Sleep apnea and hypopnea intelligent monitoring and auxiliary diagnosis system and method based on flexible sensor

Abstract

The invention provides a sleep apnea and hypopnea intelligent monitoring and auxiliary diagnosis system and method based on a flexible sensor, and belongs to the technical field of sensors and respiration monitoring equipment. The system comprises a resistive flexible strain sensor, a wireless data acquisition card and a portable data processing terminal, a resistance-type flexible strain sensor is adhered to the skin surface of a main muscle group and an auxiliary muscle group related to human respiration, resistance signals related to the respiration frequency and amplitude are obtained, respiration data of a patient are collected in real time, and further, self-adaptive amplitude alarm upper and lower limits are constructed through statistical modeling of the individual resting respiration amplitude, so that the respiration alarm accuracy is improved. Therefore, the individual adaptability and robustness of the respiration detection system are improved.

Description

Technical Field

[0001] This invention belongs to the field of sensor and respiratory monitoring equipment technology, specifically relating to the construction of an intelligent monitoring and auxiliary diagnosis system and method for sleep apnea and hypopnea based on flexible sensors. Background Technology

[0002] Sleep apnea syndrome (SAS) is a common sleep disorder characterized by repeated pauses in breathing or hypopnea during sleep. This leads to decreased blood oxygen saturation and disrupted sleep structure. In severe cases, it can induce various chronic diseases such as hypertension, coronary heart disease, arrhythmia, stroke, and type 2 diabetes, and even cause sudden death, making it a serious and often overlooked killer. Obstructive sleep apnea (OSA) is the most common type, primarily caused by repeated collapse or narrowing of the upper airway during sleep.

[0003] Currently, clinical diagnosis of sleep apnea primarily relies on polysomnography (PSG). This system monitors multiple physiological signals, including electroencephalography (EEG), eye movements, electromyography (EMG), respiratory airflow, chest and abdominal movements, and blood oxygen saturation, to comprehensively assess sleep quality and respiratory events. While considered the gold standard, this method is characterized by bulky equipment, complex operation, high testing costs, and interference with the subject's sleep environment and physical condition. It is unsuitable for home-based or long-term dynamic monitoring, limiting its widespread application in primary healthcare and personalized health management. Patents CN202411374298.2, CN202111139482.5, and CN202411172024.5 mention other auxiliary or alternative technologies, such as using microwave radar to record respiratory waveforms, using microphones to record the sound characteristics corresponding to sleep apnea within snoring, or using cameras to capture the user's sleep movement patterns. However, these technologies all suffer from significant measurement errors and the equipment is not lightweight or convenient enough. Resistive flexible strain sensors possess advantages such as being lightweight, thin, flexible, highly responsive, and conforming well to the human body. They can sense and monitor muscle or joint movements in real time, offering the potential for non-invasive, highly comfortable, and continuous monitoring. Although current research and practical applications are still in the exploratory stage, their development prospects in portable and personalized health monitoring have begun to emerge. However, existing research on respiratory monitoring based on flexible sensors mainly focuses on simple respiratory rate monitoring, lacking accurate identification and automatic diagnosis of abnormal breathing patterns (such as apnea, hypopnea, and periodic breathing), failing to meet the needs for early warning and risk screening of sleep apnea syndrome. In view of the above problems, this patent proposes the development of a wireless, non-intrusive respiratory monitoring system based on a resistive flexible sensor, and establishes an auxiliary diagnostic method for sleep apnea and hypopnea based on this system. This system, while maintaining high wearing comfort, combines signal processing and intelligent analysis algorithms to achieve real-time identification and preliminary diagnosis of apnea and hypopnea events, meeting the monitoring and management of sleep health status in a home environment, and has broad application prospects and social significance. Summary of the Invention

[0004] The sensing element in this system is a resistive flexible sensor, which is attached to the skin surface of the main and accessory muscle groups related to human respiration. It acquires resistance signals related to respiratory frequency and amplitude, collects the patient's respiratory data in real time, and further constructs adaptive upper and lower limits for amplitude alarms by statistically modeling the individual's resting respiratory amplitude, thereby improving the individual adaptability and robustness of the respiratory monitoring system.

[0005] The technical solution adopted in this invention is as follows:

[0006] A smart monitoring and auxiliary diagnostic system for sleep apnea and hypopnea based on a flexible sensor, the system comprising a resistive flexible strain sensor, a wireless data acquisition card, and a portable data processing terminal;

[0007] The flexible strain sensor described is a band-aid-type flexible strain sensor that can be directly adhered to the surface of human skin. It monitors the deformation of muscles related to respiration and converts this deformation into changes in resistance signals. The flexible strain sensor is connected to a wireless data acquisition card via a connecting cable.

[0008] The wireless data acquisition card and the portable data processing terminal are connected wirelessly.

[0009] Furthermore, the wireless data acquisition card can be hung around the neck, magnetically placed in a pajama pocket or attached to the outside of the pajama, or worn on the wrist or arm like a wristwatch.

[0010] Furthermore, the flexible strain sensor comprises, from the outside in, an adhesive layer, a force-sensitive coating, a silver paste layer, a conductive substrate layer, and an anti-interference layer.

[0011] Furthermore, the adhesive layer uses medical-grade waterproof tape.

[0012] Furthermore, the force-sensitive coating is a conductive coating composed of carbon-based materials, such as carbon black, carbon nanotubes, or graphene. Preferably, the force-sensitive coating is prepared by: dispersing the carbon-based material in a solution-based polymer matrix to obtain a conductive slurry; uniformly coating the dispersed conductive slurry onto a conductive substrate layer using a blade coating or screen printing method; and allowing it to dry naturally to obtain the force-sensitive coating. The solution-based polymer matrix is, for example, PDMS, TPU, Ecoflex, or epoxy resin; the mass fraction of the carbon-based material in the solution-based polymer matrix is ​​2-10 wt%; and the thickness of the force-sensitive coating is 10-30 μm.

[0013] Furthermore, the silver paste layer is composed of silver powder and resin, and is prepared on the conductive substrate layer by scraping or screen printing after the force-sensitive coating is applied. The silver paste layer is located at both ends of the force-sensitive coating and is used to connect the force-sensitive coating and the connecting wires; the mass ratio of silver powder is 65–70 wt%. The resin is epoxy resin or acrylate. The thickness of the silver paste layer is 10–30 μm.

