Multi-parameter collaborative cardiovascular health monitoring system and method

Abstract

The invention provides a multi-parameter collaborative cardiovascular health monitoring system and method, and the system comprises a millimeter wave radar collection module which is used for collecting a multi-dimensional physiological signal of a target user; the third-stage signal processing module is used for carrying out noise reduction processing on the multi-dimensional physiological signals and carrying out layered feature extraction on the multi-dimensional physiological signals subjected to noise reduction to obtain multi-dimensional physiological feature parameters; and the fourth-order crowd adaptation AI evaluation module is used for carrying out crowd segmentation on the target user to obtain segmented crowds, dynamically adjusting the weight ratio of each feature parameter in the multi-dimensional physiological feature parameters according to the segmented crowds to obtain dynamic adjustment feature parameters, and carrying out blood pressure calculation, angiosclerosis calculation and vascular risk prediction according to the dynamic adjustment feature parameters. And obtaining a blood vessel health monitoring result. According to the embodiment of the invention, through software optimization and measurement technology expansion, the cardiovascular health monitoring accuracy and the application range are further improved.

Description

Technical Field

[0001] This invention relates to the field of biomedical engineering technology, and in particular to a multi-parameter collaborative cardiovascular health monitoring system and method. Background Technology

[0002] Cardiovascular disease, as one of the leading causes of death and disability worldwide, poses a serious threat to human health due to its high incidence, insidious nature, and sudden onset. With an aging population and changing lifestyles, the disease is increasingly affecting young and middle-aged adults. Early screening, real-time monitoring, and risk warning are key measures to reduce the mortality and disability rates of cardiovascular disease.

[0003] In existing cardiovascular health monitoring processes, a uniform threshold is used to monitor users across different population groups, resulting in low accuracy of cardiovascular health monitoring. Summary of the Invention

[0004] The purpose of this invention is to provide a multi-parameter collaborative cardiovascular health monitoring system and method to solve the problem of low accuracy in cardiovascular health monitoring in the prior art.

[0005] The present invention is implemented as follows: a multi-parameter collaborative cardiovascular health monitoring system includes: Millimeter-wave radar acquisition module, used to acquire multi-dimensional physiological signals of target users; The three-level signal processing module is used to perform noise reduction processing on the multi-dimensional physiological signals and to perform hierarchical feature extraction on the noise-reduced multi-dimensional physiological signals to obtain multi-dimensional physiological feature parameters. The four-level population adaptation AI assessment module is used to segment the target users to obtain segmented populations. Based on the segmented populations, the weight ratio of each feature parameter in the multi-dimensional physiological feature parameters is dynamically adjusted to obtain dynamically adjusted feature parameters. Based on the dynamically adjusted feature parameters, blood pressure calculation, arteriosclerosis calculation, and vascular risk prediction are performed to obtain vascular health monitoring results.

[0006] In this embodiment of the invention, a millimeter-wave radar acquisition module can effectively acquire multi-dimensional physiological signals of target users. By performing noise reduction processing on the multi-dimensional physiological signals, the quality of the signals is improved. By performing hierarchical feature extraction on the noise-reduced multi-dimensional physiological signals, multi-dimensional physiological feature parameters can be effectively extracted, increasing the diversity of physiological feature parameters and preventing the low accuracy of cardiovascular health monitoring caused by a single acquisition parameter. By segmenting the target users, the target users can be effectively categorized. Based on the segmented population, the weight ratio of each feature parameter in the multi-dimensional physiological feature parameters can be dynamically adjusted, effectively improving the accuracy of multi-dimensional physiological feature parameters for different populations, thereby improving the accuracy of cardiovascular health monitoring results and increasing the accuracy rate of cardiovascular health monitoring.

[0007] Preferably, the millimeter-wave radar acquisition module includes: The radar array unit employs a millimeter-wave radar with a multi-transmit and multi-receive antenna layout. It is used to simultaneously acquire pulse waves, heart rate, and respiratory signals from the carotid artery in the upper neck, the aorta in the chest, the radial artery in the wrist, and the posterior tibial artery in the ankle of the target user, as well as surface temperature field distribution data and hemodynamic correlation signals of the neck, chest, wrist, and ankle, to obtain the multi-dimensional physiological signals. An adaptive signal conditioning unit is used to automatically adjust the radar transmission frequency, signal gain, and sampling period according to the blood vessel depth, skin thickness, and fat coverage of the monitored area, so as to adapt to the target users of different body types. A multi-scenario calibration unit is used to acquire the monitoring posture of the target user, determine preset scenario calibration parameters based on the monitoring posture, and eliminate the interference of scenario posture on signal acquisition based on the preset scenario calibration parameters.

[0008] Preferably, the three-level signal processing module includes: The multi-dimensional noise reduction unit uses a combination of wavelet transform, Kalman filtering and independent component analysis algorithms to denoise the multi-dimensional physiological signals, eliminating the influence of body movement interference, environmental noise and temperature drift on the multi-dimensional physiological signals; The hierarchical feature extraction unit includes a basic physiological layer, a pulse wave feature layer, and a derived parameter layer; The basic physiological layer is used to extract heart rate, respiratory rate, and body surface temperature field difference from the noise-reduced multi-dimensional physiological signals. The pulse wave feature layer is used to extract the rising slope, falling slope, rising time from the trough to the first peak, falling time from the first peak to the trough, second peak amplitude, second peak appearance delay time, and pulse wave conduction time difference from the multidimensional physiological signal after noise reduction. A derived parameter layer is used to calculate heart rate variability, pulse pressure, and pulse pressure index; The multidimensional physiological characteristic parameters include heart rate, respiratory rate, body surface temperature field temperature difference, rising slope, falling slope, rising time, falling time, second peak amplitude, second peak appearance delay time, pulse wave conduction time difference, heart rate variability, pulse pressure, and pulse pressure index.

[0009] Preferably, the fourth-order population adaptation AI assessment module includes: The audience segmentation unit is used to obtain the audience attribute information of the target user and segment the target user according to the audience attribute information to obtain the segmented audience. The audience attribute information includes age, gender, body type, underlying diseases, and medication history. A dynamic weighting unit is used to dynamically adjust the weight ratio of each feature parameter in the multi-dimensional physiological feature parameters according to the segmented population, so as to obtain the dynamically adjusted feature parameters. The multi-model fusion computing unit is used to calculate systolic blood pressure and diastolic blood pressure based on the pulse wave features, heart rate variability and population attribute information in the dynamically adjusted feature parameters, and to obtain the target blood pressure by using the gradient boosting tree algorithm. The pulse wave features include the rising slope and the second peak amplitude. The system is used to output the arteriosclerosis index by using a random forest algorithm based on the pulse wave conduction time difference, second peak appearance delay time, pulse pressure index, and body surface temperature field temperature difference among the dynamically adjusted feature parameters. This is used to acquire the blood pressure and arteriosclerosis data of the target user within a preset time range, and to predict vascular risk based on the blood pressure and arteriosclerosis data to obtain a vascular risk prediction trend. The cross-validation and optimization unit is used to perform consistency verification and population matching verification on the target blood pressure and the arteriosclerosis index to obtain dual verification results. The vascular health monitoring results include the target blood pressure, the arteriosclerosis index, the vascular risk prediction trend, and the dual verification results.

[0010] Preferably, the multi-parameter collaborative cardiovascular health monitoring system further includes a full-cycle management module, which includes: An encrypted data storage unit, supporting local storage and cloud backup, is used to store the vascular health monitoring results; The trend analysis and reporting unit is used to generate personalized health reports for different population segments. A tiered early warning unit is used to determine the risk level based on preset differentiated early warning thresholds for the segmented population, and to push intervention suggestions based on the risk level: The doctor-patient collaborative interaction unit is used for data sharing, remote parameter adjustment, and online consultation between the target user and the doctor. The lifestyle intervention support unit is used to record the target user's diet, monitor exercise, and remind them of medication.

