Blood pressure waveform estimation system based on pressurization PPG waveform

Abstract

The invention discloses a blood pressure waveform estimation system based on a pressurized PPG waveform, and the system comprises a PPG signal collection module which is used for collecting a pressurized PPG signal; the pressure applying module is used for applying external pressure to the finger tip, so that a finger tip blood vessel is located at a preset working point; the signal processing module is used for performing dynamic analysis and model construction on the PPG signal and the pressure signal and determining a continuous blood pressure waveform; the data output module is used for outputting the continuous blood pressure waveform and the blood pressure parameters calculated according to the waveform, and the stability and accuracy of blood pressure waveform detection are improved through continuous detection of the blood pressure waveform by a self-adaptive joint model.

Description

Technical Field

[0001] This application relates to the field of blood pressure measurement technology, specifically to a blood pressure waveform estimation system based on pressurized PPG waveform. Background Technology

[0002] Photoplethysmography (PPG) is a non-invasive optical technique that measures changes in light absorption on the skin surface to reflect fluctuations in blood volume within the microvascular bed. It is widely used in heart rate, blood oxygen saturation, and blood pressure monitoring. Traditional blood pressure measurement primarily relies on cuff-type blood pressure monitors, which measure systolic and diastolic blood pressure through inflation. While accurate, this method carries risks such as measurement interruptions, user discomfort, and potential vascular damage. With the rise of wearable devices, recent studies have shown that PPG signals contain rich hemodynamic information that can be used for continuous blood pressure estimation, for example, through pulse wave analysis or modeling the correlation between waveform characteristics and blood pressure.

[0003] Existing PPG-based blood pressure monitoring technologies, such as cuffless methods based on machine learning models, primarily predict blood pressure by extracting PPG waveform features and combining them with neural networks or regression algorithms. While these methods achieve non-invasive and continuous monitoring to some extent, they still have the following limitations:

[0004] Hardware limitations: Existing PPG sensors typically use a single wavelength, which is susceptible to skin type, ambient light, and motion artifacts, resulting in high signal noise and an inability to effectively capture deep blood flow information;

[0005] Algorithm-level shortcomings: Traditional methods rely heavily on fixed feature extraction or static models, which are difficult to adapt to individual physiological differences such as age and gender, as well as dynamic changes caused by posture or activity state, resulting in large estimation errors;

[0006] Robustness issues: PPG signal amplitude is weak and easily affected by breathing, temperature and light scattering. Although existing learning methods can compensate for some noise, the response is slow and frequent user calibration is required, which limits the application of long-term monitoring.

[0007] Therefore, there is an urgent need for a PPG blood pressure estimation system that simultaneously possesses pressurization control, multi-wavelength signal fusion, low-noise preprocessing capabilities, and an adaptive joint model, in order to achieve high-resolution, high-stability continuous blood pressure waveform monitoring. Summary of the Invention

[0008] This invention provides a blood pressure waveform estimation system based on pressurized PPG waveforms to solve the technical problems described in the background section. The system includes:

[0009] The PPG signal acquisition module is used to acquire pressurized PPG signals under external pressure.

[0010] The pressure application module is used to apply and regulate external pressure to the fingertip so that the blood vessels in the fingertip are locked at a preset working point;

[0011] The signal processing module is used to perform dynamic analysis and model building on the PPG signal and the external pressure signal, thereby estimating the continuous blood pressure waveform.

[0012] The data output module is used to output the continuous blood pressure waveform and the blood pressure parameters calculated based on the waveform.

[0013] Preferably, the PPG signal acquisition module specifically includes:

[0014] A multi-wavelength light source unit is used to alternately illuminate the fingertip with red and infrared LED sensors to provide a multi-wavelength light source;

[0015] The photoelectric detection unit is used to measure the light intensity changes of a multi-wavelength light source through a photoelectric detector, and generate a multi-wavelength PPG signal based on the light intensity changes;

[0016] The signal fusion filtering unit is used to adaptively filter out noise and motion artifacts in the multi-wavelength PPG signal through a filter and using the least mean square (LMS) algorithm to obtain the pressurized PPG signal.