[0014] Furthermore, the conductive substrate layer is made of medical-grade polyethylene, polypropylene, polyvinyl alcohol, and thermoplastic polyurethane film, with a thickness of 0.05-0.2 mm;

[0015] Furthermore, the anti-interference layer has a three-layer structure, comprising, from the inside out, an insulating layer, a shielding layer, and a protective layer. Preferably, the shielding layer material is PEDOT:PSS (100–300 nm thick), copper or aluminum foil (10–50 nm thick), graphene film (200–500 nm thick), or carbon nanotube conductive film (100–300 nm thick), which provides excellent electromagnetic shielding. The insulating layer is disposed between the conductive substrate layer and the shielding layer for electrical isolation and to prevent short circuits; the insulating layer material is polyimide or silicone, with a thickness of 1–5 μm. Preferably, the protective layer material is polyurethane (PU), TPU, or Ecoflex, used to enhance the wear resistance and waterproof performance of the flexible device, with a thickness of 10–50 μm.

[0016] Furthermore, the wireless data acquisition card transmits data to a portable data processing terminal via a Wi-Fi wireless acquisition module or a Bluetooth acquisition module, records and processes the respiratory data, displays it, and issues alarms for apnea and hypopnea events.

[0017] Furthermore, the portable data processing terminal includes a resting sampling module, a data preprocessing module, a peak and trough extraction module, an amplitude statistics module, an adaptive threshold calculation module, a monitoring and alarm module, and a visualization interface module.

[0018] The resting sampling module is used to guide users to collect 60 seconds of resistance-time data using a resistive flexible strain sensor while in a calm breathing state.

[0019] The data preprocessing module is used to filter, normalize, and perform differential noise reduction on the raw signal acquired by the resting sampling module.

[0020] The peak and trough extraction module is used to detect local extreme points within the respiratory cycle based on a sliding window in the preprocessed data to obtain the effective number of peaks and troughs in the resistance-time signal.

[0021] The amplitude statistics module is used to extract the resistance amplitude of each cycle based on the local extreme points within the respiratory cycle detected by the peak and trough extraction module.

[0022] The adaptive threshold calculation module is used to address the influence of different individuals' chest and abdominal compliance, body shape, and sensor fit, and to realize an individualized low ventilation detection threshold storage and retrieval module, which saves the threshold and calibration information (JSON format) using user_id;

[0023] The monitoring and alarm module is used to detect whether the current resistance amplitude exceeds the limit; if there are continuous abnormalities, an alarm will be triggered.

[0024] The information that can be input into the visualization interface module includes gender, age, BMI, measurement position (supine / lateral), breathing mode (thoracic / abdominal), and pathological state (obstructive / central).

[0025] A method for intelligent monitoring and assisted diagnosis of sleep apnea and hypoventilation based on flexible sensors, comprising the following steps:

[0026] Step 1: The user applies the flexible strain sensor to the corresponding part of the chest and abdomen according to their own needs (for chest breathing, it is generally applied between the third and fifth ribs, and for the abdomen, it is generally applied one to ten centimeters below the xiphoid process).

[0027] Step 2: Connect the portable data processing terminal to the wireless acquisition module (Wi-Fi, Bluetooth, etc.) in the wireless data acquisition card.

[0028] Step 3: Input the user's various body data through the visual interface module in the portable data processing terminal.

[0029] Step 4: Use the resting acquisition module of the portable data processing terminal to acquire initial 60 seconds of resistance-time data. Then, the data preprocessing module filters, normalizes, and performs differential noise reduction on the acquired raw signal. The preprocessed signal is then calculated by the peak and trough extraction module to obtain the number of effective resistance-time signal peaks and troughs, thereby obtaining the respiratory rate, which is displayed in real time on the visualization interface. The amplitude statistics module calculates the difference between the peaks and troughs of each resistance-time signal and converts it into respiratory amplitude, which is also displayed in real time on the visualization interface.

[0030] Step 5: Monitoring and alarm modules related to sleep apnea.

[0031] The adaptive threshold calculation module selects different adaptive adjustment methods for different types of users. The monitoring and alarm module determines whether a low ventilation event has occurred through a judgment algorithm, and counts the event when the event lasts for more than Δh seconds according to the AHI. Then, the AHI value is calculated hourly, and the alarm level is determined by combining the continuous low ventilation time and the event interval.

[0032] Furthermore, the determination algorithm is as follows:

[0033] (1) Linear model (applicable to young and middle-aged people with normal body shape, ideal situation):

[0034] F1 / F0=ΔR1 / ΔR0

[0035] Low ventilation event judgment rule: ΔR1 / ΔR0≤0.7, and if this condition is met continuously for ≥10 seconds, it is judged as a low ventilation event;

[0036] (2) Logarithmic model (suitable for children or obese individuals, where subcutaneous fat provides a large buffer and the effect of deformation on airflow is delayed):

[0037] F1 / F0 = 1 + ln(ΔR1 / ΔR0) q

[0038] Rule for determining a hypoventilation event: 1 + ln(ΔR1 / ΔR0) q If the value is ≤0.7 and this condition is met continuously for ≥10 seconds, it is considered a hypoventilation event;

[0039] (3) Power function model (considering muscle stiffness and nonlinear deformation, which is more realistic)

[0040] F1 / F0=(ΔL1 / ΔL0) q =(ΔR1 / ΔR0) q =α q

[0041] Hypoventilation event judgment rule: satisfying α q If the value is ≤0.7 and this condition is met continuously for ≥10 seconds, it is considered a hypoventilation event.

[0042] Furthermore, after determining the basic judgment algorithm based on the above formula, the targeted modification parameters required for various groups are introduced into the judgment formulas for the above three categories for adaptive adjustment.

[0043] (1) Women have smaller average chest wall size and lower vital capacity, resulting in smaller respiratory amplitude and lower ΔR0 compared to men. Therefore, the expected baseline value of ΔR0 is scaled and adjusted according to gender (e.g., ΔR0 × 0.85 for women compared to adult men), and the tolerance for amplitude fluctuation is relaxed: allowing a smaller ΔR0 as an effective respiratory signal, that is, adjusting ΔR0 to (0.8-0.9) times ΔR0.

