Vascular hemodynamic anomaly monitoring system and pulse wave analysis method thereof

Abstract

The invention relates to the technical field of medical monitoring, in particular to a vascular hemodynamics anomaly monitoring system and a pulse wave analysis method thereof.The core of the vascular hemodynamics anomaly monitoring system is that a flexible piezoelectric sensor array is used for capturing human body vascular pulse wave signals, the signals are transmitted to a wearable monitoring device for preliminary preprocessing, and the human body vascular pulse wave signals are obtained; a pulse wave analysis module in the host adopts a multi-level analysis method, and the multi-level analysis method comprises the steps of multi-scale signal decomposition and reconstruction, nonlinear feature extraction and mode recognition, multi-parameter self-adaptive judgment standard, deep learning feature automatic extraction and multi-part information fusion decision making; according to the method, the analysis accuracy and comprehensiveness are ensured, finally, a processing result is transmitted to an application platform, a blood vessel health report is generated, non-intrusive continuous monitoring is achieved, the innovation effectively fills the blank of continuous monitoring in the prior art, and the method has great significance in monitoring the blood vessel health condition in real time and preventing blood vessel diseases.

Description

Technical Field

[0001] The present invention relates to the field of medical monitoring technology, and in particular to a vascular hemodynamic abnormality monitoring system and a pulse wave analysis method thereof, which are particularly suitable for early monitoring and early warning of vascular elasticity abnormalities and deep vein thrombosis. Background Art

[0002] Abnormal vascular hemodynamics, particularly deep vein thrombosis (DVT), is a common and dangerous clinical condition. Currently, clinical methods for detecting abnormal vascular hemodynamics include Doppler ultrasound, venography, and D-dimer testing. While these methods can reflect vascular status to a certain extent, they all have significant limitations.

[0003] Ultrasound Doppler testing requires specialized medical expertise and can only provide single-point, short-term results, preventing continuous monitoring. Venography is an invasive procedure and unsuitable for frequent use. D-dimer testing is a post-hoc confirmation measure and lacks early warning capabilities. Furthermore, existing testing methods often rely on large, expensive medical equipment, making them inadequate for routine monitoring.

[0004] Furthermore, existing wearable health monitoring devices mostly focus on monitoring basic physiological parameters like heart rate and blood pressure, lacking the ability to professionally assess vascular health. This is particularly true for early warning of potentially fatal conditions like deep vein thrombosis, making it difficult to provide effective risk warnings before clinical symptoms appear.

[0005] Existing technologies also have limitations in signal processing. Traditional pulse wave analysis methods often rely on simple time-domain or frequency-domain analysis, which cannot effectively capture the subtle characteristics of vascular changes. Furthermore, they lack individualized assessment criteria and are difficult to adapt to individual physiological differences, resulting in a high rate of false positives.

[0006] Therefore, developing a system that can non-invasively and continuously monitor the hemodynamic status of blood vessels and has early abnormality warning function has important clinical significance and application value. Summary of the Invention

[0007] In response to the problems existing in the prior art, the purpose of the present invention is to provide a vascular hemodynamic abnormality monitoring system and a pulse wave analysis method thereof, which can realize non-invasive continuous monitoring of vascular health status, especially early warning of deep vein thrombosis, thereby providing a sufficient time window for clinical intervention.

[0008] The present invention proposes a vascular hemodynamic abnormality monitoring system and a pulse wave analysis method thereof, comprising:

[0009] Flexible piezoelectric sensor array, used to collect human blood vessel pulse wave signals;

[0010] a wearable monitoring device, electrically connected to the flexible piezoelectric sensor array, for receiving the pulse wave signal and performing preliminary preprocessing;

[0011] a data processing host, connected to the wearable monitoring device via wireless communication, for receiving and processing the pulse wave signal;

[0012] a pulse wave analysis module, provided in the data processing host, for performing multi-level analysis on the pulse wave signal, the pulse wave analysis module comprising a multi-scale signal decomposition and reconstruction unit, a nonlinear feature extraction and pattern recognition unit, a multi-parameter adaptive judgment standard unit, a deep learning feature automatic extraction unit, and a multi-site information fusion decision unit connected in sequence; and

[0013] The application platform is connected to the data processing host through a network, and is used to receive processing results and generate a vascular health report.

[0014] Preferably, the multi-scale signal decomposition and reconstruction unit includes:

[0015] a signal decomposition subunit, configured to receive the pulse wave signal and decompose the pulse wave signal into coefficient sets of multiple scales using a plurality of wavelet basis functions;

[0016] a denoising processing subunit, configured to perform soft threshold shrinkage denoising on the coefficient set; and

[0017] The reconstruction subunit is used to dynamically adjust the weights of coefficients at different scales according to the signal energy distribution and generate reconstructed multi-scale feature coefficients.

[0018] Preferably, the nonlinear feature extraction and pattern recognition unit includes:

[0019] An autocorrelation calculation subunit, used to calculate the pulse wave autocorrelation sequence in different time windows;

[0020] a nonlinear fitting subunit, configured to perform nonlinear fitting on the autocorrelation sequence using a Gaussian function;

[0021] a feature extraction subunit, configured to extract feature parameters reflecting vascular elasticity, blood flow resistance, cyclic stability, amplitude variability, and waveform complexity from the autocorrelation sequence and the fitting result; and

[0022] The feature optimization subunit is used to select the most discriminative features from the initial feature set based on the sample discrimination and stability scores.

[0023] Preferably, the multi-parameter adaptive judgment standard unit includes:

[0024] A static threshold subunit, used to set a fixed threshold based on physiological common sense;

[0025] The group threshold subunit is used to set the grouping threshold based on factors such as age and gender;

[0026] The personalized threshold subunit is used to learn the user's normal state range and build a personalized model;

[0027] a state recognition subunit to distinguish between resting, light activity, and vigorous exercise states; and

[0028] The abnormality classification subunit is used to classify abnormalities into five categories: abnormal vascular elasticity, abnormal blood flow resistance, abnormal vascular compliance, abnormal blood flow velocity and abnormal blood flow pattern.

[0029] Preferably, the deep learning feature automatic extraction unit includes:

[0030] Input layer, used to receive 32-channel pulse wave signals;

[0031] Feature extraction layer, including multiple convolutional layers and pooling layers, is used to extract features from different perspectives;

[0032] Attention mechanism module, used to automatically identify and enhance the features of important sensor channels;

[0033] Residual connection modules, used to address the vanishing gradient problem in deep networks; and

[0034] The output layer is used to generate vascular status classification results.

[0035] Preferably, the multi-part information fusion decision unit includes:

[0036] The weight allocation subunit is used to allocate weights to the monitoring results of different parts;

[0037] Time series analysis subunit, used to analyze short-term trend changes, medium-term cyclical changes and long-term evolution trends;

[0038] A risk assessment subunit, which determines the risk level based on the combined information; and

[0039] The report generation subunit is used to integrate information and generate a comprehensive assessment report including vascular elasticity, blood flow resistance and thrombosis risk.

[0040] Preferably, the flexible piezoelectric sensor array includes 32 piezoelectric sensors, which are made of medical-grade polyvinylidene fluoride piezoelectric film and arranged in a 4×8 matrix structure. The size of a single sensor is 3mm×3mm and the thickness is 0.2mm.

[0041] Preferably, the wearable monitoring device comprises:

[0042] Data acquisition unit, used for multi-channel synchronous acquisition control;

[0043] A signal preprocessing unit, used to perform bandpass filtering, notch filtering and outlier detection on the collected signal;

[0044] A Bluetooth communication unit is used to transmit the processed data to the data processing host; and a power management unit is used to provide a stable power supply and achieve low power consumption management.

[0045] Preferably, the application platform includes:

[0046] A data receiving module, configured to receive the processing result sent by the data processing host;

[0047] Data storage module, used to store historical monitoring data;

[0048] Analysis and display module, used to generate visual charts and trend analysis;

[0049] An early warning module, configured to issue an alarm when an anomaly is detected; and

[0050] User interaction module, used to receive user feedback and configure system parameters.