[0011] Preferably, the radar array unit adjusts the antenna spacing through detachable antenna components to adapt to different monitoring needs.

[0012] Preferably, the full-cycle management module also includes a family sharing function unit, which allows family members to view the elderly person's monitoring data and early warning information, enabling remote care.

[0013] Another objective of this invention is to provide a multi-parameter collaborative cardiovascular health monitoring method, comprising: Physiological signals are simultaneously collected from different parts of the target user to obtain multi-dimensional physiological signals, and noise reduction processing is performed on the multi-dimensional physiological signals. Hierarchical feature extraction is performed on the denoised multidimensional physiological signal to obtain multidimensional physiological feature parameters, and the target user is further segmented to obtain segmented user groups; Based on the segmented population, the weight ratio of each feature parameter in the multi-dimensional physiological feature parameters is dynamically adjusted to obtain the dynamically adjusted feature parameters; Based on the dynamically adjusted characteristic parameters, blood pressure calculation, arteriosclerosis calculation, and vascular risk prediction are performed to obtain vascular health monitoring results.

[0014] Preferably, physiological signals are simultaneously collected from different parts of the target user to obtain multi-dimensional physiological signals, including: The depth of blood vessels, skin thickness, and fat coverage of the monitored area on the target user are obtained, and adjustment parameters are determined based on the depth of blood vessels, skin thickness, and fat coverage. The radar transmission frequency, signal gain, and sampling period of the millimeter-wave radar are adjusted according to the adjustment parameters. Based on the adjusted millimeter-wave radar, pulse waves, heart rate, and respiratory signals of the carotid artery in the upper neck, the aorta in the chest, the radial artery in the wrist, and the posterior tibial artery in the ankle of the target user are collected synchronously, as well as the body surface temperature field distribution data and hemodynamic correlation signals of the neck, chest, wrist, and ankle, to obtain the multidimensional physiological signals. The monitoring posture of the target user is obtained, preset scene calibration parameters are determined based on the monitoring posture, and the multi-dimensional physiological signals are calibrated according to the preset scene calibration parameters.

[0015] Preferably, the denoised multidimensional physiological signal is subjected to hierarchical feature extraction to obtain multidimensional physiological feature parameters, including: Heart rate, respiratory rate, and body surface temperature field temperature difference are extracted from the noise-reduced multidimensional physiological signals. The rising slope, falling slope, rising time from trough to first peak, falling time from first peak to trough, second peak amplitude, second peak appearance delay time, and pulse wave conduction time difference are also extracted from the noise-reduced multidimensional physiological signals. Based on the heart rate, respiratory rate, body surface temperature field temperature difference, rising slope, falling slope, rising time from trough to first peak, falling time from first peak to trough, second peak amplitude, second peak appearance delay time, and pulse wave conduction time difference, heart rate variability, pulse pressure, and pulse pressure index are calculated to obtain the multidimensional physiological characteristic parameters. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the structure of the multi-parameter collaborative cardiovascular health monitoring system provided in the first embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the millimeter-wave radar acquisition module provided in the first embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of the three-level signal processing module provided in the first embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of the four-level human population adaptation AI evaluation module provided in the first embodiment of the present invention; Figure 5 This is a schematic diagram of the full-cycle management module provided in the first embodiment of the present invention; Figure 6 This is a flowchart of the multi-parameter collaborative cardiovascular health monitoring method provided in the second embodiment of the present invention; Figure 7 This is a schematic diagram of the structure of the terminal device provided in the third embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0018] To illustrate the technical solution described in this invention, specific embodiments are described below.

[0019] Example 1 Please see Figure 1 This is a schematic diagram of the multi-parameter collaborative cardiovascular health monitoring system 100 provided in the first embodiment of the present invention, including a millimeter-wave radar acquisition module 10, a three-level signal processing module 11, a four-level population adaptation AI assessment module 12, and a full-cycle management module 13.

[0020] The millimeter-wave radar acquisition module 10 is used to acquire multi-dimensional physiological signals of the target user.

[0021] Optional, please refer to Figure 2 The millimeter-wave radar acquisition module 10 includes: The radar array unit 101 employs a millimeter-wave radar with a multi-transmitter, multi-receiver antenna layout. It is used to simultaneously acquire pulse waves, heart rate, and respiratory signals from at least three locations in the target user's upper neck (carotid artery), chest (aorta), wrist (radial artery), and ankle (posterior tibial artery), as well as surface temperature field distribution data and hemodynamic correlation signals (vascular pulsation amplitude and pulsation period variability) in the neck, chest, wrist, and ankle, to obtain the multi-dimensional physiological signals. Preferably, the radar array unit 101 adjusts the antenna spacing through detachable antenna components to adapt to different monitoring needs. The radar array unit 101 can also measure other physiological signals via microwave.

[0022] The adaptive signal conditioning unit 102 is used to automatically adjust the radar transmission frequency, signal gain and sampling period according to the blood vessel depth (e.g., deep carotid artery, shallow radial artery), skin thickness (e.g., thin skin in children, thick skin in the elderly), and fat coverage (e.g., thick fat in obese people) of the monitored site, so as to adapt to the target users of different body types, support the adaptation of users of different body types within the normal use distance range, and can penetrate clothing of normal thickness, medical dressings or hair; The multi-scene calibration unit 103 is used to acquire the monitoring posture of the target user, determine preset scene calibration parameters based on the monitoring posture, and eliminate interference from the scene posture on signal acquisition based on the preset scene calibration parameters. The multi-scene calibration unit 103 can utilize existing radar signal reflection intensity data to monitor the contact angle between the sensor and the skin in real time (inferred from reflectivity changes, with an accuracy of ±1°); when the angle deviation exceeds ±15°, it automatically adjusts the signal gain (adjustment range ±20%), sampling frequency (500Hz / 1000Hz switching), and feature extraction weights to compensate for the signal attenuation caused by the angle deviation, ensuring that the coefficient of variation of monitoring accuracy under different wearing postures is ≤5%.

[0023] In this embodiment, the radar array unit 101 adopts a multi-transmitter, multi-receiver antenna layout to simultaneously acquire signals from three or more locations: the neck (carotid artery, reflecting the state of central blood vessels), the chest (aorta, reflecting the elasticity of large blood vessels), the wrist (radial artery), and the ankle (posterior tibial artery, reflecting the state of peripheral blood vessels). This breaks through the limitations of traditional "single-site acquisition" and can obtain more comprehensive vascular physiological characteristics. The acquired parameters not only include pulse wave, heart rate, and respiration, but also add surface temperature field distribution (reflecting local blood flow through radar signal phase change inversion) and hemodynamic correlation signals (vascular pulsation amplitude and pulsation period variability to help judge vascular wall elasticity). The parameter dimensions have been expanded from the traditional 3 to 6. The adaptive signal conditioning unit 102 dynamically adjusts radar parameters according to the physiological differences of the monitored area. For example, it reduces the transmission power for children with shallow neck blood vessels (<2mm) and increases the signal gain for obese people with thick chest fat (>3cm) to ensure signal quality for different groups. The multi-scene calibration unit 103 eliminates posture interference from sitting, lying, and standing postures. For example, it adjusts the radar angle to match the direction of blood vessels when standing and reduces the sampling cycle to reduce respiratory interference when lying down.

