Split type flexible wearable sleep respiration monitoring system

Abstract

The split type flexible wearable sleep respiration monitoring system comprises an upper computer, and a flexible respiration mobility monitoring patch, a flexible ECG monitoring patch and a PPG monitoring module which are connected with the upper computer in a wireless mode; the flexible breathing mobility monitoring patch is composed of a battery, a strain sensing unit, a temperature sensing module, a three-axis gyroscope and a communication module, wherein the strain sensing unit, the temperature sensing module, the three-axis gyroscope and the communication module are connected with the battery. According to the strain sensing unit, the PDMS substrate with the shore hardness of 10-20 A and the thickness of 0.5-2 mm and the metal conductive layer with specific parameters are adopted, and the weight lt of a single patch is matched; by means of the lightweight design of 12 g, biocompatibility is good, foreign body sensation is avoided, the adhesive force lasts for 24 h or more, and the long-term home monitoring requirement can be met; and meanwhile, the model selection standard of core components is defined, the problem of insufficient flexibility of integrated design is avoided, and the working consistency and reliability of equipment are guaranteed.

Description

Technical Field

[0001] This invention belongs to the field of medical device technology, specifically relating to a split-type flexible wearable sleep breathing monitoring system suitable for home, primary healthcare and ICU scenarios. Background Technology

[0002] Sleep-disordered breathing (SDB), a chronic disease prevalent among the elderly, affects approximately 176 million people in my country, yet its clinical diagnosis rate is less than 10%, highlighting the serious challenge of disease prevention and control. Traditional diagnostic methods rely on polysomnography (PSG) devices, which require deployment of over 20 leads in hospitals. This process is complex, uncomfortable, and costly (300,000-1,000,000 RMB), hindering widespread adoption in primary healthcare institutions and home settings, severely limiting early screening and intervention for SDB. Existing wearable monitoring devices are not adapted for ICU settings, failing to meet the asynchronous monitoring needs of patients on non-invasive / invasive ventilators. ICU patients are critically ill with unstable respiratory rhythms and require simultaneous connection to multiple devices such as ventilators and monitors. Existing devices suffer from incompatible communication protocols, weak anti-interference capabilities, insufficient biocompatibility (unable to adapt to patients with broken skin), and algorithms not optimized for ICU patients, resulting in low accuracy in asynchronous monitoring and hindering clinical adjustments to ventilator parameters.

[0003] While existing wearable sleep apnea monitoring devices have overcome some of the limitations of traditional sleep apnea (PSG) devices, they still suffer from several technical shortcomings: First, the sensors are mostly made of rigid materials, causing a strong foreign body sensation when worn and potentially leading to skin discomfort with prolonged monitoring, thus failing to meet the core need for long-term continuous monitoring at home; second, the monitoring parameters are limited, mostly focusing only on respiratory or ECG signals, lacking the ability to simultaneously collect and fuse multimodal data such as respiratory movement, sleep position, body surface temperature, ECG, and PPG, making it difficult to comprehensively assess abnormal states of the respiratory-circulatory system coupling; third, the AI ​​algorithms lack targeted optimization and do not adequately consider the needs of the elderly. The physiological differences among different groups, such as patients with chronic diseases and people living at high altitudes, result in limited accuracy in diagnosis and assessment. Fourth, most devices are integrated designs, lacking flexibility, and the selection of core components lacks clear standards, leading to poor stability and consistency of the equipment and affecting the reliability of clinical applications. Fifth, the host computer functions are incomplete, the data visualization effect is poor, the parameter configuration operation is cumbersome, and there is a lack of unified hardware adaptation standards, making it unable to adapt to the actual application scenarios of primary healthcare and home monitoring. Sixth, the application scenarios are limited, making it difficult to simultaneously cover multiple dimensions of clinical needs such as sleep apnea diagnosis, ventilator-machine asynchronous assessment, and respiratory chronic disease rehabilitation monitoring.

