A method and system for cardiovascular risk assessment based on wrist pulse wave

Abstract

This invention relates to the field of biomedical signal detection and processing, and discloses a cardiovascular risk assessment method and system based on wrist pulse waves. The method acquires radial artery main acquisition channel pulse waves, tissue reference channel pulse waves, and actual contact pressure data. Common pressure nodes are set within the individual's non-blocking pressure range, and the wrist detection component is controlled to pass through each common pressure node along the loading and unloading directions. After each common pressure node meets the stability criteria, pulse waves are acquired. Based on the low-frequency response relationship between the main acquisition channel and the tissue reference channel, the tissue transfer coefficients in the loading and unloading directions are determined. The main acquisition channel is then corrected to form loading and unloading direction response trajectories, and a radial artery dynamic response loop is constructed. A cardiovascular risk characterization value is generated by combining the zero-return residual, data reliability, and the deviation of the dynamic response characteristics from the individual's historical baseline or population reference parameters.

Description

Technical Field

[0001] This invention relates to the field of biomedical signal detection and processing, specifically to a cardiovascular risk assessment method and system based on wrist pulse waves. Background Technology

[0002] With the development of photoplethysmography (PPG) technology, wearable devices such as wristbands, smart bracelets, and smartwatches can now acquire pulse wave signals near the radial artery using photoelectric acquisition units located on the wrist. These signals can be used to analyze vascular elasticity, the degree of arteriosclerosis, or cardiovascular risk by analyzing parameters such as pulse wave amplitude, rise time, reflected wave, dicrotic notch, pulse interval, and frequency domain components. Current detection methods typically involve continuously acquiring pulse waves under a single contact pressure after the wrist detection component contacts the skin, or gradually adjusting the contact pressure to select a pressure state with a larger pulse amplitude and more stable waveform as the measurement state. This data is then combined with age, gender, blood pressure, and past health data to generate risk assessment results using statistical or machine learning models.

[0003] Wrist pulse waves are highly sensitive to the contact state between the detection component and the skin. When the contact pressure is too low, a gap may exist between the optical window and the wrist skin, resulting in a weak pulsation component. As the contact pressure increases, the skin and subcutaneous tissue compress, altering the tissue's optical properties and the propagation path of the detection light. Further increases in contact pressure may also change the local morphology of the radial artery and the pulse amplitude. Because wrist circumference, subcutaneous tissue thickness, radial artery location, and wearing tightness vary among different subjects, the same target pressure or driving displacement does not necessarily result in the same actual contact state. Current evaluation methods that use pulse waves under a single pressure, the pressure corresponding to the maximum pulse amplitude, or several discrete pressures as the basis for assessment struggle to distinguish whether waveform changes are caused by changes in vascular condition or by soft tissue compression, tissue creep, and changes in optical coupling.

[0004] Some existing multi-pressure pulse wave detection methods can establish the relationship between pressure and pulse force, pulse position, or waveform parameters, but they typically focus on determining the appropriate detection pressure or analyzing different geometric segments of a pressure change curve. Pressure-driven mechanisms may experience backlash and elastic deformation during loading and unloading, and wrist soft tissue also exhibits creep, stress relaxation, and direction-dependent viscoelastic responses. The same target pressure during loading and unloading may not correspond to the same actual contact pressure. Furthermore, wrist movement, tilting of the detection components, uneven local force, and sensor slippage can all alter the pulse wave morphology. Existing risk assessment methods lack joint judgment on the actual pressure registration of bidirectional pressure data, soft tissue direction response correction, and baseline recovery status before and after scanning. The contact pressure range and pressure nodes may also differ between different measurement batches, thus affecting the repeatability, long-term comparability, and stability of risk assessment data. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a cardiovascular risk assessment method and system based on wrist pulse waves, thereby resolving the technical problems existing in the prior art.

[0006] The above-mentioned technical objective of the present invention is achieved through the following technical solution: A cardiovascular risk assessment method and system based on wrist pulse waves includes the following steps: S1: Acquire the radial artery main acquisition channel pulse wave, tissue reference channel pulse wave, and actual contact pressure data output by the wrist detection component; S2: Based on the change of the radial artery main acquisition channel pulse wave with the actual contact pressure, determine the effective lower contact pressure and the non-blocking upper pressure, set at least three common pressure nodes between the effective lower contact pressure and the non-blocking upper pressure, and determine the effective lower contact pressure as the starting pressure; S3: Under the initial pressure, the baseline pulse wave before loading is acquired. The wrist detection component is controlled to pass through each common pressure node in sequence from low to high according to the actual contact pressure, and then return to each common pressure node in the reverse order. The actual contact pressure is used as the registration basis. The loading direction data and unloading direction data under the same common pressure node are determined as paired data. When the rate of change of the actual contact pressure, the DC component of the radial artery main acquisition channel and the DC component of the tissue reference channel do not exceed the corresponding stable threshold, the loading direction pulse wave and the unloading direction pulse wave are acquired. S4: Based on the low-frequency response relationship between the radial artery main acquisition channel and the tissue reference channel under each common pressure node, determine the tissue transmission coefficient in the loading direction and the tissue transmission coefficient in the unloading direction respectively. Correct the pressure-related low-frequency component of the radial artery main acquisition channel according to the tissue transmission coefficient in the corresponding direction to obtain the corrected pulse wave in the loading direction and the corrected pulse wave in the unloading direction. S5: Extract waveform parameters of the same type from the loading direction correction pulse wave and the unloading direction correction pulse wave corresponding to the same common pressure node. With the actual contact pressure as the independent variable and the waveform parameters as the dependent variable, the loading direction response trajectory and the unloading direction response trajectory are formed respectively. The radial artery dynamic response loop is constructed based on the two response trajectories. S6: After unloading, return the actual contact pressure to the starting pressure, collect the zero-return pulse wave, and determine the zero-return residual based on the waveform parameter difference between the zero-return pulse wave and the baseline pulse wave before loading. S7: Determine the data reliability based on the number of effective common pressure nodes, the pressure difference between paired data, the stability determination results of each common pressure node, and the zero-return residual. When the data credibility reaches the credibility judgment threshold, a cardiovascular risk characterization value is generated based on the difference in the same pressure node direction and the directional area of ​​the radial artery dynamic response loop, the zero-return residual, and the deviation of the above parameters from the individual's historical baseline or population reference parameters; when the data credibility does not reach the credibility judgment threshold, a retesting flag is output.

[0007] Preferably, the radial artery main acquisition channel and the tissue reference channel are determined from at least three photoelectric acquisition channels of the wrist detection component, specifically including: At least three photoelectric acquisition channels were sampled simultaneously within the same time window under channel screening pressure. The ratio of pulsation component to DC component and waveform similarity between adjacent cardiac cycles were calculated for each photoelectric acquisition channel. From the photoelectric acquisition channels with waveform similarity not lower than the main channel similarity threshold, the photoelectric acquisition channel with the largest ratio of pulsating component to DC component is determined as the radial artery main acquisition channel; From the photoelectric acquisition channels formed by photoelectric acquisition units set adjacent to the photoelectric acquisition unit corresponding to the radial artery main acquisition channel, select photoelectric acquisition channels whose ratio of pulsation component to DC component is not higher than the upper limit of the reference channel pulsation ratio, whose correlation coefficient between DC component and actual contact pressure is not lower than the pressure correlation threshold, and whose detection ratio of main wave peak or reflected wave peak in continuous cardiac cycles is lower than the upper limit of the reference channel wave peak detection ratio. The photoelectric acquisition channel with the highest correlation coefficient with the low-frequency component of the radial artery main acquisition channel among the screened photoelectric acquisition channels was determined as the tissue reference channel.

[0008] Preferably, determining the effective lower contact pressure and the non-blocking upper pressure includes: According to the candidate pressure order from low to high, the pulsation amplitude of the radial artery main acquisition channel and the waveform similarity between adjacent cardiac cycles are obtained sequentially at each candidate pressure. The first candidate pressure with a pulsation amplitude not lower than the effective amplitude threshold and a waveform similarity not lower than the effective similarity threshold is determined as the effective contact lower limit pressure; Determine the local maximum value of the pulsation amplitude under each candidate pressure. In two consecutive adjacent pressure intervals after the pressure corresponding to the local maximum value, when the pulsation amplitude decreases as the actual contact pressure increases, and the decrease ratio of the pulsation amplitude under the subsequent candidate pressure relative to the local maximum value reaches the compression judgment ratio, the candidate pressure before the subsequent candidate pressure is determined as the non-blocking upper limit pressure. The effective contact lower limit pressure and the non-blocking upper limit pressure are set as the first common pressure node and the last common pressure node, respectively, and at least one intermediate common pressure node is set between them; If the effective lower limit pressure or non-blocking upper limit pressure is not determined, output the wear adjustment flag and terminate the current pressure scan.

[0009] Preferably, the loading direction data and unloading direction data under the same common pressure node are determined as paired data, including: Based on actual contact pressure data, the wrist detection component is subjected to closed-loop pressure control so that the loading and unloading directions reach the common pressure nodes respectively. Calculate the pressure difference between the actual contact pressure in the loading direction and the actual contact pressure in the unloading direction at the same common pressure node; When the pressure difference is not greater than the node pressure tolerance, the corresponding loading direction data and unloading direction data are determined as paired data; When the pressure difference is greater than the node pressure tolerance, the node correction command is output to the wrist detection component according to the pressure difference, and the pulse wave in the corresponding direction is re-acquired after the actual contact pressure stabilizes after correction. If the pressure difference is still greater than the node pressure tolerance after the node correction count reaches the upper limit, the corresponding common pressure node will be marked as an invalid common pressure node; if there are fewer than three valid common pressure nodes, a retest flag will be output and the current risk assessment will be terminated.

