A brachial artery pulse waveform continuous monitoring system and layered adaptive working method

Abstract

The application discloses a brachial artery pulse waveform continuous monitoring system and a layered self-adaptive working method. The hardware is matched with a double-layer software architecture, the hardware is loaded with a selectable six-axis IMU, full-automatic / manual baseline calibration working conditions are realized, static pressure and micro-pressure dual-range parallel sampling are set, switching noise and waveform frame breaking are eliminated, an automatic baseline and a manual baseline peak value ratio fusion correction are carried out in the early morning, a five-minute sliding baseline is matched, a factory template cold start transition is inhibited, baseline drift is inhibited, software is divided into home waveform follow-up and clinical relative pressure quantitative double modes, and no systolic pressure and diastolic pressure values are output in the whole process. The system is loaded with adaptive intermittent low-power consumption sampling, multi-channel waveform fusion, hot self-checking, physiological fluctuation layered early warning and RTC+network double time sequence guarantee, is matched with a hospital PC terminal to complete waveform analysis and scientific research archiving, and is suitable for various scenes.

Description

Technical Field

[0001] This invention belongs to the field of non-invasive wearable medical devices and dynamic monitoring technology of cardiovascular hemodynamics, specifically involving a brachial artery pulse waveform continuous monitoring system and a hierarchical adaptive working method. Background Technology

[0002] Currently, clinical and home-use non-invasive cardiovascular monitoring devices are mainly divided into three categories: cuff-type ambulatory blood pressure devices, optical PPG wearable devices, and cuffless blood pressure measurement devices. All types of devices have significant technical defects and application shortcomings, and cannot meet the needs of long-term continuous monitoring, wearing comfort, low population adaptation threshold, and low registration and implementation costs.

[0003] Traditional inflatable cuff-type ambulatory blood pressure monitoring devices rely on airbag inflation to obtain blood pressure values. Prolonged wear can easily cause local pressure and soreness in the upper arm, making it impossible to achieve continuous monitoring throughout the night. At the same time, they use an intermittent single-point measurement mode with long measurement intervals, which cannot capture short-term and hidden abnormal blood flow signals such as transient hypoperfusion at night, paroxysmal vasospasm, and periodic vascular fluctuations during sleep apnea. The monitoring data is fragmented and difficult to support long-term trend assessment of cardiovascular chronic diseases.

[0004] Wearable optical PPG devices rely on the photoplethysmography principle to collect pulse signals. Although they can achieve continuous monitoring, the signal depends on the light transmittance of peripheral blood flow. In people with dark skin, diabetic patients with vascular calcification, elderly patients with arteriosclerosis, and patients with low peripheral perfusion, the optical signal is significantly attenuated and the waveform is severely distorted, resulting in a sharp decline in monitoring effectiveness.

[0005] The core of existing cuffless non-invasive monitoring solutions focuses on measuring absolute blood pressure values ​​such as SBP and DBP. To meet international medical blood pressure accuracy standards such as AAMI and BHS, stringent wearing and fixing structures, complex individual calibration algorithms, and large-scale multi-center clinical trials are required. This results in high technical barriers, high R&D costs, long medical device registration cycles, and extremely high difficulty in implementation.

[0006] Existing wearable devices based solely on waveform monitoring also suffer from numerous engineering and algorithmic shortcomings: Their baseline acquisition modes are limited, relying solely on automatic recognition or manual calibration, failing to meet the diverse needs of low-cost, posture-free sensing devices and high-precision automatic monitoring devices; the use of only a single morning static baseline makes them highly susceptible to baseline drift caused by daytime vascular tension changes, sleep position shifts, and slight wearing drift, resulting in numerous false positives; the lack of an automatic / manual dual-baseline fusion correction mechanism leads to flawed baseline arbitration logic and insufficient benchmark accuracy; the use of a fixed number of channels for fusion in multi-channel sensor arrays results in poor adaptability and makes them easily circumvented by competitors; and the measurement range switching uses a direct... The method of switching sensing ranges is prone to impact noise and waveform frame breaks; there is no cold start initial baseline mechanism, and the algorithm cannot run stably in the early stage of power-on; there is no range saturation self-recovery mechanism, and the system is prone to failure under extreme contact pressure; the uninterrupted continuous sampling mode is generally used, resulting in high device power consumption and inability to support 24-hour long-term continuous monitoring; it cannot effectively distinguish between waveform distortion caused by limb movement and poor device contact and true pathological blood perfusion attenuation, and the device has difficulty distinguishing between motion artifacts and pathological blood flow abnormalities, resulting in an excessive number of abnormality prompts; there is no local hardware clock fallback mechanism, and the waveform has no effective timing mark after network disconnection, making it impossible to carry out long-term longitudinal trend comparison analysis.

[0007] In summary, existing technologies lack a complete brachial artery continuous waveform monitoring solution that features no pressurized cuff, layered hardware and software adaptation, selectable dual-condition baseline calibration for the IMU, intelligent dual-baseline fusion correction, impact-free range switching, low-power intermittent sampling, accurate differentiation between artifacts and pathological fluctuations, and a complete fault-tolerant self-recovery mechanism. This makes it difficult to simultaneously meet the diverse needs of home-based low-cost and simple follow-up, in-hospital high-precision assessment, and clinical research big data collection, thus hindering the widespread adoption and clinical application of non-invasive continuous cardiovascular monitoring technology. Summary of the Invention

[0008] To address the industry pain points of existing monitoring devices, such as discomfort when worn, signal distortion, baseline drift, single baseline calibration mode, defects in baseline arbitration logic, distortion during range switching, failure during cold start, high false alarm rate, insufficient battery life, high registration threshold, limited scenario adaptability, and susceptibility to being circumvented by competitors, this invention provides a brachial artery pulse waveform continuous monitoring system and a hierarchical adaptive working method.