[0004] Furthermore, existing devices cannot fully meet the stringent requirements of clinical monitoring in terms of component biocompatibility, battery life, and data transmission stability. Therefore, it is imperative to develop a split-type wearable sleep apnea monitoring device that combines a flexible and comfortable wearing experience, multi-parameter fusion monitoring capabilities, personalized algorithm adaptation, clear component selection standards, and adaptability to multiple application scenarios, in order to improve the diagnosis rate of SDB, optimize the effect of ventilator therapy, and assist in the rehabilitation management of chronic respiratory diseases. Summary of the Invention

[0005] The purpose of this invention is to overcome the technical defects of existing wearable sleep breathing monitoring devices, which are poor in terms of component biocompatibility, battery life and data transmission stability, and cannot fully meet the strict requirements of clinical monitoring. This invention provides a split-type flexible wearable sleep breathing monitoring device and its monitoring system.

[0006] To achieve the above objectives, the present invention employs the following technical solution: a split-type flexible wearable sleep breathing monitoring system, comprising a host computer, and a flexible respiratory movement monitoring patch, a flexible ECG monitoring patch, and a PPG monitoring module wirelessly connected to the host computer; the flexible respiratory movement monitoring patch consists of a battery, and a strain sensing unit, a temperature sensing module, a three-axis gyroscope, and a communication module connected to the battery, wherein the strain sensing unit, the temperature sensing module, and the three-axis gyroscope are all connected to the communication module; the flexible ECG monitoring patch consists of an MCU, a signal conditioning chip connected to the MCU, and a hydrogel electrode connected to the signal conditioning chip; the PPG monitoring module consists of a signal processing chip, and a light source and a photodetector connected to the signal processing chip.

[0007] Furthermore, the strain sensing unit is composed of a PDMS substrate and a metal conductive layer disposed on the PDMS substrate.

[0008] As a preferred embodiment, the PDMS substrate has a Shore hardness of 10-20A and a thickness of 0.5-2mm; the conductive metal layer is prepared by inkjet printing using conductive inks such as silver paste ink with a preset viscosity or liquid metal with a preset melting point.

[0009] The number of flexible respiratory movement monitoring patches is more than three, and the weight of each flexible respiratory movement monitoring patch is less than 12g, and the adhesion lasts for more than 24 hours.

[0010] As another preferred embodiment of the present invention, the system also includes a cloud-based AI analysis platform, which is used to support multi-user management, encrypted data storage, historical trend analysis, report export, and data interaction with external ICU ventilators. The data collected by the flexible respiratory motion monitoring patch, the flexible ECG monitoring patch, and the PPG monitoring module are transmitted to the cloud-based AI analysis platform via a host computer.

[0011] To better realize the present invention, AI algorithms are also embedded in the cloud AI analysis platform, including a sleep-disorder breathing diagnosis AI algorithm, a ventilator human-machine asynchrony assessment algorithm, and a respiratory chronic disease rehabilitation assessment algorithm.

[0012] The AI ​​algorithm employs 8-bit fixed-point quantization and lightweight deployment, with a local inference latency of ≤80ms on the edge. It also eliminates artifacts through a dual mechanism of three-axis gyroscope motion recognition and ECG / PPG signal quality assessment, achieving robustness of ≥88% in complex environments.

[0013] The communication module supports Bluetooth, Wi-Fi, and hospital intranet communication protocols, enabling real-time data synchronization with ICU ventilators.

[0014] The flexible respiratory movement monitoring patch and the flexible ECG monitoring patch are equipped with a sweat-proof and waterproof encapsulation structure.

[0015] Compared with the prior art, the beneficial effects of the present invention are:

[0016] (1) The strain sensing unit of the present invention adopts a PDMS substrate with a Shore hardness of 10-20A and a thickness of 0.5-2mm and a metal conductive layer with specific parameters. Combined with a lightweight design with a single patch weight of <12g, it has good biocompatibility and no foreign body sensation. The adhesion lasts for ≥24h, which can meet the needs of long-term home monitoring. At the same time, the selection criteria of core components are clearly defined to avoid the problem of insufficient flexibility of integrated design and ensure the consistency and reliability of equipment operation.

[0017] (2) This invention can simultaneously acquire multimodal data such as respiratory movement, electrocardiogram, blood oxygen, body surface temperature, and sleep position through the coordinated acquisition of flexible respiratory movement monitoring patch, ECG monitoring patch and PPG monitoring module, filling the technical gap of single parameter monitoring in existing equipment, and can comprehensively assess the abnormal state of respiratory-circulatory system coupling.