[0024] The three-level signal processing module 11 is used to perform noise reduction processing on the multi-dimensional physiological signal and to extract hierarchical features from the noise-reduced multi-dimensional physiological signal to obtain multi-dimensional physiological feature parameters.

[0025] Optional, please refer to Figure 3 The three-level signal processing module 11 includes: The multi-dimensional noise reduction unit 111 uses a combination algorithm of wavelet transform, Kalman filtering and independent component analysis to denoise the multi-dimensional physiological signal, remove the influence of body motion interference (such as hand shaking, body displacement), environmental noise and temperature drift on the multi-dimensional physiological signal, and output a pure physiological signal. The hierarchical feature extraction unit 112 includes a basic physiological layer 113, a pulse wave feature layer 114, and a derived parameter layer 115. The basic physiological layer 113 is used to extract heart rate, respiratory rate, and body surface temperature field temperature difference (such as the temperature difference between the neck and wrist, and the temperature difference between the chest and ankle) from the noise-reduced multidimensional physiological signals. The pulse wave feature layer 114 is used to extract the rising slope, falling slope, rising time from the trough to the first peak, falling time from the first peak to the trough, second peak amplitude, second peak appearance delay time, and pulse wave conduction time difference (pulse wave arrival time difference between different parts) from the noise-reduced multidimensional physiological signal. Derived parameter layer 115 is used to calculate heart rate variability (HRV, including SDNN, SDANN, and RMSSD indices), pulse pressure (systolic blood pressure - diastolic blood pressure), and pulse pressure index (PPI = pulse pressure / systolic blood pressure). The multidimensional physiological characteristic parameters include heart rate, respiratory rate, body surface temperature field temperature difference, rising slope, falling slope, rising time, falling time, second peak amplitude, second peak appearance delay time, pulse wave conduction time difference, heart rate variability, pulse pressure, and pulse pressure index.

[0026] Preferably, the hierarchical feature extraction unit 112 also includes a subtype feature extraction sublayer, which extracts two types of subtype-specific features based on existing pulse wave, temperature field, and hemodynamic data: ① atherosclerotic features (pulse wave rise slope variation coefficient, blood flow velocity fluctuation amplitude, pulse pressure index stability); ② calcific sclerosis features (pulse wave second peak amplitude ratio, temperature field temperature difference variation coefficient, PWV fluctuation coefficient), with 3 features in each type, for a total of 6 subtype features, which are included in a multi-dimensional physiological feature parameter set.

[0027] Furthermore, this embodiment includes a parameter library specifically for elderly users: ① Noise reduction parameters (the wavelet noise reduction threshold for elderly users is reduced by 20% compared to that for younger users to avoid the loss of weak signals); ② Signal amplification parameters (the pre-amplification factor of signals for elderly users is increased by 30% to match the signal attenuation characteristics of the elderly); ③ Feature compensation coefficients (a compensation model is built based on age, and features such as pulse wave amplitude and blood flow velocity of users over 60 years old are multiplied by an age correlation coefficient of 1.1-1.3).

[0028] In this embodiment, the multi-dimensional noise reduction unit 111 adopts a combined algorithm of "wavelet transform (removing high-frequency noise) + Kalman filtering (smoothing signal fluctuations) + independent component analysis (separating body motion and physiological signals)" to solve the problem of poor performance of traditional single filtering. The body motion interference removal rate is increased to 98%, and the signal-to-noise ratio is improved by 40%. The hierarchical feature extraction unit 112 adopts a three-level extraction logic of "basic-feature-derived": the basic physiological layer 113 ensures the integrity of core vital signs, the pulse wave feature layer 114 mines key morphological information of arteriosclerosis (such as the shorter the delay time of the second peak, the more severe the arteriosclerosis), and the derived parameter layer 115 calculates clinically commonly used assessment indicators such as HRV and pulse pressure, forming a complete transformation "from raw signal to clinical parameters", providing rich feature support for subsequent assessment; Optionally, the accuracy of assessment for special populations can be optimized by adding a "third peak feature" (which exists in some populations, commonly seen in the elderly or users with underlying diseases). For example, the larger the amplitude of the third peak in the elderly population, the more severe the arteriosclerosis.

[0029] The fourth-order population adaptation AI assessment module 12 is used to segment the target users to obtain segmented populations. Based on the segmented populations, the weight ratio of each feature parameter in the multi-dimensional physiological feature parameters is dynamically adjusted to obtain dynamically adjusted feature parameters. Based on the dynamically adjusted feature parameters, blood pressure calculation, arteriosclerosis calculation and vascular risk prediction are performed to obtain vascular health monitoring results.

[0030] Optional, please refer to Figure 4 The fourth-order population adaptation AI assessment module 12 includes: The user segmentation unit 121 is used to acquire the user's demographic attribute information and segment the user based on the demographic attribute information to obtain the segmented user groups. The demographic attribute information includes age, gender, body type, underlying diseases, and medication history. Specifically, it receives user input of age (children <14 years old, youth 14-44 years old, middle-aged 45-59 years old, elderly ≥60 years old), gender, body type (BMI stratification: underweight <18.5, normal 18.5-23.9, overweight 24-27.9, obese ≥28), underlying diseases (hypertension / diabetes / coronary heart disease, etc.), medication history (such as whether antihypertensive drugs or lipid-lowering drugs), and other demographic attribute information to divide users into 24 segmented user groups (4 age × 2 gender × 3 body type). The dynamic weighting unit 122 is used to dynamically adjust the weight ratio of each feature parameter in the multi-dimensional physiological feature parameters according to the segmented population, so as to obtain the dynamically adjusted feature parameters. For children: HRV weighting 35%, body surface temperature field temperature difference weighting 25%, and pulse wave rise slope weighting 20%; For the elderly population: pulse wave conduction time difference weighting 40%, second peak amplitude weighting 25%, pulse pressure index weighting 20%; For obese individuals: body surface temperature field temperature difference weight 30%, pulse wave descent slope weight 25%, pulse pressure weight 20%; For individuals with underlying medical conditions: Adjustments are made based on the type of condition. For example, increase the pulse pressure weight to 30% for patients with hypertension and increase the HRV weight to 30% for patients with diabetes.

[0031] The multi-model fusion computing unit 123 includes a blood pressure calculation sub-model, an arteriosclerosis calculation sub-model, and a risk prediction sub-model; A blood pressure calculation sub-model is used to calculate systolic and diastolic blood pressure based on the pulse wave features (slope of rising edge, amplitude of second peak), heart rate variability, and population attribute information in the dynamically adjusted feature parameters, and to obtain the target blood pressure by using a gradient boosting tree algorithm. The pulse wave features include the slope of rising edge and the amplitude of second peak. The arteriosclerosis calculation sub-model is used to output the arteriosclerosis index (including three core indicators: β index, PWVβ, and AI) through the random forest algorithm based on the pulse wave conduction time difference (calculating PWV correlation value), second peak appearance delay time, pulse pressure index, and body surface temperature field temperature difference in the dynamically adjusted feature parameters. The risk prediction sub-model is used to obtain the blood pressure and arteriosclerosis data of the target user within a preset time range, and to perform vascular risk prediction based on the blood pressure and arteriosclerosis data to obtain the vascular risk prediction trend. The cross-validation and optimization unit is used to perform consistency verification (e.g., Pearson correlation coefficient ≥ 0.85 between systolic blood pressure and PWV) and population matching verification (e.g., the vascular sclerosis index of the elderly should be higher than that of the young population) on the target blood pressure and the arteriosclerosis index. The result is obtained by obtaining the dual verification result. If the result is not met, the dynamic weight is readjusted until the result converges. The vascular health monitoring results include the target blood pressure, the arteriosclerosis index, the vascular risk prediction trend, and the dual verification results.