[0017] Preferably, the weight update formula for the filter is as follows:

[0018] ;

[0019] in, For filter weights, Step size factor For error signals, Let be the input signal. This formula is used to adaptively update the filter weights by minimizing the mean square error, thereby improving the signal-to-noise ratio of the pressurized PPG signal and suppressing motion artifacts.

[0020] Preferably, the pressure application module specifically includes:

[0021] A finger-type pressure actuator for applying the external pressure to the fingertip;

[0022] A closed-loop controller is used to lock the initial pressure on the fingertip and ensure that the fingertip blood vessels are at the preset operating point.

[0023] Preferably, vascular compliance is determined by the following formula:

[0024] ;

[0025] Wherein, C represents vascular compliance, ΔV represents the change in blood volume in the fingertip vessels, and ΔP represents the change in external pressure; the vascular compliance is used to guide the pressure application module to lock the fingertip vessels at the point of maximum compliance as the preset working point, and to provide physiological prior constraints for the subsequent joint model.

[0026] Preferably, the signal processing module specifically includes:

[0027] The feature extraction unit is used to divide the preprocessed signal into cardiac cycles based on the cardiac cycle segmentation algorithm and extract time-domain features from each cardiac cycle. The time-domain features include pulse peak amplitude, rising slope, systolic area integral, time and amplitude of diabetic notch, and pressure-amplitude curve features.

[0028] The joint model unit is used to construct a joint model that includes a parameterized sub-model and a learning sub-model, and to estimate a continuous blood pressure waveform based on the time-domain features, the pressurized PPG signal, and the external pressure signal.

[0029] The parameter calibration unit is used to perform real-time calibration and slow calibration on the continuous blood pressure waveform and its model parameters. The real-time calibration is mainly for calibrating vascular compliance, physiological mapping parameters in the joint model and regression coefficients, while the slow calibration is used to compensate for long-term physiological drift.

[0030] The data fusion unit is used to perform state fusion and smoothing on the continuous blood pressure waveform and related parameters of multiple heartbeat cycles using extended Kalman filtering, and output the final high-precision continuous blood pressure waveform.

[0031] The data fusion unit is used to perform state fusion and smoothing on the continuous blood pressure waveforms and related model parameters of multiple heartbeat cycles using extended Kalman filtering, thereby outputting a final continuous blood pressure waveform with higher stability and accuracy.

[0032] Preferably, the joint model unit specifically includes:

[0033] The parameterized sub-model, specifically determined through the arterial physiological model, is used to establish a priori mapping relationship between external pressure, intra-arterial pressure and PPG waveform based on vascular compliance, and the output of the parameterized sub-model is used as a priori constraint for the joint model.

[0034] The learning sub-model is used to take the pressurized PPG signal, the external pressure signal and the time-domain features as inputs, output the instantaneous pressure residual, and obtain the continuous blood pressure waveform by fitting it with a linear regression model.

[0035] The joint model combines the theoretically predicted waveform of the parameterized sub-model with the instantaneous pressure residual of the learned sub-model to obtain the final continuous blood pressure waveform.

[0036] Preferably, the signal preprocessing unit is specifically used for:

[0037] The pressurized PPG signal PPG(t) and the external pressure signal Pactuator(t) are sampled and synchronized, bandpass filtered, baseline drift removed, and artifacts detected and eliminated. Specifically, a discrete wavelet transform with Daubechies 4 as the mother wavelet is used to simultaneously suppress noise and remove baseline drift. The threshold denoising formula is as follows:

[0038] ;

[0039] Where σ is the noise standard deviation and N is the signal length, in order to improve the waveform signal-to-noise ratio and preserve dynamic characteristics;

[0040] Among them, artifact removal is performed based on acceleration sensor signals and PPG peak abrupt changes.

[0041] Preferably, the parameter calibration unit is specifically used for:

[0042] The least squares method is used to fit multi-period waveform data, and the error parameters are adjusted using an error parameter adjustment formula.