[0018] (3) The cloud AI analysis platform of the present invention embeds three types of core AI algorithms, supports personalized adjustment for healthy elderly, chronic disease comorbidity, and high-altitude populations, and combines lightweight deployment with 8-bit fixed-point quantization and dual-mechanism artifact removal technology. The edge inference latency is ≤80ms, the robustness in complex environments is ≥88%, the accuracy of respiratory event detection and the consistency of human-machine asynchronous assessment are both at a high level, and the diagnostic and rehabilitation assessment results are more clinically valuable.

[0019] (4) The device of the present invention adopts BLE5.3 wireless connection method, and is equipped with a cross-platform host computer and cloud multi-user management function. It supports multiple scenarios such as sleep breathing disorder diagnosis, ventilator human-machine asynchronous assessment, and respiratory chronic disease rehabilitation monitoring. It also simplifies the data visualization, parameter configuration and report export process, and is suitable for different usage scenarios such as home, nursing home, and primary medical institutions, reducing the threshold and cost of medical monitoring.

[0020] (5) This invention is adapted to the ICU scenario and can be precisely connected with non-invasive / invasive ventilators, filling the technological gap of wearable respiratory monitoring devices in the ICU.

[0021] (6) The present invention has strong anti-interference, high biocompatibility and anti-drop design, which meets the complex environment of ICU and the special needs of patients, and improves the accuracy of human-machine asynchronous monitoring to more than 95%. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0023] The reference numerals in the above figures are named as follows:

[0024] 1-Host computer, 2-Flexible respiratory motion monitoring patch, 3-Flexible ECG monitoring patch, 4-PPG monitoring module, 5-Cloud AI analysis platform, 6-ICU ventilator, 21-Battery, 22-Strain sensing unit, 23-Temperature sensing module, 24-Three-axis gyroscope, 25-Communication module, 31-MCU, 32-Signal conditioning chip, 33-Hydrogel electrode, 41-Signal processing chip, 42-Light source, 43-Photodetector. Detailed Implementation

[0025] The technical solutions of 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.

[0026] Example

[0027] like Figure 1As shown, the split-type flexible wearable sleep apnea monitoring system described in this embodiment includes a host computer 1, and a flexible respiratory movement monitoring patch 2, a flexible ECG monitoring patch 3, and a PPG monitoring module 4 that are wirelessly connected to the host computer 1 for data interaction. Furthermore, as an extended application, this embodiment's split-type flexible wearable sleep apnea monitoring system also includes a cloud-based AI analysis platform 5 and an ICU ventilator 6. The host computer 1 connects to the cloud-based AI analysis platform 5 via Wi-Fi (802.11b / g / n) or cellular network (4G / 5G) for data interaction. The ICU ventilator 6 includes both non-invasive and invasive types. The non-invasive ventilator uses a nasal mask / oronasal mask connection, is non-invasive and well-tolerated, and is suitable for mild to moderate respiratory failure and transitional respiratory support scenarios. The invasive ventilator establishes an artificial airway through endotracheal intubation / tracheostomy, is highly invasive, and is suitable for severely ill patients with severe respiratory failure or unconsciousness.

[0028] The host computer 1 is the core control and data hub of this system. Its function is to connect the flexible respiratory motion monitoring patch 2, the flexible ECG monitoring patch 3, and the PPG monitoring module 4 with the cloud-based AI analysis platform 5, realizing full-process control of "data reception-processing-feedback-upload". At the same time, the host computer 1 also supports data interaction with the ICU ventilator 6 and has the function of real-time analysis of human-machine asynchrony.

[0029] The host computer 1 consists of at least a main control chip, a storage module connected to the main control chip, and a display screen. The main control chip is preferably implemented using the Qualcomm Snapdragon W5 Gen2, a system-on-a-chip (SoC) platform for wearable devices launched by Qualcomm, targeting products such as smartwatches and medical wearable devices. It features satellite communication, high-precision positioning, low power consumption, miniaturization, and support for NB-NTN satellite connectivity. The storage module has ≥128GB of storage and supports UFS 3.1; the display screen is preferably a 1.3-1.5 inch AMOLED with a resolution of 320×320.