[0009] This invention employs an innovative architecture combining general-purpose hardware with layered software algorithms. It features optional IMU configuration, adapting to both manual static calibration without an IMU and automatic static calibration with an IMU, balancing low-cost mass production with high-precision automated monitoring needs. It enables simple home waveform follow-up and in-hospital quantitative vascular assessment without hardware modifications. A local + network multi-mode timing assurance system and a bidirectional fusion correction mechanism combining automatic nighttime baseline and manual standard baseline are constructed, clearly defining peak scaling correction logic. Combined with short-term sliding dynamic baseline and cold-start transition baseline, it completely solves the baseline drift problem throughout the entire time period. An innovative dual-range parallel sampling shockless switching technology solves the waveform distortion problem during range switching. Through dynamic multi-channel optimization, range saturation self-recovery, and multi-level layered early warning, it accurately distinguishes interference signals from pathological abnormalities. It does not output absolute blood pressure values ​​throughout the entire process, circumventing stringent accuracy standards and clinical review thresholds for blood pressure devices. While ensuring the clinical value of monitoring, it significantly reduces product development, registration, and deployment costs.

[0010] This invention discloses a continuous monitoring system for brachial artery pulse waveforms, comprising a flexible MEMS patch hardware terminal, mobile monitoring software, and a hospital PC clinical analysis terminal. The hardware operates entirely without an airbag, air pump, or pressure cuff, and is encapsulated in ultra-thin, flexible medical materials. It achieves high-precision acquisition of brachial artery pulse signals through a 3×3 MEMS piezoresistive sensor array. A six-axis IMU attitude sensor is an optional hardware configuration to adapt to different monitoring scenarios and cost requirements. The hardware undergoes factory calibration for consistency across all sensor units and features a dual-range parallel sample-and-hold circuit. Two ranges are independently acquired, with one range selected for output. Combined with end-diastolic smooth switching logic, this completely solves the physical problems of high static pressure masking weak pulse signals and switching noise distortion. The hardware locally embeds a complete set of waveform preprocessing, dynamic channel fusion, artifact recognition, and pressure detection algorithms to optimize the raw signal beforehand.

[0011] At the software level, it innovatively sets up dual independent working modes, allowing a single set of general-purpose hardware to switch usage scenarios without structural modifications; it innovates the automatic baseline + manual baseline fusion correction arbitration logic, abandoning the simple two-choice mode, and explicitly uses a peak amplitude ratio coefficient to complete the baseline proportional scaling correction, taking into account both physiological baseline and daytime calibration accuracy; it adapts to optional IMU hardware architectures, distinguishing between fully automatic static identification baseline acquisition and manual active static baseline acquisition; it is equipped with a cold start temporary baseline and a dynamic sliding baseline all-time baseline guarantee mechanism; it adopts an adaptive intermittent periodic sampling strategy to balance monitoring accuracy and device endurance; it uses a dual mechanism of array bonding thermal self-check and attitude detection to eliminate invalid sampling samples from the source; it has a built-in periodic physiological fluctuation feature library to accurately distinguish between physiological fluctuations and pathological abnormalities; it sets up multi-level hierarchical early warning logic to significantly reduce the false alarm rate; it relies on the RTC local clock as a backup + multi-mode network time synchronization dual time sequence mechanism to ensure the integrity and validity of time sequence data in the event of network outages; and it dynamically selects effective acquisition channels to avoid the adaptation defects of a fixed number of channels and mitigate risks.

[0012] The hospital's PC clinical terminal enables batch waveform playback, quantitative analysis of hemodynamic parameters, long-term baseline longitudinal comparison, standardized follow-up report output, and batch archiving of research data, adapting to the needs of hospital clinical diagnosis and treatment and cohort research.

[0013] This invention includes a hierarchical adaptive monitoring method that adapts to both IMU-equipped and non-equipped hardware operating conditions. It sequentially completes the entire process of mode selection, time-series synchronization, dual-condition baseline acquisition, baseline fusion correction, intermittent sampling, waveform preprocessing, artifact filtering, feature comparison, graded early warning, long-term trend analysis, and hierarchical differential calculation. Throughout the process, only individual longitudinal time-series waveform comparisons are performed; blood pressure values ​​are not measured or output. This approach balances compliance, stability, and clinical applicability, and possesses a complete fault tolerance, self-recovery, and adaptive mechanism.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows.

[0015] 1. Hardware signal acquisition performance advantages (1) This invention sets up two parallel sampling and holding circuits for static pressure (0~10kPa) and micro pressure (0~2kPa). The two channels synchronously latch the signal throughout the entire process, and only select the end after the pulse reaches a stable end-diastolic window. This eliminates the transient noise of capacitor charging and discharging and waveform frame breakage distortion caused by traditional single-range switching. It can synchronously and completely acquire the static pressure of the upper arm and the weak pulse fluctuation signal at the 50Pa level during diastole, and completely preserve all the subtle waveform features of the pulse rising limb, dicrotic notch, and diastolic attenuation. (2) The present invention adopts a 3×3 multi-channel dynamic optimization waveform fusion mechanism to calculate the signal-to-noise ratio and pulsation correlation of each channel in real time, without fixing the number of effective acquisition channels; it is suitable for scenarios such as wearing displacement and differences in arm thickness, while avoiding the design loophole of competitors simply circumventing the patent by fixing the number of channels; (3) The present invention is configured with range saturation self-recovery logic. When the static bonding pressure exceeds the upper limit of 10 kPa, the device automatically enters the silent acquisition pause state. After the pressure drops back to the effective range, the acquisition is automatically restarted to avoid signal saturation cutoff and acquisition interruption, and improve the device operation stability under extreme conditions such as strenuous exercise and muscle congestion. (4) The signal link is equipped with a 24-bit Σ-Δ analog-to-digital converter and a low-noise instrumentation amplifier. The hardware ENOB is ≥19-bit, and it is equipped with a 0.5Hz~20Hz physiological-specific bandpass filter and a 50Hz power frequency notch filter. The weak pulse signal recognition capability is better than that of conventional single-range MEMS acquisition equipment.