[0044] (2) In the elderly, the skin is loose and the muscle tension is reduced, the deformation signal amplitude is weakened and the period is slowed down. The exponent q of the modified power function model increases with age (reflecting the decrease in stiffness).

[0045] (3) In children and people with high BMI, subcutaneous fat buffers deformation, weakens the electrical response, and reduces deformation conduction. Therefore, q should also be increased.

[0046] (4) Obstructive type: The patient has effortful breathing behavior but insufficient airflow, ΔR1 is normal but airflow decreases; therefore, for obstructive patients, analyze the synchronicity of ΔR1 and SpO2 (blood oxygen saturation); if ΔR1 is normal but blood oxygen does not recover, it can be judged as hypoventilation.

[0047] (5) Central type: The patient does not exert breathing effort, and ΔR1 decreases or disappears. Therefore, for central type patients, apnea is directly judged by a continuous decrease in ΔR1 or cessation for more than 10 seconds.

[0048] Furthermore, calculating the apnea-hypopnea index (AHI) requires dividing the number of hypopnea events by the effective sleep time. This system determines sleep time based on respiratory signals; therefore, AHI is calculated based on the total sleep time (TRT). Assume the number of detected hypopnea events is E. low The total time is T total If (TRT), then the formula for calculating AHI is:

[0049]

[0050] Among them: E low This refers to the number of hypoventilation events. T total It is the total sleep time after the breathing signal stabilizes, expressed in hours.

[0051] The beneficial effects of this invention are:

[0052] (1) By automatically modeling the initial respiratory data of each user, setting an individual-specific respiratory amplitude alarm threshold, it can effectively adapt to people of different genders, ages and body types; and overcome the false alarm and missed alarm problems caused by the fixed threshold setting of the traditional system.

[0053] (2) The sensor has good zero-bias robustness and does not rely on the initial absolute resistance value of the flexible strain sensor. It only focuses on the relative change amplitude caused by breathing. The automatic normalization mechanism cancels the initial value difference caused by the sensor manufacturing and attachment method, and enhances the system's versatility and portability.

[0054] (3) It has a simple structure and high computational efficiency. It is implemented using lightweight statistical algorithms (mean, standard deviation, extreme value detection), which do not require model training and large-scale data support. It is easy to deploy in embedded hardware, wearable devices and low-power systems, and has strong real-time performance.

[0055] (4) It is easy to expand and adapt to a wide range of scenarios. It supports multi-user configuration storage and is compatible with data interaction methods on multiple platforms such as LabVIEW, MCU, and Bluetooth. It is suitable for clinical and home monitoring and can also be extended to multiple application scenarios such as sleep health, remote rehabilitation, and child safety. Detailed Implementation

[0056] A smart monitoring and auxiliary diagnostic system for sleep apnea and hypopnea based on a flexible sensor, the system comprising a resistive flexible strain sensor, a wireless data acquisition card, and a portable data processing terminal;

[0057] The flexible strain sensor described is a band-aid-type flexible strain sensor that can be directly adhered to the surface of human skin. It monitors the deformation of muscles related to respiration and converts this deformation into changes in resistance signals. The flexible strain sensor is connected to a wireless data acquisition card via a connecting cable.

[0058] The wireless data acquisition card and the portable data processing terminal are connected wirelessly.

[0059] Furthermore, the wireless data acquisition card can be hung around the neck, magnetically placed in a pajama pocket or attached to the outside of the pajama, or worn on the wrist or arm like a wristwatch.

[0060] Furthermore, the flexible strain sensor comprises, from the outside in, an adhesive layer, a force-sensitive coating, a silver paste layer, a conductive substrate layer, and an anti-interference layer.

[0061] Furthermore, the adhesive layer uses medical-grade waterproof tape.

[0062] Furthermore, the force-sensitive coating is a conductive coating composed of carbon-based materials, such as carbon black, carbon nanotubes, or graphene. Preferably, the force-sensitive coating is prepared by: dispersing the carbon-based material in a solution-based polymer matrix to obtain a conductive slurry; uniformly coating the dispersed conductive slurry onto a conductive substrate layer using a blade coating or screen printing method; and allowing it to dry naturally to obtain the force-sensitive coating. The solution-based polymer matrix is, for example, PDMS, TPU, Ecoflex, or epoxy resin; the mass fraction of the carbon-based material in the solution-based polymer matrix is ​​2-10 wt%; and the thickness of the force-sensitive coating is 10-30 μm.

[0063] Furthermore, the silver paste layer is composed of silver powder and resin, and is prepared on the conductive substrate layer by scraping or screen printing after the force-sensitive coating is applied. The silver paste layer is located at both ends of the force-sensitive coating and is used to connect the force-sensitive coating and the connecting wires; the mass ratio of silver powder is 65–70 wt%. The resin is epoxy resin or acrylate. The thickness of the silver paste layer is 10–30 μm.

[0064] Furthermore, the conductive substrate layer is made of medical-grade polyethylene, polypropylene, polyvinyl alcohol, and thermoplastic polyurethane film, with a thickness of 0.05-0.2 mm;

[0065] Furthermore, the anti-interference layer has a three-layer structure, comprising, from the inside out, an insulating layer, a shielding layer, and a protective layer. Preferably, the shielding layer material is PEDOT:PSS (100–300 nm thick), copper or aluminum foil (10–50 nm thick), graphene film (200–500 nm thick), or carbon nanotube conductive film (100–300 nm thick), which provides excellent electromagnetic shielding. The insulating layer is disposed between the conductive substrate layer and the shielding layer for electrical isolation and to prevent short circuits; the insulating layer material is polyimide or silicone, with a thickness of 1–5 μm. Preferably, the protective layer material is polyurethane (PU), TPU, or Ecoflex, used to enhance the wear resistance and waterproof performance of the flexible device, with a thickness of 10–50 μm.

[0066] Furthermore, the wireless data acquisition card transmits data to a portable data processing terminal via a Wi-Fi wireless acquisition module or a Bluetooth acquisition module, records and processes the respiratory data, displays it, and issues alarms for apnea and hypopnea events.