[0051] The pulse wave analysis method applied to the vascular hemodynamic abnormality monitoring system comprises the following steps:

[0052] Collecting pulse wave signals and sending the collected pulse wave signals to a data processing host;

[0053] Performing multi-scale decomposition on the pulse wave signal to obtain characteristic coefficient sets at different scales;

[0054] Extracting nonlinear features from the characteristic coefficient set, including vascular elasticity, blood flow resistance, periodic stability, amplitude variability, and waveform complexity features;

[0055] According to individual differences and physiological states, a multi-level adaptive judgment standard is established to evaluate the nonlinear characteristics;

[0056] Use deep learning networks to automatically extract deep features from raw signals and classify vascular status;

[0057] Integrate assessment results from multiple monitoring sites to conduct temporal pattern analysis and comprehensive risk assessment; and

[0058] Generates vascular health reports including abnormality type, risk level, and health recommendations.

[0059] This invention uses a flexible piezoelectric sensor array to collect pulse wave signals, combined with multi-level signal processing and analysis techniques, to comprehensively assess vascular elasticity, blood flow resistance, and thrombosis risk. The system employs a bottom-up signal processing and analysis strategy, significantly improving the sensitivity and specificity of vascular anomaly detection through innovative technologies such as multi-scale signal decomposition, nonlinear feature extraction, adaptive threshold determination, deep learning feature extraction, and multi-site information fusion.

[0060] The beneficial effects of the present invention are mainly reflected in:

[0061] 1. It realizes non-invasive and continuous monitoring of vascular health status, filling the gap in existing technologies in continuous monitoring;

[0062] 2. Through multi-level signal processing and analysis technology, the detection sensitivity is significantly improved, and the rising risk of thrombosis can be detected 48-72 hours before clinical symptoms appear;

[0063] 3. The use of personalized adaptive threshold judgment criteria effectively solves the problem of misjudgment caused by individual differences, reducing the false positive rate from 25% of traditional methods to 8%;

[0064] 4. Innovatively combines monitoring results from multiple locations and improves overall detection reliability through information fusion;

[0065] 5. The system has high integration and is easy to operate, suitable for daily wear and long-term monitoring, which improves user compliance;

[0066] 6. The cost is significantly lower than traditional hospital examination equipment and has wide application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 Schematic diagram of the overall structure of the vascular hemodynamic abnormality monitoring system of the present invention;

[0068] Figure 2 Schematic diagram of the structure of the flexible piezoelectric sensor array of the present invention;

[0069] Figure 3 Schematic diagram of the structure of the wearable monitoring device of the present invention;

[0070] Figure 4 Schematic diagram of the structure of the pulse wave analysis module of the present invention;

[0071] Figure 5 This is a signal processing flow chart of the multi-scale signal decomposition and reconstruction unit of the present invention;

[0072] Figure 6 This is a functional block diagram of the nonlinear feature extraction and pattern recognition unit of the present invention;

[0073] Figure 7 This is a judgment flow chart of the multi-parameter adaptive judgment standard unit of the present invention;

[0074] Figure 8 This is a network structure diagram of the deep learning feature automatic extraction unit of the present invention;

[0075] Figure 9 This is a workflow diagram of the multi-part information fusion decision-making unit of the present invention;

[0076] Figure 10 Flowchart of the pulse wave analysis method of the present invention. DETAILED DESCRIPTION

[0077] Please refer to the attached Figure 1-10 The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. These embodiments are only used to illustrate the present invention and should not be understood as limiting the present invention.

[0078] Reference Figure 1 The present invention provides a vascular hemodynamic abnormality monitoring system, including a flexible piezoelectric sensor array 1, a wearable monitoring device 2, a data processing host 3, a pulse wave analysis module 4 and an application platform 5.

[0079] A flexible piezoelectric sensor array 1 is used to collect pulse wave signals from human blood vessels. In one embodiment of the present invention, the flexible piezoelectric sensor array 1 senses the pulsating pressure changes in subcutaneous blood vessels by contacting the surface of human skin, converting mechanical energy into electrical signals to achieve non-invasive collection of pulse wave signals. The array is made of flexible materials, suitable for long-term attachment to the skin surface, improving user comfort. Preferably, when the pulsating pressure of the blood vessels is transmitted through the skin to the sensor surface, the voltage signal generated by the flexible piezoelectric sensor is proportional to the pressure change within the blood vessel, thereby accurately capturing the dynamic characteristics of the blood vessels.

[0080] The wearable monitoring device 2 is electrically connected to the flexible piezoelectric sensor array 1, and is used to receive the pulse wave signal and perform preliminary preprocessing. Preferably, the wearable monitoring device 2 adopts a watchband design and can be worn on the wrist, ankle and other blood vessel-rich areas for daily use. The wearable monitoring device 2 has a built-in microprocessor and a wireless communication module, which can preprocess the collected signals and transmit them to the data processing host 3. In practical applications, for example, for deep vein thrombosis monitoring of long-term bedridden patients, the wearable monitoring device 2 can be worn on the lower limbs to continuously monitor venous return; and for patients who need to monitor the blood flow of the carotid artery, the device can be worn on the neck to achieve real-time monitoring of the blood supply to the brain.

[0081] The data processing host 3 is connected to the wearable monitoring device 2 via wireless communication and is used to receive and process pulse wave signals. In one embodiment of the present invention, the data processing host 3 can be a smartphone, tablet computer, or dedicated processing device, responsible for running complex signal processing and analysis algorithms. Preferably, the data processing host 3 adopts a multi-threaded parallel processing architecture to process monitoring data from different parts in real time, ensuring that the response speed is maintained when a large amount of data is simultaneously input. In addition, the data processing host 3 also has a caching function, which can temporarily store data when the wireless connection is interrupted, and automatically synchronize after the connection is restored to ensure data integrity.

[0082] The pulse wave analysis module 4, housed within the data processing host 3, is the core component of the system, performing multi-level analysis of pulse wave signals. It comprises a sequentially connected multi-scale signal decomposition and reconstruction unit 41, a nonlinear feature extraction and pattern recognition unit 42, a multi-parameter adaptive judgment criterion unit 43, a deep learning feature automatic extraction unit 44, and a multi-site information fusion decision unit 45. These five functional units form a complete signal processing chain, progressively extracting valuable features from the raw pulse wave signal and ultimately enabling accurate assessment of vascular status. For example, in the early stages of detecting deep vein thrombosis in the lower extremities, the system can identify changes in pulse wave characteristics caused by increased venous return resistance and issue an early warning 48-72 hours before clinical symptoms appear, providing a valuable window for clinical intervention.

[0083] Application platform 5 is connected to data processing host 3 via a network to receive processing results and generate vascular health reports. Application platform 5 displays monitoring results to users through a visual interface and issues warning alerts when abnormalities are detected. Furthermore, application platform 5 can store data in the cloud, supporting remote medical consultations and long-term health management. In clinical practice, medical staff can remotely monitor patients' vascular health status through application platform 5 and adjust treatment plans in a timely manner. For patients recovering at home, the system can provide continuous monitoring and health guidance, reducing the frequency of hospital visits.

[0084] Reference Figure 2The flexible piezoelectric sensor array 1 of the present application comprises 32 piezoelectric sensors arranged in a 4x8 matrix structure. Each piezoelectric sensor is made of medical-grade polyvinylidene fluoride piezoelectric film, with a single sensor size of 3mmx3mm and a thickness of 0.2mm. The spacing between the piezoelectric sensors is 2mm, ensuring that the array can cover a large enough area while maintaining overall flexibility. The design sensitivity of the flexible piezoelectric sensor array 1 is greater than 5mV / g, which can accurately capture weak pulse wave signals. This sensitivity design is based on clinical research, which shows that the pressure change generated by normal adult arterial pulsation is about 0.2-0.8g, while the pressure change caused by early vascular abnormalities is only 0.05-0.1g. Therefore, higher sensitivity is required to capture these small changes.