[0010] Preferably, pulse waves in the loading or unloading direction are collected at each common pressure node, including: After the wrist detection component reaches the common pressure node, it enters the pressure holding state, and calculates the actual contact pressure change rate, the change rate of the DC component of the radial artery main acquisition channel, and the change rate of the DC component of the tissue reference channel within a continuous stable detection time window. When the absolute value of the actual contact pressure change rate is not greater than the pressure stabilization threshold, the absolute value of the change rate of the DC component of the radial artery main acquisition channel is not greater than the main channel stabilization threshold, the absolute value of the change rate of the DC component of the tissue reference channel is not greater than the reference channel stabilization threshold, and the above conditions are maintained for at least two consecutive stable detection time windows, pulse waves of no less than three complete cardiac cycles are collected. If the pressure holding time reaches the node holding time limit and the common pressure node still fails to meet the above conditions, the corresponding common pressure node will be marked as an invalid common pressure node.

[0011] Preferably, determining the tissue transfer coefficients in the loading direction and unloading direction and constructing the radial artery dynamic response loop includes: Low-frequency components that vary with actual contact pressure were extracted from data from the radial artery main acquisition channel and the tissue reference channel, respectively. Select a local pressure window that includes the current common pressure node. Based on the regression relationship between the change in the low-frequency component of the radial artery main acquisition channel and the change in the low-frequency component of the tissue reference channel within the local pressure window, determine the tissue transfer coefficient in the loading direction and the tissue transfer coefficient in the unloading direction, respectively. The low-frequency component offset of the tissue reference channel relative to the initial pressure state is mapped to the radial artery main acquisition channel according to the tissue transmission coefficient in the corresponding direction. The mapping result is subtracted from the low-frequency component of the radial artery main acquisition channel. The pulsating component is normalized with the corrected DC component to obtain the loading direction corrected pulse wave and the unloading direction corrected pulse wave. Extract at least two waveform parameters from the loading-direction corrected pulse wave and the unloading-direction corrected pulse wave: main wave rise time, maximum rise rate, reflected wave amplitude ratio, and diastolic area. For each waveform parameter, a loading direction response trajectory and an unloading direction response trajectory are formed. The region enclosed by the loading direction response trajectory, the unloading direction response trajectory, and the boundary segments connecting the endpoints of the two response trajectories at the effective contact lower limit pressure and the non-blocking upper limit pressure is defined as the radial artery dynamic response loop for the corresponding waveform parameter. Radial artery dynamic response loop groups are formed by the radial artery dynamic response loops corresponding to different waveform parameters.

[0012] Preferably, determining data reliability and generating cardiovascular risk characterization values ​​includes: In the same risk assessment, two loading and unloading processes are executed consecutively. The scan repeatability consistency is determined based on the waveform parameter differences of the response trajectories obtained from the two loading and unloading processes at each common pressure node and the dispersion of the direction difference with the same pressure node. When the number of effective common pressure nodes is not less than the lower limit of effective nodes, the pressure difference between each pair of data is not greater than the node pressure tolerance, each effective common pressure node meets the stability judgment condition, the scan repeatability consistency is not lower than the repeatability consistency threshold and the zero-return residual is not greater than the zero-return residual threshold, the data credibility is judged to reach the credibility judgment threshold; when any condition is not met, the data credibility is judged not to reach the credibility judgment threshold. The individual historical baseline consists of historical assessment data from no fewer than three tests on the same subject, in which the data reliability reached the reliability judgment threshold. The baseline median and baseline dispersion range of the directed area, the difference in direction of the same pressure node, the slope difference, and the zero-return residual of each radial artery dynamic response loop are recorded. When there is an individual historical baseline corresponding to the current test object, the individual deviation is determined based on the differences between the oriented area, the difference in the direction of the same pressure node, the slope difference, and the zero-return residual obtained in this assessment and the median value of the corresponding baseline, as well as the baseline dispersion range. When no individual historical baseline is stored, the initial cardiovascular risk characterization value is generated based on the deviation of the directional area, the difference in direction of the same pressure node, the slope difference, and the zero-return residual obtained in this assessment from the population reference parameters. After obtaining no less than three assessment data with data credibility reaching the credibility judgment threshold, the individual historical baseline is established. Based on labeled samples with cardiovascular risk stratification labels, a mapping relationship is established between directed area, difference in direction of the same pressure node, slope difference, zero-return residual and individual deviation and cardiovascular risk characterization value, and the cardiovascular risk characterization value corresponding to this assessment is generated according to the mapping relationship. The data from this assessment will be written into the individual's historical baseline only if the reliability of the data reaches the reliability threshold and the deviations of the directional area, the difference in direction of the same pressure node, the slope difference, and the zero-return residual obtained in this assessment from the corresponding baseline median value are all no greater than the baseline update deviation limit; otherwise, the individual's historical baseline will remain unchanged.

[0013] A cardiovascular risk assessment system based on wrist pulse waves includes a wrist detection component, a pressure regulation component, a processor, and a memory; The wrist detection component includes a radial artery main acquisition unit, a tissue reference acquisition unit, and a contact pressure detection unit. The radial artery main acquisition unit is used to output the radial artery main acquisition channel pulse wave, the tissue reference acquisition unit is used to output the tissue reference channel pulse wave, and the contact pressure detection unit is used to output the actual contact pressure data. The pressure adjustment component is connected to the wrist detection component and is used to change the contact pressure between the wrist detection component and the wrist skin according to the load control command and unload control command output by the processor. The processor is connected to the wrist detection component, the pressure regulation component, and the memory, and is configured as follows: Determine the effective contact lower limit pressure and the non-blocking upper limit pressure, set at least three common pressure nodes between the two, and determine the effective contact lower limit pressure as the starting pressure; Baseline pulse waves were acquired before loading under initial pressure. The pressure regulation component was controlled to pass through each common pressure node according to the loading and unloading directions. The actual contact pressure was used as the registration basis. The loading direction data and unloading direction data under the same common pressure node were determined as paired data. Pulse waves were acquired when the rate of change of the actual contact pressure, the DC component of the radial artery main acquisition channel and the DC component of the tissue reference channel did not exceed the corresponding stable threshold. Based on the low-frequency response relationship between the radial artery main acquisition channel and the tissue reference channel, the tissue transmission coefficients in the loading direction and unloading direction are determined respectively. The radial artery main acquisition channel is corrected according to the tissue transmission coefficients in the corresponding directions to obtain the corrected pulse wave in the loading direction and the corrected pulse wave in the unloading direction. The radial artery dynamic response loop is constructed based on the corrected pulse wave in the loading direction and the corrected pulse wave in the unloading direction corresponding to the same common pressure node. After unloading is completed, the control pressure regulating component returns to the starting pressure, the zero-return pulse wave is collected, and the zero-return residual is determined based on the zero-return pulse wave and the baseline pulse wave before loading. Data reliability is determined based on the number of effective common pressure nodes, the pressure difference between paired data, the stability determination results of each common pressure node, and the zero-return residual. When the data reliability reaches the reliability determination threshold, a cardiovascular risk characterization value is generated based on the radial artery dynamic response loop, the zero-return residual, and the deviation of the above parameters from the individual's historical baseline or population reference parameters. When the data reliability does not reach the reliability determination threshold, a retest flag is output.

[0014] Preferably, the wrist detection component includes a pressure contact and a pressure-distributing flexible layer disposed on the side of the pressure contact facing the wrist. At least three photoelectric acquisition units are disposed laterally on the pressure contact along the wrist, and the optical window of each photoelectric acquisition unit is flush with the contact surface of the pressure-distributing flexible layer facing the wrist. The radial artery master acquisition unit and the tissue reference acquisition unit are formed by corresponding photoelectric acquisition units from at least three photoelectric acquisition units; The contact pressure detection unit includes at least two pressure detection units respectively arranged on both sides of the pressure contact along the arrangement direction of the photoelectric acquisition unit. The processor determines the actual contact pressure based on the average value of the pressure data output by the at least two pressure detection units, and determines the pressure uniformity based on the difference between the maximum pressure data and the minimum pressure data. The pressure regulating component includes a driving element and a pressure transmitting element. One end of the pressure transmitting element is connected to the driving element, and the other end is connected to the pressure contact. The driving element drives the pressure contact to reciprocate along the normal direction of the wrist contact surface according to the load control command and unload control command output by the processor. When the pressure uniformity exceeds the pressure uniformity threshold, the processor stops collecting pulse waves from the corresponding common pressure node and marks the corresponding common pressure node as an invalid common pressure node.

[0015] Preferably, the memory stores node pressure tolerance, pressure stability threshold, main channel stability threshold, reference channel stability threshold, confidence judgment threshold, repeatability consistency threshold, zero-return residual threshold, baseline update deviation limit, population reference parameters, and individual historical baselines. The processor is also configured to control the pressure regulation component to perform two loading and unloading processes consecutively in the same risk assessment, and to determine the scan repeatability consistency based on the waveform parameter differences of the response trajectories obtained from the two loading and unloading processes at each common pressure node and the dispersion of the direction difference with the same pressure node. When the number of effective common pressure nodes is not less than the lower limit of effective nodes, the pressure difference between each pair of data is not greater than the node pressure tolerance, each effective common pressure node meets the stability judgment condition, the scan repeatability consistency is not lower than the repeatability consistency threshold, the zero-return residual is not greater than the zero-return residual threshold and the pressure uniformity is not greater than the pressure uniformity threshold, the processor determines that the data credibility reaches the credibility judgment threshold. When an individual historical baseline corresponding to the current subject is stored, the processor generates a cardiovascular risk characterization value based on the deviation of the radial artery dynamic response loop and zero-return residual obtained in this assessment from the individual historical baseline; when no individual historical baseline is stored, the processor generates an initial cardiovascular risk characterization value based on the deviation of the radial artery dynamic response loop and zero-return residual obtained in this assessment from the population reference parameter, and establishes an individual historical baseline after obtaining no less than three assessment data with a confidence level reaching the confidence judgment threshold. The processor will write the current assessment data into the individual's historical baseline only if the reliability of the current data reaches the reliability judgment threshold, and the deviation of the radial artery dynamic response loop and the zero-return residual obtained in this assessment from the corresponding baseline median value is not greater than the baseline update deviation limit; otherwise, the individual's historical baseline will remain unchanged.