[0067] Furthermore, the portable data processing terminal includes a resting sampling module, a data preprocessing module, a peak and trough extraction module, an amplitude statistics module, an adaptive threshold calculation module, a monitoring and alarm module, and a visualization interface module.

[0068] The resting sampling module is used to guide users to collect 60 seconds of resistance-time data using a resistive flexible strain sensor while in a calm breathing state.

[0069] The data preprocessing module is used to filter, normalize, and perform differential noise reduction on the raw signal acquired by the resting sampling module.

[0070] The peak and trough extraction module is used to detect local extreme points within the respiratory cycle based on a sliding window in the preprocessed data to obtain the effective number of peaks and troughs in the resistance-time signal.

[0071] The amplitude statistics module is used to extract the resistance amplitude of each cycle based on the local extreme points within the respiratory cycle detected by the peak and trough extraction module.

[0072] The adaptive threshold calculation module is used to address the influence of different individuals' chest and abdominal compliance, body shape, and sensor fit, and to realize an individualized low ventilation detection threshold storage and retrieval module, which saves the threshold and calibration information (JSON format) using user_id;

[0073] The monitoring and alarm module is used to detect whether the current resistance amplitude exceeds the limit; if there are continuous abnormalities, an alarm will be triggered.

[0074] The information that can be input into the visualization interface module includes gender, age, BMI, measurement position (supine / lateral), breathing mode (thoracic / abdominal), and pathological state (obstructive / central).

[0075] A method for intelligent monitoring and assisted diagnosis of sleep apnea and hypoventilation based on flexible sensors, comprising the following steps:

[0076] Step 1: The user applies the flexible strain sensor to the corresponding part of the chest and abdomen according to their own needs (for chest breathing, it is generally applied between the third and fifth ribs, and for the abdomen, it is generally applied one to ten centimeters below the xiphoid process).

[0077] Step 2: Connect the portable data processing terminal to the wireless acquisition module (Wi-Fi, Bluetooth, etc.) in the wireless data acquisition card.

[0078] Step 3: Input the user's various body data through the visual interface module in the portable data processing terminal.

[0079] Step 4: Use the resting acquisition module of the portable data processing terminal to acquire initial 60 seconds of resistance-time data. Then, the data preprocessing module filters, normalizes, and performs differential noise reduction on the acquired raw signal. The preprocessed signal is then calculated by the peak and trough extraction module to obtain the number of effective resistance-time signal peaks and troughs, thereby obtaining the respiratory rate, which is displayed in real time on the visualization interface. The amplitude statistics module calculates the difference between the peaks and troughs of each resistance-time signal and converts it into respiratory amplitude, which is also displayed in real time on the visualization interface.

[0080] Step 5: Monitoring and alarm modules related to sleep apnea.

[0081] The monitoring and alarm module uses a judgment algorithm to determine whether a low ventilation event has occurred, and counts the event as an event if the event lasts for more than Δh seconds according to the AHI (Advanced Ventilation Hierarchy Index). Then, it calculates the AHI value on an hourly basis and combines the continuous low ventilation time and the event interval to comprehensively determine the alarm level.

[0082] The derivation process and design idea of ​​the decision algorithm are as follows:

[0083] Resistive flexible strain sensors detect changes in resistance caused by the deformation of respiratory muscles, thereby obtaining respiratory event information through the resistance signal. To obtain characteristic information of apnea and hypoventilation events, the change in respiratory airflow is correlated with the deformation of respiratory muscles, and further linked to the resistance signal directly output by the flexible sensor.

[0084] The deformation amplitude of the respiratory muscles in the chest / abdomen can be regarded as tidal volume V. T (t)(tidal volume, VT) or ventilation velocity An approximate reflection is given by setting the relationship between tidal volume and ventilation velocity as follows: instantaneous ventilation velocity value The muscle deformation variable ΔL has a certain functional relationship with tidal volume (lung volume index). Considering the influence of muscle and thoracic compliance and individual body shape differences, let:

[0085] V T (t)=k1×[ΔL(t)]n

[0086] Where: ΔL(t) represents the deformation amplitude measured by the resistive flexible strain sensor; k1 represents the proportionality constant related to the body shape, which is obtained by fitting the data obtained from relevant experiments using the resistive flexible strain sensor and the airflow meter through the above formula; n represents the empirical coefficient, which takes into account nonlinear expansion (can be set to 1~1.5).

[0087] Instantaneous value of ventilation velocity

[0088] During normal breathing: the peak airflow is F0, and the corresponding peak chest and abdominal deformation is ΔL0; when the deformation during a single breath is detected to be lower than normal, let the deformation be ΔL1, and the corresponding airflow rate be F. 1， When F1 decreases to F1≤0.7×F0, that is, the airflow decreases by more than 30%, it can be identified as a low ventilation event. Based on this, a clear judgment criterion should be established.

[0089] The first scenario: A linear model is suitable for young and middle-aged adults with normal body types.

[0090] F1 / F0=ΔL1 / ΔL0

[0091] The percentage decrease in deformation is the same as the percentage decrease in airflow velocity. If ΔL1 / ΔL 0＝ If γ < 0.7, meaning a 30% decrease in deformation equals a 30% decrease in ventilation rate, a hypoventilation event can be considered to have occurred. γ is a proportionality coefficient for the first scenario, used to characterize the relationship between thoracic and abdominal deformation and ventilation rate.

[0092] The second scenario: For children, obese individuals, or those with large subcutaneous fat buffering, whose deformation changes have a delayed impact on airflow, a logarithmic model is used.

[0093] F1 / F0=ln(1+ΔL1 / ΔL0)

[0094] The change in deformation has a logarithmic relationship with the ventilation rate. If F1 / F0 = ln(1 + ΔL1 / ΔL0) = β < 0.7, a low ventilation event can be considered to have occurred. β represents the logarithmic change in the current peak ventilation rate compared to the baseline peak ventilation rate in the second case, which can be regarded as the logarithmic mapping result of the degree of deformation decrease.

[0095] The third case: When considering muscle stiffness and nonlinear deformation, a power function model is used.