[0030] The main functions of the host computer 1 are as follows: (1) Wireless communication, that is, receiving the collected data from the flexible respiratory motion patch 2, the flexible ECG monitoring patch 3 and the PPG monitoring module 4 through the BLE 5.3 protocol, and issuing configuration instructions such as sampling rate and alarm threshold; (2) Data processing and visualization, used for real-time analysis of multimodal data, and displaying indicators such as respiratory motion, ECG, PPG, body position and temperature in waveform diagrams, with a response delay of ≤50ms; (3) Function control, that is, providing data storage, CSV / EDF format export, abnormal alarm (sound + vibration + pop-up), and parameter dynamic adjustment functions; (4) Data transfer, that is, uploading the received data to the cloud AI analysis platform 5 through Wi-Fi / 4G / 5G, and supporting local storage and remote synchronization dual modes.

[0031] The flexible respiratory motion monitoring patch 2 is used to accurately capture multi-dimensional data related to breathing, providing a basis for subsequent analysis. There are more than three patches. In actual use, the three flexible respiratory motion monitoring patches 2 are respectively attached to the left thorax (4th-5th intercostal space at the anterior axillary line), the right thorax (4th-5th intercostal space at the anterior axillary line), and below the xiphoid process to achieve comprehensive coverage of respiratory signals.

[0032] The flexible respiratory dynamics monitoring patch 2 specifically comprises a battery 21 and a strain sensing unit 22, a temperature sensing module 23, a three-axis gyroscope 24, and a communication module 25 connected to the battery 21. To ensure the effectiveness of this embodiment, the strain sensing unit 22 is composed of a PDMS substrate and a metal conductive layer disposed on the PDMS substrate. The PDMS substrate has a Shore hardness of 10-20A and a thickness of 0.5-2mm; while the metal conductive layer is prepared by inkjet printing using conductive inks such as silver paste ink with a preset viscosity or liquid metal with a preset melting point.

[0033] As a preferred embodiment, the viscosity of the silver paste ink is 5000-8000 mPa·s, and the melting point of the liquid metal with the preset melting point is <30°C.

[0034] The temperature sensing module 23 preferably uses an NTC thermistor, specifically the MF52-103F3950. Here, MF52 represents the product series, a common industry identifier for NTC thermistors; 103 indicates a nominal resistance of 10kΩ (10×10³Ω) at 25℃; F indicates a resistance tolerance of ±1%, signifying high accuracy; and 3950 represents a B value (temperature coefficient between 25℃ and 85℃) of 3950K, reflecting the sensitivity of the resistance to temperature changes.

[0035] The three-axis gyroscope 24 is implemented using STMicroelectronics LSM6DSO, with a sampling rate of 100Hz and a range of ±2000° / s.

[0036] The communication module 25 supports Bluetooth, Wi-Fi and hospital intranet communication protocols, and can achieve real-time data synchronization with the ICU ventilator 6. The communication module 25 also supports the BLE 5.3 protocol.

[0037] Battery 21 is preferably a lithium polymer battery with a capacity of 100mAh and a voltage of 3.7V. Battery 21 is used to provide power for strain sensing unit 22, temperature sensing module 23, three-axis gyroscope 24 and communication module 25.

[0038] The flexible respiratory motion monitoring patch 2 can collect the following data: respiratory motion signals, that is, the respiratory deformation of the chest and abdomen is captured by the strain sensing unit 22, reflecting the respiratory amplitude and frequency; synchronous acquisition of auxiliary data, that is, the body surface temperature is obtained by the temperature sensing module 23, and the sleep position and body movement are identified by the three-axis gyroscope 24 to eliminate artifact interference; data transmission and adaptation, that is, the multi-dimensional data collected is transmitted to the host computer 1 in real time via the built-in communication module 25 through the BLE 5.3 protocol.

[0039] Each flexible respiratory movement monitoring patch 2 weighs less than 12g and has a adhesion lasting ≥24h, which can meet the continuous monitoring needs in home, rehabilitation and other scenarios and is suitable for long-term monitoring.