[0016] In summary, the present invention has the following main beneficial effects: By setting a common pressure node between the effective lower limit pressure of the test object and the upper limit pressure of the non-blocking pressure, and controlling the wrist detection component to pass through the same actual contact pressure node along the loading and unloading directions, the actual contact pressure measured by the contact pressure detection unit is used as the registration basis, rather than the displacement of the drive component or the target pressure command as the data correspondence basis. This achieves the effect of reducing the influence of drive component return clearance, elastic deformation of pressure transmission component, and wrist wearing tightness on the correspondence of bidirectional pulse wave data. At the same time, pulse waves are collected only after the actual contact pressure, the DC component of the radial artery main acquisition channel, and the DC component of the tissue reference channel at each common pressure node have stabilized. This avoids mistaking wrist soft tissue creep, stress relaxation, and instantaneous contact state changes for vascular dynamic response.

[0017] By determining the radial artery main acquisition channel and tissue reference channel from adjacent photoelectric acquisition channels at the wrist, and determining the tissue transmission coefficient based on the low-frequency response relationship of the two channels under loading and unloading directions, the directional correction of the pressure-related tissue response in the radial artery main acquisition channel is achieved, thus separating the optical changes caused by pressure on the skin and subcutaneous tissue from the radial artery pulsation changes. On this basis, based on the waveform parameter differences between the pulse wave corrected under the loading direction and the pulse wave corrected under the same actual contact pressure node, a radial artery dynamic response loop is constructed. This achieves the effect of obtaining the vascular directional response characteristics that cannot be provided by a single fixed pressure measurement or unidirectional pressure scan, and avoids mistaking the geometric rising and falling segments of a static pressure pulse wave curve as vascular loading and unloading responses.

[0018] By returning to the initial pressure and performing zero-return verification after the pressure scan, and combining the effective number of common pressure nodes, node pressure difference, node stability, tissue transmission coefficient validity, pressure uniformity, and consistency of repeated scans, the reliability of the data is determined. This achieves the effect of eliminating invalid data generated by sensor slippage, wrist movement, uneven local force, and changes in state before and after measurement. Only when the data reliability meets the judgment criteria is a cardiovascular risk characterization value generated based on the radial artery dynamic response loop, zero-return residual, and its deviation from the individual's historical baseline or population reference parameters. This achieves the effect of taking into account inter-individual differences and long-term changing trends of the same subject, and improving the stability and interpretability of risk assessment data. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1 refer to Figure 1 A cardiovascular risk assessment method and system based on wrist pulse waves includes the following steps: S1: Acquire the radial artery main acquisition channel pulse wave, tissue reference channel pulse wave, and actual contact pressure data output by the wrist detection component; S2: Based on the change of the radial artery main acquisition channel pulse wave with the actual contact pressure, determine the effective lower contact pressure and the non-blocking upper pressure, set at least three common pressure nodes between the effective lower contact pressure and the non-blocking upper pressure, and determine the effective lower contact pressure as the starting pressure; S3: Under the initial pressure, the baseline pulse wave before loading is acquired. The wrist detection component is controlled to pass through each common pressure node in sequence from low to high according to the actual contact pressure, and then return to each common pressure node in the reverse order. The actual contact pressure is used as the registration basis. The loading direction data and unloading direction data under the same common pressure node are determined as paired data. When the rate of change of the actual contact pressure, the DC component of the radial artery main acquisition channel and the DC component of the tissue reference channel do not exceed the corresponding stable threshold, the loading direction pulse wave and the unloading direction pulse wave are acquired. S4: Based on the low-frequency response relationship between the radial artery main acquisition channel and the tissue reference channel under each common pressure node, determine the tissue transmission coefficient in the loading direction and the tissue transmission coefficient in the unloading direction respectively. Correct the pressure-related low-frequency component of the radial artery main acquisition channel according to the tissue transmission coefficient in the corresponding direction to obtain the corrected pulse wave in the loading direction and the corrected pulse wave in the unloading direction. S5: Extract waveform parameters of the same type from the loading direction correction pulse wave and the unloading direction correction pulse wave corresponding to the same common pressure node. With the actual contact pressure as the independent variable and the waveform parameters as the dependent variable, the loading direction response trajectory and the unloading direction response trajectory are formed respectively. The radial artery dynamic response loop is constructed based on the two response trajectories. S6: After unloading, return the actual contact pressure to the starting pressure, collect the zero-return pulse wave, and determine the zero-return residual based on the waveform parameter difference between the zero-return pulse wave and the baseline pulse wave before loading. S7: Determine the data reliability based on the number of effective common pressure nodes, the pressure difference between paired data, the stability determination results of each common pressure node, and the zero-return residual. When the data credibility reaches the credibility judgment threshold, a cardiovascular risk characterization value is generated based on the difference in the same pressure node direction and the directional area of ​​the radial artery dynamic response loop, the zero-return residual, and the deviation of the above parameters from the individual's historical baseline or population reference parameters; when the data credibility does not reach the credibility judgment threshold, a retesting flag is output.

[0022] The system is used to perform pressure scans in the loading and unloading directions within the individual non-blocking pressure range of the wrist of the test subject, construct a radial artery dynamic response loop based on the pulse wave differences under the same actual contact pressure, and generate cardiovascular risk characterization values, data reliability, and retest markers.

[0023] This embodiment does not directly output risk results based on the pulse wave morphology under a single contact pressure, nor does it use the pressure with the largest pulse amplitude as the sole measurement pressure. Instead, it uses local contact pressure as a controllable excitation quantity to obtain the difference in dynamic response of the radial artery under loading and unloading directions. The output cardiovascular risk characterization value is used to indicate the degree of deviation of the current wrist vascular dynamic response from the individual's historical baseline or population reference parameters, and is not directly used as a diagnostic conclusion for specific cardiovascular diseases.

[0024] The cardiovascular risk assessment system includes a wrist detection component, a pressure regulation component, a processor, and a memory.

[0025] The wrist detection component includes a pressure contact, a pressure-distributing flexible layer, at least three photoelectric acquisition units, and a contact pressure detection unit. This embodiment employs five photoelectric acquisition units, which are arranged sequentially and at intervals along the transverse direction of the wrist to cover the possible passage of the radial artery and the adjacent tissue areas on both sides of the radial artery. The transverse direction of the wrist refers to the direction in which the wrist detection component intersects with the extension direction of the radial artery after it is worn.

[0026] Each photoelectric acquisition unit includes a light-emitting part and a photoelectric receiving part. The light-emitting part emits detection light towards the wrist tissue, and the photoelectric receiving part receives the detection light reflected or scattered by the wrist tissue and outputs a photoelectric signal. Different photoelectric acquisition units form different photoelectric acquisition channels.

[0027] A pressure-distributing flexible layer is disposed on the side of the pressure contact head facing the wrist, and the optical windows of each photoelectric acquisition unit are flush with the contact surface of the pressure-distributing flexible layer facing the wrist. The pressure-distributing flexible layer is made of a flexible material with elastic recovery capability that allows detection light to pass through, in order to reduce local stress concentration at the edge of the pressure contact head and improve the contact between each photoelectric acquisition unit and the skin of the wrist.

[0028] The contact pressure detection unit includes at least two pressure detection units. The two pressure detection units are located on both sides of the pressure contact along the arrangement direction of the photoelectric acquisition unit, and are arranged in the force transmission path between the pressure contact and the pressure regulating component, so that the normal force applied to the pressure contact by the pressure regulating component passes through the two pressure detection units respectively.

[0029] The processor determines the actual contact pressure based on the average of the pressure values ​​output by the two pressure detection units, and determines the pressure uniformity based on the difference between the two pressure values. When the pressure uniformity is greater than the pressure uniformity threshold, it indicates that the pressure contact is tilted, locally raised, or that the force on both sides is significantly inconsistent. The processor then stops collecting the pulse wave of the current common pressure node and marks the common pressure node as an invalid common pressure node.

[0030] The pressure uniformity threshold was determined through prototype calibration. During calibration, data from two pressure detection units were collected under different wrist curvature models, different pressure contact tilt angles, and different offset positions. States that maintained the repeatability of waveforms in each photoelectric acquisition channel were marked as valid states, while states with pressure contact tilting, channel saturation, or acquisition position slippage were marked as invalid states. The pressure uniformity threshold was determined based on the pressure difference distribution corresponding to the two types of states.

[0031] The pressure regulating assembly includes a drive element and a pressure transmission element. One end of the pressure transmission element is connected to the drive element, and the other end is connected to the pressure contact. The drive element drives the pressure contact to reciprocate along the normal direction of the wrist contact surface according to the load control command or unload control command output by the processor.

[0032] The pressure regulation component employs closed-loop control based on actual contact pressure. The processor compares the target common pressure node with the actual contact pressure output by the contact pressure detection unit; when the pressure difference between the two is greater than the node pressure tolerance, the processor controls the drive to continue loading or unloading based on the direction and magnitude of the pressure difference; when the pressure difference is not greater than the node pressure tolerance, the processor controls the drive to enter a pressure holding state.

[0033] Therefore, whether the common pressure node has reached the pressure actually measured by the contact pressure detection unit is determined, instead of replacing the actual contact pressure with the displacement of the drive component, the number of drive pulses, or the target pressure command. This reduces the impact of the drive component return clearance and the elastic deformation of the pressure transmission component on the bidirectional pressure data registration.