[0085] Preferably, the flexible piezoelectric sensor array 1 is made of flexible printed circuit board technology, which has good ductility and adhesion, and is suitable for long-term attachment to the skin surface without causing discomfort. At the same time, the sensor surface is coated with a biocompatible material to avoid skin allergic reactions caused by long-term contact. In clinical applications, for example, for obese patients, traditional rigid sensors often have difficulty adhering to the skin surface, resulting in a decrease in signal quality. The flexible design of the present application can well adapt to the needs of patients of different body types, ensuring signal acquisition quality.

[0086] Reference Figure 3 The wearable monitoring device 2 of the present application comprises a data acquisition unit 21, a signal preprocessing unit 22, a Bluetooth communication unit 23 and a power management unit 24.

[0087] The data acquisition unit 21 is used for multi-channel synchronous acquisition control and can simultaneously read the signals of 32 piezoelectric sensors with a sampling frequency of 200Hz and a sampling accuracy of 12 bits. The data acquisition unit 21 uses a high-precision analog-to-digital converter to achieve accurate digitization of the signal. To ensure acquisition accuracy, the crosstalk suppression between channels of the data acquisition unit 21 is greater than 60dB, effectively avoiding signal interference between channels. In actual monitoring processes, for example, when a patient needs to monitor multiple sites (such as both lower extremities) simultaneously, the high-precision multi-channel synchronous acquisition capability ensures the time alignment of data from different parts, providing a basis for subsequent information fusion analysis.

[0088] The signal preprocessing unit 22 is used to perform bandpass filtering, notch filtering, and outlier detection on the collected signal. Preferably, the bandpass filter has a passband range of 0.5-40 Hz, effectively preserving the effective components of the pulse wave signal while filtering out baseline drift and high-frequency noise. The notch filter center frequency is 50 Hz (or 60 Hz) to suppress power frequency interference. Outlier detection, based on a moving median algorithm, can identify and correct abnormal data points during the acquisition process. The passband range is selected based on the spectral characteristics of the human pulse wave signal: the fundamental frequency (heart rate) is typically in the range of 0.8-2 Hz, with harmonic components reaching 10-15 Hz, while subtle changes related to vascular elasticity and compliance are reflected in frequencies up to 40 Hz. Selecting 0.5 Hz as the low-frequency cutoff point can filter out baseline drift caused by respiration and body position changes. Selecting 40 Hz as the high-frequency cutoff point preserves high-frequency components related to vascular status while filtering out high-frequency noise such as electromyographic interference.

[0089] The Bluetooth communication unit 23 utilizes Bluetooth 5.0 low-power technology to transmit processed data to the data processing host 3. The transmission rate exceeds 2 Mbps, meeting real-time data transmission requirements. Data packets are 256 bytes each, and a CRC16 checksum algorithm is used to ensure transmission accuracy. In everyday use scenarios, such as when users are at home, the Bluetooth communication unit 23 ensures a stable connection with data processing hosts such as smartphones. If transmission quality degrades, the system automatically adjusts the sampling rate and compression ratio to prioritize the real-time transmission of critical data.

[0090] The Power Management Unit 24 uses a rechargeable 200mAh lithium polymer battery to provide stable power and achieve low-power management. In continuous monitoring mode, the device can operate for over 24 hours; in intermittent monitoring mode, the operating time can be extended to over 72 hours. The Power Management Unit 24 also supports wireless charging and USB Type-C charging, making it convenient for daily use. For patients with chronic diseases requiring long-term monitoring, the Power Management Unit 24 uses an intelligent power scheduling strategy to dynamically adjust the sampling frequency and processing depth based on user activity status and signal complexity, ensuring monitoring quality while maximizing battery life.

[0091] Reference Figure 4 The pulse wave analysis module 4 of the present invention is the core of the system and includes five functional units. These five units form a progressive signal processing chain, with the output of each unit serving as the input of the next, forming a complete information flow and ultimately achieving accurate assessment of vascular status.

[0092] Reference Figure 5 The multi-scale signal decomposition and reconstruction unit 41 includes a signal decomposition subunit 411 , a denoising processing subunit 412 and a reconstruction subunit 413 .

[0093] Signal decomposition subunit 411 is used to receive the pulse wave signal and decompose it into coefficient sets of multiple scales using a variety of wavelet basis functions. In one embodiment of the present invention, signal decomposition subunit 411 constructs a library of wavelet basis functions containing eight different characteristics, each optimized for different frequency components of the pulse wave signal. The pulse wave signal S(t) can be expressed through wavelet transform as:

[0094] C j,k =∫S(t)·ψ j，k (t)dt,

[0095] Among them, C j，k is the wavelet coefficient, i represents the scale parameter, k represents the translation parameter, ψ j，k (t) is the wavelet basis function, defined as:

[0096] ψ j，k (t) = 2 -j / 2 ·ψ(2 -j tk),

[0097] Among them, ψ(t) is the mother wavelet function; 2 -j / 2 is the normalization factor to ensure energy conservation at different scales; 2 -j is the scaling factor, controlling the expansion and contraction of the wavelet; k is the translation factor, controlling the translation of the wavelet. The present invention preferably uses the Daubechies wavelet (db4) as the mother wavelet due to its excellent localization and energy concentration, making it particularly suitable for capturing transient changes in pulse wave signals. In practical applications of vascular monitoring, such as detecting the early stages of deep vein thrombosis, this multiscale decomposition can simultaneously capture changes in blood flow velocity (low-frequency components) and changes in vessel wall stiffness (high-frequency components), providing a comprehensive assessment of vascular status.

[0098] The signal decomposition subunit 411 performs an 8-layer progressive decomposition on the original pulse wave signal. The first layer captures the characteristics of the 20-40Hz frequency band, while the eighth layer captures the characteristics of the 0.5-1Hz frequency band. This multi-scale decomposition strategy can comprehensively capture the characteristics of the pulse wave signal in different frequency bands, providing a rich source of information for subsequent analysis. In clinical applications, different frequency band characteristics are associated with different pathological conditions: the 0.5-1Hz frequency band mainly reflects the basic frequency of the heart rate, and abnormalities may indicate arrhythmia; the 1-5Hz frequency band reflects vascular compliance, and abnormalities may indicate arteriosclerosis; the 5-15Hz frequency band reflects vascular resistance, and abnormalities may indicate thrombosis; and the 15-40Hz frequency band may contain micro-vibration characteristics of the blood vessel wall, and abnormalities may indicate endothelial damage. Through full-band analysis, the system can provide a more comprehensive assessment of vascular health.

[0099] The denoising subunit 412 is used to perform soft threshold shrinkage denoising on the coefficient set. The soft threshold shrinkage function is defined as:

[0100]

[0101] Among them, δ λ (x) is the coefficient value after soft threshold shrinkage; λ is the threshold, which determines the denoising strength; x is the original wavelet coefficient; sgn(x) is the sign function, which takes the value of 1 when x>0, takes the value of -1 when x<0, and takes the value of 0 when x=0; |x| represents the absolute value of x. The threshold is usually set to σ is the standard deviation of noise, and N is the signal length. In practical applications, σ is usually estimated by the median absolute deviation of the minimum scale (highest frequency) subband of the wavelet coefficients: σ = median(|C 1，k |) / 0.6745, where 0.6745 is the normalization factor for the median absolute deviation of the standard normal distribution. The denoising subunit 412 applies different thresholds to the coefficients of each scale, with larger thresholds applied to smaller scales (high frequencies) and smaller thresholds applied to larger scales (low frequencies), effectively suppressing noise while preserving the key features of the signal. In practical applications of vascular monitoring, denoising is particularly important because factors such as user activity, breathing, and environmental vibrations can introduce noise, affecting detection accuracy. For example, in elderly patients, decreased skin elasticity leads to increased sensor signal noise. In this case, soft threshold denoising can effectively improve signal quality without losing key physiological features.