[0016] 2. Advantages of baseline correction and graded early warning algorithms (1) The present invention constructs an automatic baseline + artificial baseline peak ratio fusion mechanism in the early morning, using the automatic baseline that is undisturbed while lying flat at night as the waveform morphology benchmark, and uses K = artificial peak / automatic peak ratio coefficient to scale the benchmark waveform as a whole; unlike the traditional single baseline two-choice arbitration mode, it simultaneously preserves the basic waveform morphology of human physiology and the calibration amplitude during the day, eliminating the baseline drift caused by the deviation of vascular tension and strap tightness during day and night. (2) The present invention sets a cold start factory template transition baseline. When there is no personalized sampling data in the first five minutes before the equipment is powered on, waveform comparison can be completed. After accumulating sufficient data, the five-minute dynamic sliding baseline is automatically switched to eliminate the algorithm blank period in the power-on stage of the equipment and realize the monitoring of the equipment at all times. (3) The present invention has a built-in motion artifact recognition + sleep apnea physiological fluctuation feature library dual filtering logic, which can automatically distinguish limb movement distortion, physiological periodic blood flow fluctuation, and continuous pathological blood flow attenuation; single waveform abnormalities are only stored locally, and only long-term continuous waveform degradation outputs follow-up prompts, reducing invalid abnormal records.

[0017] 3. Advantages of generalized hardware architecture and market adaptability (1) The hardware of this invention supports optional six-axis IMU attitude sensor: low-cost models without IMU rely on the user to manually complete the baseline acquisition while resting, and high-end models with IMU can automatically identify the lying still posture; a single hardware structure can cover the two product gradients of civilian home and hospital clinical use without mold modification, reducing the cost of R&D, mold opening and mass production of multiple models; (2) The whole device uses a flexible silicone patch with a thickness of <450μm without airbags. The weight of the whole device is ≤7g. There is no rigid pressure structure, which can be worn continuously all night. With the adaptive intermittent sampling strategy, the single sampling time is 30s for 2min during the day and the single sampling time is extended at night. The 120mAh lithium battery can support 24 hours of intermittent monitoring and reduce the average power consumption of the whole device.

[0018] 4. Advantages of Medical Device Registration and Compliance (1) This system only compares the morphology of a single individual's longitudinal pulse waveform throughout the entire process. Neither the hardware nor the software calculates or outputs the absolute values ​​of systolic and diastolic blood pressure and blood pressure units. It does not need to meet the AAMI and BHS blood pressure measurement accuracy standards, nor does it need to conduct large-scale human clinical trials corresponding to blood pressure devices, thus shortening the registration cycle of Class II medical devices and reducing the cost of clinical validation. (2) The system only outputs relative quantitative indicators of vascular elasticity and peripheral perfusion, and does not generate disease diagnosis conclusions. It is only used as auxiliary reference data for clinical follow-up to avoid the review restrictions of medical devices for disease diagnosis. (3) Only the longitudinal comparison of individual time-series waveforms is used, without establishing a unified judgment threshold across populations, thus eliminating the assessment bias caused by differences in age, physical condition, and vascular basis, resulting in higher consistency in blood flow status assessment results.

[0019] 5. Advantages of time-series data storage and patent protection (1) The present invention adopts a dual timing architecture of RTC local independent clock + NTP / GPS / BeiDou multi-network timing. The waveform is still bound to a valid timestamp in the state of network disconnection, and the long-term longitudinal blood flow trend data is complete and traceable. (2) All automatic baselines, manual baselines and raw waveform data are encrypted and permanently stored. The device will not automatically overwrite or delete historical reference waveforms. Long-term follow-up can retrieve baselines from many years for longitudinal comparison. (3) The dual working mode of hardware and software layering, dynamic multi-channel fusion, dual baseline ratio fusion, and parallel impact-free sampling form a multi-layered technical innovation barrier. Competitors can not fully reproduce all the monitoring functions of this invention by only deleting a single module, and the patent protection is more stable.

[0020] The term "impactless switching" refers to the fact that the dual-range parallel sample-and-hold circuit completely avoids two types of signal degradation caused by traditional single-range switching at the physical level: First, by using two independent sample-and-hold circuits to latch the static and micro-voltage signals respectively, the back-end analog switch only selects the voltage that has been stabilized and held, and the ADC input is not affected by the transient charging and discharging of the parasitic capacitance of the front-end analog switch; Second, through pulse cycle synchronization control, the channel switching action is strictly constrained to the stable waveform window at the end of diastole, avoiding key characteristic periods such as the rising branch and the contraction peak, thereby eliminating waveform frame breaks and feature point truncation. The two front-end samplings are uninterrupted throughout the entire process, with only the back-end selection logic switching, achieving truly blind-zone-free and transient noise-free switching. Attached Figure Description

[0021] Figure 1 Overall system architecture block diagram of this invention; Figure 2 A schematic diagram of a flexible MEMS 3×3 sensor array patch structure; Figure 3 This is a schematic diagram of a hierarchical adaptive complete workflow; Figure 4 Block diagram for dual baseline peak ratio fusion correction logic; Figure 5 This is a timing logic diagram for dual-range parallel shockless switching.