[0040] The flexible ECG monitoring patch 3 is a precise ECG signal acquisition unit used to capture high-quality ECG data and assist in the assessment of the respiratory-circulatory system coupling status. The flexible ECG monitoring patch 3 consists of an MCU 31, a signal conditioning chip 32 connected to the MCU 31, and a hydrogel electrode 33 connected to the signal conditioning chip 32. Simultaneously, a miniature lithium polymer battery is also installed inside the flexible ECG monitoring patch 3 to power the MCU 31, signal conditioning chip 32, and hydrogel electrode 33, supporting 24-hour continuous monitoring.

[0041] The hydrogel electrode 33 is preferably implemented using 3M 2228, a commercially available gel-type ECG (electrocardiogram) monitoring electrode from 3M, primarily used for clinical and routine ECG signal acquisition in adults. It features both excellent fit and signal stability. It utilizes an adhesive foam tape backing and a dedicated conductive hydrogel, connected with a stainless steel press-fit joint; its latex-free design ensures biocompatibility with normal skin types, leaving no gel residue on the skin after use.

[0042] The signal conditioning chip 32 is preferably implemented using the TI ADS1292, which has a resolution of 24 bits and a sampling rate of 125Hz. The TI ADS1292 is a high-precision analog front-end signal conditioning chip specifically designed for electrocardiogram (ECG) signals. Its function is to process the weak raw ECG signals collected by the hydrogel electrode 33 into pure and accurate digital signals, which are adapted for subsequent transmission by the MCU 31 and the cloud AI analysis platform 5. The signal conditioning chip 32 has the following features: First, signal amplification and noise reduction. It receives the microvolt (μV) raw ECG signal captured by the hydrogel electrode 33, amplifies the signal amplitude through its built-in instrumentation amplifier, and suppresses power frequency interference (50 / 60Hz) and electromyographic noise, achieving a common-mode rejection ratio of ≥110dB to ensure clear QRS complexes and key waveforms. Second, high-precision acquisition and conversion. 24-bit high resolution ensures no loss of signal details, and the 125Hz sampling rate accurately matches the frequency range of ECG signals (0.5-100Hz), converting analog ECG signals into digital data to provide a precise data source for heart rate variability and respiratory event correlation analysis. Third, low-power adaptation. It supports low-power operating modes and is compatible with the micro battery power supply of the flexible ECG monitoring patch 3, meeting the power requirements for 24-hour continuous monitoring. Fourth, signal adaptation and transmission. It seamlessly connects with the MCU 31 through the SPI / I2C interface, outputting standardized digital signals, laying the foundation for subsequent BLE 5.3 wireless transmission to the host computer 1.

[0043] The MCU 31 is the core control and data processing hub of the flexible ECG monitoring patch 3, and it is primarily implemented using an STM32L476RG. Leveraging its low power consumption and high performance, the STM32L476RG handles the entire process management between the signal conditioning chip 32 and the host computer 1. Its specific functions are as follows: First, data reception and preprocessing: receiving 24-bit high-precision digitized ECG data converted by the signal conditioning chip 32 via the SPI / I2C digital interface, performing data verification, format standardization, and preliminary artifact screening (such as removing obvious electromyographic interference data) to ensure the integrity and validity of the data transmitted to the host computer 1; Second, wireless communication management: integrating BLE 5.3 wireless communication functionality to encrypt and transmit the preprocessed ECG data to the host computer 1; simultaneously receiving and executing commands issued by the host computer 1; Third, low-power power management: the STM32L476RG... The series is an MCU designed by ST for low-power scenarios. It supports multiple low-power modes such as sleep, stop, and standby. It can intelligently manage the power supply of each component in the surface mount and is compatible with surface mount micro lithium polymer battery power supply scenarios, ensuring more than 24 hours of continuous monitoring. Fourth, it features hardware collaborative control, sending precise control commands to the signal conditioning chip 32 and adjusting the working state of the signal conditioning chip 32 in real time to adapt to the ECG signal acquisition needs of different users (such as the elderly and patients with chronic diseases).

[0044] The PPG monitoring module 4 is the core monitoring unit for blood oxygen and heart rate. It consists of a signal processing chip 41, a light source 42, and a photodetector 43 connected to the signal processing chip 41. The PPG monitoring module 4 also incorporates a miniature lithium polymer battery to power the signal processing chip 41, the light source 42, and the photodetector 43, meeting the requirements for long-term wear.