[0043] The empirical mode decomposition formula was used to slowly adjust the parameters of the continuous blood pressure waveform;

[0044] The error adjustment formula is as follows:

[0045] ;

[0046] in, This is an estimated value. For reference only. The number of samples is denoted by . This formula is used to quantify the mean squared error between the estimated blood pressure and the reference blood pressure, and to guide model parameter calibration.

[0047] The empirical mode decomposition formula is as follows:

[0048] ;

[0049] in For modal functions, The residual is used to decompose the blood pressure waveform into components at different time scales to compensate for slow physiological drift.

[0050] Preferably, the data output module specifically includes:

[0051] The wireless communication unit is used to transmit the continuous blood pressure waveform and the blood pressure parameters calculated from the waveform to a mobile device or the cloud in real time via wireless communication.

[0052] A visualization display unit is used to display the continuous blood pressure waveform and blood pressure parameters on a display device;

[0053] The storage and recalibration unit is used to trigger an automatic recalibration procedure when the signal quality score is lower than a preset threshold. This includes updating the joint model parameters using a stepwise pressure scan or a small-range pressure oscillation method, wherein the signal quality score is determined by the signal-to-noise ratio.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] This invention discloses a blood pressure waveform estimation system based on pressurized PPG waveforms. The system includes: a PPG signal acquisition module for acquiring pressurized PPG signals; a pressure application module for applying external pressure to the fingertip to bring the fingertip blood vessels to a preset working point; a signal processing module for dynamically analyzing and modeling the PPG and pressure signals, and determining a continuous blood pressure waveform; and a data output module for outputting the continuous blood pressure waveform and blood pressure parameters calculated from the waveform. By using an adaptive joint model for continuous detection of the blood pressure waveform, the system improves the stability and accuracy of blood pressure waveform detection. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 The diagram shows a schematic of a blood pressure waveform estimation system based on pressurized PPG waveform proposed in an embodiment of the present invention. Detailed Implementation

[0058] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0059] Combination Figure 1 The blood pressure waveform estimation system based on pressurized PPG waveform proposed in this embodiment of the invention is as follows:

[0060] The PPG signal acquisition module is used to acquire pressurized PPG signals under external pressure.

[0061] The pressure application module is used to apply and regulate external pressure to the fingertip so that the blood vessels at the fingertip are locked at a preset working point;

[0062] The signal processing module is used to perform dynamic analysis and model building on the PPG signal and the external pressure signal, thereby estimating the continuous blood pressure waveform.

[0063] The data output module is used to output the continuous blood pressure waveform and the blood pressure parameters calculated based on the waveform.

[0064] In a preferred embodiment of this application, the PPG signal acquisition module specifically includes:

[0065] A multi-wavelength light source unit is used to alternately illuminate the fingertip with red and infrared LED sensors to provide a multi-wavelength light source;

[0066] The photoelectric detection unit is used to measure the light intensity changes of a multi-wavelength light source through a photoelectric detector, and generate a multi-wavelength PPG signal based on the light intensity changes;

[0067] The signal fusion filtering unit is used to adaptively filter out noise and motion artifacts in the multi-wavelength PPG signal through a filter and using the least mean square (LMS) algorithm to obtain the pressurized PPG signal.

[0068] In a preferred embodiment of this application, the weight update formula for the filter is specifically as follows:

[0069] ;

[0070] in, For filter weights, Step size factor For error signals, Let be the input signal. This formula is used to adaptively update the filter weights by minimizing the mean square error, thereby improving the signal-to-noise ratio of the pressurized PPG signal and suppressing motion artifacts.

[0071] Specifically, the PPG signal acquisition module acquires the raw pressurized PPG signal from the fingertip. The module uses red and infrared LED sensors to alternately illuminate the fingertip, and a photodetector measures the change in transmitted light intensity to generate a multi-wavelength PPG signal containing both AC and DC components. An adaptive fusion filtering algorithm is used to fuse the red and infrared signals, leveraging the sensitivity of red light to superficial blood vessels and the penetrability of infrared light to deep tissues to improve the overall signal-to-noise ratio, thereby obtaining a higher-quality pressurized PPG signal.