[0096] F1 / F0=(ΔL1 / ΔL0) q, Where q∈[1.1,1.4]

[0097] If we assume that the peak value of chest and abdominal deformation detected at a certain moment decreases as follows:

[0098] ΔL1 / ΔL0 = α, where α represents the third case, indicating the ratio of the respiratory deformation amplitude at a certain moment (or a certain cycle) to the normal inspiratory deformation amplitude under resting conditions.

[0099] For the third case, we can deduce the corresponding decrease in airflow velocity as follows:

[0100] F1 / F0=(ΔL1 / ΔL0) q =α q

[0101] Therefore: if α q If the value is ≤0.7, it can be determined that the ventilation velocity has decreased by ≥30%, and thus it can be considered that a hypoventilation event has occurred.

[0102] Based on preliminary experimental results, the sensitivity coefficient (GF) of the resistive flexible strain sensor used in sleep monitoring is approximately 10-50 within its operating range. Where R0 is the initial resistance of the sensor, R is the resistance after the external force is applied, ε is the strain, and ΔR is the change in resistance.

[0103] Combining the formula V mentioned above T (t)=k1×[ΔL(t)]n; This explanation uses the third case, the power function model, as an example, where the initial criterion is F1 / F0 = (ΔL1 / ΔL0). q ≤α q ΔL can be replaced with the resistance signal ΔR / R0, which can be directly measured by the sensor. First, express the deformation ΔL(t) as strain ε(t) and establish the resistance-strain relationship:

[0104]

[0105] in: ε represents the relative resistance change rate; GF represents the sensitivity coefficient (10–50); ε represents the strain; ΔL(t) represents the deformation amplitude measured by the resistive flexible strain sensor; L0 represents the initial length of the sensor.

[0106] Substituting the tidal volume expression into the expression for ΔL(t), we can find the value of V. T(t): A function that yields tidal volume ΔR / R.

[0107]

[0108] Where k2 = k1·(L0 / GF·R0) n , which is the final scaling factor.

[0109] Differentiating the above equation, we obtain the instantaneous airflow velocity:

[0110]

[0111] Let the resistance change during normal inspiratory peak be ΔR0, and the detected inspiratory peak be ΔR1. α still represents the ratio of the respiratory deformation amplitude at a certain moment (or a certain cycle) to the normal inspiratory deformation amplitude at rest. The only difference is that the deformation represented by the airflow sensor is replaced by resistance, and the relative decrease ratio α is defined as α = ΔR1 / ΔR0.

[0112]

[0113] (1) Linear model (applicable to young and middle-aged people with normal body shape, ideal situation):

[0114] F1 / F0=ΔR1 / ΔR0

[0115] Low ventilation event judgment rule: ΔR1 / ΔR0≤0.7, and if this condition is met continuously for ≥10 seconds, it is judged as a low ventilation event;

[0116] (2) Logarithmic model (suitable for children or obese individuals, where subcutaneous fat provides a large buffer and the effect of deformation on airflow is delayed):

[0117] F1 / F0 = 1 + ln(ΔR1 / ΔR0) q

[0118] Rule for determining a hypoventilation event: 1 + ln(ΔR1 / ΔR0) q If the value is ≤0.7 and this condition is met continuously for ≥10 seconds, it is considered a hypoventilation event;

[0119] (3) Power function model (considering muscle stiffness and nonlinear deformation, which is more realistic)

[0120] F1 / F0=(ΔL1 / ΔL0) q =(ΔR1 / ΔR0) q =α q

[0121] Hypoventilation event judgment rule: satisfying α q If the value is ≤0.7 and this condition is met continuously for ≥10 seconds, it is considered a hypoventilation event.

[0122] Furthermore, after determining the basic judgment algorithm based on the above formula, the targeted modification parameters required for various groups are introduced into the judgment formulas for the above three categories for adaptive adjustment.

[0123] (1) Women have smaller average chest wall size and lower vital capacity, resulting in smaller respiratory amplitude and lower ΔR0 compared to men. Therefore, the expected baseline value of ΔR0 is scaled and adjusted according to gender (e.g., ΔR0 × 0.85 for women compared to adult men), and the tolerance for amplitude fluctuation is relaxed: allowing a smaller ΔR0 as an effective respiratory signal, that is, adjusting ΔR0 to (0.8-0.9) times ΔR0.

[0124] (2) In the elderly, the skin is loose and the muscle tension is reduced, the deformation signal amplitude is weakened and the period is slowed down. The exponent q of the modified power function model increases with age (reflecting the decrease in stiffness).

[0125] (3) In children and people with high BMI, subcutaneous fat buffers deformation, weakens the electrical response, and reduces deformation conduction. Therefore, q should also be increased.

[0126] (4) Obstructive type: The patient has effortful breathing behavior but insufficient airflow, ΔR1 is normal but airflow decreases; therefore, for obstructive patients, analyze the synchronicity of ΔR1 and SpO2 (blood oxygen saturation); if ΔR1 is normal but blood oxygen does not recover, it can be judged as hypoventilation.

[0127] (5) Central type: The patient does not exert breathing effort, and ΔR1 decreases or disappears. Therefore, for central type patients, apnea is directly judged by a continuous decrease in ΔR1 or cessation for more than 10 seconds.

[0128] Calculating the apnea-hypopnea index (AHI) requires dividing the number of hypopnea events by the effective sleep time. This system determines sleep time based on respiratory signals; therefore, AHI is calculated based on total sleep time (TRT). Assume the number of detected hypopnea events is E. low The total time is T total If (TRT), then the formula for calculating AHI is:

[0129]

[0130] Among them: E low This refers to the number of hypoventilation events. T total It is the total sleep time after the breathing signal stabilizes, expressed in hours.

[0131] The proportionality coefficient α, derived from the respiratory deformation ΔL(t) and the resting reference value ΔL0, is used in conjunction with the AHI to determine hypoventilation or apnea events, classify their severity, and provide prompts or interventions to users / family members. Based on the AHI calculation formula, the number of consecutive ineffective respiratory cycles N... invalidContinuous alarm counter N alert The alarm determination logic is as follows:

[0132] The algorithm described above is used to determine whether a low ventilation event has occurred, and the event count is recorded when the event lasts for more than Δh seconds according to the AHI. The AHI value is then calculated hourly, and the alarm level is determined by combining the continuous low ventilation time and the event interval.