[0045] The PPG monitoring module 4 collects key physiological data based on the photoplethysmography (PPG) principle, providing core support for respiratory-circulatory system assessment. Its specific functions are as follows: First, it accurately collects core indicators. Light source 42 (Osram SFH7050, red + infrared dual light source) emits light, photodetector 43 receives reflected light from subcutaneous blood vessels, and signal processing chip 41 (MaximMAX30102) converts the light signal into an electrical signal to calculate heart rate and blood oxygen saturation, while simultaneously capturing pulse waveform characteristics. Second, it performs multimodal data fusion. The collected data is linked with ECG and respiratory movement data to assist in the identification of sleep apnea events (such as sleep apnea) and heart rate variability analysis, while also participating in dual-mechanism artifact removal (PPG). (Signal quality assessment) to improve overall monitoring accuracy; third, it is suitable for wearable scenarios. It adopts a finger or ring design, weighs less than 5g, is convenient and unrestricted to wear, and is suitable for multiple scenarios such as sleep and rehabilitation training. It has good biocompatibility and does not cause skin discomfort when worn for a long time; fourth, it transmits data in real time. After filtering and digitization by the signal processing chip 41, it is transmitted to the host computer 1 through the BLE 5.3 protocol to provide key data such as blood oxygen stability and heart rate trend for AI algorithms.

[0046] The cloud-based AI analysis platform 5 supports multi-user management (including tiered access permissions for medical staff), encrypted data storage, historical trend analysis, report export, and data integration with the ICU ventilator 6. Data collected by the flexible respiratory motion monitoring patch 2, the flexible ECG monitoring patch 3, and the PPG monitoring module 4 is transmitted to the host computer 1 via the BLE 5.3 protocol. The host computer 1 then uploads the data to the cloud-based AI analysis platform 5 via Wi-Fi (802.11b / g / n) or cellular network (4G / 5G), supporting both local storage and remote synchronization modes.

[0047] In this embodiment, AI algorithms are embedded in the cloud-based AI analysis platform 5. These AI algorithms include a sleep-disorder breathing diagnosis algorithm, a ventilator-human-machine asynchrony assessment algorithm, and a respiratory chronic disease rehabilitation assessment algorithm.

[0048] Among them, the sleep-disordered breathing diagnosis algorithm is built based on a CNN+Transformer hybrid model. It integrates multimodal data such as respiratory dynamics, ECG, PPG, sleep position, and body surface temperature, and can accurately identify respiratory events such as obstructive / central sleep apnea and hypoventilation. It calculates the AHI (apnea-hypopnea index) and the number and type of respiratory events, outputs the lowest blood oxygen saturation value, assists in the graded diagnosis of SDB (sleep-disordered breathing) (mild / moderate / severe), and has a respiratory event detection accuracy of ≥92%. It supports personalized adjustment for healthy elderly people, people with chronic diseases, and people at high altitudes.

[0049] The ventilator-patient asynchrony assessment algorithm synchronously analyzes respiratory dynamics waveforms and ventilator pressure / flow rate data collected by a flexible sensing module. It identifies six typical asynchrony types, including invalid triggering, insufficient inspiratory flow, dual triggering, and delayed switching. The algorithm statistically analyzes their frequency and calculates the asynchrony index (AI), providing data-driven guidance for physicians to adjust ventilator trigger sensitivity, inspiratory flow rate, and inspiratory time. The algorithm's consistency with the clinical gold standard (PSG + ventilator data) is ≥96%. This algorithm is adaptable to both non-invasive and invasive ventilators and can optimize analysis parameters based on the physiological characteristics of ICU patients.

[0050] In this embodiment, the flexible sensing module is a separate assembly consisting of the flexible respiratory motion monitoring patch 2, the flexible ECG monitoring patch 3, and the PPG monitoring module 4.

[0051] The respiratory chronic disease rehabilitation assessment algorithm is designed for patients with chronic respiratory diseases such as COPD and heart failure. It continuously monitors core indicators such as respiratory rate, respiratory amplitude variation coefficient, heart rate variability (HRV), and blood oxygen saturation stability. Combined with the patient's rehabilitation training plan, it generates a weekly / monthly rehabilitation effect score (0-100 points) and indicator trend report, which intuitively reflects the rehabilitation progress and provides guidance for rehabilitation therapists to adjust the intensity and frequency of training. The score has a consistency of ≥85% with clinical rehabilitation assessment.