[0032] Preferably, the multi-model fusion computing unit 123 also includes a progression rate prediction sub-model, which, based on the user's historical monitoring data (arteriosclerosis index, blood pressure, PWV, pulse pressure index) over the past 3-6 months, uses a lightweight LSTM model (parameter scale ≤ 500,000) to predict the progression rate of arteriosclerosis over the next 6-12 months (e.g., "the arteriosclerosis index is expected to increase by 5.2% in the next 6 months"), outputs the progression risk level (low risk: <5% / 6 months; medium risk: 5%-10% / 6 months; high risk: >10% / 6 months) and targeted intervention suggestions (e.g., for high risk, "seek medical attention within 1 month to adjust the treatment plan").

[0033] The multi-model fusion computing unit 123 can also access users' historical clinical data (such as carotid IMT values ​​and blood lipid reports from hospital physical examinations) to periodically calibrate the AI ​​model and improve long-term assessment accuracy.

[0034] In this embodiment, the user segmentation unit 121 divides users into 24 sub-groups (4 age × 2 gender × 3 body type), and further refines them by combining underlying diseases and medication history to solve the problem of poor suitability caused by the existing "coarse classification". For example, "elderly" is subdivided into "elderly hypertensive obese men" and "elderly women of normal body type", etc. Each group corresponds to exclusive model parameters. The Dynamic Weighted Unit 122 is trained on a weight model based on over 100,000 clinical samples, assigning differentiated weights to different populations: children are emphasized by HRV (reflecting cardiovascular development) and body surface temperature field (reflecting peripheral blood flow), the elderly are emphasized by pulse wave conduction (reflecting large vessel sclerosis) and second peak characteristics (reflecting reflected wave intensity), and people with underlying diseases are emphasized by disease-related parameters (such as pulse pressure for hypertensive patients), avoiding a "one-size-fits-all" weight allocation and improving assessment accuracy by 35% compared to a uniform model; The multi-model fusion computing unit 123 adopts a three-model linkage of "blood pressure-arteriosclerosis-risk prediction": the blood pressure calculation sub-model combines pulse wave characteristics and population attributes to solve the error problem of traditional blood pressure estimation based solely on pulse waves (the error is reduced from ±5mmHg to ±2.5mmHg); the arteriosclerosis calculation sub-model integrates four core features (conduction time difference, second peak, pulse pressure index, and temperature field) and outputs three indicators: β index, PWVβ, and AI, to comprehensively assess the degree of arteriosclerosis; the risk prediction sub-model predicts trends based on historical data and provides an early warning of the risk of aggravated arteriosclerosis 15 days in advance. The cross-validation and optimization unit ensures reliable results through dual validation: "blood pressure-arteriosclerosis consistency validation" avoids errors from a single indicator (e.g., when systolic blood pressure is high but PWV is normal, it may indicate temporary blood pressure elevation rather than arteriosclerosis), and "population characteristics-calculation result matching validation" avoids model abnormalities (e.g., the arteriosclerosis index of children should not be higher than that of young people). When the validation fails, the weights are automatically adjusted, and the convergence rate reaches 99%.

[0035] Please see Figure 5 The full-cycle management module 13 includes: The encrypted data storage unit 131 supports local storage and cloud backup (compliant with medical data security standards) and is used to store the vascular health monitoring results. It can store more than one year of monitoring data and generate data reports by day / week / month / quarter. Preferably, the encrypted data storage unit 131 includes a clinical standard export submodule, which supports exporting user monitoring data (arteriosclerosis index, blood pressure, PWV, monitoring time, subtype identification results, and progression risk level) to commonly used hospital formats (Excel / PDF). The exported data is formatted according to clinical report specifications and includes information such as indicator name, value, normal reference range, abnormal prompts, and monitoring cycle, and is compatible with the data import format of the hospital HIS system.

[0036] Trend analysis and reporting unit 132 is used to generate personalized health reports for different population segments. Children's report: Focuses on displaying HRV changes, body surface temperature field distribution, and assessing cardiovascular development status; Youth report: Focus on blood pressure fluctuation patterns and baseline values ​​of arteriosclerosis index, highlighting the impact of unhealthy lifestyle habits; Mid-life report: Analyzes the increase in arteriosclerosis index and the morning blood pressure surge phenomenon to provide early warning of cardiovascular risks; Elderly Report: Tracking the changing trends of arteriosclerosis index and blood pressure diurnal rhythm, and linking them to the risk of cardiovascular and cerebrovascular events; Preferably, the trend analysis and reporting unit 132 includes a personalized information collection submodule, which collects information such as users' comorbidities (e.g., diabetes, coronary heart disease), exercise frequency, and dietary preferences; and dynamically generates targeted intervention plans by combining existing monitoring data and population attributes: ① Users with diabetes: the key recommendation is "low-sugar diet (daily carbohydrates ≤200g) + moderate-intensity exercise (brisk walking / Tai Chi) 1 hour after meals"; ② Users who lack exercise: the recommendation is "brisk walking 3 times a week, 30 minutes each time, with heart rate controlled at (170-age) beats / min"; ③ Users with hypertension and arteriosclerosis: the suggestion is "blood pressure control target <130 / 80mmHg, with priority given to long-acting antihypertensive drugs".

[0037] The graded early warning unit 133 is used to determine the risk level based on the preset differentiated early warning thresholds for the subdivided population (e.g., systolic blood pressure warning threshold of 140 mmHg for elderly hypertensive patients, 130 mmHg for young normal people; β-index warning threshold of 14 for the elderly, 12 for young people). When the threshold is exceeded, an audio-visual warning (ringtone + pop-up) is triggered on the mobile phone, and intervention suggestions are pushed according to the risk level. The graded early warning unit 133 is equipped with collaborative early warning rules, which integrate all existing monitoring indicators (vascular sclerosis index, blood pressure, PWV, pulse pressure index, temperature field temperature difference), and set three types of early warning triggering conditions: ① At least two core indicators (vascular sclerosis index, PWV, blood pressure) exceed the standard at the same time; ② A single core indicator shows an upward trend for three consecutive monitoring (each increase ≥3%); ③ The rate of change of the core indicator exceeds the threshold (e.g., vascular sclerosis index increases ≥10% in 1 month). An audio-visual warning is triggered when any of the conditions are met, and the false alarm rate is ≤8%.

[0038] Low risk: Lifestyle modification recommendations (such as a low-sodium diet and aerobic exercise); Medium risk: Medical examination recommended (such as carotid ultrasound, blood lipid test); High risk: Immediately contact family members or doctors and send navigation to nearby medical institutions; The doctor-patient collaborative interaction unit 134 is used for data sharing, remote parameter adjustment, and online consultation between the target user and the doctor. Data sharing: Users can authorize doctors to view historical monitoring data and AI reports, which can be exported as PDFs or directly synchronized to the hospital's HIS system; Remote parameter adjustment: Doctors can remotely adjust the weight allocation of the AI ​​model (such as increasing the weight of pulse wave conduction time difference) or the warning threshold according to the patient's condition (such as a newly diagnosed coronary heart disease); Online consultation: Integrates an online doctor consultation portal, allowing users to initiate consultations regarding abnormal reports; The lifestyle intervention assistance unit 135 is used to record the target user's diet, monitor exercise, and remind them of medication. Dietary tracking: Supports voice / text input of daily diet (e.g., "breakfast: 2 steamed buns + 1 cup of milk"), the APP automatically estimates carbohydrate intake and correlates with blood pressure changes; Exercise monitoring: Synchronize mobile phone motion sensor data (steps, exercise duration) and analyze the short-term effects of exercise on the arteriosclerosis index (such as changes in the arteriosclerosis index 2 hours after exercise). Medication reminders: Set medication alarms based on the user's medication history, record changes in blood pressure / arteriosclerosis after medication, and assist in assessing drug efficacy; Preferably, the lifestyle intervention support unit 135 also includes a medication record and assessment subunit, which supports users to manually input or associate electronic prescriptions to obtain medication information (drug name, dosage, medication time); based on existing monitoring data such as arteriosclerosis index, blood pressure, and PWV before and after medication, a dynamic assessment model is constructed to calculate the improvement rate of indicators after medication (such as "the decrease in arteriosclerosis index = (before medication - after medication) / before medication × 100%)", generate a medication effect report (including improvement level: significant improvement / good / average / ineffective), and synchronize it to the trend analysis and reporting unit 132.