[0072] In a preferred embodiment of this application, the pressure application module specifically includes:

[0073] A finger-type pressure actuator for applying the external pressure to the fingertip;

[0074] A closed-loop controller is used to dynamically adjust and lock the external pressure on the fingertip based on changes in vascular compliance, ensuring that the fingertip blood vessels are at a preset operating point.

[0075] In a preferred embodiment of this application, vascular compliance is specifically determined by the following formula:

[0076] ;

[0077] Wherein, C represents vascular compliance, ΔV represents the change in blood volume in the fingertip vessels, and ΔP represents the change in external pressure; the vascular compliance is used to guide the pressure application module to lock the fingertip vessels at the point of maximum compliance as the preset working point, and to provide physiological prior constraints for the subsequent joint model.

[0078] The specific pressure application module uses a finger-type pressure actuator to apply controllable external pressure to the fingertip, and monitors pressure changes in real time through a closed-loop controller. The change in blood volume ΔV can be approximately characterized by the amplitude change of the pressurized PPG signal. By scanning different external pressure levels, the curve of vascular compliance as a function of pressure is calculated, and the point of maximum compliance is selected as the preset operating point to ensure that the blood vessel is in a state that is most sensitive to and stable in response to pressure changes.

[0079] In a preferred embodiment of this application, the signal processing module specifically includes:

[0080] A signal preprocessing unit is used to preprocess the pressurized PPG signal and the pressure signal to obtain a preprocessed signal;

[0081] The feature extraction unit is used to divide the preprocessed signal into cardiac cycles based on the cardiac cycle segmentation algorithm and extract time-domain features from each cardiac cycle. The time-domain features include pulse peak amplitude, rising slope, systolic area integral, time and amplitude of diabetic notch, and pressure-amplitude curve features.

[0082] The joint model unit is used to construct a joint model that includes a parameterized sub-model and a learning sub-model, and to estimate a continuous blood pressure waveform based on the time-domain features, the pressurized PPG signal, and the external pressure signal.

[0083] The parameter calibration unit is used to perform real-time calibration and slow calibration on the continuous blood pressure waveform and its model parameters. The real-time calibration is mainly for calibrating vascular compliance, physiological mapping parameters in the joint model and regression coefficients, while the slow calibration is used to compensate for long-term physiological drift.

[0084] The data fusion unit is used to perform state fusion and smoothing on the continuous blood pressure waveform and related parameters of multiple heartbeat cycles using extended Kalman filtering, and output the final high-precision continuous blood pressure waveform.

[0085] In a preferred embodiment of this application, the signal preprocessing unit is specifically used for:

[0086] The pressurized PPG signal PPG(t) and the external pressure signal Pactuator(t) are sampled and synchronized, bandpass filtered, baseline drift removed, and artifacts detected and eliminated. Specifically, a discrete wavelet transform with Daubechies 4 as the mother wavelet is used to simultaneously suppress noise and remove baseline drift. The threshold denoising formula is as follows:

[0087] ;

[0088] Where σ is the noise standard deviation and N is the signal length, in order to improve the waveform signal-to-noise ratio and preserve dynamic characteristics;

[0089] Among them, artifact removal is performed based on acceleration sensor signals and PPG peak abrupt changes.

[0090] Preferably, a discrete wavelet transform with Daubechies 4 as the mother wavelet and a decomposition level of 5 is used to remove noise and baseline drift. The threshold denoising formula is as follows:

[0091] ;

[0092] Where σ is the noise standard deviation and N is the signal length. Simultaneously, motion artifact removal is performed based on the abrupt changes in the PPG peak shape of the accelerometer signal.