[0022] Among them, 1-flexible substrate, 2-3×3 sensor array, 3-connection line, 4-sensor unit. Detailed Implementation

[0023] Example 1: Specific Implementation of Flexible MEMS Patch Hardware Terminal This invention's flexible MEMS patch hardware terminal uses medical-grade Ecoflex layered flexible silicone with a thickness of <450μm for encapsulation. The entire device weighs ≤7g, and it has no air pump, airbag, or rigid compression structure. Combined with an elastic flexible strap, it can adaptively fit users with different upper arm circumferences. The core sensing area uses a 3×3 MEMS piezoresistive sensor array, with a single-point sensing size of 6mm×6mm and an overall array size of 38mm×14mm, accurately covering the core monitoring area of ​​the brachial artery and ensuring complete signal acquisition. A six-axis IMU attitude sensor is optional and can be selectively installed according to product positioning, reducing the hardware cost of entry-level devices.

[0024] Each MEMS sensor unit is independently configured with a voltage divider calibration circuit. Before the device leaves the factory, the sensitivity, temperature drift coefficient, and zero-point offset parameters of each channel are calibrated. The calibration parameters are stored in the main control MCU. The device automatically loads the compensation parameters upon power-up to offset the ±15% individual consistency deviation of the MEMS units. The hardware has a built-in NTC miniature temperature measurement unit, which achieves accurate temperature drift compensation within the entire operating temperature range of 5℃ to 45℃, ensuring the stability of signal acquisition under different ambient temperatures.

[0025] The hardware configuration features a dual-range parallel sample-and-hold circuit with two independent signal acquisition links. The static bonding pressure range is 0~10kPa, which monitors the tightness of the patch on the upper arm and the local pressure distribution in real time. The dynamic pulse micro-pressure range is 0~2kPa, specifically amplifying and capturing weak pulse fluctuation signals at the 50Pa level during diastole. The two ranges are sampled synchronously and latched separately. The backend uses an analog switch to select the channel before sending it to the ADC, rather than directly switching the sensor unit range. Furthermore, the channel switching action is strictly controlled by the pulse cycle, and switching is only performed during the time window when the waveform is most stable and the signal fluctuation is minimal at the end of diastole, completely avoiding the capacitor charging and discharging impact noise, waveform frame breaks, and signal distortion problems of traditional switching methods.

[0026] A range saturation self-recovery mechanism is set up for extreme working conditions: When the user engages in strenuous exercise or the upper arm muscles become congested, causing the static bonding pressure to continuously exceed the upper limit of the 10kPa range, the system actively enters a safe silent mode, suspends waveform acquisition, and guides the user to readjust the patch wearing position and tightness through the APP. After the bonding pressure drops back to the effective range of 0~10kPa, the device automatically restarts the monitoring function to avoid signal saturation truncation, data abnormalities, and equipment signal saturation acquisition stagnation, ensuring the long-term stable operation of the system.

[0027] The signal acquisition link uses a 24-bit Σ-Δ analog-to-digital converter paired with a pre-amplifier with low noise, providing excellent weak signal recognition capability. The measured effective number of bits (ENOB) is ≥19 bits. The hardware has built-in dedicated bandpass filtering for physiological signals from 0.5Hz to 20Hz and 50Hz power frequency notch filtering to accurately filter out high-frequency environmental noise and power frequency interference. The hardware has a fixed sampling frequency of 200Hz, which fully preserves the core details of the pulse waveform, such as the rising limb, dicrotic notch, and diastolic attenuation.

[0028] The device features a built-in optional six-axis IMU attitude sensor, an independent local RTC hardware clock, a BLE5.3 low-power wireless transmission module, and a 120mAh micro-pack lithium battery. The RTC clock operates independently and continuously, automatically synchronizing with network standard time such as NTP, GPS, and BeiDou when connected to the network. It can also independently complete time stamping in the absence of network access, ensuring the integrity and validity of all monitoring data. The BLE5.3 module enables low-latency, low-power wireless data transmission. When fully charged, it can stably support 24 hours of uninterrupted sampling and monitoring.

[0029] The hardware-locally embedded array consistency correction, multi-channel dynamic waveform fusion optimization, five-minute sliding baseline correction, and bonding pressure calculation algorithms enable real-time preprocessing of raw signals, reducing the computational burden on mobile devices and improving data real-time performance and accuracy. The hardware version with an IMU additionally supports motion posture recognition and automatic stillness determination algorithms, while the version without an IMU retains only basic signal processing algorithms to suit differentiated product positioning.

[0030] In this embodiment, the 3×3 MEMS piezoresistive sensing array includes two types of range sensing units. The preferred arrangement is four range A (0~10kPa) hydrostatic MEMS units arranged at the four corners of the array, and five range B (0~2kPa) micropulse MEMS units arranged in the central cross region. The array row and column arrangement is as follows: First row: Unit A, Unit B, Unit A Second row: Unit B, Unit B, Unit B Third row: Unit A, Unit B, Unit A The four corner A units are used to collect static bonding pressure across the entire patch area, generate bonding thermal distribution maps, and monitor pressure overload status; the five central B units cover the brachial artery region, and the dynamic multi-channel algorithm optimizes and fuses waveforms to improve the fidelity of weak pulse signals.