[0034] The processor is connected to the photoelectric acquisition unit, the contact pressure detection unit, the pressure regulation component, and the memory. Each photoelectric acquisition unit and the contact pressure detection unit uses the same sampling clock, or records a sampling time stamp separately. The processor aligns the pulse wave data and the actual contact pressure data according to the time stamp.

[0035] The memory is used to store node pressure tolerance, pressure stability threshold, main channel stability threshold, reference channel stability threshold, pressure uniformity threshold, node retention time limit, repeatability consistency threshold, zero-return residual threshold, baseline update deviation limit, population reference parameters, risk assessment model parameters, and individual historical baselines.

[0036] The wrist detection component is worn on the wrist of the subject, so that five photoelectric acquisition units cover the area that the radial artery may pass through and its adjacent tissue areas.

[0037] The processor first controls the pressure contacts to make contact with the wrist skin using channel screening pressure. The channel screening pressure should ensure continuous optical contact between each optical window and the wrist skin, but not enough to cause a sustained decrease in the amplitude of the radial artery pulsation.

[0038] The channel screening pressure is determined through device calibration. During calibration, the contact pressure is gradually increased starting from the state where the pressure contact is not in contact with the wrist simulator. The pressure at which each photoelectric acquisition unit changes from having no stable DC signal to continuously outputting a stable DC signal is recorded, and the channel screening pressure is selected between this pressure and the effective pulse wave measurement pressure.

[0039] The processor samples each photoelectric acquisition channel simultaneously within a time window under channel screening pressure, and the sampling time window covers at least three complete cardiac cycles.

[0040] The processor divides the cardiac cycle based on adjacent main peaks or the position of maximum rise. For each complete cardiac cycle, the average value of the original photoelectric signal within that cycle is taken as the periodic DC value, and the portion obtained by subtracting the periodic DC value from the original photoelectric signal is taken as the periodic pulsation component.

[0041] The DC component of a channel within a sampling time window is determined using the median of the DC values ​​of each effective cardiac cycle, and the channel pulsation amplitude is determined using the median of the peak-to-valley difference of each effective cardiac cycle. Using the median can reduce the impact of a single abnormal cardiac cycle or transient motion interference on the channel screening results.

[0042] The processor uses the ratio of channel ripple amplitude to channel DC component as the ripple percentage. The same ripple amplitude and DC component calculation method should be used for device calibration, model training, and practical evaluation.

[0043] The processor aligns the complete cardiac cycles of the same photoelectric acquisition channel according to the start point of the cycle and the position of the main peak, and resamples them to the same number of sampling points.

[0044] The processor calculates the waveform correlation coefficient between adjacent cardiac cycles and uses the median of the correlation coefficients of all adjacent cardiac cycles within a sampling time window as the waveform similarity of the corresponding photoelectric acquisition channel.

[0045] From the photoelectric acquisition channels with waveform similarity not lower than the main channel similarity threshold, the photoelectric acquisition channel with the largest proportion of pulsation is determined as the radial artery main acquisition channel.

[0046] The main channel similarity threshold is determined based on valid pulse wave samples and motion interference samples. During device calibration, waveform similarity is calculated for valid and invalid samples that have been manually verified or confirmed by synchronous reference equipment. The boundary value that can exclude obvious motion interference and retain valid pulse waves is determined based on the distribution of the two types of samples.

[0047] When there is no photoelectric acquisition channel whose waveform similarity reaches the main channel similarity threshold, the processor outputs a wearing adjustment flag and stops performing pressure scanning.

[0048] The tissue reference channel is selected from the channels formed by photoelectric acquisition units located adjacent to the photoelectric acquisition unit corresponding to the radial artery main acquisition channel.

[0049] Candidate tissue reference channels should simultaneously meet the following conditions: their pulsation ratio is not higher than the upper limit of the reference channel pulsation ratio; the correlation coefficient between their DC component and the actual contact pressure is not lower than the pressure correlation threshold; the detection ratio of the main wave peak or reflected wave peak in continuous cardiac cycles is lower than the upper limit of the reference channel wave peak detection ratio; and their low-frequency component is correlated with the low-frequency component of the radial artery main acquisition channel.

[0050] The main peak should be a local maximum within the corresponding cardiac cycle that satisfies the amplitude and interpeak duration conditions. The reflected peak should be located after the main peak and before the start of the next cardiac cycle, and should also satisfy the local maximum and peak prominence conditions.

[0051] The main peak detection ratio is the ratio of the number of cardiac cycles with valid main peaks detected to the total number of complete cardiac cycles; the reflection peak detection ratio is the ratio of the number of cardiac cycles with valid reflection peaks detected to the total number of complete cardiac cycles.

[0052] The upper limit of the reference channel pulsation ratio, the pressure-related threshold, and the upper limit of the reference channel peak detection ratio are determined based on the sample distribution of the confirmed radial artery covered channels and non-arterial adjacent tissue channels.

[0053] Among the candidate channels that meet the above conditions, the channel with the highest correlation coefficient with the low-frequency component of the radial artery main acquisition channel is determined as the tissue reference channel.

[0054] The tissue reference channel is used to characterize pressure-related optical changes in skin and subcutaneous tissue under similar pressure conditions to the main radial artery acquisition location. It is not used to acquire the second arterial pulse wave, nor is it directly subtracted from the original pulse wave of the main radial artery acquisition channel.

[0055] When no suitable tissue reference channel is available, the processor outputs a wear adjustment indicator and prompts the user to readjust the position of the wrist detection component.

[0056] After determining the radial artery main acquisition channel and tissue reference channel, the processor controls the pressure regulation component to be loaded step by step in the order of candidate pressure from low to high.

[0057] The candidate pressure step size is determined based on the measurement resolution of the contact pressure detection unit, the closed-loop control error of the pressure regulation component, and the waveform parameters' ability to distinguish pressure changes. The candidate pressure step size should be greater than the overall pressure error of the device to ensure that adjacent candidate pressures can be reliably distinguished in actual measurements.

[0058] Under each candidate pressure, the processor waits for the actual contact pressure, the DC component of the radial artery main acquisition channel, and the DC component of the tissue reference channel to stabilize before acquiring the pulsation amplitude and waveform similarity of the radial artery main acquisition channel.

[0059] The first candidate pressure with a pulsation amplitude not lower than the effective amplitude threshold and a waveform similarity not lower than the effective similarity threshold is determined as the effective contact lower limit pressure.

[0060] The effective amplitude threshold is determined based on the device noise floor. Photoelectric signals are acquired on a pulsation-free input or tissue simulator, the upper limit of the noise pulsation amplitude is calculated, and a signal recognition margin is added to this upper limit to obtain the effective amplitude threshold.

[0061] Continue increasing the candidate pressure and record the pulsation amplitude of the radial artery main acquisition channel. After obtaining the local maximum value of the pulsation amplitude, in two consecutive adjacent pressure intervals after the pressure corresponding to the local maximum value, if the pulsation amplitude decreases with the increase of the actual contact pressure, and the decrease ratio of the pulsation amplitude at the latter candidate pressure relative to the local maximum value reaches the compression determination ratio, then the candidate pressure adjacent to the latter candidate pressure and with a pressure value lower than the latter candidate pressure is determined as the non-blocking upper limit pressure.

[0062] The compression judgment ratio is determined by pressure scan data during the device verification phase. It is used to identify the state in which local contact pressure begins to cause significant pulsation attenuation, and is not used as a disease diagnosis threshold.

[0063] The upper limit of the device's safe pressure is determined by a combination of the rated load of the pressure regulating component, the range of the contact pressure detection unit, the effective contact area of ​​the pressure contact, the allowable load of the pressure-distributed flexible layer, and actual wearing safety verification, and adopts the lowest allowable pressure among the above limits.

[0064] If the effective contact lower limit pressure cannot be determined, or if the non-blocking upper limit pressure cannot be determined even after reaching the device safety pressure upper limit, the processor outputs a wear adjustment flag and terminates the current pressure scan.

[0065] The non-blocking upper limit pressure is the upper limit of local contact pressure that still allows for a continuous and effective pulse wave and does not enter a state of continuous pulsation disappearance. It is not the pressure that blocks arterial blood flow through the upper arm cuff.

[0066] The effective contact lower limit pressure is set as the first common pressure node, the non-blocking upper limit pressure is set as the last common pressure node, and at least one intermediate common pressure node is set between the two, so that the number of common pressure nodes is not less than three.

[0067] Common pressure nodes can be uniformly set according to the actual contact pressure. Alternatively, initial common pressure nodes can be set first, and when the pulsation amplitude or target waveform parameter change between adjacent common pressure nodes exceeds the node densification threshold, common pressure nodes can be added in the corresponding pressure range.

[0068] The node densification threshold is determined based on the repeatability error of the target waveform parameters under the same pressure state. Only when the parameter variation between adjacent nodes exceeds the repeatability error are common pressure nodes added, thus avoiding the increase of invalid nodes due to noise.

[0069] The effective lower limit of contact pressure is determined as the starting pressure. Under the starting pressure, after the actual contact pressure, the DC component of the radial artery main acquisition channel, and the DC component of the tissue reference channel meet the stability criteria, baseline pulse waves before loading are acquired for no less than three complete cardiac cycles.

[0070] The processor outputs a load control command to the pressure regulation component, causing the actual contact pressure to start from the initial pressure and arrive at each common pressure node in ascending order.

[0071] After completing the data acquisition of the last common pressure node, the processor outputs an unloading control command, causing the actual contact pressure to return to each common pressure node in reverse order, until it returns to the starting pressure.

[0072] For the same common pressure node, the processor calculates the pressure difference between the actual contact pressure in the loading direction and the actual contact pressure in the unloading direction.