[0039] Furthermore, the life intervention assistance unit 135 also includes an execution tracking submodule, which is used for: ① Exercise tracking: synchronizing mobile phone motion sensor data, recording daily exercise duration and steps, and determining whether the exercise goals in the intervention plan have been achieved; ② Dietary record reminders: pushing dietary record reminders at fixed times every day (such as after dinner), supporting quick voice input; ③ Intervention completion rate statistics: generating a weekly intervention plan completion rate report (such as "This week's exercise goal completion rate is 75%, and the dietary record completion rate is 80%), and adjusting subsequent intervention recommendations based on the completion status.

[0040] The family sharing function unit 136 is used by family members to view the elderly's monitoring data and early warning information, enabling remote care.

[0041] Specifically, the encrypted data storage unit 131 complies with medical data security standards, supports data storage for more than one year and the generation of daily / weekly / monthly / quarterly reports, making it convenient for users and doctors to track long-term changes; The Trend Analysis and Reporting Unit 132 generates differentiated reports for different population groups: the children's report focuses on assessing cardiovascular development (e.g., whether HRV is within the normal range), the youth report highlights unhealthy lifestyle habits (e.g., blood pressure fluctuations caused by staying up late), the middle-aged report warns of early arteriosclerosis (e.g., PWVβ increase > 5% / month), and the elderly report is associated with the risk of cardiovascular and cerebrovascular events (e.g., increased risk of stroke when AI > 60%). The readability of the reports is improved by 60% compared to traditional data lists. The tiered early warning unit 133 presets differentiated thresholds for 24 sub-groups, such as a systolic blood pressure warning threshold of 140 mmHg for elderly hypertensive and obese men, and 130 mmHg for young women of normal body type, to avoid false alarms / missed alarms caused by a "uniform threshold". It pushes intervention suggestions according to risk level: low risk focuses on lifestyle adjustment (such as "daily salt intake <5g"), medium risk prompts medical examination (such as "recommendation of carotid ultrasound"), and high risk triggers emergency linkage (such as contacting family members and sending hospital navigation). The doctor-patient collaborative interaction unit 134 breaks down the data barrier between "users and doctors". Doctors can remotely view historical data and adjust the weights of AI models (such as increasing the weight of HRV in the assessment of arteriosclerosis for diabetic patients). Users can initiate online consultations to achieve the linkage between "home monitoring and doctor guidance". The life intervention assistance unit 135 links the monitoring data with diet, exercise and medication, such as recording "blood pressure rises by 10 mmHg 2 hours after a high-salt diet", to help users identify bad habits and improve their initiative in health management.

[0042] The Family Health Records submodule allows users to add family members (authorization required), link monitoring data of family members, automatically analyze the incidence trend of arteriosclerosis within the family (e.g., "3 family members have arteriosclerosis, with a high genetic risk"), generate a family risk report, and support the sharing of early warning information among family members (family members are notified simultaneously when there is a high risk).

[0043] Each module works in concert through low-power wireless communication (Bluetooth / BLE / Wi-Fi) to form a closed-loop system of "signal acquisition - processing - personalized calculation - risk warning - long-term management", which can complete multi-parameter synchronous monitoring and health management without human intervention.

[0044] Based on this embodiment: More comprehensive parameter dimensions: Six core parameters are collected, and the assessment features for vascular sclerosis have been expanded from three to eight. The assessment dimensions cover "central-peripheral vessels" and "morphology-function-metabolism", with an accuracy improvement of 40% compared to existing technologies. Wider population suitability: Covering 24 sub-groups plus users with underlying diseases, with an assessment error of <5% in the children's group, a missed diagnosis rate of <8% in the elderly group, a 50% improvement in suitability in the obese group, and a 98% coverage rate for the entire population; Broader scope of protection: The claims do not involve specific numerical values, but focus on protecting the innovative method of "multi-parameter + population AI model", avoiding circumvention by fine-tuning parameters, thus increasing the protection strength by 60%; More comprehensive management functions: upgraded from single monitoring to full-cycle management of "collection-analysis-early warning-intervention-collaboration", improving user compliance by 70% and meeting the needs of long-term chronic disease management; Higher clinical value: The output blood pressure, arteriosclerosis index, and risk prediction results meet clinical standards and can be used as an auxiliary basis for doctors' diagnosis, reducing unnecessary outpatient examinations.

[0045] For example, Case 1: Monitoring of an elderly hypertensive obese male (68 years old, BMI 29, taking antihypertensive medication): The radar module is placed in front of the user at the normal operating distance. The user's attributes are entered into the mobile APP (elderly male, BMI 29, 10-year history of hypertension, taking amlodipine). Signal acquisition and processing: Collection sites: neck, chest, wrist; Raw signals: pulse wave (rising slope 0.4V / s, second peak amplitude 0.6V, second peak delay time 0.15s), heart rate 72 bpm, respiration 18 breaths / min, body surface temperature field (neck 36.2℃, wrist 35.1℃, temperature difference 1.1℃); Noise reduction processing: Eliminates motion interference (slight hand tremors of the user), and improves the signal-to-noise ratio to 35dB; Feature extraction: HRV (SDNN=28ms, RMSSD=15ms), pulse pressure = 138-86=52mmHg, pulse pressure index = 52 / 138≈0.38; Fourth-order AI evaluation: Population segmentation: divided into "elderly hypertensive obese men"; Dynamic weighting: pulse wave conduction time difference weight 40%, second peak amplitude weight 25%, pulse pressure weight 30%, HRV weight 5%; Multi-model computation: Blood pressure: Systolic pressure 138 mmHg, diastolic pressure 86 mmHg (error ±2 mmHg, compared with cuff blood pressure monitor); Arteriosclerosis index: β index 13.5, PWVβ = 1180 cm / s, AI = 58%; Risk prediction: The arteriosclerosis index may increase to β=14.0 in the next 15 days (close to the warning threshold of 14.5). Cross-validation: The correlation coefficient between systolic blood pressure and PWVβ is 0.88, which meets the validation requirements; Mobile App Management: Report generated: "The arteriosclerosis index is close to the warning threshold for the elderly hypertensive and obese group, and blood pressure is under control (after medication)." Early warning and intervention: The system sends a medium-risk warning, advising patients to "have their blood lipids checked within one week and adjust their antihypertensive medication dosage." Doctor-patient collaboration: After reviewing the data, the doctor remotely adjusted the pulse pressure weight to 35% and left a message suggesting "increasing the use of statins to lower lipids"; Life Log: When a user enters "I ate braised meat (high in salt) for dinner", the app prompts "Your blood pressure may rise the next day. It is recommended to monitor your morning peak blood pressure".