[0031] In addition to the preferred 4A+5B arrangement mentioned above, the array of the present invention is not limited to this ratio, and other arrangement schemes such as 5A+4B and 8A+1 B can also be used. As long as the 3×3 array is configured with two types of MEMS channels, namely static pressure and micro pressure, and two independent parallel sampling channels, it falls within the protection scope of the present invention.

[0032] Example 2: Implementation of Mobile Monitoring Software

[0033] After the device successfully connects via Bluetooth, the mobile software automatically synchronizes the local RTC clock with the network standard time. All monitored waveforms and baseline data are bound to a unique and valid timestamp, supporting one-click switching between dual modes to adapt to different usage scenarios.

[0034] In pure waveform follow-up mode, the system internally retains the original pressure amplitude signal for low-level calculations of waveform normalization and morphology correction. It does not output or display any physical pressure amplitude or unit; it only normalizes and outputs the pulse waveform outline. The system automatically eliminates the overall amplitude scaling of the waveform caused by the user's slight daily limb movements and minor changes in the tightness of the bandage through dynamic baseline normalization, recognizing only substantial changes in waveform outline, diastolic decay pattern, and pulsatility rhythm. This mode is suitable for simple health follow-up at home, before and after exercise, and throughout the night.

[0035] In the relative pressure quantitative clinical mode, the system calls the factory-fixed calibration coefficients and, combined with the user's standardized resting baseline, calculates the standardized relative pressure difference between the peak and trough. It automatically extracts quantitative hemodynamic parameters such as pulse rise duration, dicrotic wave relative height ratio, diastolic attenuation index, PEP ejection time, and pulse rate variability. All parameters are only used by clinicians to assist in assessing vascular elasticity and peripheral perfusion status. The software interface does not display blood pressure-related values ​​and units such as SBP, DBP, and mmHg throughout the process.

[0036] The baseline acquisition and fusion correction logic adapts to dual hardware conditions: Baseline acquisition is automatically triggered at 1:00 AM daily. Devices equipped with an IMU continuously monitor human posture for five minutes. After confirming that the user is lying flat without significant turning over or arm raising, a 30-second standard waveform is acquired and stored as the automatic baseline. If the posture does not meet the requirements, acquisition is abandoned. Devices without an IMU prompt the user to remain lying flat and still via software push notifications. Baseline acquisition is completed after the user confirms stillness. Users can manually trigger the calibration process at any time. A pop-up window guides the user to sit still with upper limbs flat for five minutes to complete a 30-second standardized manual baseline acquisition. Multiple acquisitions are supported per day. The system automatically filters and removes discrete, distorted, and abnormal waveforms, retaining the most consistent valid manual baseline. To address the dual baseline deviation caused by differences in physiological state between day and night, the system abandons the simple binary choice logic. Using the automatic baseline at dawn as the waveform skeleton, it extracts the peak amplitude ratio coefficient between the manual and automatic baselines and performs an overall proportional scaling correction on the automatic baseline. This preserves the true physiological waveform structure while correcting for wear fit deviations during the day, generating a precise fused baseline. All original baseline data is permanently encrypted and archived, and is not automatically deleted or overwritten.

[0037] Cold start baseline transition mechanism: During the first five minutes after the device is powered on for the first time, after being reset while wearing the device, or after a long period of shutdown and restart, the system does not have sufficient personalized sliding baseline data. At this time, the factory-built-in general standard pulse waveform template is used as a temporary transition baseline to ensure that the device can be compared and run normally as soon as it is powered on. After the device accumulates five minutes of valid personalized monitoring data, it automatically switches to the user-exclusive five-minute adaptive sliding baseline to eliminate the algorithm gap period and achieve stable monitoring at all times.

[0038] Dynamic channel fusion and fit self-test logic: The system dynamically calculates the signal-to-noise ratio and pulsation correlation of the 9 sensor channels in real time. It does not fix the number of effective channels and dynamically selects at least one of the best high-quality channels to reconstruct the waveform. When the number of effective high-quality channels is less than two and the waveform signal-to-noise ratio does not meet the standard, the system actively pushes wearing posture adjustment and patch fit self-test prompts to guide the user to re-fix the patch and ensure the waveform acquisition quality.

[0039] Intermittent sampling and preprocessing operation logic: During the daytime, sampling is triggered every two minutes, with each sample acquiring a 30-second valid waveform; during the sleep period from midnight to 6 AM, the sampling duration is extended to 60 seconds to improve the stability of weak signal acquisition at night. Before each sampling, the system automatically detects the contact pressure of the 9-channel sensor units, generates a visual heatmap, and identifies localized suspension, overpressure, and poor fit. In case of abnormalities, the sample is directly discarded, and the user is prompted to adjust the fit. After acquisition, the system automatically performs array deviation compensation, multi-channel dynamic optimal waveform fusion, and five-minute sliding baseline normalization to completely eliminate signal interference caused by hardware errors, wear drift, and physiological vascular tension fluctuations.