[0073] When the pressure difference is not greater than the node pressure tolerance, the loading direction data and unloading direction data are determined as the paired data of the common pressure node.

[0074] When the pressure difference exceeds the node pressure tolerance, the processor outputs a node correction command to the pressure regulation component based on the pressure difference. After node correction, the stability determination is re-executed, and complete cardiac cycle data in the corresponding direction are re-acquired only after the stability determination conditions are met again. Unregistered pulse wave data before correction does not participate in the construction of the radial artery dynamic response loop.

[0075] When the number of node corrections reaches the upper limit and the pressure difference is still greater than the node pressure tolerance, the common pressure node is marked as an invalid common pressure node.

[0076] The node pressure tolerance is determined jointly based on the measurement error of the contact pressure detection unit, the closed-loop control error, and the influence of pressure deviation on waveform parameters. During device calibration, the target pressure is repeatedly reached under a stable load, and the pressure range that ensures the waveform parameter variation caused by pressure deviation does not exceed the allowable repeatable measurement error is selected as the node pressure tolerance.

[0077] This embodiment does not directly interpolate the original pulse waves under different actual contact pressures. The cardiac cycle length and waveform phase may differ under different pressures, and interpolating the original pulse waves can easily result in inaccurate waveforms.

[0078] When there are fewer than three valid common pressure nodes, the processor outputs a retest flag and terminates the current risk assessment.

[0079] After the wrist detection component reaches the common pressure node, it enters the pressure holding state.

[0080] The length of the stable detection time window is not shorter than one complete cardiac cycle currently estimated. The length of the current cardiac cycle is determined based on the median value of the intervals between the most recent effective main wave peaks. When a stable cardiac cycle estimate cannot be obtained, the upper limit of the stable detection time window, determined based on the lowest heart rate of the target user population, is used.

[0081] The processor compares the average actual contact pressure, the average DC component of the radial artery main acquisition channel, and the average DC component of the tissue reference channel within the current stable detection time window and the previous stable detection time window, and determines the corresponding rate of change based on the difference between the average values ​​of two adjacent windows and the time window length.

[0082] When the absolute value of the actual contact pressure change rate is not greater than the pressure stability threshold, the absolute value of the change rate of the DC component of the radial artery main acquisition channel is not greater than the main channel stability threshold, and the absolute value of the change rate of the DC component of the tissue reference channel is not greater than the reference channel stability threshold, and the above conditions are maintained for at least two consecutive stable detection time windows, the current common pressure node is determined to meet the stability determination conditions.

[0083] The pressure stabilization threshold is determined based on the noise and drift of the contact pressure detection unit under constant load. The main channel stabilization threshold and the reference channel stabilization threshold are determined based on the natural fluctuation of the DC component of the corresponding photoelectric acquisition channel under constant pressure.

[0084] After the common pressure node meets the stability criteria, pulse waves in the loading or unloading direction are collected for no less than three complete cardiac cycles.

[0085] If the pressure holding time reaches the node holding time limit and the current common pressure node still does not meet the stability determination condition, the common pressure node is marked as an invalid common pressure node.

[0086] By simultaneously judging pressure changes and DC component changes in both channels, pulse waves can be avoided when the pressure has stabilized but the soft tissue of the wrist is still in a creep or stress relaxation process.

[0087] This embodiment uses a method based on the DC value of the cardiac cycle to determine the pressure-related low-frequency components, in order to avoid inconsistent processing results caused by using different digital filter cutoff frequencies.

[0088] For a common pressure node, the average value of the original photoelectric signal within each complete cardiac cycle is determined as the periodic DC value of that cardiac cycle, and the median of the periodic DC values ​​of all effective cardiac cycles is determined as the low-frequency component of that common pressure node.

[0089] The low-frequency components were extracted using the methods described above for the radial artery main acquisition channel and the tissue reference channel.

[0090] The low-frequency component reflects the slow optical baseline change that occurs when moving from one common pressure node to another, and does not include the pulsatile waveform component within a single cardiac cycle.

[0091] For each valid common pressure node, the processor selects a local pressure window that includes the current common pressure node.

[0092] When the number of effective common pressure nodes is no less than five, the local pressure window located in the middle of the pressure range includes the current common pressure node and one effective common pressure node before and after it. For the common pressure node located at the effective lower contact pressure limit end, three effective common pressure nodes are selected consecutively from that end towards the middle of the pressure range. For the common pressure node located at the non-blocking upper pressure limit end, three effective common pressure nodes are selected consecutively from that end towards the middle of the pressure range.

[0093] When the total number of effective common pressure nodes is three or four, all effective common pressure nodes together form the local pressure window of each common pressure node.

[0094] The processor establishes a linear regression relationship between the change in the low-frequency component of the radial artery main acquisition channel and the change in the low-frequency component of the tissue reference channel within the local pressure window based on the loading direction data and the unloading direction data, respectively, and determines the regression slope as the tissue transmission coefficient in the loading direction and the tissue transmission coefficient in the unloading direction, respectively.

[0095] Taking the loading direction as an example, the first The tissue transmission coefficients corresponding to the common pressure nodes are: In the formula, Indicates the first Local pressure windows corresponding to common pressure nodes; Indicates the tissue reference channel under the loading direction at the 1st Low-frequency components of each pressure node; Indicates the radial artery main acquisition channel under the loading direction at the [number]th [position]. Low-frequency components of each pressure node; and These represent the average values ​​of the corresponding low-frequency components within the local pressure window; To prevent positive numbers with a denominator of zero.

[0096] The unloading direction organization transfer coefficient is determined independently based on the unloading direction data in the same way.

[0097] Before calculating the tissue transfer coefficient, the processor first determines the range of variation of the low-frequency components of the tissue reference channel within the local pressure window. When the difference between the maximum and minimum values ​​of the low-frequency components of the tissue reference channel is less than the effective variation threshold of the reference channel, it is determined that the tissue reference channel has not formed a discernible pressure response, and the tissue transfer coefficient obtained from that local pressure window is not used.

[0098] The effective variation threshold of the reference channel is determined based on the low-frequency component noise and drift range of the tissue reference channel under constant pressure, and is higher than the statistical upper limit of the natural fluctuation range.

[0099] The processor also calculates the determination coefficient of the regression relationship. When the determination coefficient is lower than the regression validity threshold, the corresponding common pressure node is marked as an invalid common pressure node.

[0100] The effective threshold for regression was determined based on the distribution of the determination coefficients of the validated valid tissue stress response samples and invalid samples.

[0101] The tissue transfer coefficients are determined separately for the loading and unloading directions because the skin and subcutaneous tissue of the wrist may exhibit different viscoelastic responses during loading and unloading. If a fixed coefficient is shared for both directions, tissue directional responses may remain in the radial artery main acquisition channel.

[0102] For each effective common pressure node, the low-frequency component offset of the tissue reference channel relative to the initial pressure state is mapped to the radial artery main acquisition channel according to the tissue transmission coefficient in the corresponding direction, and the mapping result is subtracted from the low-frequency component of the radial artery main acquisition channel to obtain the corrected DC component of the main acquisition channel.

[0103] Subsequently, the corresponding pulsation component is normalized using the corrected DC component of the main acquisition channel to obtain the loading direction corrected pulse wave and the unloading direction corrected pulse wave.

[0104] This processing does not directly subtract the original waveform of the tissue reference channel from the original waveform of the radial artery main acquisition channel, but rather corrects for pressure-related low-frequency optical changes caused by tissue compression.

[0105] When the corrected DC component is lower than the lower limit of the DC component corresponding to the linear response range of the photoelectric receiver, or when the tissue transfer coefficient does not meet the effective change condition or the effective regression condition of the reference channel, the corresponding common pressure node will be marked as an invalid common pressure node.

[0106] The processor performs cardiac cycle segmentation on the loading direction-corrected pulse wave and the unloading direction-corrected pulse wave, respectively.

[0107] Cardiac cycles that include signal saturation, sudden changes in motion, or a cycle length that deviates significantly from the median length of the current cardiac cycle are marked as invalid cardiac cycles.

[0108] The retained cardiac cycles are time-aligned according to the cycle start point and the main peak, and then resampled to the same number of sampling points. The median value of each sampling position is then calculated to form the node representative waveform corresponding to the common pressure node.

[0109] At least two waveform parameters are extracted from the node-represented waveform. In this embodiment, at least two of the following are used: main wave rise time, maximum rise rate, reflected wave amplitude ratio, and diastolic area.

[0110] The rise time of the main wave is the time interval from the beginning of the cardiac cycle to the peak of the main wave. The maximum rate of ascent is the maximum value of the first derivative of the waveform represented by the node within the rise interval of the main wave. The reflected wave amplitude ratio is the ratio of the amplitude of the reflected wave peak to the amplitude of the main wave peak. The diastolic area is the integral area of ​​the waveform from the dicrotic notch or the end of systole to the end of the cardiac cycle.

[0111] When the detection rate of the reflected wave peak in the effective cardiac cycle is lower than the effective detection rate of the reflected wave, the reflected wave amplitude ratio of this node is not used.

[0112] When the diabetic notch cannot be stably identified, the first and second derivatives are used together to determine the end position of the systolic phase, and the same feature point determination rules are used for model training and actual evaluation.

[0113] For any waveform parameter, with the actual contact pressure as the independent variable and the value of the waveform parameter as the dependent variable, the loading direction response trajectory and the unloading direction response trajectory are formed respectively.

[0114] Under the same common pressure node, the difference between the waveform parameters in the loading direction and the waveform parameters in the unloading direction is the difference in the direction of the same pressure node.

[0115] The region enclosed by the loading direction response trajectory, the unloading direction response trajectory, and the boundary segments connecting the endpoints of the two response trajectories at the effective contact lower limit pressure and the non-blocking upper limit pressure is defined as the radial artery dynamic response loop corresponding to the waveform parameter.