[0133] Example:

[0134] (1) Application of Resistive Flexible Strain Sensors and Registration of User Information: Multiple resistive flexible strain sensors were simultaneously attached to the subject's primary and accessory respiratory muscle groups (located 1-10 cm below the xiphoid process, parallel to the fifth rib, etc.). The other end of each resistive flexible strain sensor was connected to a wireless data acquisition card, which was then connected to a portable data processing terminal (host computer) via Bluetooth. The connection stability was checked to ensure there were no loose connections or breaks. After system startup, user identification information (user_id, gender, age, BMI, measurement posture, etc.) was input through the visual interface module in the portable data processing terminal, and the data buffer was initialized to store sampled values ​​and timestamps. User information is used for subsequent binding and retrieval of individual threshold configuration files, and is used based on the user's resting respiratory sampling results.

[0135] (2) Adjusting the user's state: Guide the user to maintain a side-lying or supine position (back straight, shoulders relaxed, hands naturally placed on the thighs), avoid strenuous body movements, and maintain natural breathing. Slowly enter a sleep state, and the system collects resistance data at a fixed sampling frequency (preferably 20Hz, dynamically adjusted according to the user's breathing state, such as increasing to 50Hz when rapid breathing is detected, to balance data accuracy and storage efficiency).

[0136] (3) The initial 60-second resistance-time data is collected using the resting acquisition module of a portable data processing terminal. Then, the raw signal is filtered, normalized, and differentially denoised by the data preprocessing module. The preprocessed signal is then calculated by the peak and trough extraction module to obtain the number of effective resistance-time signal peaks and troughs, thereby obtaining the respiratory rate and displaying it in real time on the visualization interface. The amplitude statistics module calculates the difference between the peaks and troughs of each resistance-time signal and converts it into respiratory amplitude, which is also displayed in real time on the visualization interface. The system uses a nonlinear mapping model to convert the chest and abdominal deformation ΔL(t) into tidal volume VT(t), and then further estimates the air flow rate V(t). ΔL0 is measured according to the above algorithm (the maximum reference deformation obtained from resting acquisition); ΔL1 is the low ventilation judgment threshold derived from the model (e.g., ΔL1≤0.7×ΔL0 or ΔL1≤ΔL0×0.7^(1 / n)). This threshold will be used as the real-time reference baseline during the night monitoring process to determine events of weakened or stopped breathing. After the user falls asleep, the system enters sleep monitoring mode, collects data from the chest / abdomen patch, and calculates the deformation ΔL of each breath in real time. _peak Whether the temperature continuously drops below ΔL1 for ≥10 seconds; whether it meets the criteria for a "hypoventilation event" or "apnea". Further, based on:

[0137]

[0138] Automatically count the number of all hypoventilation or apnea events at night and calculate the standard apnea index (AHI).

[0139] (4) Alarm Response: The system responds in stages based on the detected respiratory abnormalities. The host computer software will trigger different alarm prompts according to the stage. When AHI is less than 5 and the duration of hypopnea is less than 20 seconds, the system is in normal condition and no alarm is triggered. When AHI is between 5 and 15 or the cumulative hypopnea exceeds 60 seconds within 1 hour, it is judged as a moderate alarm, triggering a local vibration alert and logging. When AHI is greater than or equal to 15, or a single hypopnea lasts more than 30 seconds, or the interval between multiple events is less than 60 seconds, it is judged as a serious alarm. The system will trigger voice and vibration alarms simultaneously, record the serious event, and send a notification to the nursing staff's APP, forming a complete intelligent early warning system for sleep breathing abnormalities. If a serious alarm is triggered, the host computer software will set the emergency contact's mobile phone number / account (e.g., binding a family member's APP account). When it is judged as a serious alarm (Alert_Level = 2), the host computer will send an HTTPS POST request / MQTT message carrying the current user ID, alarm type, timestamp, and brief data summary (e.g., AHI = 18, ΔL decreases, 3 pauses).

[0140] (5) Data recording and storage: Throughout the monitoring period, the system continuously records the following information: ΔL waveform for each breath; start and end time of each hypoventilation event; hourly AHI value change; total number of alarms and their classification; user's current position, channel selection and threshold parameters (written to the log). The data is saved as structured files (such as .json, .csv) and visualized for use by users or doctors for subsequent diagnostic analysis.

Claims

1. A smart monitoring and assisted diagnostic system for sleep apnea and hypopnea based on flexible sensors, characterized in that, The system includes a resistive flexible strain sensor, a wireless data acquisition card, and a portable data processing terminal; The flexible strain sensor is a band-aid-type flexible strain sensor that can be directly adhered to the surface of human skin. It monitors the deformation of muscles related to breathing and converts it into changes in resistance signals. The flexible strain sensor is connected to a wireless data acquisition card via a connecting cable. The wireless data acquisition card and the portable data processing terminal are connected wirelessly. The portable data processing terminal includes a resting sampling module, a data preprocessing module, a peak and trough extraction module, an amplitude statistics module, an adaptive threshold calculation module, a monitoring and alarm module, and a visualization interface module.

2. The intelligent monitoring and auxiliary diagnosis system for sleep apnea and hypopnea based on a flexible sensor according to claim 1, characterized in that, The wireless data acquisition card can be hung around the neck, magnetically placed in a pajama pocket or attached to the outside of the pajama, or worn on the wrist or arm like a wristwatch.