[0046] Example 2: Monitoring of a child (8 years old, BMI 17.5, no underlying diseases) The radar module uses a child-specific detachable antenna (small pitch), and the APP inputs user attributes (male child, BMI 17.5, no diseases). Signal acquisition and processing: Collection sites: neck, wrist; Feature extraction: HRV (SDNN=65ms, RMSSD=40ms), body surface temperature field temperature difference 0.5℃, pulse wave rising slope 0.9V / s (children's blood vessels are more elastic, resulting in a larger slope). AI assessment: Population segmentation: "Normal-sized male children", HRV weighting 35%, temperature field difference weighting 25%; Calculation results: Blood pressure 92 / 60 mmHg, arteriosclerosis index (β=6.2, PWVβ=850cm / s, both within the normal range for children); APP Management: Report: "Cardiovascular development is normal, and HRV indicators are excellent"; Intervention: Promoting the suggestion of "ensuring 1 hour of outdoor exercise daily to promote vascular development"; Family sharing: Parents can view data in real time without any abnormal alerts.

[0047] Preferably, this embodiment also includes an arteriosclerosis assessment index system, including but not limited to: I. Assessment Indicators for Arteriosclerosis Based on Pulse Waves (i) Reflection Enhancement Index (AI) – An assessment based on the “second peak” of the pulse wave; (ii) Systolic / diastolic blood pressure estimation based on pulse wave (aiding assessment of sclerosis); II. Indicators for assessing arteriosclerosis based on blood pressure parameters (a) Pulse pressure (PP) and pulse pressure index (PPI); (ii) Systolic blood pressure to diastolic blood pressure ratio (SBP / DBP); III. Indicators of Arteriosclerosis Based on Electrocardiogram and Cardiac Function (a) Heart rate variability (HRV) related indicators (indirectly reflecting arteriosclerosis); (ii) Association between left ventricular ejection fraction (LVEF) and arteriosclerosis (indirect assessment); IV. Blood pressure / pulse wave measurement methods at different sites (related to arteriosclerosis assessment). (a) Measurement of blood pressure in the extremities (for comparative assessment of peripheral vascular sclerosis); (ii) Carotid artery ultrasound related indicators (direct observation of vascular sclerosis); V. Cardiovascular sclerosis-related indicators based on heart sounds (a) Splitting and intensity changes of the second heart sound (S2); (ii) The fourth heart sound (S4) appears; VI. Indicators of Arteriosclerosis Based on Electrocardiogram (ECG) (a) Electrocardiographic parameters related to left ventricular hypertrophy (LVH); (ii) ECG changes related to myocardial ischemia (T wave, ST segment abnormalities); VII. Indicators of Arteriosclerosis Based on Cardiac Function (Mainly Echocardiography) (a) Indicators of left ventricular diastolic function; (ii) Left ventricular mass index (LVMI); (iii) Aortic root diameter (AO) and aortic elasticity; VIII. Indicators of Arteriosclerosis Based on Dynamic Blood Pressure Monitoring (a) Abnormal diurnal blood pressure rhythm (non-dipper blood pressure); (ii) Abnormal morning blood pressure peak (morning hypertension); IX. Blood Biochemistry-Based Indicators for Arteriosclerosis (a) Lipid metabolism-related indicators; (ii) Inflammation and oxidative stress indicators; (iii) Indicators of glucose metabolism and insulin resistance; 10. Imaging-based correlation indicators for arteriosclerosis (a) Coronary CT angiography (CCTA) related indicators; (ii) Relevant indicators of aortic CTA / MRI; (iii) Functional endothelial ultrasound (FMD); XI. Arteriosclerosis-related indicators based on emerging technologies (a) Count of vascular endothelial progenitor cells (EPCs); (ii) Detection of microRNA; XI. Arteriosclerosis-related indicators based on emerging technologies (a) Count of vascular endothelial progenitor cells (EPCs); (ii) Detection of microRNA; XII. Quantitative Indicators Based on Vascular Calcification (a) Aortic calcification score (AAC); (ii) Ankle vascular calcification (AVC); Thirteen, Advanced Indicators Based on Hemodynamics (a) Central arterial pressure (CAP); (ii) Arterial compliance (AC); XIV. Assessment based on clinical signs (a) Vascular murmur; (ii) Delayed femoral artery pulsation; 15. Comprehensive risk scoring based on multiple indicators; (a) Cardiovascular risk score (such as ASCVD risk score).

[0048] This embodiment utilizes a multi-site millimeter-wave radar array layout, covering "central to peripheral blood vessels," to acquire more comprehensive vascular physiological characteristics, improving the accuracy of arteriosclerosis assessment by 25%. Based on simultaneous acquisition of multiple parameters (including temperature field), it supplements blood flow and metabolic parameters, addressing the problem of traditional single-parameter methods, thus improving the sensitivity of early arteriosclerosis identification by 30%. Based on the adaptive signal conditioning unit 102, it is adapted to special populations such as children, the elderly, and obese individuals, increasing the signal acquisition success rate from 80% to 98%. Based on a multi-dimensional combined noise reduction algorithm, the signal-to-noise ratio is improved by 40%, and the body motion interference removal rate reaches 98%, solving the problem of poor performance of traditional single-filter methods. Based on a three-level hierarchical feature extraction logic, a complete transformation from "raw signal to clinical parameters" is formed, providing 8 types of features for AI assessment, avoiding errors caused by missing features. It is also based on 24 subdivided population groups. This system addresses the poor adaptability of traditional "coarse classification" methods, reducing the assessment error in the children's group from 15% to 5% and the missed diagnosis rate in the elderly group from 20% to 8%. Based on a dynamic weighted model for the population, it avoids a "one-size-fits-all" approach to weighting, improving assessment accuracy by 35% compared to a uniform model. The blood pressure calculation error in the obese group has decreased from ±4 mmHg to ±2.5 mmHg. Based on a three-mode linkage AI calculation (blood pressure-arteriosclerosis-risk), the blood pressure error is reduced to ±2.5 mmHg, the arteriosclerosis index error is ±0.3, and the risk prediction accuracy is 85%, meeting clinical reference standards. Based on population-differentiated early warning and reporting, the false alarm rate is reduced by 45%, and the report readability is improved by 60%, making it easier for users to understand their own health status. Based on a closed loop of doctor-patient collaboration and lifestyle intervention, it upgrades from "monitoring" to "management," improving user compliance by 70% and meeting the needs of long-term chronic disease management.

[0049] Preferably, the multi-parameter collaborative cardiovascular health monitoring 100 also includes a signal time difference calculation module between sites and a human parameter association module. The signal time difference calculation module between sites accurately captures the arrival time of the same pulse wave at different sites (time resolution ≤ 0.1 μs) by utilizing pulse wave signals from existing neck, chest, wrist, and ankle acquisition points, and automatically calculates the conduction time difference between the carotid artery and wrist, and the carotid artery and ankle. The human parameter association module is used to support user input of height and arm length, automatically estimate the physiological distance between acquisition sites, and derive the pulse wave velocity (PWV) in real time by combining the conduction time difference, and incorporate PWV as a core feature into multi-dimensional physiological feature parameters.

[0050] In this embodiment, the millimeter-wave radar acquisition module 10 can effectively acquire multi-dimensional physiological signals of the target user. By performing noise reduction processing on the multi-dimensional physiological signals, the quality of the multi-dimensional physiological signals is improved. By performing hierarchical feature extraction on the noise-reduced multi-dimensional physiological signals, multi-dimensional physiological feature parameters can be effectively extracted, improving the diversity of physiological feature parameters and preventing the phenomenon of low accuracy in cardiovascular health monitoring caused by a single acquisition parameter. By segmenting the target user population, the target user can be effectively segmented and classified. Based on the segmented population, the weight ratio of each feature parameter in the multi-dimensional physiological feature parameters can be dynamically adjusted, which effectively improves the accuracy of multi-dimensional physiological feature parameters for different populations, thereby improving the accuracy of cardiovascular health monitoring results and increasing the accuracy rate of cardiovascular health monitoring.