[0052] To ensure practical effectiveness, the AI ​​algorithm described above employs 8-bit fixed-point quantization and lightweight deployment, with local inference latency on the edge ≤80ms. It also eliminates artifacts through a dual mechanism of three-axis gyroscope motion recognition and ECG / PPG signal quality assessment, achieving robustness ≥88% in complex environments.

[0053] Among them, 8-bit fixed-point quantization and lightweight deployment refer to "compressing and optimizing" the originally complex AI model, using 8-bit binary numbers to store data (replacing high-precision floating-point data), making the model smaller, consuming less computing power / power, and able to run directly on the device without relying on the cloud, thus achieving "local rapid analysis".

[0054] The local inference latency on the device side is ≤80ms, where the device side refers to the local location of the host computer 1 (wearable terminal / mobile phone). "Inference latency" refers to the time from receiving sensor data to the AI ​​calculating the result. With the above settings, the host computer 1 can instantly identify and trigger alarms without missing critical anomalies due to latency.

[0055] The aforementioned dual-mechanism artifact removal mechanism refers to interference data (such as sensor displacement caused by turning over in sleep, or noise generated by poor electrode contact). The dual mechanism refers to dual filtering: first, the three-axis gyroscope identifies body movements (such as data during turning over is marked as invalid), and then the quality of the ECG / PPG signal is evaluated (such as removing signals that are too messy), to ensure that the data input to the AI ​​is pure and valid, and to avoid misjudgment.

[0056] Robustness in complex environments ≥88%, where robustness refers to the algorithm's ability to resist interference. When robustness is ≥88%, it means that even in complex scenarios such as turning over in sleep, sweating, or slight sensor displacement, the algorithm still has at least an 88% probability of outputting accurate results, so as to ensure that the diagnosis and evaluation results are still reliable when monitoring sleep (the user will not deliberately remain still).

[0057] Based on the above structure, the split-type flexible wearable sleep breathing monitoring system of this embodiment has the following application scenarios.

[0058] Firstly, sleep apnea diagnosis: In nursing homes or home environments, users wear a flexible respiratory movement monitoring patch 2, a flexible ECG monitoring patch 3, and a PPG monitoring module 4 for 1-3 consecutive nights. The system outputs diagnostic indicators such as the AHI index, respiratory event type and frequency, and lowest blood oxygen saturation value to assist clinicians in grading SDB (mild: AHI 5-15, moderate: 15-30, severe: >30).

[0059] Secondly, assessment of ventilator-patient asynchrony: The mechanically ventilated patient wears a flexible respiratory motion monitoring patch 2, a flexible ECG monitoring patch 3, and a PPG monitoring module 4. The host computer 1 synchronizes the ventilator pressure / flow rate data via Bluetooth. The system identifies the type and frequency of asynchrony in real time and outputs the asynchrony index (AI), providing a basis for doctors to adjust ventilator parameters (trigger sensitivity, inspiratory flow rate, inspiratory time).

[0060] Third, respiratory chronic disease rehabilitation monitoring: During the rehabilitation period of patients with chronic obstructive pulmonary disease (COPD), heart failure, etc., flexible respiratory motion monitoring patch 2, flexible ECG monitoring patch 3 and PPG monitoring module 4 are worn for 2-3 nights per week. The system generates trend reports of indicators such as respiratory rate, amplitude, heart rate variability and blood oxygen stability to guide rehabilitation therapists to adjust training programs (such as respiratory training intensity and frequency).

[0061] Fourth, in the ICU invasive ventilator patient monitoring scenario: at this time, the flexible respiratory motion patch 2 is attached to the patient's chest (avoiding drainage tubes and monitor electrodes) and fixed with medical-grade anti-fall-off tape; the flexible ECG monitoring patch 3 is attached below the clavicle (compatible with skin dressings); the PPG monitoring module 4 adopts a finger sleeve design and is fixed to the finger that is not receiving intravenous fluids.