3. The intelligent monitoring and auxiliary diagnosis system for sleep apnea and hypopnea based on a flexible sensor according to claim 1, characterized in that, The flexible strain sensor comprises, from the outside to the inside, an adhesive layer, a force-sensitive coating layer, a silver paste layer, a conductive substrate layer, and an anti-interference layer. The adhesive layer uses medical-grade waterproof tape. The force-sensitive coating is a conductive coating composed of carbon-based materials, such as carbon black, carbon nanotubes, or graphene. Preferably, the force-sensitive coating is prepared by: dispersing the carbon-based material in a solution-based polymer matrix to obtain a conductive slurry; uniformly coating the dispersed conductive slurry onto a conductive substrate layer using a scraping or screen printing method; and allowing it to dry naturally to obtain the force-sensitive coating. The solution-based polymer matrix is, for example, PDMS, TPU, Ecoflex, or epoxy resin; the mass fraction of the carbon-based material in the solution-based polymer matrix is ​​2-10 wt%; and the thickness of the force-sensitive coating is 10-30 μm. The silver paste layer is composed of silver powder and resin, and is prepared on the conductive substrate by scraping or screen printing after the force-sensitive coating is applied. The silver paste layer is located at both ends of the force-sensitive coating and is used to connect the force-sensitive coating and the connecting wire. The mass ratio of silver powder is 65-70 wt%. The resin is epoxy resin or acrylate. The thickness of the silver paste layer is 10-30 μm. The conductive substrate layer is made of medical-grade polyethylene, polypropylene, polyvinyl alcohol, and thermoplastic polyurethane film, with a thickness of 0.05-0.2mm. The anti-interference layer has a three-layer structure, consisting of an insulating layer, a shielding layer, and a protective layer from the inside out. The shielding layer is made of PEDOT:PSS, copper or aluminum foil with a thickness of 10–50 nm, graphene film with a thickness of 200–500 nm, or carbon nanotube conductive film with a thickness of 100–300 nm, providing excellent electromagnetic shielding. The insulating layer is disposed between the conductive substrate layer and the shielding layer for electrical isolation and to prevent short circuits. The insulating layer is made of polyimide or silicone and has a thickness of 1–5 μm. The protective layer is made of polyurethane, TPU, or Ecoflex and is used to enhance the wear resistance and waterproof performance of the flexible device. It has a thickness of 10–50 μm.

4. The intelligent monitoring and auxiliary diagnosis system for sleep apnea and hypopnea based on a flexible sensor according to claim 1, characterized in that, The wireless data acquisition card transmits data to a portable data processing terminal via a Wi-Fi or Bluetooth module, records, processes, and displays respiratory data, and issues alarms for apnea and hypopnea events.

5. The intelligent monitoring and auxiliary diagnosis system for sleep apnea and hypopnea based on a flexible sensor according to claim 1, characterized in that, The resting sampling module is used to guide users to collect 60 seconds of resistance-time data using a resistive flexible strain sensor while in a calm breathing state. The data preprocessing module is used to filter, normalize, and perform differential noise reduction on the raw signal acquired by the resting sampling module. The peak and trough extraction module is used to detect local extreme points within the respiratory cycle based on a sliding window in the preprocessed data to obtain the effective number of peaks and troughs in the resistance-time signal. The amplitude statistics module is used to extract the resistance amplitude of each cycle based on the local extreme points within the respiratory cycle detected by the peak and trough extraction module. The adaptive threshold calculation module is used to address the influence of different individuals' chest and abdominal compliance, body shape, and sensor fit, and to realize an individualized low ventilation detection threshold storage and retrieval module. The monitoring and alarm module is used to detect whether the current resistance amplitude exceeds the limit; if there are continuous abnormalities, an alarm will be triggered. The information that can be input into the visualization interface module includes gender, age, BMI, measurement position, breathing mode, and pathological condition.

6. A method for intelligent monitoring and auxiliary diagnosis of sleep apnea and hypopnea based on a flexible sensor using the intelligent monitoring and auxiliary diagnosis system for sleep apnea and hypopnea based on any one of claims 1-5, characterized in that, The steps are as follows: Step 1: The user applies the flexible strain sensor to the corresponding area on the chest and abdomen according to their own needs; Step 2: Connect the portable data processing terminal to the wireless acquisition module in the wireless data acquisition card; Step 3: Input the user's various body data through the visual interface module in the portable data processing terminal; Step 4: Use the resting acquisition module of the portable data processing terminal to acquire the initial 60 seconds of resistance-time data. Then, the data preprocessing module filters, normalizes, and performs differential noise reduction on the acquired raw signal. The preprocessed signal is then calculated by the peak and trough extraction module to obtain the number of effective resistance-time signal peaks and troughs, thereby obtaining the respiratory rate and displaying it in real time on the visualization interface. The amplitude statistics module calculates the difference between the peaks and troughs of each resistance-time signal and converts it into respiratory amplitude, which is also displayed in real time on the visualization interface. Step 5: For the monitoring and alarm module involved in sleep apnea: The adaptive threshold calculation module selects different adaptive adjustment methods for different types of users. The monitoring and alarm module determines whether a hypoventilation event has occurred through a judgment algorithm, and counts the event count when the event lasts for more than Δh seconds according to the AHI. Then, the AHI value is calculated hourly, and the alarm level is determined by combining the continuous hypoventilation time and the event interval.