[0102] The reconstruction subunit 413 is used to dynamically adjust the weights of coefficients at different scales according to the signal energy distribution and generate reconstructed multi-scale feature coefficients. The signal reconstruction process can be expressed as:

[0103] S′(t)=∑ j，k w j ·C j，k ·ψ j，k (t),

[0104] Among them, S′(t) is the reconstructed signal, which is a function of time t; w j is the weight coefficient of the jth scale, which is used to adjust the contribution of different scales to the reconstructed signal; C j，k is the wavelet coefficient; ψ j，k (t) is the wavelet basis function; the summation symbol ∑ j，k Indicates the accumulation of all scales j and translations k. Weight coefficient w j Dynamically adjust according to the energy proportion of the scale signal:

[0105]

[0106] Among them, E j=∑ k |C j，k | 2 is the energy of the jth scale, calculated as the sum of the squares of all wavelet coefficients at that scale; is the sum of all scale energies; J is the total number of scales, and in this embodiment, J=8. This dynamic weight adjustment strategy can enhance the frequency band containing pathological features and improve the sensitivity of abnormality detection.

[0107] Preferably, the reconstruction subunit 413 updates the reconstruction weights every 5 seconds to adapt to different physical states and activity levels. The processing window length is selected to be 8 seconds to ensure that it contains at least 6-10 complete pulse cycles, providing sufficient information for feature extraction and analysis. The selection of the window length takes into account the range of heart rate changes: the normal resting heart rate of an adult is about 60-80 beats / minute, which can reach 100-120 beats / minute during exercise, while the resting heart rate of some elderly people or athletes can be as low as 40-50 beats / minute. Therefore, the 8-second window can contain a sufficient number of complete cardiac cycles in various situations to ensure the reliability of the analysis.

[0108] Reference Figure 6 The nonlinear feature extraction and pattern recognition unit 42 includes an autocorrelation calculation subunit 421 , a nonlinear fitting subunit 422 , a feature extraction subunit 423 and a feature optimization subunit 424 .

[0109] The autocorrelation calculation subunit 421 is used to calculate the pulse wave autocorrelation sequence in different time windows. The autocorrelation function R(τ) is defined as:

[0110]

[0111] Where R(τ) is the autocorrelation function, which indicates the correlation between the signal and its own time delay τ; S(t) is the pulse wave signal; T is the total length of the signal in seconds; τ is the time delay in seconds; the integral symbol represents the integral from 0 to T-τ; is a normalization factor to ensure that the results of different delays τ are comparable. The autocorrelation calculation subunit 421 selects four different time windows (0.5 seconds, 1 second, 2 seconds and 4 seconds) to calculate the autocorrelation sequence to capture the signal characteristics at different time scales. In vascular monitoring, autocorrelation analysis of different window lengths is targeted at different vascular pathological states: the 0.5-second window mainly captures changes within a single cardiac cycle and is suitable for detecting heart valve disease; the 1-second window captures the relationship between adjacent heartbeats and is suitable for detecting arrhythmias; the 2-second window is suitable for detecting changes in vascular compliance; and the 4-second window is suitable for detecting longer-term hemodynamic changes, such as venous return disorders. Through multi-window analysis, the system can comprehensively evaluate different types of vascular abnormalities.

[0112] The nonlinear fitting subunit 422 is used to perform nonlinear fitting on the autocorrelation sequence using a Gaussian function. The Gaussian fitting function is defined as:

[0113]

[0114] Where R′(τ) is the fitting function, which is a function of the time delay π; A is the amplitude parameter, reflecting the signal strength; α is the attenuation coefficient, the unit is s -2 , reflecting the signal attenuation speed; τ0 is the time delay, in seconds; f b is the pulse wave frequency in Hz; e is the base of the natural logarithm, approximately 2.71828; cos is the cosine function. Nonlinear fitting uses the least squares method to solve the parameters:

[0115]

[0116] in, Indicates the sum of N sampling points; R(τ i ) is the actual autocorrelation value; R′(τ i ) is the fitting value; N is the number of sampling points. In the actual application of vascular monitoring, the fitting parameters have clear physiological significance: the amplitude parameter A reflects the strength of the pulse wave signal and is related to the level of vascular perfusion; the attenuation coefficient α reflects the elasticity of the blood vessels and the blood flow resistance, and an increase in α indicates a decrease in elasticity or an increase in resistance; the pulse frequency f b This nonlinear fitting method captures the nonlinear characteristics of the pulse wave signal and is more sensitive to subtle changes in vascular elasticity and blood flow resistance. For example, in the early stages of deep vein thrombosis, a slight increase in blood flow resistance can cause the attenuation coefficient α to increase by 15-20%, a subtle change that traditional linear analysis methods may have difficulty detecting.

[0117] The feature extraction subunit 423 is used to extract characteristic parameters reflecting vascular elasticity, blood flow resistance, cyclic stability, amplitude variability, and waveform complexity from the autocorrelation sequence and fitting results. The main extracted features include:

[0118] 1. Attenuation coefficient α: reflects vascular elasticity and blood flow resistance, usually 0.2-0.8s in a healthy state -2 The α value is usually within the range of 0.2-0.4s. -2 In the range of 0.4s to 0.6s, the α value of the elderly is usually 0.4s to 0.6s due to the reduced elasticity of blood vessels. -2 In the early stages of deep vein thrombosis, the a value may rise to 0.6-0.8s -2 .

[0119] 2. Autocorrelation Amplitude (A): This value reflects the stability of the pulse wave amplitude. In a healthy state, it is typically greater than 0.85 and decreases when the pulse is unstable. Amplitude stability is related to cardiac ejection function and peripheral vascular resistance. In patients with heart failure or peripheral vascular disease, the A value may drop below 0.7.

[0120] 3. Cycle stability index: calculated by taking the value of the autocorrelation function at the cycle point, it is in the range of 0.75-0.95 in a healthy state. Reduced cycle stability may indicate arrhythmia or autonomic dysfunction. The specific calculation is Where T p is the pulse period, R(T p ) is the autocorrelation value at one period interval, and R(0) is the autocorrelation value at zero delay (equal to the signal energy).

[0121] 4. Spectral entropy: reflects the complexity of the waveform. In a healthy state, it is usually in the range of 1.2-2.0. The complexity increases with vascular disease. Spectral entropy calculation is based on the energy distribution of wavelet decomposition coefficients: in is the energy proportion of the jth scale. The turbulence caused by vascular lesions increases signal complexity and increases spectral entropy.

[0122] 5. Transit time difference: The difference in the arrival time of the pulse wave at different locations reflects vascular compliance and is typically within the range of 0.1-0.3 seconds. For example, the transit time from the heart to the wrist is approximately 0.2 seconds, and to the ankle approximately 0.3 seconds. Prolonged transit time may indicate decreased vascular compliance or obstruction.

[0123] The feature selection subunit 424 is used to select the most discriminative features from the initial feature set based on the sample discrimination and stability scores. The discrimination score is defined as:

[0124]

[0125] Among them, Score disc is the discrimination score, with larger values ​​indicating stronger discrimination ability; μ1 and μ2 are the characteristic means of the healthy group and the abnormal group, respectively; σ1 and σ2 are the standard deviations of the two groups, respectively; |μ1-μ2| represents the absolute value of the mean difference between the two groups. The stability score is defined as:

[0126]

[0127] Among them, Score stab is the stability score, the larger the value, the more stable the feature; CV is the coefficient of variation, which represents the ratio of the standard deviation to the mean; μ is the feature mean; σ is the feature standard deviation. In practical applications, the comprehensive score is calculated as the weighted sum of discrimination and stability: Score total=0.7·Scoree disc +0.3 Score stab The weight setting is based on clinical validation results, reflecting the fact that discrimination is relatively more important in vascular anomaly detection. The feature optimization subunit 424 comprehensively considers discrimination and stability, selects the 20 features with the highest comprehensive scores, and forms the feature vector F:

[0128] F=[f1,f2,...,f 20 ] T ,

[0129] Among them, F is the eigenvector, which is a 20×1 column vector; f i represents the i-th eigenvalue; the superscript T represents the vector transpose. Finally, the feature optimization subunit 424 performs Z-score normalization on the features to make features of different dimensions comparable:

[0130]

[0131] Among them, f i ′ is the standardized eigenvalue; f i is the original eigenvalue; μ i and σ i are the mean and standard deviation of the feature in the training dataset. This standardization ensures that different features can be compared on the same scale, such as the attenuation coefficient α (unit: s -2 ) and cyclic stability index (dimensionless).