[0040] Anomaly Identification and Graded Early Warning Operation Logic: Devices equipped with an IMU can automatically identify artifacts caused by large-amplitude limb movements. Waveform anomalies during movement are directly marked as invalid and not included in the anomaly count. Devices without an IMU identify invalid data through waveform distortion and abnormal fit pressure. Single valid waveform feature deviations are only recorded and archived in the background without triggering any prompts. If three consecutive samples show persistent diastolic attenuation or waveform morphology degradation, wear correction and baseline retest reminders are pushed step by step. After the user triggers manual baseline retest calibration with one click, the anomaly record is cleared once the waveform returns to normal. The system has a built-in periodic physiological fluctuation feature library, which can accurately identify periodic vascular fluctuations during nocturnal sleep apnea and transient physiological perfusion fluctuations during the day. These are uniformly classified, statistically archived, and only follow-up trend records are kept, without pushing pathological abnormalities or medical attention prompts. The system automatically summarizes baseline data weekly to generate a long-term longitudinal trend curve. Only when persistent waveform attenuation or vascular degradation is detected for several consecutive weeks will outpatient vascular examination follow-up suggestions be pushed to avoid misjudgments caused by single physiological fluctuations.

[0041] Example 3: Implementation of a Hospital PC Clinical Analysis Terminal

[0042] The hospital's clinical PC terminal connects to the hardware's original monitoring data and the mobile terminal's archived data, supports the binding and management of patients' basic medical records, and enables simultaneous monitoring and analysis across multiple windows, including a real-time raw pulse waveform window, a 24-hour blood flow amplitude trend window, a nighttime sleep waveform detail magnification window, and a 9-channel pressure and heat window.

[0043] The terminal automatically quantifies and outputs professional hemodynamic indicators, including pulse rise time, relative height ratio of dicrotic wave, diastolic decay index, pulse rate variability, frequency of nocturnal periodic vascular pulsations, and ejection time; it automatically matches the periodic physiological fluctuation feature library, classifies and labels three types of abnormal waveforms: persistent hypoperfusion, periodic fluctuations of sleep apnea, and abnormal pulse; and supports one-click longitudinal comparison and analysis of the current monitoring data with historical hospital baseline waveforms.

[0044] The system can automatically generate standardized clinical follow-up PDF reports. The reports only objectively describe the vascular elasticity status, peripheral blood flow fluctuation trends, and abnormal event statistics, without any descriptions related to blood pressure diagnosis or disease confirmation. It supports batch encrypted archiving of raw waveform data, adapting to the research needs of cardiovascular clinical cohort studies, big data collection, and algorithm training.

[0045] Example 4: Implementation of the Complete Hierarchical Adaptive Monitoring Method

[0046] S1. The user wears the flexible patch and completes the Bluetooth connection. The system automatically synchronizes the local RTC clock with the network standard time and completes the full timing calibration. The user can choose the pure waveform follow-up mode or the relative pressure quantitative clinical mode according to the usage scenario. S2. The hardware RTC local clock keeps a continuous countdown. When connected to the network, it synchronizes with network standard time such as NTP, GPS, and Beidou in real time. All baseline data and monitoring waveforms are bound to a unique time sequence marker. The time sequence data is complete and valid even when the network is disconnected. S3. The system automatically starts the baseline acquisition process at 1:00 AM every day: The hardware version with IMU continuously detects no large body movements for five minutes. After determining that the body is still, it acquires a 30-second standard brachial artery baseline waveform and stores it as the automatic baseline for the day. If the posture is abnormal, the acquisition will be automatically terminated. The hardware version without IMU prompts the user to keep lying flat and still through software push operation. The baseline acquisition is completed after the user confirms that the body is still. If the body is not still in a standardized manner, the acquisition for the day will be abandoned. S4. Users can manually trigger the baseline calibration function as needed. The software guides users to maintain a standardized resting state and remain still for five minutes to complete a 30-second manual baseline acquisition. After multiple acquisitions per day, the system automatically removes discrete abnormal waveforms and retains valid baseline data. S5. The system automatically compares all valid manual baselines and automatic baselines for the day, extracts the peak amplitude ratio coefficients of the manual and automatic baselines, uses the automatic baseline as the morphological skeleton to complete the overall proportional scaling correction, realizes bidirectional fusion optimization, generates a unique valid comparison baseline for the day, and encrypts and permanently archives all original baseline data. S6. The system enters the adaptive intermittent cyclic sampling mode, collecting a 30-second waveform every two minutes during the day, and extending the single sampling duration to 60 seconds from midnight to 6 am at night; before each sampling, the array bonding status and pressure range status are checked, and if the bonding is poor or the range is saturated, the sampling data is directly discarded and the corresponding prompt is triggered. S7. The acquired raw waveforms are sequentially processed by array consistency compensation, multi-channel dynamic optimal waveform fusion, and baseline normalization preprocessing. During the initial cold start, the factory template baseline is used for transition. After accumulating sufficient data, it automatically switches to a personalized sliding baseline to eliminate various signal interferences. S8. Devices equipped with IMU filter out invalid waveforms of large-amplitude limb movements through IMU posture data, while devices without IMU filter valid data through waveform and fit features, extract the core features of valid waveforms and make accurate comparisons with the valid baseline of the day. S9. Single waveform feature differences are only archived in the background; three consecutive valid waveforms with continuous attenuation are pushed with adaptation prompts step by step, supporting manual baseline retesting and verification, and the abnormal count is cleared after the retest is normal; waveforms that match the periodic physiological fluctuation feature library are marked and archived separately, and no severe case warning is pushed. S10. The system automatically summarizes the effective baseline data weekly and generates a longitudinal blood flow trend curve. Only when the vascular condition is continuously deteriorated for several consecutive weeks will the system push out a follow-up examination recommendation to the outpatient department. S11. The system performs differentiated calculations and outputs based on the user-selected working mode: the pure waveform follow-up mode retains the amplitude signal internally for normalization calculations, does not output or display the physical pressure amplitude, only compares the waveform contour morphology changes, and is suitable for long-term home follow-up; the relative pressure quantitative clinical mode outputs standardized vascular elasticity and peripheral perfusion quantitative indicators, which is suitable for clinical auxiliary assessment, and does not display or output any absolute blood pressure values ​​throughout the process.