7. A method for intelligent monitoring and assisted diagnosis of sleep apnea and hypoventilation based on a flexible sensor according to any one of claims 6, characterized in that, The resistive flexible strain sensor used senses the resistance change of the respiratory muscle group by sensing the muscle deformation, and then obtains respiratory event information through the resistance signal; in order to obtain the characteristic information of apnea and hypoventilation events, the change in respiratory airflow is correlated with the muscle deformation of the respiratory muscle group, and further linked with the resistance signal directly output by the flexible sensor. The deformation amplitude of the respiratory muscles in the chest / abdomen is considered as tidal volume V. T (t) or airflow velocity An approximate reflection is given by setting the relationship between tidal volume and ventilation velocity as follows: instantaneous ventilation velocity value The muscle deformation ΔL has a certain functional relationship with tidal volume. Considering the influence of muscle and thoracic compliance and individual body shape differences, let: V T (t)=k1×[ΔL(t)]n Where: ΔL(t) represents the deformation amplitude measured by the resistive flexible strain sensor; k1 represents the proportionality constant related to the body shape, which is obtained by fitting the data obtained from relevant experiments using the resistive flexible strain sensor and the airflow meter through the above formula; n represents the empirical coefficient, taking into account nonlinear expansion; Instantaneous value of ventilation velocity During normal breathing, the peak airflow is F0, and the corresponding peak chest and abdominal deformation is ΔL0. When the deformation of a single breath is detected to be lower than the normal level, let the deformation be ΔL1, and the corresponding ventilation rate be F1. When F1 drops to F1≤0.7×F0, that is, the airflow drops by more than 30%, it can be determined as a hypoventilation event. Based on this, a clear judgment criterion is established. The first scenario: A linear model is suitable for young and middle-aged adults with normal body types. F1 / F0=ΔL1 / ΔL0 The decrease in deformation is proportional to the decrease in airflow velocity; if ΔL1 / ΔL 0＝ If γ < 0.7, that is, a 30% decrease in deformation equals a 30% decrease in ventilation rate, a hypoventilation event is considered to have occurred; γ is a proportionality coefficient for the first case, used to characterize the relationship between chest and abdominal deformation and ventilation rate; The second scenario: For children, obese individuals, or those with large subcutaneous fat buffering, whose deformation changes have a delayed impact on airflow, a logarithmic model is used. F1 / F0=ln(1+ΔL1 / ΔL0) The change in deformation has a logarithmic relationship with the ventilation rate; if F1 / F0=ln(1+ΔL1 / ΔL0)=β<0.7, a low ventilation event is considered to have occurred; β represents the logarithmic change of the current ventilation peak compared to the baseline ventilation peak in the second case, which can be regarded as the logarithmic mapping result of the degree of deformation decrease; The third case: When considering muscle stiffness and nonlinear deformation, a power function model is used. F1 / F0=(ΔL1 / ΔL0) q Where q∈[1.1,1.4] If we assume that the peak value of chest and abdominal deformation detected at a certain moment decreases as follows: ΔL1 / ΔL0=α, where α represents the third case, indicating the ratio of the respiratory deformation amplitude at a certain moment to the normal inspiratory deformation amplitude at rest. For the third case, the corresponding decrease in airflow velocity is derived as follows: F1 / F0=(ΔL1 / ΔL0) q =a q Therefore: if α q If the value is ≤0.7, it can be determined that the ventilation velocity has decreased by ≥30%, and thus a hypoventilation event has occurred. Based on preliminary experimental results, the sensitivity coefficient GF of the resistive flexible strain sensor used in sleep monitoring within its operating range is [value missing]. Where R0 is the initial resistance of the sensor, R is the resistance after the external force is applied, ε is the strain, and ΔR is the change in resistance. Combining the formula V mentioned above T (t)=k1×[ΔL(t)]n; This explanation uses the third case, the power function model, as an example, where the initial criterion is F1 / F0 = (ΔL1 / ΔL0). q ≤α q Replace ΔL with the resistance signal ΔR / R0 that can be directly measured by the sensor; first, express the deformation ΔL(t) as strain ε(t) and establish the resistance-strain relationship: in: GF represents the relative resistance change rate; ε represents the sensitivity coefficient; ΔL(t) represents the strain amplitude measured by the resistive flexible strain sensor; L0 represents the initial length of the sensor. Substituting the tidal volume expression into the expression for ΔL(t), we can find the value of V. T (t): A function that yields tidal volume ΔR / R. Where k2 = k1·(L0 / GF·R0) n , which is the final scaling factor; Differentiating the above equation, we obtain the instantaneous airflow velocity: Let the resistance change at the normal inspiratory peak be ΔR0, and the detected inspiratory peak be ΔR1. α still represents the ratio of the respiratory deformation amplitude at a certain moment to the normal inspiratory deformation amplitude at rest; however, the determination criterion is changed from the deformation represented by the airflow sensor to resistance, and the relative decrease ratio α = ΔR1 / ΔR0 is defined. The determination algorithm is as follows: (1) Linear model, applicable to young and middle-aged people with normal body shape, ideal situation: F1 / F0=ΔR1 / ΔR0 Low ventilation event judgment rule: ΔR1 / ΔR0≤0.7, and if this condition is met continuously for ≥10 seconds, it is judged as a low ventilation event; (2) Logarithmic model, suitable for children or obese body types, where subcutaneous fat provides a large buffer and the effect of deformation changes on airflow is delayed: F1 / F0=1+ln(ΔR1 / ΔR0) q Rule for determining a hypoventilation event: 1 + ln(ΔR1 / ΔR0) q If the value is ≤0.7 and this condition is met continuously for ≥10 seconds, it is considered a hypoventilation event; (3) Power function model, which considers muscle stiffness and nonlinear deformation, is more realistic: F1 / F0=(ΔL1 / ΔL0) q =(ΔR1 / ΔR0) q =a q Hypoventilation event judgment rule: satisfying α q If the value is ≤0.7 and this condition is met continuously for ≥10 seconds, it is considered a hypoventilation event.

8. The method for intelligent monitoring and assisted diagnosis of sleep apnea and hypoventilation based on a flexible sensor according to claim 7, characterized in that, After determining the basic judgment algorithm based on the above formula, the targeted modification parameters required for various groups of people are introduced into the judgment formulas for the above three categories for adaptive adjustment. Among them: (1) Women have smaller average chest size and lower vital capacity, and their breathing amplitude is smaller than that of men, resulting in a lower ΔR0. Therefore, the expected baseline value of ΔR0 is scaled and adjusted according to gender (e.g., ΔR0 × 0.85 for women compared to adult men), and the tolerance for amplitude fluctuation is relaxed: allowing a smaller ΔR0 as an effective breathing signal, that is, adjusting ΔR0 to (0.8-0.9) times ΔR0. (2) Elderly people have loose skin and decreased muscle tension, resulting in weaker deformation signal amplitude and slower cycle. The exponent q of the corrected power function model increases with age. (3) Subcutaneous fat in children and people with high BMI buffers deformation, weakens the electrical response, and reduces the conduction of deformation; therefore, q should also be increased. (4) Obstructive type: The patient has effortful breathing behavior but insufficient airflow, ΔR1 is normal but airflow decreases; therefore, for obstructive patients, analyze the synchronicity of ΔR1 and blood oxygen saturation; if ΔR1 is normal but blood oxygen does not recover, it can be judged as hypoventilation. (5) Central type: The patient does not exert breathing behavior, and ΔR1 decreases or disappears; therefore, for central type patients, apnea is directly judged by a continuous decrease in ΔR1 or cessation for more than 10 seconds.

Images