[0116] The directed area of ​​the radial artery dynamic response loop is: ; In the formula, This represents the directed area of ​​the dynamic response loop corresponding to the waveform parameters; Indicates the number of effective shared pressure nodes; and This represents the actual contact pressure at adjacent common pressure nodes; and This indicates the directional difference between corresponding nodes with common pressure.

[0117] A radial artery dynamic response loop is constructed for at least two waveform parameters, and a radial artery dynamic response loop group is formed by the radial artery dynamic response loops corresponding to different waveform parameters.

[0118] In this embodiment, the two response trajectories are derived from the actual loading process and the actual unloading process, respectively, and are registered according to the same actual contact pressure. Instead, a static pressure pulse wave curve is divided into rising and falling segments according to its geometric shape.

[0119] The effective lower contact pressure and non-blocking upper pressure determined for the same subject on different measurement dates or under different wearing conditions may be different. Therefore, the pressure coordinates need to be normalized before comparing with individual historical baselines or population reference parameters.

[0120] No. The normalized pressure coordinates of the common pressure nodes are: ; In the formula, Indicates the first Normalized pressure coordinates of a common pressure node; This indicates the actual contact pressure at the common pressure node; This indicates the lower limit pressure of the effective contact. This indicates the upper limit pressure for this non-blocking event.

[0121] The normalized pressure coordinate ranges from 0 to 1.

[0122] Within this assessment, data registration and dynamic response loop construction for loading and unloading directions were still performed based on actual contact pressure. When comparing with historical assessment data or population reference parameters, the extracted waveform parameter trajectories that had passed quality assessment were resampled into a uniformly set normalized pressure coordinate set in memory.

[0123] Resampling uses piecewise linear interpolation between adjacent effective common pressure nodes, interpolating only the waveform parameter trajectory and not the original pulse wave.

[0124] The differences in the orientation of the same pressure nodes, the slope of the response trajectory, and the area of ​​the dynamic response loop for different measurement batches are all compared under a unified normalized pressure coordinate set.

[0125] After completing the data acquisition of all common pressure nodes in the unloading direction, the processor controls the pressure regulation component to return to the starting pressure.

[0126] Under initial pressure, reassess whether the actual contact pressure, the DC component of the radial artery main acquisition channel, and the DC component of the tissue reference channel meet the stability criteria. If the stability criteria are met, acquire the zero pulse wave.

[0127] Baseline waveform parameters of the same type are extracted from the pre-load baseline pulse wave and the zero-return pulse wave, and the zero-return residual is determined based on the relative difference of each waveform parameter before and after the scan.

[0128] For waveform parameters used in zero-return verification, those with minimal variation in repeated measurements under steady-state conditions should be selected first. The weights of each parameter are determined based on the degree of variation in repeated measurements under steady-state conditions; if reliable weighted data has not yet been formed, equal weights are used.

[0129] The zero-return residual threshold is determined based on the distribution of zero-return residuals obtained from repeated scans of the same subject under stable wearing conditions. If the zero-return residual is greater than the threshold, it indicates that wrist movement, probe slippage, significant heart rate changes, or changes in local blood flow may have occurred during the scan, and the data in this case will not be used for risk assessment.

[0130] Two loading and unloading processes are executed consecutively within the same risk assessment. Between the two scans, the actual contact pressure is brought back to the initial pressure, and the initial pressure state is reconfirmed to meet the stability criteria.

[0131] If the zero-return residual after the first scan is greater than the zero-return residual threshold, the second scan will not be performed, and the retest flag will be output directly.

[0132] After two scans are completed, the waveform parameter trajectories of the two scans are resampled to the same normalized pressure coordinate set, and the waveform parameter difference, the dispersion of the direction difference of the same pressure node, and the difference of the dynamic response loop area at each coordinate position of the two scans are calculated to determine the scan repeatability consistency.

[0133] The repeatability threshold is determined based on data from repeated measurements of the same subject over a short period of time. The calibration data includes the device's inherent noise, normal heart rate variations, and minor wrist disturbances within the allowable range.

[0134] This embodiment uses binary data confidence. The data confidence level is set to 1 when the following conditions are met simultaneously: The number of effective common pressure nodes shall be no less than three; The pressure difference between all pairs of data involved in constructing the dynamic response loop is no greater than the node pressure tolerance. All effective common pressure nodes meet the stability criteria. The transmission coefficients of each organization all satisfy the reference channel's effective variation condition and the regression's effective condition; The scan repeatability consistency is not lower than the repeatability consistency threshold; The zero-return residual is not greater than the zero-return residual threshold. Pressure uniformity is not greater than the pressure uniformity threshold. The radial artery main acquisition channel and tissue reference channel did not fail or switch during this evaluation.

[0135] If any condition is not met, the data credibility is set to 0.

[0136] The confidence threshold is set to 1. When the data confidence is 1, cardiovascular risk characterization values ​​continue to be generated; when the data confidence is 0, a retest flag is output, and the current data is not input into the risk assessment model, nor is the current data used to update the individual's historical baseline.

[0137] An individual's historical baseline consists of at least three historical assessment data points for the same subject whose data reliability reached the reliability threshold.

[0138] Each valid historical assessment should record at least the loading direction waveform parameter trajectory, unloading direction waveform parameter trajectory, direction difference of the same pressure node, directed area of ​​the dynamic response loop, slope difference of the two response trajectories, and zero-return residual under a unified normalized pressure coordinate set.

[0139] Individual historical baselines are represented by the median and discrete range of each dynamic response characteristic. The discrete range can be the median absolute deviation, and a minimum discrete value can be set in conjunction with the device's repeated measurement error to avoid the denominator being zero when historical characteristics are exactly the same.

[0140] The deviation of the current dynamic response characteristic from the individual's historical baseline is determined based on the difference between the current characteristic value and the baseline median value, as well as the corresponding baseline dispersion range.

[0141] Individual historical baselines are not unconditionally updated after each assessment. Data from the current assessment is only written into the individual historical baseline if the data's reliability meets the reliability threshold, and the deviations of the current dynamic response loop's directed area, the difference in direction at the same pressure nodes, the slope difference, and the zero-return residual from the corresponding baseline median are all no greater than the baseline update deviation limit.

[0142] When the deviation of any dynamic response feature exceeds the baseline update deviation limit, the individual historical baseline is kept unchanged to avoid measurement interference or possible state changes being written into the normal baseline.

[0143] The baseline update deviation limit is determined based on intraday and interday repeated measurement data of the same subject during the period when no significant changes in condition are confirmed. It is used to define the range of natural fluctuations in an individual and is not used as a disease diagnosis threshold.

[0144] If the subject is using this system for the first time, or if there are fewer than three historical assessments that meet the data credibility requirements, there will be no complete individual historical baseline in the memory.

[0145] In this state, individual deviations are not calculated; instead, initial cardiovascular risk characterization values ​​are generated based on the deviations of the dynamic response characteristics from the population reference parameters.

[0146] The population reference parameters are established from the labeled samples collected during the development phase. The labeled samples should have reference risk stratification results that are independent of the output of this system. The reference risk stratification process does not use the dynamic response loop, zero-regression residual, or cardiovascular risk characterization values ​​output by this system.

[0147] The median and discrete range of dynamic response characteristics at different risk levels are statistically analyzed to form corresponding population reference parameters. These parameters can be stratified according to age ranges, gender, or other necessary physiological factors, but risk outcomes cannot be generated solely based on demographic information.

[0148] The risk assessment model sets a baseline status flag. When there is no individual historical baseline, the baseline status flag is set to "no baseline" and the individual deviation input is processed according to the missing value handling rules determined during the model training phase. When there is an individual historical baseline, the baseline status flag is set to "with baseline" and the actual determined individual deviation is input.

[0149] After obtaining assessment data that reaches the credibility threshold at least three times, establish the individual historical baseline of the current test subject.

[0150] The inputs to the risk assessment model include at least the directional area of ​​the dynamic response loop, the difference in orientation between nodes under the same pressure under the unified normalized pressure coordinates, the slope difference between the response trajectory in the loading direction and the response trajectory in the unloading direction, the zero-return residual, the baseline status indicator, and the individual deviation when an individual historical baseline exists.

[0151] Risk assessment models can employ logistic regression, support vector machines, gradient boosting models, or neural networks. The model type is not central to this application; the same defined data processing procedures and dynamic response characteristics must be used in both the model training and actual evaluation phases.

[0152] Model parameters are determined based on the training set, and the validation set is used to evaluate the model's discriminative power, calibration error, and repeated measures stability. Data from the same subject must not be included in both the training and validation sets simultaneously.

[0153] The risk stratification labels for the labeled samples are derived from clinical evaluations, vascular status tests, or validated risk assessments independent of wrist pulse wave testing. Applicants should record the testing items, evaluation dates, and label formation rules upon which the labels are based based on actual R&D data, and should not use the output of this system to reverse-generate training labels.

[0154] The processor outputs a cardiovascular risk characterization value between 0 and 100 based on the risk assessment model. This cardiovascular risk characterization value represents the relative position of the current wrist vascular dynamic response within the risk stratification of the training samples, and does not directly indicate that the subject has a specific cardiovascular disease.

[0155] The node pressure tolerance, pressure stability threshold, pressure uniformity threshold, and upper limit of node correction times are determined based on the performance of the contact pressure detection unit, the closed-loop control error of the pressure regulating component, the force calibration of the pressure contact, and the repeatability test of the device.

[0156] The main channel similarity threshold, the upper limit of the reference channel pulsation ratio, the pressure-related threshold, the upper limit of the reference channel peak detection ratio, the effective amplitude threshold, and the effective similarity threshold are determined based on wrist pulse wave samples that have been marked with effective and invalid labels.

[0157] The effective variation threshold of the reference channel is determined based on the noise and drift range of the low-frequency component of the tissue reference channel under constant pressure.