Claims

1. A continuous monitoring system for brachial artery pulse waveforms, comprising a flexible MEMS patch hardware terminal, mobile monitoring software, and a hospital PC clinical analysis terminal; the flexible MEMS patch hardware terminal is used to acquire human brachial artery pulse pressure signals, the mobile monitoring software is used for signal preprocessing, human-computer interaction, and baseline data management, and the hospital PC clinical analysis terminal is used for batch waveform analysis and data archiving; characterized in that: (1) The flexible MEMS patch hardware terminal is a flexible, pressure-free wearable structure without an air pump, airbag, or pressure cuff. It includes a 3×3 MEMS piezoresistive sensor array, a dual-range parallel sampling and holding circuit, an NTC temperature compensation module, a 24-bit Σ-Δ analog-to-digital converter and a pre-amplifier with low noise, an optional six-axis IMU attitude sensor, an RTC local real-time clock, a low-power main control MCU, a BLE5.3 wireless communication unit, and a micro lithium battery. In the dual-range parallel sampling and holding circuit, the static bonding pressure range is 0~10kPa, and the dynamic pulse micro-pressure range is 0~2kPa. The two channels are independently sampled and held, and the channel switching selection is only completed in the stable waveform range at the end of diastole. The hardware is calibrated and stored in the main control unit after the factory to complete the sensitivity, temperature drift coefficient, and zero-point offset parameters of each channel. The array consistency correction, multi-channel waveform fusion, five-minute sliding baseline correction, motion artifact recognition, and bonding pressure calculation algorithm are locally stored. (2) The mobile monitoring software includes a dual-mode switching module, a baseline intelligent management module, an intermittent sampling control module, a fit quality visualization self-check module, a timing synchronization module, and a multi-level abnormal warning module; the dual-mode switching module is configured with two independent working modes: pure waveform follow-up mode and relative pressure quantitative clinical mode; the pure waveform follow-up mode only normalizes the output of the pulse waveform contour shape, and does not output or display the physical pressure amplitude; the relative pressure quantitative clinical mode calls the factory calibration parameters to calculate the standardized relative pressure difference, automatically extracts vascular elasticity and peripheral perfusion-related hemodynamic indicators, and the interface does not display or output systolic pressure, diastolic pressure values ​​and blood pressure units throughout the process; The baseline intelligent management module includes an automatic baseline acquisition unit, a manual baseline acquisition unit, and a baseline fusion correction determination unit; The automatic baseline acquisition unit triggers reference waveform acquisition at 1:00 AM daily. When equipped with an IMU, it automatically determines the static state for baseline acquisition based on human posture recognition. When not equipped with an IMU, it outputs instructions to guide the user to manually maintain a static state for baseline acquisition. The manual baseline acquisition unit supports one-click calibration. The baseline fusion correction judgment unit is used to compare and optimize the consistency of multiple sets of manual and automatic baselines for the day. All original baseline waveform data is encrypted and permanently stored. The intermittent sampling control module sets an adaptive periodic acquisition strategy, using a fixed intermittent sampling interval during the day and extending the duration of a single waveform acquisition at night; the fit quality visualization self-check module collects sensor array pressure data in real time, generates an upper arm fit pressure thermal distribution map, and identifies abnormal fit conditions; the timing synchronization module adopts a dual timing mechanism of RTC local clock backup and automatic network standard time synchronization, with all waveforms bound to a unique timestamp, and timing data remaining valid even when the network is disconnected; the multi-level abnormality early warning module distinguishes between motion artifacts, physiological periodic fluctuations of sleep apnea, and pathological blood flow attenuation, recording only a single waveform abnormality in the background, and outputting follow-up and re-examination suggestions only for continuous waveform deterioration over multiple weeks. (3) The hospital PC clinical analysis terminal is used to realize batch playback of original waveforms, quantitative analysis of hemodynamic parameters, longitudinal comparison of historical baselines, generation of standardized clinical follow-up reports, and encrypted archiving of cardiovascular research data. (4) The system only uses individual longitudinal time-series waveform comparison to assess hemodynamic changes, does not calculate or output absolute blood pressure values ​​of systolic and diastolic pressure, and does not output disease diagnosis conclusions.

2. The continuous monitoring system according to claim 1, characterized in that: The dual-range parallel sampling and holding circuit has a range saturation self-recovery mechanism. When the static bonding pressure continuously exceeds the upper limit of the 10kPa range, the system actively enters the safe silent mode and outputs the wearing adjustment command. After the pressure drops back to the effective range, the monitoring will automatically resume.

3. The continuous monitoring system according to claim 1, characterized in that: The 3×3 MEMS piezoresistive sensing array adopts a static pressure unit and a micro pressure unit arrangement structure; the static pressure unit collects the contact pressure over the entire area to generate a heat map, and the micro pressure unit collects the weak pulse signal of the brachial artery.

4. The continuous monitoring system according to claim 1, characterized in that: The multi-channel dynamic optimization signal-to-noise ratio waveform fusion algorithm calculates the waveform signal-to-noise ratio and pulsation correlation of the 9 sensing channels in real time, and dynamically optimizes at least one or more optimal effective channels to reconstruct the final pulse waveform, without fixing the number of channels; when the number of effective high-quality channels is less than two, it outputs a self-test command for wearing posture adjustment and patch fit.