[0093] The feature extraction unit, based on a heartbeat segmentation algorithm, extracts time-domain features from each heartbeat cycle, including: pulse amplitude, rise slope, area integral of systole (SAI), temporal position and amplitude of the dicrotic notch, and pressure-amplitude curve features. The area integral of systole is defined as:

[0094] ;

[0095] Among them, t s t is the start time of the contraction period. c The end of systole is the time when this feature is used to reflect the hemodynamic characteristics during systole.

[0096] In a preferred embodiment of this application, the joint model unit specifically includes:

[0097] The parameterized sub-model, specifically determined through the arterial physiological model, is used to establish a priori mapping relationship between external pressure, intra-arterial pressure and PPG waveform based on vascular compliance, and the output of the parameterized sub-model is used as a priori constraint for the joint model.

[0098] The learning sub-model is used to take the pressurized PPG signal, the external pressure signal and the time-domain features as inputs, output the instantaneous pressure residual, and obtain the continuous blood pressure waveform by fitting it with a linear regression model.

[0099] The joint model combines the theoretically predicted waveform of the parameterized sub-model with the instantaneous pressure residual of the learned sub-model to obtain the final continuous blood pressure waveform.

[0100] Construct a joint model to estimate the continuous blood pressure waveform P art (t), the joint model comprises a parameterized sub-model based on an arterial physiological model and a data-driven learning sub-model. The parameterized sub-model establishes a priori mapping relationships between applied pressure, intra-arterial pressure, and PPG waveforms based on vascular compliance C, and uses the output of the physical model as a priori constraint for the joint model.

[0101] The learning sub-model is used to output the instantaneous pressure residual by taking the pressurized PPG signal, external pressure signal, heart rate and the time-domain features as inputs.

[0102] Continuous blood pressure waveforms are obtained by combining the theoretical predictions of the parameterized sub-model with the instantaneous pressure residuals output by the learned sub-model. The learned sub-model uses a linear regression model to fit the continuous blood pressure waveforms.

[0103] ;

[0104] in, , , , The regression coefficients are obtained by least squares optimization fitting. This represents the error term; the model is used to compensate for individual differences and dynamic changes under physiological prior constraints, achieving continuous blood pressure waveform estimation.

[0105] In a preferred embodiment of this application, the parameter calibration unit is specifically used for:

[0106] The least squares method is used to fit multi-period waveform data, and the error parameters are adjusted using an error parameter adjustment formula.

[0107] The error adjustment formula is as follows:

[0108] ;

[0109] in, This is an estimated value. For reference only. The number of samples;

[0110] Slow calibration employs the Empirical Mode Decomposition (EMD) method, with the following decomposition formula:

[0111] ;

[0112] in For modal functions, It represents the residual. It is used to compensate for long-term physiological drift and environmental disturbances.

[0113] The data fusion submodule integrates extended Kalman filtering to fuse multi-period data, and the state update formula is:

[0114] ;

[0115] ;

[0116] Where f and h are nonlinear functions, and w and v are noise, used to improve the robustness of continuous blood pressure waveforms under different individuals and different working conditions.

[0117] In a preferred embodiment of this application, the wireless communication unit is used to transmit a continuous blood pressure waveform and blood pressure parameters calculated based on the waveform to a mobile device or the cloud in real time via wireless communication.

[0118] A visualization display unit is used to display the continuous blood pressure waveform and blood pressure parameters on a display device;

[0119] The storage and recalibration unit is used to trigger an automatic recalibration procedure when the signal quality score is lower than a preset threshold. This includes updating the joint model parameters using a stepwise pressure scan or a small-range pressure oscillation method, wherein the signal quality score is determined by the signal-to-noise ratio.

[0120] The principle of this solution will now be further explained with reference to specific application examples.

[0121] Example 1:

[0122] In this embodiment, a specific acquisition and preprocessing process for blood pressure waveform estimation based on pressurized PPG waveform is provided.

[0123] The PPG signal acquisition module uses red LEDs (wavelength 660 nm) and infrared LEDs (wavelength 940 nm) as light sources, and alternately irradiates the fingertip tissue in a time-division multiplexing manner, with a single irradiation time of 0.5 ms and an irradiation interval of 0.2 ms; the signal of transmitted light intensity change is acquired through silicon photodiodes.