[0158] The effective threshold for regression is determined based on the distribution of regression determination coefficients between confirmed effective tissue stress response samples and invalid samples.

[0159] The repeatability consistency threshold and the zero-regression residual threshold are determined based on repeated stress scan data of the same test object under steady-state conditions.

[0160] The baseline update deviation limit is determined based on the intraday and interday natural fluctuation range of the same subject during the period when no significant state change is confirmed.

[0161] The effective contact lower limit pressure and the non-blocking upper limit pressure are determined in real time for each assessment based on the pulsation amplitude, waveform similarity and pressure changes of the current subject, and a uniform fixed population pressure value is not used.

[0162] The population reference parameters and risk assessment model parameters are derived from labeled samples with independent risk stratification labels and are not determined temporarily based on a single subject or a single assessment result.

[0163] In this embodiment, the radial artery main acquisition channel and tissue reference channel are first determined based on the current wearing position. Then, the effective lower limit pressure and non-blocking upper limit pressure are determined based on the change of the pulse wave of the current subject with contact pressure.

[0164] A common pressure node is set within the individual non-blocking pressure range, and the wrist detection component is controlled to pass through the same common pressure node in both the loading and unloading directions. Both the loading and unloading direction data are registered according to the actual measured contact pressure to avoid misinterpreting drive component backlash or displacement errors as vascular direction responses.

[0165] At each common pressure node, pulse waves are only acquired after the actual contact pressure, the DC component of the radial artery main acquisition channel, and the DC component of the tissue reference channel have all stabilized, thus avoiding the inclusion of optical changes in the wrist soft tissue during creep or stress relaxation processes into the dynamic response loop.

[0166] The tissue reference channel is used to characterize the pressure-related optical changes of adjacent non-arterial tissues and to determine the tissue transfer coefficients in the loading and unloading directions, thereby correcting the direction-related tissue response in the main acquisition channel.

[0167] The corrected pulse waves in the loading and unloading directions form paired data under the same actual contact pressure node, and the loading and unloading response trajectories are constructed respectively. The two response trajectories form a dynamic response loop of the radial artery.

[0168] After completing the pressure scan, the wrist detection component and local blood flow status are judged to have returned to the pre-scan state by returning to zero residual, and accidental movement or contact status changes are further ruled out by repeating the scan consistency.

[0169] Only data that has passed pressure node registration, node stability determination, tissue transmission coefficient validity judgment, zero-return verification, and repeatability consistency judgment can enter the risk assessment model, thus forming a complete technical closed loop consisting of pressure control, pulse wave acquisition, tissue response correction, dynamic response feature construction, data credibility judgment, and risk characterization output.

[0170] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A cardiovascular risk assessment method based on wrist pulse waves, characterized in that, Includes the following steps: S1: Acquire the radial artery main acquisition channel pulse wave, tissue reference channel pulse wave, and actual contact pressure data output by the wrist detection component; S2: Based on the change of the radial artery main acquisition channel pulse wave with the actual contact pressure, determine the effective lower contact pressure and the non-blocking upper pressure, set at least three common pressure nodes between the effective lower contact pressure and the non-blocking upper pressure, and determine the effective lower contact pressure as the starting pressure; S3: Under the initial pressure, the baseline pulse wave before loading is acquired. The wrist detection component is controlled to pass through each common pressure node in sequence from low to high according to the actual contact pressure, and then return to each common pressure node in the reverse order. The actual contact pressure is used as the registration basis. The loading direction data and unloading direction data under the same common pressure node are determined as paired data. When the rate of change of the actual contact pressure, the DC component of the radial artery main acquisition channel and the DC component of the tissue reference channel do not exceed the corresponding stable threshold, the loading direction pulse wave and the unloading direction pulse wave are acquired. S4: Based on the low-frequency response relationship between the radial artery main acquisition channel and the tissue reference channel under each common pressure node, determine the tissue transmission coefficient in the loading direction and the tissue transmission coefficient in the unloading direction respectively. Correct the pressure-related low-frequency component of the radial artery main acquisition channel according to the tissue transmission coefficient in the corresponding direction to obtain the corrected pulse wave in the loading direction and the corrected pulse wave in the unloading direction. S5: Extract waveform parameters of the same type from the loading direction correction pulse wave and the unloading direction correction pulse wave corresponding to the same common pressure node. With the actual contact pressure as the independent variable and the waveform parameters as the dependent variable, the loading direction response trajectory and the unloading direction response trajectory are formed respectively. The radial artery dynamic response loop is constructed based on the two response trajectories. S6: After unloading, return the actual contact pressure to the starting pressure, collect the zero-return pulse wave, and determine the zero-return residual based on the waveform parameter difference between the zero-return pulse wave and the baseline pulse wave before loading. S7: Determine the data reliability based on the number of effective common pressure nodes, the pressure difference between paired data, the stability determination results of each common pressure node, and the zero-return residual. When the data credibility reaches the credibility judgment threshold, a cardiovascular risk characterization value is generated based on the difference in the same pressure node direction and the directional area of ​​the radial artery dynamic response loop, the zero-return residual, and the deviation of the above parameters from the individual's historical baseline or population reference parameters; when the data credibility does not reach the credibility judgment threshold, a retesting flag is output.

2. The cardiovascular risk assessment method based on wrist pulse wave as described in claim 1, characterized in that, The radial artery main acquisition channel and tissue reference channel are determined from at least three photoelectric acquisition channels of the wrist detection component, specifically including: At least three photoelectric acquisition channels were sampled simultaneously within the same time window under channel screening pressure. The ratio of pulsation component to DC component and waveform similarity between adjacent cardiac cycles were calculated for each photoelectric acquisition channel. From the photoelectric acquisition channels with waveform similarity not lower than the main channel similarity threshold, the photoelectric acquisition channel with the largest ratio of pulsating component to DC component is determined as the radial artery main acquisition channel; From the photoelectric acquisition channels formed by photoelectric acquisition units set adjacent to the photoelectric acquisition unit corresponding to the radial artery main acquisition channel, select photoelectric acquisition channels whose ratio of pulsation component to DC component is not higher than the upper limit of the reference channel pulsation ratio, whose correlation coefficient between DC component and actual contact pressure is not lower than the pressure correlation threshold, and whose detection ratio of main wave peak or reflected wave peak in continuous cardiac cycles is lower than the upper limit of the reference channel wave peak detection ratio. The photoelectric acquisition channel with the highest correlation coefficient with the low-frequency component of the radial artery main acquisition channel among the screened photoelectric acquisition channels was determined as the tissue reference channel.

3. The cardiovascular risk assessment method based on wrist pulse wave according to claim 2, characterized in that, Determine the effective lower contact pressure and the non-blocking upper pressure, including: According to the candidate pressure order from low to high, the pulsation amplitude of the radial artery main acquisition channel and the waveform similarity between adjacent cardiac cycles are obtained sequentially at each candidate pressure. The first candidate pressure with a pulsation amplitude not lower than the effective amplitude threshold and a waveform similarity not lower than the effective similarity threshold is determined as the effective contact lower limit pressure; Determine the local maximum value of the pulsation amplitude under each candidate pressure. In two consecutive adjacent pressure intervals after the pressure corresponding to the local maximum value, when the pulsation amplitude decreases as the actual contact pressure increases, and the decrease ratio of the pulsation amplitude under the subsequent candidate pressure relative to the local maximum value reaches the compression judgment ratio, the candidate pressure before the subsequent candidate pressure is determined as the non-blocking upper limit pressure. The effective contact lower limit pressure and the non-blocking upper limit pressure are set as the first common pressure node and the last common pressure node, respectively, and at least one intermediate common pressure node is set between them; If the effective lower limit pressure or non-blocking upper limit pressure is not determined, output the wear adjustment flag and terminate the current pressure scan.

4. The cardiovascular risk assessment method based on wrist pulse wave according to claim 3, characterized in that, Loading direction data and unloading direction data under the same common pressure node are identified as paired data, including: Based on actual contact pressure data, the wrist detection component is subjected to closed-loop pressure control so that the loading and unloading directions reach the common pressure nodes respectively. Calculate the pressure difference between the actual contact pressure in the loading direction and the actual contact pressure in the unloading direction at the same common pressure node; When the pressure difference is not greater than the node pressure tolerance, the corresponding loading direction data and unloading direction data are determined as paired data; When the pressure difference is greater than the node pressure tolerance, the node correction command is output to the wrist detection component according to the pressure difference, and the pulse wave in the corresponding direction is re-acquired after the actual contact pressure stabilizes after correction. If the pressure difference is still greater than the node pressure tolerance after the node correction count reaches the upper limit, the corresponding common pressure node will be marked as an invalid common pressure node; if there are fewer than three valid common pressure nodes, a retest flag will be output and the current risk assessment will be terminated.

5. A cardiovascular risk assessment method based on wrist pulse waves according to claim 4, characterized in that, At each common pressure node, pulse waves in the loading or unloading direction are collected, including: After the wrist detection component reaches the common pressure node, it enters the pressure holding state, and calculates the actual contact pressure change rate, the change rate of the DC component of the radial artery main acquisition channel, and the change rate of the DC component of the tissue reference channel within a continuous stable detection time window. When the absolute value of the actual contact pressure change rate is not greater than the pressure stabilization threshold, the absolute value of the change rate of the DC component of the radial artery main acquisition channel is not greater than the main channel stabilization threshold, the absolute value of the change rate of the DC component of the tissue reference channel is not greater than the reference channel stabilization threshold, and the above conditions are maintained for at least two consecutive stable detection time windows, pulse waves of no less than three complete cardiac cycles are collected. If the pressure holding time reaches the node holding time limit and the common pressure node still fails to meet the above conditions, the corresponding common pressure node will be marked as an invalid common pressure node.