[0132] Reference Figure 7 The multi-parameter adaptive judgment standard unit 43 includes a static threshold subunit 431 , a group threshold subunit 432 , a personalized threshold subunit 433 , a state recognition subunit 434 and an abnormality classification subunit 435 .

[0133] Static threshold subunit 431 is used to set fixed thresholds based on common physiological knowledge. These fixed thresholds are suitable for identifying extreme abnormalities, such as a sudden drop of more than 50% in pulse wave amplitude or complete disappearance, which generally indicates severe blood flow obstruction and requires immediate intervention. Static thresholds are generally determined based on clinical experience, for example:

[0134] Pulse wave frequency: The normal range for adults is 0.8-2Hz (corresponding to a heart rate of 48-120 beats / minute). Anything outside this range is considered abnormal.

[0135] Pulse wave amplitude: A decrease of >50% relative to baseline is considered abnormal and may indicate acute vascular occlusion;

[0136] Pulse wave transit time: A sudden prolongation of >50% is considered abnormal and may indicate acute thrombosis;

[0137] These static thresholds are universal and suitable for all user groups, mainly for capturing acute and severe vascular abnormal events. In clinical applications, for example, for high-risk patients who may have lacunar cerebral infarction, when the carotid artery pulse amplitude suddenly drops by more than 40%, the system will immediately issue an emergency warning, indicating that a vascular obstruction may occur.

[0138] The group threshold subunit 432 is used to set the group threshold according to factors such as age, gender, etc. The group threshold is based on large-scale population data statistics, taking into account physiological differences in different age groups and genders. For example, the vascular elasticity of the elderly group is usually lower than that of young people, so the upper limit of the normal range of the decay coefficient a is adjusted to 1.0 s -2 (compared to 0.8 s -2 for young people); the pulse amplitude of women is usually slightly lower than that of men, so an adjustment factor of 0.85 is considered when setting the threshold. In clinical applications, for example, for people over 70 years old, the system will automatically adjust the judgment threshold of the vascular elasticity-related parameters to reduce false positives due to physiological reduction in vascular elasticity; similarly, for child users, the system will adjust to a threshold range more suitable for their physiological characteristics.

[0139] The personalized threshold subunit 433 is used to learn the user's normal state range and establish a personalized model. The personalized threshold is calculated based on the user's historical data:

[0140] Threshold personal = μ personal ± 1.96·σper sonal ,

[0141] where Threshold personal is the personalized threshold, which can be an upper threshold (plus) or a lower threshold (minus); μ personal is the average value of the feature in the user's normal state; σ personal is the standard deviation; 1.96 corresponds to a 95% confidence interval, meaning that under normal conditions, the feature value has a 95% probability of falling within this interval. Preferably, the personalized threshold subunit 433 needs to accumulate at least 7 days of data to form a preliminary personalized model, and more than 30 days of data to form a stable personalized threshold. The application of personalized thresholds greatly improves detection accuracy, especially for people with special physiological characteristics. For example, some professional athletes may have lower resting heart rates and higher vascular elasticity, which may lead to false judgments if using universal thresholds; while personalized thresholds can adapt to these individual differences and reduce false positive rates.

[0142] The state identification subunit 434 is used to distinguish between resting, light activity, and vigorous exercise states. State identification is based on accelerometer data and heart rate variability analysis, and different judgment criteria are applied to different states. The state classification criteria are:

[0143] Resting state: acceleration root mean square <0.1g, heart rate variability SDNN <50ms;

[0144] Light activity: acceleration root mean square 0.1-0.5g, heart rate variability SDNN50-100ms;

[0145] Vigorous exercise: acceleration root mean square > 0.5g, heart rate variability SDNN > 100ms;

[0146] Among them, SDNN is the standard deviation of heart rate interval, in milliseconds (ms), which is a common indicator of heart rate variability; g is the acceleration due to gravity, which is about 9.8m / s 2 . In the resting state, the judgment update frequency is once a minute; in the active state, the frequency increases to once every 30 seconds to adapt to the rapidly changing physiological state. In actual application, state recognition greatly improves the detection accuracy of the system. For example, after strenuous exercise, the user's vascular elasticity and hemodynamic parameters will temporarily change significantly. At this time, the system will not misjudge it as a pathological state, but will adjust the judgment criteria according to the activity level; when similar changes occur in the user's resting state, it may indicate vascular abnormalities, and the system will issue an early warning.

[0147] The abnormality classification subunit 435 is used to classify abnormalities into five categories: abnormal vascular elasticity, abnormal blood flow resistance, abnormal vascular compliance, abnormal blood flow velocity, and abnormal blood flow pattern. Abnormality classification uses a decision tree method to determine the abnormality type based on a combination of different features. For example:

[0148] The attenuation coefficient α increases (>0.8s -2 ) + shortened propagation time (<0.08s) → abnormal vascular elasticity, suggesting arteriosclerosis;

[0149] The attenuation coefficient α increases (>0.8s -2) + Prolonged propagation time (>0.3s) → abnormal blood flow resistance, suggesting thrombosis;

[0150] Decreased pulse wave amplitude (<50% of baseline) + increased spectral entropy (>2.2) → abnormal vascular compliance, suggesting vascular stenosis;

[0151] Changes in pulse wave velocity (>±30%) → abnormal blood flow velocity, indicating abnormal blood pressure;

[0152] Increased waveform complexity (spectral entropy > 2.5) + decreased cyclic stability (< 0.7) → abnormal blood flow pattern, suggesting turbulent flow;

[0153] Different types of abnormalities correspond to different clinical significance and intervention measures: abnormal vascular elasticity indicates the possible presence of atherosclerosis, which requires lipid-lowering treatment; abnormal blood flow resistance indicates the possible presence of thrombosis, which requires anticoagulant treatment; abnormal vascular compliance indicates the possible presence of vascular stenosis, which requires vasodilation treatment; abnormal blood flow velocity indicates the possible presence of abnormal blood pressure, which requires adjustment of antihypertensive drugs; abnormal blood flow pattern indicates the possible presence of heart valve disease or large vessel disease, which requires further examination.

[0154] Optimally, abnormality confirmation utilizes a "triple confirmation" mechanism: a single threshold-exceeding detection triggers observation, while three consecutive threshold-exceeding detections trigger an alert, effectively reducing the false positive rate. In clinical applications, such as monitoring deep vein thrombosis in patients confined to bed after surgery, when the system detects a possible abnormal blood flow resistance, it performs three consecutive confirmations. Only if all three exceed the threshold does an alert sound, thus avoiding false alarms caused by temporary factors such as changes in body position.

[0155] Reference Figure 8 The deep learning feature automatic extraction unit 44 includes an input layer 441, a feature extraction layer 442, an attention mechanism module 443, a residual connection module 444 and an output layer 445.

[0156] Input layer 441 is used to receive 32-channel pulse wave signals, with each channel containing 128 data points (corresponding to 0.64 seconds and a sampling frequency of 200 Hz). The input data is organized as a 32×128 two-dimensional matrix, with each row representing the timing signal of a sensor channel. In practical applications of vascular monitoring, the 32-channel design can cover a sufficiently large monitoring area to capture signal characteristics at different vascular locations. For example, when monitoring deep vein thrombosis in the lower extremities, the sensor array can cover multiple key locations such as the posterior tibial vein, peroneal vein, and great saphenous vein, providing a more comprehensive assessment of vascular status.

