequency heart rate variability power exceeds the baseline threshold, the window positioning and clipping module confirms that an obstructive sleep apnea event has occurred, records the start time of the blood oxygen decrease and the end time of the blood oxygen recovery to a stable baseline, and defines the time from the start time to the end time as the candidate time window.

4. The OSA myocardial load assessment system based on multimodal physiological signal fusion according to claim 3, characterized in that, The window positioning and cropping module applies a bandpass digital filter to the chest surface electrocardiogram signal within the candidate time window to separate the trunk muscle group contraction vibration component, and uses a sliding time window to calculate the short-time root mean square value of the trunk muscle group contraction vibration component to generate a motion energy envelope. The window positioning and cropping module performs a reverse time axis scan from the end time to the start time within the candidate time window, and determines the moment when the motion energy envelope first crosses the dynamic threshold of signal quality as the micro-awakening trigger point. The window positioning and cropping module uses the start time and the micro-awakening trigger point as boundaries to obtain the effective exciton window.

5. The OSA myocardial load assessment system based on multimodal physiological signal fusion according to claim 2, characterized in that, The feature extraction module locates the aortic valve opening feature point and the aortic valve closing feature point for each cardiac cycle within the effective exciton window, calculates the ratio of the pre-ejection period to the left ventricular ejection time, and performs a differential square operation on the ratio and the synchronous ventricular ejection work reference surface to extract the myocardial mechanical load component. The feature extraction module traverses all cardiac cycles within the effective exciton window and records the absolute timestamp of the specific cardiac cycle corresponding to when the myocardial mechanical load component reaches its maximum value as the first timestamp.

6. The OSA myocardial load assessment system based on multimodal physiological signal fusion according to claim 5, characterized in that, The feature extraction module locates the ST segment interval of each cardiac cycle of the single-lead ECG signal within the effective exciton window, calculates the absolute amplitude integral of the voltage deviation of the real signal waveform from the ST segment isoelectric reference voltage, and multiplies the absolute amplitude integral by the rate of decrease of the synchronous blood oxygen saturation signal to extract the ECG ischemic component. The feature extraction module finds the cardiac cycle corresponding to the maximum extreme value of the ischemic component of the electrocardiogram, and records the absolute timestamp of the cardiac cycle in which the maximum extreme value occurs as the second timestamp; The feature extraction module calculates the absolute difference between the first timestamp and the second timestamp on the time axis to generate the force-electric delay decoupling time difference.

7. The OSA myocardial load assessment system based on multimodal physiological signal fusion according to claim 6, characterized in that, The evaluation and judgment module extracts the maximum extreme value of the myocardial mechanical load component and the maximum extreme value of the electrocardiographic ischemia component within the effective exciton window, respectively. The assessment and judgment module uses static reference values ​​of the patient in a conscious, lying position to convert the maximum extreme value of the myocardial mechanical load component into a normalized mechanical extreme value through baseline division, and converts the maximum extreme value of the electrocardiographic ischemic component into a normalized electrical extreme value.

8. The OSA myocardial load assessment system based on multimodal physiological signal fusion according to claim 7, characterized in that, The evaluation and judgment module sums the squares of the normalized mechanical extreme value and the squares of the normalized electrical extreme value, and then performs a square root operation to obtain the comprehensive absolute overload. The evaluation and judgment module multiplies the preset time penalty weight coefficient by the force-electric delay decoupling time difference and uses the result as the exponent of the natural exponential function to obtain the penalty coefficient. The assessment and judgment module calculates the myocardial mechanical ischemic stress index by multiplying the comprehensive absolute overload by the penalty coefficient.

9. The OSA myocardial load assessment system based on multimodal physiological signal fusion according to claim 8, characterized in that, The assessment and judgment module has built-in safety judgment thresholds and decoupling critical thresholds; When the myocardial mechanical ischemic stress index is determined to be less than the safety threshold, it is determined to be a low-risk compensatory phenotype. When the myocardial mechanical ischemic stress index is greater than or equal to the safety judgment threshold and the electromechanical delay decoupling time difference is greater than the decoupling critical threshold, it is judged as a high-risk phenotype of electromechanical decompensation. When the myocardial mechanical ischemic stress index is greater than or equal to the safety judgment threshold and the mechanical-electrical delay decoupling time difference is less than or equal to the decoupling critical threshold, if the ratio of the normalized mechanical extreme value to the normalized electrical extreme value is greater than or equal to a preset ratio, it is determined to be a mechanical overload dominant phenotype; if the ratio of the normalized electrical extreme value to the normalized mechanical extreme value is greater than or equal to the preset ratio, it is determined to be an ischemic vulnerability dominant phenotype. If the ratio of the normalized mechanical extreme value to the normalized electrical extreme value is within the range of the reciprocal of the preset ratio and the preset ratio, it is determined to be a phenotype of mixed mechanical and electrical impairment.

10. A method for assessing OSA myocardial load based on multimodal physiological signal fusion, applied to the system described in any one of claims 1-9, characterized in that, Includes the following steps: Acquire blood oxygen saturation signals, single-lead electrocardiogram signals, and chest surface electrocardiogram signals; Based on the blood oxygen saturation signal and the single-lead electrocardiogram signal, obstructive sleep apnea events are determined and candidate time windows are defined. The trunk muscle contraction and vibration components of the chest surface electrocardiogram signal are extracted to determine the micro-arousal trigger point. The candidate time window is truncated to obtain the effective exciton window. Within the effective exciton window, the myocardial mechanical load component and the first timestamp of the chest surface electrocardiogram signal are extracted, and the electrocardiographic ischemic component and the second timestamp of the single-lead electrocardiogram signal are extracted. The electromechanical delay decoupling time difference is calculated based on the first timestamp and the second timestamp. The myocardial mechanical ischemia stress index is calculated based on the myocardial mechanical load component, the electrocardiographic ischemia component, and the time difference between the electromechanical delay and decoupling. Based on the myocardial mechanical ischemia stress index, the myocardial load assessment result of obstructive sleep apnea with myocardial vulnerability phenotype classification is output.

Images