5. The continuous monitoring system according to claim 1, characterized in that: When there is insufficient sliding data in the first five minutes of cold start, the factory-built-in general waveform template is used as a temporary baseline. After accumulating five minutes of valid personalized data, it automatically switches to the adaptive sliding baseline.

6. The continuous monitoring system according to claim 1, characterized in that: The hardware version equipped with an IMU identifies large-amplitude limb movements through continuous posture data and marks the corresponding waveforms as invalid samples, excluding them from anomaly statistics; the version without an IMU relies on the characteristics of contact pressure distortion and waveform distortion to identify invalid data.

7. The continuous monitoring system according to claim 1, characterized in that: The multi-level abnormal warning classification judgment logic is as follows: single waveform feature deviation is only archived locally; after three consecutive valid waveforms with blood flow attenuation, wear self-test and baseline retest instructions are output; after the retest calibration is completed, the waveform returns to normal and the abnormal count is reset; waveforms matching the physiological fluctuation feature library are archived separately and no serious illness consultation reminders are output.

8. A hierarchical adaptive working method based on the continuous monitoring system according to any one of claims 1 to 7, characterized in that, Includes the following steps: S1. Mode selection steps: Users select either the pure waveform follow-up mode or the relative pressure quantitative clinical mode on the mobile monitoring software according to the usage scenario. S2. Timing synchronization steps: The hardware RTC local clock keeps counting continuously. When the device is connected to the network, it automatically synchronizes with the network standard time. All baseline waveforms and monitoring waveforms are bound to independent and unique timing markers. When the network is disconnected, the timing data is complete and valid. S3. Automatic Baseline Acquisition Procedure: The baseline acquisition process is automatically started at 1:00 AM daily. When the hardware is equipped with an IMU, the IMU continuously detects for five minutes that the human body does not undergo significant turning over, arm raising, or limb swaying movements. A 30-second standard brachial artery baseline waveform is acquired and stored as the automatic baseline for the day. If the human body posture does not meet the stillness condition, the automatic baseline acquisition process for the day is automatically terminated. When the hardware is not equipped with an IMU, the software outputs instructions to prompt the user to maintain a supine and still state. After the user confirms that the user is in a properly still state, a 30-second baseline acquisition is started. If the user does not maintain stillness, the baseline acquisition for the day is abandoned. S4. Artificial baseline acquisition steps: The user clicks the calibration button as needed. The software outputs instructions prompting the user to remain seated with upper limbs flat and still for five minutes. Then, a 30-second standardized artificial baseline waveform is acquired. Multiple manual acquisitions are supported per day. The system automatically removes discrete, distorted, and abnormal invalid artificial waveforms and retains highly consistent valid baseline data. S5. Daily valid baseline correction steps: The software automatically compares all valid manual baselines and automatic baselines for the day, extracts the peak amplitude ratio coefficient between the manual and automatic baselines, uses the automatic baseline as the morphological skeleton, and performs overall proportional scaling correction of the automatic baseline through the ratio coefficient to complete the bidirectional fusion optimization, generating a unique corrected valid comparison baseline for the day, and encrypts and permanently archives all original baseline data. S6. Intermittent Cyclic Sampling Steps: The system starts an adaptive intermittent sampling mechanism, performing waveform acquisition every two minutes during the day, with a single acquisition duration of thirty seconds; during the nighttime sleep period from midnight to six o'clock, the single acquisition duration is extended to sixty seconds; before each sampling, the array bonding pressure data is read synchronously, and if the bonding status is abnormal, the waveform sampling sample is directly discarded; S7. Waveform preprocessing steps: The acquired raw waveforms are sequentially processed by array consistency compensation, multi-channel dynamic optimal waveform fusion, and five-minute sliding baseline normalization. During the initial cold start, the factory template baseline is used for transition to eliminate signal interference caused by hardware unit deviation, wearing fit drift, and physiological vascular tension fluctuations. S8. Artifact Filtering and Feature Comparison Steps: When the hardware is equipped with an IMU, large-amplitude limb movement waveforms are identified through IMU posture data and marked as invalid data, which are not included in the anomaly statistics; when the hardware is not equipped with an IMU, invalid data are identified through abnormal contact pressure and waveform distortion features; core features such as rising limb duration, diphtheria amplitude, and diastolic decay rate are extracted from valid waveforms and accurately matched and compared with the valid baseline of the day. S9. Graded Abnormal Handling Steps: Single waveform feature differences are only recorded and archived in the background without triggering prompts; if three consecutive valid waveforms show continuous blood flow attenuation, instructions for wearing self-check and baseline retest are output step by step; if the waveform returns to normal after the user completes the retest calibration, the abnormal count is cleared; waveforms matching the periodic physiological fluctuation feature library are archived and marked separately, without outputting reminders for severe illness consultation; S10, Long-term trend assessment steps: The software automatically summarizes the effective baseline data of the day every week and generates a longitudinal blood flow trend curve. It only outputs vascular re-examination and follow-up suggestions when the waveform continues to decay for several consecutive weeks and the vascular condition deteriorates. S11, Layered Differentiation Calculation Steps: In the pure waveform follow-up mode, the internal pressure amplitude signal is retained for waveform normalization calculation. It does not output or display the physical pressure amplitude and unit. It only compares the changes in waveform contour shape and is suitable for daily home follow-up. In the relative pressure quantitative clinical mode, the factory calibration coefficient is called to calculate the standardized relative pressure amplitude and extract vascular elasticity and peripheral perfusion quantitative indicators. The software does not display or output any blood pressure values ​​throughout the process.

Images