[0157] The feature extraction layer 442 includes multiple convolutional layers and pooling layers for feature extraction from different perspectives. The present invention adopts an improved LeNet architecture, specifically designed as follows:

[0158] First convolutional layer: 8 3×1 convolution kernels (short-time convolution), stride 1, output size 32×126×8;

[0159] First pooling layer: 2×1 max pooling, stride 2, output size 32×63×8;

[0160] Second convolutional layer: 16 5×1 convolution kernels (medium-time convolution), stride 1, output size 32×59×16;

[0161] Second pooling layer: 2×1 max pooling, stride 2, output size 32×29×16;

[0162] The third convolutional layer: 32 9×1 convolution kernels (long-term convolution), stride 1, output size 32×21×32;

[0163] Third pooling layer: 3×1 max pooling, stride 3, output size 32×7×32;

[0164] This multi-view convolution design can simultaneously capture local features, medium-range features, and long-term changes, forming a comprehensive understanding of the pulse wave signal. All convolutional layers use the ReLU activation function, which is defined as:

[0165] ReLU(x)=max(0,x),

[0166] Here, ReLU(x) is the activation function output; x is the input value; and max represents the larger of the two. The ReLU function has advantages such as simple computation and stable gradients, making it suitable for deep network training. In vascular monitoring applications, convolution kernels of different sizes target different types of vascular abnormalities: short-term convolution (3×1) is suitable for capturing pulse wave spikes and is sensitive to changes in vascular resistance; medium-term convolution (5×1) is suitable for capturing pulse wave waveform changes and is sensitive to changes in vascular elasticity; and long-term convolution (9×1) is suitable for capturing changing patterns across multiple cardiac cycles and is sensitive to cardiac rhythm and hemodynamic stability. This multi-scale design enables the network to comprehensively analyze different types of vascular abnormalities.

[0167] The attention mechanism module 443 is used to automatically identify and enhance the features of important sensor channels. The channel attention mechanism calculates the importance weight of each channel:

[0168] w c =σ(W2·ReLU(W1·GAP(F c ))),

[0169] Among them, w c is the attention weight of the c-th channel, with a value range of [0,1]; F c is the feature map of the cth channel, with a dimension of L×D (L is the sequence length and D is the feature dimension); GAP represents the global average pooling operation, which compresses the L×D feature map into a 1×D vector; W1 and W2 are learnable weight matrices, which map the feature dimension from D to D / r and then back to D, respectively, where r is the dimensionality reduction ratio, usually set to 16; ReLU is the activation function; σ is the Sigmoid activation function:

[0170]

[0171] Where σ(x) is the output value of the Sigmoid function, with a range of (0, 1); x is the input value; and e is the base of the natural logarithm, which is approximately 2.71828. The attention weight is applied to the original feature to obtain the weighted feature:

[0172] F c ′=w c ·F c ,

[0173] Among them, F′ c is the weighted feature map; w c is the attention weight; F c is the original feature map; · represents the scalar multiplication by the matrix. This attention mechanism automatically identifies and strengthens sensor channels containing useful information, suppresses the influence of noise or irrelevant channels, and improves the effectiveness of feature extraction. In practical applications of vascular monitoring, the attention mechanism is particularly useful because not all sensor channels can capture valuable signals. For example, when monitoring deep vein thrombosis, only sensor channels located above the major veins contain critical information. The attention mechanism can automatically identify these channels and strengthen their weights, while suppressing interference from other channels and improving detection accuracy.

[0174] The residual connection module 444 is used to solve the gradient vanishing problem in deep networks. The residual connection is defined as:

[0175] H(x)=F(x)+x,

[0176] Here, H(x) is the output of this layer; F(x) is the output of a regular convolutional layer; x is the input; and + represents element-by-element addition. By adding skip connections, gradients can be directly propagated from deep layers to shallow layers, effectively solving the vanishing gradient problem in deep neural network training. In vascular monitoring applications, residual connections ensure that deep networks can learn effectively. This is especially true when analyzing complex vascular abnormality patterns, such as combined lesions of atherosclerosis and venous thrombosis, which require the powerful representation capabilities of deep networks.

[0177] The output layer 445 is used to generate the vascular status classification results. The output layer includes two fully connected layers:

[0178] First fully connected layer: 128 neurons, ReLU activation;

[0179] Second fully connected layer (output layer): 2 neurons, Softmax activation;

[0180] The Softmax function is defined as:

[0181]

[0182] Among them, Softmax(z_i) is the output probability of the i-th category; z iis the original output of the i-th category; K is the total number of categories. In this example, K = 2, indicating normal and abnormal states; e is the base of the natural logarithm, which is approximately 2.71828; The Softmax function converts the original output into a probability distribution, and the output value represents the probability that the sample belongs to each category. In the vascular monitoring application, the final output is [p_normal, p_abnormal], which represents the probability of normal and abnormal vascular status respectively. The system usually sets p_abnormal>0.7 as the abnormal alarm threshold, and 0.5 <p_abnormal<0.7则视为需观察状态。

[0183] Preferably, the deep learning feature automatic extraction unit 44 uses a cross entropy loss function to perform model training:

[0184]

[0185] Among them, Loss is the loss function value; N is the number of samples; K is the number of categories, in this case K = 2; y ij is the true label (0 or 1) of sample i belonging to category j; p ij is the predicted probability; log is the natural logarithm function; Indicates the sum of all N samples; represents the sum over all K categories. Training utilizes mini-batch gradient descent with a batch size of 32, an initial learning rate of 0.001, and a 10% learning rate decay after every 50 epochs. To prevent overfitting, the model utilizes dropout with a ratio of 0.5 and L2 regularization with a weight decay coefficient of 0.0001. In practical applications for vascular monitoring, model training utilizes an annotated dataset from clinical studies, including data from healthy controls and patients with various types of vascular abnormalities. Using transfer learning techniques, the model can quickly adapt to the physiological characteristics of new users, improving personalized detection accuracy.

[0186] Reference Figure 9 The multi-part information fusion decision unit 45 includes a weight allocation subunit 451 , a time series analysis subunit 452 , a risk assessment subunit 453 and a report generation subunit 454 .

[0187] The weight allocation subunit 451 is used to allocate weights to the monitoring results of different parts. The weight allocation adopts a combination of static weights and dynamic weights:

[0188] Static weights are preset based on medical knowledge, for example:

[0189] Lower limbs (ankles): weight 0.4, most prone to deep vein thrombosis;

[0190] Upper limb (wrist): weight 0.3, vascular condition is good and monitoring is stable;

[0191] Neck: weight 0.3, reflecting the blood supply to the brain;

[0192] Dynamic weights are adjusted based on signal quality and historical accuracy:

[0193]

[0194] Among them, w′ i is the adjusted weight; w i is the static weight of part i; q i is the signal quality score, the value range is [0,1], 1 represents the best quality; a i is the historical accuracy, with a value range of [0,1]; and are the average signal quality and average accuracy respectively; λ and μ are adjustment coefficients, usually set to 0.2 and 0.3; is the signal quality adjustment factor; is the accuracy adjustment factor. Finally, the weights are normalized:

[0195]

[0196] Among them, w″ i is the normalized weight, satisfying M is the total number of monitored sites; is the sum of all adjusted weights. In practical applications of vascular monitoring, dynamic weight adjustment can adapt to individual differences among users. For example, for obese patients, the ankle signal quality may be poor. The system will automatically reduce the weight there and increase the weight of areas with better signal quality to ensure the accuracy of the overall assessment.

[0197] The time series analysis subunit 452 is used to analyze short-term trend changes, medium-term periodic changes, and long-term evolution trends. Short-term trend analysis detects the trend of changes in 3-5 consecutive measurements and defines the trend index:

[0198]

[0199] Among them, Trend is the trend index, the value range is [-1,1]; N is the number of consecutive measurements, usually 5; x i is the result of the i-th measurement; sign is the sign function, which returns 1 when the input is positive, -1 when it is negative, and 0 when it is zero; It means summing all i from 1 to N-1; is the normalization factor. |Trend| > 0.6 indicates a clear trend; positive values ​​indicate an upward trend, and negative values ​​indicate a downward trend. In vascular monitoring applications, short-term trend analysis can capture acute changes, such as a rapid increase in deep vein thrombosis risk in postoperative patients, enabling timely intervention by medical staff.