[0051] Example 2 Please see Figure 6 This is a flowchart of a multi-parameter collaborative cardiovascular health monitoring method provided in the second embodiment of the present invention. This multi-parameter collaborative cardiovascular health monitoring method can be applied to any device or system, and includes the following steps: Step S10: Simultaneously collect physiological signals from different parts of the target user to obtain multi-dimensional physiological signals, and perform noise reduction processing on the multi-dimensional physiological signals. Among them, a combination algorithm of wavelet transform, Kalman filtering and independent component analysis is used to denoise multidimensional physiological signals, eliminating the influence of body movement interference, environmental noise and temperature drift on multidimensional physiological signals.

[0052] Optionally, physiological signals can be simultaneously collected from different parts of the target user to obtain multi-dimensional physiological signals, including: The depth of blood vessels, skin thickness, and fat coverage of the monitored area on the target user are obtained, and adjustment parameters are determined based on the depth of blood vessels, skin thickness, and fat coverage. The radar transmission frequency, signal gain, and sampling period of the millimeter-wave radar are adjusted according to the adjustment parameters. Based on the adjusted millimeter-wave radar, pulse waves, heart rate, and respiratory signals of the carotid artery in the upper neck, the aorta in the chest, the radial artery in the wrist, and the posterior tibial artery in the ankle of the target user are collected synchronously, as well as the body surface temperature field distribution data and hemodynamic correlation signals of the neck, chest, wrist, and ankle, to obtain the multidimensional physiological signals. The monitoring posture of the target user is obtained, preset scene calibration parameters are determined based on the monitoring posture, and the multi-dimensional physiological signals are calibrated according to the preset scene calibration parameters.

[0053] Step S20: Perform hierarchical feature extraction on the denoised multidimensional physiological signal to obtain multidimensional physiological feature parameters, and segment the target user to obtain segmented user groups; Optionally, hierarchical feature extraction is performed on the denoised multidimensional physiological signal to obtain multidimensional physiological feature parameters, including: Heart rate, respiratory rate, and body surface temperature field temperature difference are extracted from the noise-reduced multidimensional physiological signals. The rising slope, falling slope, rising time from trough to first peak, falling time from first peak to trough, second peak amplitude, second peak appearance delay time, and pulse wave conduction time difference are also extracted from the noise-reduced multidimensional physiological signals. Based on the heart rate, respiratory rate, body surface temperature field temperature difference, rising slope, falling slope, rising time from trough to first peak, falling time from first peak to trough, second peak amplitude, second peak appearance delay time, and pulse wave conduction time difference, heart rate variability, pulse pressure, and pulse pressure index are calculated to obtain the multidimensional physiological characteristic parameters.

[0054] Step S30: Based on the subdivided population, dynamically adjust the weight ratio of each feature parameter in the multi-dimensional physiological feature parameters to obtain the dynamically adjusted feature parameters; This process involves acquiring the demographic attribute information of the target users, segmenting the target users based on the demographic attribute information, obtaining the segmented population, and dynamically adjusting the weight ratio of each feature parameter in the multi-dimensional physiological feature parameters based on the segmented population to obtain the dynamically adjusted feature parameters. The demographic attribute information includes age, gender, body type, underlying diseases, and medication history.

[0055] Step S40: Based on the dynamically adjusted characteristic parameters, calculate blood pressure, calculate arteriosclerosis, and predict vascular risk to obtain vascular health monitoring results; Specifically, based on pulse wave characteristics, heart rate variability, and population attribute information in the dynamically adjusted feature parameters, systolic and diastolic blood pressure are calculated using the gradient boosting tree algorithm to obtain the target blood pressure. The pulse wave characteristics include the rising slope and the amplitude of the second peak. Based on the pulse wave conduction time difference, the delay time of the second peak, the pulse pressure index, and the temperature difference of the body surface temperature field in the dynamically adjusted feature parameters, the arteriosclerosis index is output using the random forest algorithm. Blood pressure and arteriosclerosis data of the target user within a preset time range are obtained, and vascular risk is predicted based on the blood pressure and arteriosclerosis data to obtain the vascular risk prediction trend. Consistency verification and population matching verification are performed on the target blood pressure and arteriosclerosis index to obtain dual verification results. The vascular health monitoring results include target blood pressure, arteriosclerosis index, vascular risk prediction trend, and dual verification results.

[0056] In this embodiment, noise reduction processing of multi-dimensional physiological signals improves their quality. Layered feature extraction of the denoised multi-dimensional physiological signals effectively extracts multi-dimensional physiological feature parameters, increasing their diversity and preventing low accuracy in cardiovascular health monitoring caused by single-parameter acquisition. Furthermore, user segmentation effectively categorizes target users, allowing for dynamic adjustment of the weight of each feature parameter based on the segmented population. This improves the accuracy of multi-dimensional physiological feature parameters for different populations, thereby enhancing the accuracy of vascular health monitoring results and improving the overall accuracy of cardiovascular health monitoring.

[0057] Example 3 Figure 7 This is a structural block diagram of a terminal device 2 provided in the third embodiment of this application. For example... Figure 7 As shown, the terminal device 2 in this embodiment includes a processor 20, a memory 21, and a computer program 22 stored in the memory 21 and executable on the processor 20, such as a program for a multi-parameter collaborative cardiovascular health monitoring method. When the processor 20 executes the computer program 22, it implements the steps in the various embodiments of the multi-parameter collaborative cardiovascular health monitoring method described above.

[0058] For example, the computer program 22 may be divided into one or more modules, which are stored in the memory 21 and executed by the processor 20 to complete this application. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program 22 in the terminal device 2. The terminal device may include, but is not limited to, the processor 20 and the memory 21.

[0059] The processor 20 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0060] The memory 21 can be an internal storage unit of the terminal device 2, such as a hard drive or memory of the terminal device 2. The memory 21 can also be an external storage device of the terminal device 2, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the terminal device 2. Furthermore, the memory 21 can include both internal and external storage units of the terminal device 2. The memory 21 is used to store the computer program and other programs and data required by the terminal device. The memory 21 can also be used to temporarily store data that has been output or will be output.

[0061] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0062] If an integrated module is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. This computer-readable storage medium can be non-volatile or volatile. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable storage medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the contents of a computer-readable storage medium may be appropriately added to or subtracted from the contents as required by the legislation and patent practice in a jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, a computer-readable storage medium may not include electrical carrier signals and telecommunication signals.

[0063] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A multi-parameter collaborative cardiovascular health monitoring system, characterized in that, The system includes: Millimeter-wave radar acquisition module, used to acquire multi-dimensional physiological signals of target users; The three-level signal processing module is used to perform noise reduction processing on the multi-dimensional physiological signals and to perform hierarchical feature extraction on the noise-reduced multi-dimensional physiological signals to obtain multi-dimensional physiological feature parameters. The four-level population adaptation AI assessment module is used to segment the target users to obtain segmented populations. Based on the segmented populations, the weight ratio of each feature parameter in the multi-dimensional physiological feature parameters is dynamically adjusted to obtain dynamically adjusted feature parameters. Based on the dynamically adjusted feature parameters, blood pressure calculation, arteriosclerosis calculation, and vascular risk prediction are performed to obtain vascular health monitoring results.