[0062] During operation, the host computer 1 first establishes a connection with the invasive ventilator through the hospital intranet, synchronizing airway pressure (range: 0-80cmH2O), tidal volume (50-1500mL), and positive end-expiratory pressure (0-20cmH2O) data. Then, the cloud-based AI analysis platform 5 calls the ICU-specific human-machine asynchrony algorithm, integrating respiratory dynamics, ECG, PPG data and ventilator parameters to identify three types of asynchrony unique to invasive ventilation: high-pressure airway triggering, dual triggering, and delayed switching, with an inference delay of ≤60ms. Finally, the host computer 1 displays the asynchrony type and frequency in real time and pushes it synchronously to the ICU nurse station monitoring screen, triggering audible and visual alarms when abnormalities occur (e.g., dual triggering frequency ≥5 times / hour).

[0063] As described above, the present invention can be well implemented.

Claims

1. A split-type flexible wearable sleep breathing monitoring system, characterized in that, The system includes a host computer (1), a flexible respiratory motion monitoring patch (2), a flexible ECG monitoring patch (3), and a PPG monitoring module (4) connected to the host computer (1) wirelessly. The flexible respiratory motion monitoring patch (2) consists of a battery (21), a strain sensing unit (22), a temperature sensing module (23), a three-axis gyroscope (24), and a communication module (25) connected to the battery (21). The strain sensing unit (22), the temperature sensing module (23), and the three-axis gyroscope (24) are all connected to the communication module (25). The flexible ECG monitoring patch (3) consists of an MCU (31), a signal conditioning chip (32) connected to the MCU (31), and a hydrogel electrode (33) connected to the signal conditioning chip (32). The PPG monitoring module (4) consists of a signal processing chip (41), a light source (42), and a photodetector (43) connected to the signal processing chip (41).

2. The split-type flexible wearable sleep breathing monitoring system according to claim 1, characterized in that, The strain sensing unit (22) is composed of a PDMS substrate and a metal conductive layer disposed on the PDMS substrate.

3. The split-type flexible wearable sleep breathing monitoring system according to claim 2, characterized in that, The PDMS substrate has a Shore hardness of 10-20A and a thickness of 0.5-2mm; the conductive metal layer is prepared by inkjet printing using conductive inks such as silver paste ink with a preset viscosity or liquid metal with a preset melting point.

4. The split-type flexible wearable sleep breathing monitoring system according to claim 1, characterized in that, The number of flexible respiratory motion monitoring patches (2) is more than 3, and the weight of each flexible respiratory motion monitoring patch (2) is <12g and the adhesion lasts for ≥24h.

5. A split-type flexible wearable sleep breathing monitoring system according to any one of claims 1 to 4, characterized in that, The system also includes a cloud AI analysis platform (5), which is used to support multi-user management, encrypted data storage, historical trend analysis, report export and data interaction with the ICU ventilator (6) outside the system. The data collected by the flexible respiratory motion monitoring patch (2), flexible ECG monitoring patch (3) and PPG monitoring module (4) are transmitted to the cloud AI analysis platform (5) after passing through the host computer (1).

6. A split-type flexible wearable sleep breathing monitoring system according to claim 5, characterized in that, AI algorithms are also embedded in the cloud-based AI analysis platform (5), including a sleep-disordered breathing diagnosis AI algorithm, a ventilator-human-machine asynchrony assessment algorithm, and a respiratory chronic disease rehabilitation assessment algorithm.

7. A split-type flexible wearable sleep breathing monitoring system according to claim 6, characterized in that, The AI ​​algorithm employs 8-bit fixed-point quantization and lightweight deployment, with a local inference latency of ≤80ms on the edge. It also eliminates artifacts through a dual mechanism of three-axis gyroscope motion recognition and ECG / PPG signal quality assessment, achieving robustness of ≥88% in complex environments.

8. A split-type flexible wearable sleep breathing monitoring system according to claim 6, characterized in that, The communication module (25) supports Bluetooth, Wi-Fi and hospital intranet communication protocols, and can achieve real-time data synchronization with the ICU ventilator (6).

9. A split-type flexible wearable sleep breathing monitoring system according to claim 5, characterized in that, The flexible respiratory movement monitoring patch (2) and the flexible ECG monitoring patch (3) are equipped with a sweat-proof and waterproof encapsulation structure.

Images