[0200] Medium-term pattern recognition analyzes periodic changes within 24 hours, using autocorrelation analysis to detect periodicity:

[0201]

[0202] Among them, R day (τ) is the autocorrelation function of the 24-hour data; X(t) is the sequence of measurement results within 24 hours; T is the total time, which is 24 hours; τ is the time delay; represents the integral from 0 to T-τ; is the normalization factor. The peak position indicates the cycle length, and the peak height indicates the periodic intensity. In clinical applications, mid-term pattern recognition can identify patterns of vascular status changes associated with daily activities, such as slowed lower limb venous return after prolonged sedentary work or postprandial vasodilation, providing users with personalized lifestyle recommendations.

[0203] Long-term evolution tracking monitors slow changing trends over 7-30 days using linear regression analysis:

[0204] X(t)=β0+β1·t+∈,

[0205] Where X(t) is the measurement result at time t; β0 is the intercept, representing the initial value; β1 is the slope, representing the rate of change; t is the time variable; and ∈ is the error term, which follows a normal distribution with mean 0. A significance of P < 0.05 for β1 indicates a clear long-term trend. In vascular monitoring applications, long-term trend analysis is particularly valuable for assessing treatment efficacy or disease progression. For example, in patients with hypertension, monitoring the long-term trend of vascular elasticity can assess the effectiveness of antihypertensive medications; in patients with atherosclerosis, it can assess the effectiveness of lipid-lowering therapy.

[0206] The risk assessment subunit 453 is used to determine the risk level based on the comprehensive information. The risk assessment adopts a five-level risk level:

[0207] 1. Caution level: single abnormality, risk index 0.2-0.4, no immediate intervention required;

[0208] 2. Warning level: Continuous abnormalities but mild degree, risk index 0.4-0.6, recommended to reduce activities;

[0209] 3. Alert level: Continuous and obvious abnormalities, risk index 0.6-0.7, rest and monitoring are recommended;

[0210] 4. Emergency: Severe abnormality, risk index 0.7-0.8, immediate medical attention is recommended;

[0211] 5. Critical level: Extreme abnormality, risk index > 0.8, automatic notification of emergency contacts;

[0212] The risk index R is calculated as a weighted average:

[0213]

[0214] Among them, R is the comprehensive risk index, and its value range is [0,1]; r i is the risk score of site i, with a value range of [0,1]; w i ” is the normalized weight; represents the sum of all M monitored sites. In practical applications of vascular monitoring, risk levels are associated with specific interventions. For example, for warning-level risks, the system recommends reducing sedentary time, increasing activity, or elevating the lower limbs; for alert-level risks, the system recommends bed rest and continuous monitoring; and for emergency and critical-level risks, the system recommends immediate medical attention and can automatically notify pre-defined emergency contacts.

[0215] The report generation subunit 454 is used to integrate information and generate a comprehensive assessment report including vascular elasticity, blood flow resistance, and thrombosis risk. The report content includes: current risk level and trend; abnormality type and characteristic description; detailed data of each monitored site; historical data comparison analysis and personalized health recommendations;

[0216] Optimally, reports present results in visual charts for easy user understanding. Routine reports are generated daily, with abnormalities notified in real time, ensuring users are kept informed of their health status. In clinical applications, for hospitalized patients, reports can be directly integrated into the electronic medical record system for review by medical staff. For home users, reports are displayed via mobile apps and can be optionally shared with family doctors, forming a complete health management closed loop.

[0217] Reference Figure 10 The present invention also provides a pulse wave analysis method applied to the above-mentioned vascular hemodynamic abnormality monitoring system, comprising the following steps:

[0218] Step S1: Collecting a pulse wave signal, and sending the collected pulse wave signal to a data processing host.

[0219] In this step, the flexible piezoelectric sensor array collects human vascular pulse wave signals at a sampling frequency of 200Hz. After the wearable monitoring device pre-processes the signal, it transmits the data to the data processing host via Bluetooth 5.0. The transmitted data includes the original pulse wave signal and sensor position information. To ensure data quality, the system will evaluate the signal quality in real time, and trigger a re-collection or position adjustment prompt when the quality score is lower than 0.6 (out of 1.0). In actual application scenarios, for example, for critically ill patients who have been bedridden for a long time, the system will automatically perform a complete collection cycle every hour to monitor the risk of deep vein thrombosis; for people who sit in the office for a long time, the collection frequency can be automatically increased after sitting for 2 hours.

[0220] Step S2: Perform multi-scale decomposition on the pulse wave signal to obtain characteristic coefficient sets at different scales.

[0221] In this step, the multiscale signal decomposition and reconstruction unit receives the pulse wave signal and decomposes it into coefficient sets at multiple scales through wavelet transform. An 8-layer decomposition is performed using the Daubechies wavelet (db4) to capture signal features within the 0.5-40 Hz range. The coefficients are subjected to soft threshold denoising, and the weights are dynamically adjusted based on the energy distribution to generate reconstructed multiscale feature coefficients. This step effectively separates the features of different frequency components. For example, when monitoring the early stages of deep vein thrombosis, changes in energy distribution in the 4-8 Hz frequency band are a key indicator. Multiscale decomposition can effectively extract the features of this frequency band without interference from signals in other frequency bands.

[0222] Step S3: extracting nonlinear features from the characteristic coefficient set, including vascular elasticity, blood flow resistance, periodic stability, amplitude variability and waveform complexity features.

[0223] In this step, the nonlinear feature extraction and pattern recognition unit calculates the pulse wave autocorrelation sequence in different time windows and performs nonlinear fitting using Gaussian functions. Feature parameters reflecting various aspects of vascular status are extracted from the autocorrelation sequence and fitting results, including attenuation coefficient, autocorrelation amplitude, periodic stability index, spectral entropy, and propagation time difference. The feature optimization subunit selects the 20 most discriminative features based on the discrimination and stability scores, and performs Z-score normalization. In clinical applications, different types of vascular abnormalities exhibit different feature combinations. For example, venous thrombosis is usually manifested as increased blood flow resistance and prolonged propagation time, while atherosclerosis is mainly manifested as decreased vascular elasticity and pulse wave deformation. By extracting a comprehensive feature set, the system is able to distinguish different types of vascular abnormalities.

[0224] Step S4: Establish a multi-level adaptive judgment standard based on individual differences and physiological status to evaluate nonlinear characteristics.

[0225] In this step, the multi-parameter adaptive judgment standard unit establishes a three-level judgment standard system: static threshold, group threshold, and personalized threshold. The state recognition subunit distinguishes between resting, light activity, and intense exercise states, and applies different standards in different states. The abnormality classification subunit divides abnormalities into five categories and adopts a "three-time confirmation" mechanism to reduce the false positive rate. In actual application, for example, for elderly diabetic patients over 65 years old, the system will automatically adjust the group threshold of vascular elasticity parameters and increase the normal upper limit of the attenuation coefficient α from 0.8s to 1. -2 Adjust to 1.0s -2 , to adapt to the physiological characteristics of naturally reduced vascular elasticity in elderly diabetic patients; at the same time, in order to account for individual differences, the system will establish a personalized baseline based on at least 7 days of historical data to make the threshold more accurate.

[0226] Step S5: Use the deep learning network to automatically extract deep features from the original signal and perform vascular status classification.

[0227] In this step, the deep learning feature automatic extraction unit receives the 32-channel pulse wave signal and performs feature learning and classification through the improved LeNet network. The network includes a multi-view convolution design, an attention mechanism, and a residual connection, which can capture complex time and frequency domain features. The output layer uses a Softmax function to generate the vascular status classification results and their confidence. In vascular monitoring applications, deep learning networks can identify complex patterns that are difficult to capture with traditional feature extraction methods, especially when multiple pathological conditions coexist. For example, when a patient has both hypertension and venous insufficiency, the signal characteristics become complex, and traditional methods may be difficult to distinguish. However, deep learning networks can effectively identify these complex patterns by learning from a large amount of labeled data.