[0124] When performing spectral separation processing on the acquired signal, the AC component is filtered using a third-order Butterworth bandpass filter with a filtering frequency range of 0.6–6 Hz. Its transfer function is expressed as:

[0125] ;

[0126] in, The cutoff angular frequency is 0.05 Hz. The DC component is extracted using a low-pass filter with a cutoff frequency of 0.05 Hz. The red and infrared signals are fused using an adaptive minimum mean square algorithm, with the weight update formula as follows:

[0127] ;

[0128] Where w(n) is the filter weight, μ is the step size factor, preferably 0.03, e(n) is the error signal, and x(n) is the input signal vector.

[0129] This filter removes noise and motion artifacts, ensuring a signal-to-noise ratio greater than 28 dB.

[0130] The preprocessing submodule processes PPG(t) and the applied pressure signal P. actuator (t) performs sampling synchronization, with the sampling rate set to 500 Hz.

[0131] Subsequently, wavelet transform was used to remove high-frequency noise and baseline drift. The wavelet mother function was selected as Daubechies 4, and the wavelet decomposition level was 5. The threshold denoising formula is as follows:

[0132] ;

[0133] in, The standard deviation of noise. The signal length is given. Further bandpass filtering is used to remove low-frequency drift components below 0.5 Hz and high-frequency interference components above 5 Hz. Motion artifact detection is performed using accelerometer data, and abnormal heartbeats are identified and removed based on peak abrupt change thresholds.

[0134] Example 2:

[0135] This embodiment mainly focuses on feature extraction methods and blood pressure waveform estimation process.

[0136] The feature extraction submodule extracts multidimensional temporal and morphological features from the preprocessed PPG signal. Heartbeat segmentation employs a gradient threshold-based detection algorithm, where the threshold is set to 0.3 times the corresponding peak heartbeat amplitude to identify the start and end points of systole.

[0137] The pulse amplitude is calculated as the peak value minus the trough value, and the rising slope is obtained using the finite difference method.

[0138] ;

[0139] The area integral during contraction (SAI) is:

[0140] ;

[0141] Where t s and t c Let represent the start and end times of the contraction period, respectively. The integral is approximated using the trapezoidal integral method.

[0142] The location of the diabetic notch is detected using the Hilbert transform envelope, and the envelope calculation formula is as follows:

[0143] ;

[0144] in, This is the result of the Hilbert transform.

[0145] During the external pressure scan, the external pressure was gradually increased from 50 mmHg to 150 mmHg within a 5-second time window, with a step size of 10 mmHg, to construct the pressure-amplitude relationship curve, which was then fitted using a quadratic polynomial.

[0146] The blood pressure estimation submodule employs a joint modeling approach combining a parametric model and a learned model. The parametric submodel is constructed based on vascular compliance C, and its mapping relationship is solved using the finite difference method.

[0147] The learning sub-model adopts a combined CNN and LSTM network structure. During training, the Adam optimizer is used with a learning rate of 0.001. Data augmentation is performed by superimposing Gaussian noise with an amplitude of 10% of the original signal into the training samples.

[0148] Blood pressure waveform estimation uses a linear regression model to fit P. art (t), the regression formula is:

[0149] ;

[0150] in, i The regression coefficients are obtained using the least squares method. This is the residual term.

[0151] Finally, systolic blood pressure (SBP), diastolic blood pressure (DBP), and mean arterial pressure (MAP) are calculated using the peak, trough, and average values ​​of the pressure waveform, with the estimation error preferably controlled within ±4 mmHg.

[0152] Example 3:

[0153] This embodiment focuses on parameter calibration, data fusion, and system application.

[0154] The parameter calibration submodule calibrates compliance parameters in real time, fitting 10 cardiac cycles of data using the least squares method. The error function is:

[0155] ;

[0156] in, For model estimation, For calibration reference, This represents the number of samples.

[0157] The calibration parameter update cycle is set to be performed once per heartbeat cycle.