2. The multi-parameter collaborative cardiovascular health monitoring system as described in claim 1, characterized in that, The millimeter-wave radar acquisition module includes: The radar array unit employs a millimeter-wave radar with a multi-transmit and multi-receive antenna layout. It is used to simultaneously acquire pulse waves, heart rate, and respiratory signals from the carotid artery in the upper neck, the aorta in the chest, the radial artery in the wrist, and the posterior tibial artery in the ankle of the target user, as well as surface temperature field distribution data and hemodynamic correlation signals of the neck, chest, wrist, and ankle, to obtain the multi-dimensional physiological signals. An adaptive signal conditioning unit is used to automatically adjust the radar transmission frequency, signal gain, and sampling period according to the blood vessel depth, skin thickness, and fat coverage of the monitored area, so as to adapt to the target users of different body types. A multi-scenario calibration unit is used to acquire the monitoring posture of the target user, determine preset scenario calibration parameters based on the monitoring posture, and eliminate the interference of scenario posture on signal acquisition based on the preset scenario calibration parameters.

3. The multi-parameter collaborative cardiovascular health monitoring system as described in claim 1, characterized in that, The three-level signal processing module includes: The multi-dimensional noise reduction unit uses a combination of wavelet transform, Kalman filtering and independent component analysis algorithms to denoise the multi-dimensional physiological signals, eliminating the influence of body movement interference, environmental noise and temperature drift on the multi-dimensional physiological signals; The hierarchical feature extraction unit includes a basic physiological layer, a pulse wave feature layer, and a derived parameter layer; The basic physiological layer is used to extract heart rate, respiratory rate, and body surface temperature field difference from the noise-reduced multi-dimensional physiological signals. The pulse wave feature layer is used to extract the rising slope, falling slope, rising time from the trough to the first peak, falling time from the first peak to the trough, second peak amplitude, second peak appearance delay time, and pulse wave conduction time difference from the multidimensional physiological signal after noise reduction. A derived parameter layer is used to calculate heart rate variability, pulse pressure, and pulse pressure index; The multidimensional physiological characteristic parameters include heart rate, respiratory rate, body surface temperature field temperature difference, rising slope, falling slope, rising time, falling time, second peak amplitude, second peak appearance delay time, pulse wave conduction time difference, heart rate variability, pulse pressure, and pulse pressure index.

4. The multi-parameter collaborative cardiovascular health monitoring system as described in claim 1, characterized in that, The fourth-order population adaptation AI assessment module includes: The audience segmentation unit is used to obtain the audience attribute information of the target user and segment the target user according to the audience attribute information to obtain the segmented audience. The audience attribute information includes age, gender, body type, underlying diseases, and medication history. A dynamic weighting unit is used to dynamically adjust the weight ratio of each feature parameter in the multi-dimensional physiological feature parameters according to the segmented population, so as to obtain the dynamically adjusted feature parameters. The multi-model fusion computing unit is used to calculate systolic blood pressure and diastolic blood pressure based on the pulse wave features, heart rate variability and population attribute information in the dynamically adjusted feature parameters, and to obtain the target blood pressure by using the gradient boosting tree algorithm. The pulse wave features include the rising slope and the second peak amplitude. The system is used to output the arteriosclerosis index by using a random forest algorithm based on the pulse wave conduction time difference, second peak appearance delay time, pulse pressure index, and body surface temperature field temperature difference among the dynamically adjusted feature parameters. This is used to acquire the blood pressure and arteriosclerosis data of the target user within a preset time range, and to predict vascular risk based on the blood pressure and arteriosclerosis data to obtain a vascular risk prediction trend. The cross-validation and optimization unit is used to perform consistency verification and population matching verification on the target blood pressure and the arteriosclerosis index to obtain dual verification results. The vascular health monitoring results include the target blood pressure, the arteriosclerosis index, the vascular risk prediction trend, and the dual verification results.

5. The multi-parameter collaborative cardiovascular health monitoring system as described in claim 1, characterized in that, The multi-parameter collaborative cardiovascular health monitoring system also includes a full-cycle management module, which includes: An encrypted data storage unit, supporting local storage and cloud backup, is used to store the vascular health monitoring results; The trend analysis and reporting unit is used to generate personalized health reports for different population segments. A tiered early warning unit is used to determine the risk level based on the preset differentiated early warning thresholds of the subdivided population, and to push intervention suggestions based on the risk level. The doctor-patient collaborative interaction unit is used for data sharing, remote parameter adjustment, and online consultation between the target user and the doctor. The lifestyle intervention support unit is used to record the target user's diet, monitor exercise, and remind them of medication.

6. The multi-parameter collaborative cardiovascular health monitoring system as described in claim 3, characterized in that, The radar array unit adjusts the antenna spacing through detachable antenna components to adapt to different monitoring needs.

7. The multi-parameter collaborative cardiovascular health monitoring system as described in claim 5, characterized in that, The full-cycle management module also includes a family sharing function unit, which allows family members to view the elderly person's monitoring data and early warning information, enabling remote care.

8. A multi-parameter collaborative cardiovascular health monitoring method, characterized in that, The method includes: Physiological signals are simultaneously collected from different parts of the target user to obtain multi-dimensional physiological signals, and noise reduction processing is performed on the multi-dimensional physiological signals. Hierarchical feature extraction is performed on the denoised multidimensional physiological signal to obtain multidimensional physiological feature parameters, and the target user is further segmented to obtain segmented user groups; Based on the segmented population, the weight ratio of each feature parameter in the multi-dimensional physiological feature parameters is dynamically adjusted to obtain the dynamically adjusted feature parameters; Based on the dynamically adjusted characteristic parameters, blood pressure calculation, arteriosclerosis calculation, and vascular risk prediction are performed to obtain vascular health monitoring results.

9. The multi-parameter collaborative cardiovascular health monitoring method as described in claim 8, characterized in that, Physiological signals are simultaneously collected from different parts of the target user to obtain multi-dimensional physiological signals, including: The depth of blood vessels, skin thickness, and fat coverage of the monitored area on the target user are obtained, and adjustment parameters are determined based on the depth of blood vessels, skin thickness, and fat coverage. The radar transmission frequency, signal gain, and sampling period of the millimeter-wave radar are adjusted according to the adjustment parameters. Based on the adjusted millimeter-wave radar, pulse waves, heart rate, and respiratory signals of the carotid artery in the upper neck, the aorta in the chest, the radial artery in the wrist, and the posterior tibial artery in the ankle of the target user are collected synchronously, as well as the body surface temperature field distribution data and hemodynamic correlation signals of the neck, chest, wrist, and ankle, to obtain the multidimensional physiological signals. The monitoring posture of the target user is obtained, preset scene calibration parameters are determined based on the monitoring posture, and the multi-dimensional physiological signals are calibrated according to the preset scene calibration parameters.

10. The multi-parameter collaborative cardiovascular health monitoring method as described in claim 8, characterized in that, Hierarchical feature extraction is performed on the denoised multidimensional physiological signal to obtain multidimensional physiological feature parameters, including: Heart rate, respiratory rate, and body surface temperature field temperature difference are extracted from the noise-reduced multidimensional physiological signals. The rising slope, falling slope, rising time from trough to first peak, falling time from first peak to trough, second peak amplitude, second peak appearance delay time, and pulse wave conduction time difference are also extracted from the noise-reduced multidimensional physiological signals. Based on the heart rate, respiratory rate, body surface temperature field temperature difference, rising slope, falling slope, rising time from trough to first peak, falling time from first peak to trough, second peak amplitude, second peak appearance delay time, and pulse wave conduction time difference, heart rate variability, pulse pressure, and pulse pressure index are calculated to obtain the multidimensional physiological characteristic parameters.

Images