[0228] Step S6: Fusion of assessment results from multiple monitoring sites to conduct temporal pattern analysis and comprehensive risk assessment.

[0229] In this step, the multi-site information fusion decision-making unit assigns weights to the monitoring results of different sites, and analyzes short-term trend changes, medium-term cyclical changes, and long-term evolution trends. The risk assessment subunit determines the risk level based on the comprehensive information, using a five-level risk level system. In clinical applications, multi-site monitoring and information fusion significantly improve detection accuracy. For example, when monitoring the risk of deep vein thrombosis in the lower limbs, the ankle and knee positions are monitored simultaneously. If an abnormality occurs in a single site, it may be caused by local factors. However, when signals of increased blood flow resistance appear at multiple sites at the same time, it is more likely that the real risk of thrombosis is increased, and timely intervention is required.

[0230] Step S7: Generate a vascular health report, including abnormality type, risk level, and health recommendations.

[0231] In this step, the report generation subunit integrates various types of information to generate a comprehensive assessment report. The report content includes the current risk level and trend, abnormal type and its characteristic description, detailed data of each monitoring site, historical data comparison and analysis, and personalized health recommendations. Routine reports are generated once a day, and abnormal situations are pushed in real time. In clinical applications, the timeliness and pertinence of reports are crucial. For example, for patients at high risk of postoperative venous thrombosis, when the system detects an alarm-level risk, it will immediately notify the nursing station and provide specific risk sites and characteristic descriptions in the report to guide nursing staff to carry out targeted interventions; for home users, the report will provide easy-to-understand health recommendations, such as "slowed venous return in the lower limbs is detected, it is recommended to avoid maintaining the same posture for a long time, stand and move for 5 minutes every hour, and raise the lower limbs to rest when necessary".

[0232] Through the above steps, the present invention enables accurate assessment of vascular health and early warning of abnormalities, particularly the early identification of deep vein thrombosis. Compared with existing technologies, this invention offers significant advantages, including non-invasiveness, continuous monitoring, personalized assessment, and multi-site collaborative assessment, providing an effective tool for the prevention and early intervention of vascular diseases.

[0233] The present invention is not limited to the above embodiments, and those skilled in the art may make adjustments to the above embodiments without violating the spirit of the present invention. The scope of protection of the present invention shall be subject to the claims.

Claims

1. A vascular hemodynamic abnormality monitoring system, characterized in that: include: Flexible piezoelectric sensor array, used to collect human blood vessel pulse wave signals; a wearable monitoring device, electrically connected to the flexible piezoelectric sensor array, for receiving the pulse wave signal and performing preliminary preprocessing; a data processing host, connected to the wearable monitoring device via wireless communication, for receiving and processing the pulse wave signal; A pulse wave analysis module is provided in the data processing host and is used to perform multi-level analysis on the pulse wave signal. The pulse wave analysis module includes a multi-scale signal decomposition and reconstruction unit, a nonlinear feature extraction and pattern recognition unit, a multi-parameter adaptive judgment standard unit, a deep learning feature automatic extraction unit, and a multi-site information fusion decision unit connected in sequence; as well as The application platform is connected to the data processing host through a network, and is used to receive processing results and generate a vascular health report.

2. The vascular hemodynamic abnormality monitoring system according to claim 1, characterized in that: The multi-scale signal decomposition and reconstruction unit includes: a signal decomposition subunit, configured to receive the pulse wave signal and decompose the pulse wave signal into coefficient sets of multiple scales using a plurality of wavelet basis functions; a denoising processing subunit, configured to perform soft threshold shrinkage denoising on the coefficient set; and The reconstruction subunit is used to dynamically adjust the weights of coefficients at different scales according to the signal energy distribution and generate reconstructed multi-scale feature coefficients.

3. The vascular hemodynamic abnormality monitoring system according to claim 1, characterized in that: The nonlinear feature extraction and pattern recognition unit includes: An autocorrelation calculation subunit, used to calculate the pulse wave autocorrelation sequence in different time windows; a nonlinear fitting subunit, configured to perform nonlinear fitting on the autocorrelation sequence using a Gaussian function; a feature extraction subunit, configured to extract feature parameters reflecting vascular elasticity, blood flow resistance, cyclic stability, amplitude variability, and waveform complexity from the autocorrelation sequence and the fitting result; and The feature optimization subunit is used to select the most discriminative features from the initial feature set based on the sample discrimination and stability scores.

4. The vascular hemodynamic abnormality monitoring system according to claim 1, characterized in that: The multi-parameter adaptive judgment standard unit includes: A static threshold subunit, used to set a fixed threshold based on physiological common sense; The group threshold subunit is used to set the grouping threshold based on factors such as age and gender; The personalized threshold subunit is used to learn the user's normal state range and build a personalized model; a state recognition subunit to distinguish between resting, light activity, and vigorous exercise states; and The abnormality classification subunit is used to classify abnormalities into five categories: abnormal vascular elasticity, abnormal blood flow resistance, abnormal vascular compliance, abnormal blood flow velocity and abnormal blood flow pattern.

5. The vascular hemodynamic abnormality monitoring system according to claim 1, characterized in that: The deep learning feature automatic extraction unit includes: Input layer, used to receive 32-channel pulse wave signals; Feature extraction layer, including multiple convolutional layers and pooling layers, is used to extract features from different perspectives; Attention mechanism module, used to automatically identify and enhance the features of important sensor channels; Residual connection modules, used to address the vanishing gradient problem in deep networks; and The output layer is used to generate vascular status classification results.

6. The vascular hemodynamic abnormality monitoring system according to claim 1, characterized in that: The multi-part information fusion decision unit includes: The weight allocation subunit is used to allocate weights to the monitoring results of different parts; Time series analysis subunit, used to analyze short-term trend changes, medium-term cyclical changes and long-term evolution trends; A risk assessment subunit, which determines the risk level based on the combined information; and The report generation subunit is used to integrate information and generate a comprehensive assessment report including vascular elasticity, blood flow resistance and thrombosis risk.

7. The vascular hemodynamic abnormality monitoring system according to claim 1, characterized in that: The flexible piezoelectric sensor array includes 32 piezoelectric sensors, which are made of medical-grade polyvinylidene fluoride piezoelectric film and arranged in a 4×8 matrix structure. The size of a single sensor is 3mm×3mm and the thickness is 0.2mm.

8. The vascular hemodynamic abnormality monitoring system according to claim 1, characterized in that: The wearable monitoring device comprises: Data acquisition unit, used for multi-channel synchronous acquisition control; A signal preprocessing unit, used to perform bandpass filtering, notch filtering and outlier detection on the collected signal; a Bluetooth communication unit, configured to transmit the processed data to the data processing host; and The power management unit is used to provide stable power supply and achieve low power consumption management.

9. The vascular hemodynamic abnormality monitoring system according to claim 1, characterized in that: The application platform includes: A data receiving module, configured to receive the processing result sent by the data processing host; Data storage module, used to store historical monitoring data; Analysis and display module, used to generate visual charts and trend analysis; An early warning module, configured to issue an alarm when an anomaly is detected; and User interaction module, used to receive user feedback and configure system parameters.

10. The pulse wave analysis method applied to the vascular hemodynamic abnormality monitoring system of claim 1, characterized in that: The following steps are involved: Collecting pulse wave signals and sending the collected pulse wave signals to a data processing host; Performing multi-scale decomposition on the pulse wave signal to obtain characteristic coefficient sets at different scales; Extracting nonlinear features from the characteristic coefficient set, including vascular elasticity, blood flow resistance, periodic stability, amplitude variability, and waveform complexity features; According to individual differences and physiological states, a multi-level adaptive judgment standard is established to evaluate the nonlinear characteristics; Use deep learning networks to automatically extract deep features from raw signals and classify vascular status; Integrate assessment results from multiple monitoring sites to conduct temporal pattern analysis and comprehensive risk assessment; and generate vascular health reports, including abnormality types, risk levels, and health recommendations.

Images