6. The cardiovascular risk assessment method based on wrist pulse wave according to claim 5, characterized in that, Determine the tissue transfer coefficients in the loading and unloading directions and construct the radial artery dynamic response loop, including: Low-frequency components varying with actual contact pressure were extracted from data from the radial artery main acquisition channel and the tissue reference channel, respectively. Select a local pressure window that includes the current common pressure node. Based on the regression relationship between the change in the low-frequency component of the radial artery main acquisition channel and the change in the low-frequency component of the tissue reference channel within the local pressure window, determine the tissue transfer coefficient in the loading direction and the tissue transfer coefficient in the unloading direction, respectively. The low-frequency component offset of the tissue reference channel relative to the initial pressure state is mapped to the radial artery main acquisition channel according to the tissue transmission coefficient in the corresponding direction. The mapping result is subtracted from the low-frequency component of the radial artery main acquisition channel. The pulsating component is normalized with the corrected DC component to obtain the loading direction corrected pulse wave and the unloading direction corrected pulse wave. Extract at least two waveform parameters from the loading-direction corrected pulse wave and the unloading-direction corrected pulse wave: main wave rise time, maximum rise rate, reflected wave amplitude ratio, and diastolic area. For each waveform parameter, a loading direction response trajectory and an unloading direction response trajectory are formed. The region enclosed by the loading direction response trajectory, the unloading direction response trajectory, and the boundary segments connecting the endpoints of the two response trajectories at the effective contact lower limit pressure and the non-blocking upper limit pressure is defined as the radial artery dynamic response loop for the corresponding waveform parameter. Radial artery dynamic response loop groups are formed by the radial artery dynamic response loops corresponding to different waveform parameters.

7. A cardiovascular risk assessment method based on wrist pulse waves according to claim 6, characterized in that, Determining data reliability and generating cardiovascular risk characterization values ​​includes: In the same risk assessment, two loading and unloading processes are executed consecutively. The scan repeatability consistency is determined based on the waveform parameter differences of the response trajectories obtained from the two loading and unloading processes at each common pressure node and the dispersion of the direction difference with the same pressure node. When the number of effective common pressure nodes is not less than the lower limit of effective nodes, the pressure difference between each pair of data is not greater than the node pressure tolerance, each effective common pressure node meets the stability judgment condition, the scan repeatability consistency is not lower than the repeatability consistency threshold and the zero-return residual is not greater than the zero-return residual threshold, the data credibility is judged to reach the credibility judgment threshold; when any condition is not met, the data credibility is judged not to reach the credibility judgment threshold. The individual historical baseline consists of historical assessment data from no fewer than three tests on the same subject, in which the data reliability reached the reliability judgment threshold. The baseline median and baseline dispersion range of the directed area, the difference in direction of the same pressure node, the slope difference, and the zero-return residual of each radial artery dynamic response loop are recorded. When there is an individual historical baseline corresponding to the current test object, the individual deviation is determined based on the differences between the oriented area, the difference in the direction of the same pressure node, the slope difference, and the zero-return residual obtained in this assessment and the median value of the corresponding baseline, as well as the baseline dispersion range. When no individual historical baseline is stored, the initial cardiovascular risk characterization value is generated based on the deviation of the directional area, the difference in direction of the same pressure node, the slope difference, and the zero-return residual obtained in this assessment from the population reference parameters. After obtaining no less than three assessment data with data credibility reaching the credibility judgment threshold, the individual historical baseline is established. Based on labeled samples with cardiovascular risk stratification labels, a mapping relationship is established between directed area, difference in direction of the same pressure node, slope difference, zero-return residual and individual deviation and cardiovascular risk characterization value, and the cardiovascular risk characterization value corresponding to this assessment is generated according to the mapping relationship. The data from this assessment will be written into the individual's historical baseline only if the reliability of the data reaches the reliability threshold and the deviations of the directional area, the difference in direction of the same pressure node, the slope difference, and the zero-return residual obtained in this assessment from the corresponding baseline median value are all no greater than the baseline update deviation limit; otherwise, the individual's historical baseline will remain unchanged.

8. A cardiovascular risk assessment system based on wrist pulse waves, applicable to the cardiovascular risk assessment method based on wrist pulse waves as described in any one of claims 1-7, characterized in that, Includes wrist detection components, pressure regulation components, a processor, and memory; The wrist detection component includes a radial artery main acquisition unit, a tissue reference acquisition unit, and a contact pressure detection unit. The radial artery main acquisition unit is used to output the radial artery main acquisition channel pulse wave, the tissue reference acquisition unit is used to output the tissue reference channel pulse wave, and the contact pressure detection unit is used to output the actual contact pressure data. The pressure adjustment component is connected to the wrist detection component and is used to change the contact pressure between the wrist detection component and the wrist skin according to the load control command and unload control command output by the processor. The processor is connected to the wrist detection component, the pressure regulation component, and the memory, and is configured as follows: Determine the effective contact lower limit pressure and the non-blocking upper limit pressure, set at least three common pressure nodes between the two, and determine the effective contact lower limit pressure as the starting pressure; Baseline pulse waves were acquired before loading under initial pressure. The pressure regulation component was controlled to pass through each common pressure node according to the loading and unloading directions. The actual contact pressure was used as the registration basis. The loading direction data and unloading direction data under the same common pressure node were determined as paired data. Pulse waves were acquired when the rate of change of the actual contact pressure, the DC component of the radial artery main acquisition channel and the DC component of the tissue reference channel did not exceed the corresponding stable threshold. Based on the low-frequency response relationship between the radial artery main acquisition channel and the tissue reference channel, the tissue transmission coefficients in the loading direction and unloading direction are determined respectively. The radial artery main acquisition channel is corrected according to the tissue transmission coefficients in the corresponding directions to obtain the corrected pulse wave in the loading direction and the corrected pulse wave in the unloading direction. The radial artery dynamic response loop is constructed based on the corrected pulse wave in the loading direction and the corrected pulse wave in the unloading direction corresponding to the same common pressure node. After unloading is completed, the control pressure regulating component returns to the starting pressure, the zero-return pulse wave is collected, and the zero-return residual is determined based on the zero-return pulse wave and the baseline pulse wave before loading. Data reliability is determined based on the number of effective common pressure nodes, the pressure difference between paired data, the stability determination results of each common pressure node, and the zero-return residual. When the data reliability reaches the reliability determination threshold, a cardiovascular risk characterization value is generated based on the radial artery dynamic response loop, the zero-return residual, and the deviation of the above parameters from the individual's historical baseline or population reference parameters. When the data reliability does not reach the reliability determination threshold, a retest flag is output.

9. A cardiovascular risk assessment system based on wrist pulse waves according to claim 8, characterized in that, The wrist detection component includes a pressure contact and a pressure-distributing flexible layer disposed on the side of the pressure contact facing the wrist. At least three photoelectric acquisition units are disposed laterally on the pressure contact along the wrist, and the optical window of each photoelectric acquisition unit is flush with the contact surface of the pressure-distributing flexible layer facing the wrist. The radial artery master acquisition unit and the tissue reference acquisition unit are formed by corresponding photoelectric acquisition units from at least three photoelectric acquisition units; The contact pressure detection unit includes at least two pressure detection units respectively arranged on both sides of the pressure contact along the arrangement direction of the photoelectric acquisition unit. The processor determines the actual contact pressure based on the average value of the pressure data output by the at least two pressure detection units, and determines the pressure uniformity based on the difference between the maximum pressure data and the minimum pressure data. The pressure regulating component includes a driving element and a pressure transmitting element. One end of the pressure transmitting element is connected to the driving element, and the other end is connected to the pressure contact. The driving element drives the pressure contact to reciprocate along the normal direction of the wrist contact surface according to the load control command and unload control command output by the processor. When the pressure uniformity exceeds the pressure uniformity threshold, the processor stops collecting pulse waves from the corresponding common pressure node and marks the corresponding common pressure node as an invalid common pressure node.

10. A cardiovascular risk assessment system based on wrist pulse waves according to claim 9, characterized in that, The memory stores node pressure tolerance, pressure stability threshold, main channel stability threshold, reference channel stability threshold, confidence judgment threshold, repeatability consistency threshold, zero-return residual threshold, baseline update deviation limit, population reference parameters, and individual historical baselines. The processor is also configured to control the pressure regulation component to perform two loading and unloading processes consecutively in the same risk assessment, and to determine the scan repeatability consistency based on the waveform parameter differences of the response trajectories obtained from the two loading and unloading processes at each common pressure node and the dispersion of the direction difference with the same pressure node. When the number of effective common pressure nodes is not less than the lower limit of effective nodes, the pressure difference between each pair of data is not greater than the node pressure tolerance, each effective common pressure node meets the stability judgment condition, the scan repeatability consistency is not lower than the repeatability consistency threshold, the zero-return residual is not greater than the zero-return residual threshold and the pressure uniformity is not greater than the pressure uniformity threshold, the processor determines that the data credibility reaches the credibility judgment threshold. When an individual historical baseline corresponding to the current subject is stored, the processor generates a cardiovascular risk characterization value based on the deviation of the radial artery dynamic response loop and zero-return residual obtained in this assessment from the individual historical baseline; when no individual historical baseline is stored, the processor generates an initial cardiovascular risk characterization value based on the deviation of the radial artery dynamic response loop and zero-return residual obtained in this assessment from the population reference parameter, and establishes an individual historical baseline after obtaining no less than three assessment data with a confidence level reaching the confidence judgment threshold. The processor will write the current assessment data into the individual's historical baseline only if the reliability of the current data reaches the reliability judgment threshold, and the deviation of the radial artery dynamic response loop and the zero-return residual obtained in this assessment from the corresponding baseline median value is not greater than the baseline update deviation limit; otherwise, the individual's historical baseline will remain unchanged.

Images