[0158] Slow calibration employs the Empirical Mode Decomposition (EMD) method, performed once every 30 heartbeat cycles. Its signal decomposition expression is as follows:

[0159] ;

[0160] in For modal functions, The residuals are used. The IMF component with the largest energy proportion is selected to update the regression coefficients, and the results are smoothed by median filtering with a window length of 5.

[0161] The data fusion submodule integrates an extended Kalman filter, and the state vector includes stress and compliance. The update formula is:

[0162] ;

[0163] ;

[0164] The covariance matrix of process noise w and measurement noise v is initially set as a diagonal matrix, and its diagonal elements are preferably 0.1.

[0165] After fusion, the output is continuous P art (t), and cardiac output was calculated using the pulse contour integral method.

[0166] The system is implemented as a finger-shaped device with a power consumption of less than 50mW. It supports Bluetooth 5.0 communication protocol and connects to mobile terminal applications to achieve real-time data display and cloud storage. The system was validated in 20 subjects, with an average error of less than 5mmHg, meeting the requirements of ISO81060-2 standard, and is suitable for home health monitoring and clinical auxiliary diagnostic scenarios.

[0167] In summary, this invention achieves high-precision blood pressure waveform estimation through sophisticated signal processing and model optimization, providing a reliable solution for non-invasive cardiovascular monitoring.

[0168] Those skilled in the art will understand that the modules in the device can be distributed within the device of the implementation scenario as described, or they can be located in one or more devices different from this implementation scenario, with corresponding changes. The modules of the above-mentioned implementation scenario can be combined into one module, or they can be further divided into multiple sub-modules.

[0169] The serial numbers of the present invention mentioned above are for descriptive purposes only and do not represent the superiority or inferiority of the implementation scenarios.

[0170] The above-disclosed examples are only a few specific implementation scenarios of the present invention. However, the present invention is not limited thereto, and any variations that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A blood pressure waveform estimation system based on pressurized PPG waveform, characterized in that, The system includes: The PPG signal acquisition module is used to acquire PPG signals under pressure. The pressure application module is used to apply and monitor external pressure to the fingertip to keep the blood vessels in the fingertip at a preset working point and to output the corresponding external pressure signal. The signal processing module is used to perform dynamic analysis and model building on the PPG signal and the external pressure signal, thereby estimating the continuous blood pressure waveform. The data output module is used to output the continuous blood pressure waveform and the blood pressure parameters calculated based on the waveform.

2. The blood pressure waveform estimation system based on pressurized PPG waveform as described in claim 1, characterized in that, The PPG signal acquisition module specifically includes: A multi-wavelength light source unit is used to alternately illuminate the fingertip with red and infrared LED sensors to provide a multi-wavelength light source; The photoelectric detection unit is used to measure the light intensity changes of a multi-wavelength light source through a photoelectric detector, and generate a multi-wavelength PPG signal based on the light intensity changes; The signal fusion filtering unit is used to adaptively filter out noise and motion artifacts in the multi-wavelength PPG signal by passing it through a filter and using the least mean square algorithm to obtain the pressurized PPG signal.

3. The blood pressure waveform estimation system based on pressurized PPG waveform as described in claim 2, characterized in that, The specific formula for updating the weights of the filter is as follows: ； in, For filter weights, Step size factor For error signals, This is the input signal.

4. The blood pressure waveform estimation system based on pressurized PPG waveform as described in claim 1, characterized in that, The pressure application module specifically includes: A finger-type pressure actuator for applying the external pressure to the fingertip; A closed-loop controller is used to lock the external pressure on the fingertip based on vascular compliance and ensure that the fingertip blood vessels are at a preset operating point.

5. The blood pressure waveform estimation system based on pressurized PPG waveform as described in claim 1, characterized in that, Vascular compliance is specifically determined by the following formula: ； Wherein, C represents vascular compliance, ΔV represents the change in blood volume in the fingertip vessels, and ΔP represents the change in external pressure; the vascular compliance is used to guide the pressure application module to lock the fingertip vessels at the point of maximum compliance as the preset working point, and to provide physiological prior constraints for the subsequent joint model.

6. The blood pressure waveform estimation system based on pressurized PPG waveform as described in claim 5, characterized in that, The signal processing module specifically includes: A signal preprocessing unit is used to preprocess the pressurized PPG signal and the pressure signal to obtain a preprocessed signal; The feature extraction unit is used to divide the preprocessed signal into cardiac cycles based on the cardiac cycle segmentation algorithm and extract time-domain features from each cardiac cycle. The time-domain features include pulse peak amplitude, rising slope, systolic area integral, time and amplitude of diabetic notch, and pressure-amplitude curve features. The joint model unit is used to construct a joint model that includes a parameterized sub-model and a learning sub-model, and to estimate a continuous blood pressure waveform based on the time-domain features, the pressurized PPG signal, and the external pressure signal. The parameter calibration unit is used to perform real-time calibration and slow calibration on the continuous blood pressure waveform and its model parameters. The real-time calibration is mainly for calibrating vascular compliance, physiological mapping parameters in the joint model and regression coefficients, while the slow calibration is used to compensate for long-term physiological drift. The data fusion unit is used to perform state fusion and smoothing on the continuous blood pressure waveform and related parameters of multiple heartbeat cycles using extended Kalman filtering, and output the final high-precision continuous blood pressure waveform.

7. The blood pressure waveform estimation system based on pressurized PPG waveform as described in claim 6, characterized in that, The joint model unit specifically includes: The parameterized sub-model, specifically determined through the arterial physiological model, is used to establish a priori mapping relationship between external pressure, intra-arterial pressure and PPG waveform based on vascular compliance, and the output of the parameterized sub-model is used as a priori constraint for the joint model. The learning sub-model is used to take the pressurized PPG signal, the external pressure signal and the time-domain features as inputs, output the instantaneous pressure residual, and obtain the continuous blood pressure waveform by fitting it with a linear regression model. The joint model combines the theoretically predicted waveform of the parameterized sub-model with the instantaneous pressure residual of the learned sub-model to obtain the final continuous blood pressure waveform.

8. The blood pressure waveform estimation system based on pressurized PPG waveform as described in claim 7, characterized in that, The signal preprocessing unit is specifically used for: The pressurized PPG signal PPG(t) and the external pressure signal Pactuator(t) are sampled and synchronized, bandpass filtered, baseline drift removed, and artifacts detected and eliminated. Specifically, a discrete wavelet transform with Daubechies 4 as the mother wavelet is used to simultaneously suppress noise and remove baseline drift. The threshold denoising formula is as follows: ； Where σ is the noise standard deviation and N is the signal length, in order to improve the waveform signal-to-noise ratio and preserve dynamic characteristics; Among them, artifact removal is performed based on acceleration sensor signals and PPG peak abrupt changes.

9. The blood pressure waveform estimation system based on pressurized PPG waveform as described in claim 7, characterized in that, The parameter calibration unit is specifically used for: The least squares method is used to fit multi-period waveform data, and the error parameters are adjusted using an error parameter adjustment formula. The empirical mode decomposition formula was used to slowly adjust the parameters of the continuous blood pressure waveform; The error adjustment formula is specifically as follows: ； in, This is an estimated value. For reference only. The number of samples; The empirical mode decomposition formula is as follows: ； in For modal functions, It represents the residual.

10. The blood pressure waveform estimation system based on pressurized PPG waveform as described in claim 1, characterized in that, The data output module specifically includes: The wireless communication unit is used to transmit the continuous blood pressure waveform and the blood pressure parameters calculated from the waveform to a mobile device or the cloud in real time via wireless communication. A visualization display unit is used to display the continuous blood pressure waveform and blood pressure parameters on a display device; The storage and recalibration unit is used to trigger an automatic recalibration procedure when the signal quality score is lower than a preset threshold. This includes updating the joint model parameters using a stepwise pressure scan or a small-range pressure oscillation method, wherein the signal quality score is determined by the signal-to-noise ratio.

Images