Pulse wave envelope reconstruction method and device, storage medium and electronic equipment

Abstract

The application relates to the field of blood pressure measurement, and discloses a pulse wave envelope reconstruction method and device, a storage medium and an electronic device. Pulse wave signals of multiple heart cycles are collected in a holding stage of each pressure step of step-by-step pressure reduction; outlying signals of the same step are filtered out based on a statistical outlying criterion of multidimensional morphological time sequence characteristics to obtain an effective pulse wave set; waveforms in the set are subjected to similarity clustering to obtain dominant amplitudes; the dominant amplitudes of all effective steps are arranged in descending order of pressure values to form an envelope skeleton sequence; according to the spindle physiological law that the amplitude of a pulse wave first increases and then decreases with the cuff pressure, abnormal skeleton points that violate the trend are removed or corrected to obtain a verified skeleton sequence; and the verified skeleton sequence is subjected to interpolation smoothing fitting reconstruction to obtain a continuous and stable pulse wave envelope. The application can remove abnormal waveforms and ensure the physiological reasonableness of the envelope, and improve the blood pressure measurement accuracy.

Description

Technical Field

[0001] This application relates to the field of biological signal processing, and in particular to a method, apparatus, storage medium and electronic device for pulse wave envelope reconstruction. Background Technology

[0002] Non-invasive electronic blood pressure devices generally use the oscillometric method to detect blood pressure. This method collects pulse wave signals during deflation, constructs a pulse wave envelope based on the pulse wave amplitude corresponding to different pressures, and then calculates the blood pressure value by identifying characteristic points within the envelope. This measurement method can obtain relatively stable and accurate results in people with sinus rhythm.

[0003] For individuals with atrial fibrillation and premature beats, the heart rhythm is irregular, and the amplitude and interval of the pulse wave fluctuate significantly. While some existing blood pressure measurement methods use a stepwise reduction approach to acquire pulse waves, they do not perform reconstruction processing on the acquired pulse waves. The resulting pulse wave envelope is prone to distortion and does not conform to the "spindle-shaped" model of oscillometric blood pressure measurement, leading to large blood pressure recognition errors and poor measurement stability. Currently, there is a lack of targeted processing solutions to perform statistical outlier filtering, extract representative amplitudes, and reconstruct a high-quality pulse wave envelope from discrete pulse samples, thereby improving the insufficient accuracy of blood pressure measurement in atrial fibrillation scenarios. Summary of the Invention

[0004] Based on this, it is necessary to address the technical problems of pulse wave envelope distortion during blood pressure measurement in atrial fibrillation rhythm in existing technologies, which leads to large blood pressure measurement errors and poor stability. Therefore, a pulse wave envelope reconstruction method is proposed.

[0005] In a first aspect, a pulse wave envelope reconstruction method is provided, the method comprising: During the non-invasive blood pressure measurement process of gradually reducing blood pressure according to a preset pressure gradient, pulse wave signals of multiple consecutive cardiac cycles are collected sequentially for each pressure step during the pressure holding phase of that pressure step. Based on the statistical outlier criterion of multi-dimensional morphological-temporal features, outlier samples of all pulse wave signals collected within the same pressure step are filtered out based on statistical distribution to obtain the effective pulse wave set corresponding to the pressure step. Based on the waveform characteristics of each pulse wave in the effective pulse wave set, similarity clustering is performed. The cluster with the largest number of samples is determined as the dominant pulse wave set of the pressure step. The cluster center of the dominant pulse wave set in the clustering feature space is mapped to the amplitude dimension, and the amplitude dimension is used as the dominant amplitude of the pressure step. The dominant amplitudes corresponding to the pressure steps of all the collected effective pulse waves are arranged in descending order of pressure value to form an envelope skeleton sequence. The envelope skeleton sequence is subjected to trend consistency verification based on the prior law of physiological envelope in oscillometric blood pressure measurement. The prior law of physiological envelope is as follows: during step-low blood pressure reduction, the pulse wave amplitude changes with the cuff pressure from high to low, showing an overall spindle-shaped curve characteristic of first rising and then falling, with a unique main peak. The amplitude on the left side of the main peak shows an upward trend, and the amplitude on the right side of the main peak shows a downward trend. Abnormal skeleton points in the envelope skeleton sequence that violate the trend of spindle-shaped curve are removed or corrected to obtain the verified envelope skeleton sequence. The verified envelope skeleton sequence is subjected to interpolation, smoothing, and fitting reconstruction to obtain a continuous and stable pulse wave envelope for blood pressure identification.

[0006] Secondly, a pulse wave envelope reconstruction device is provided, the device comprising: The acquisition module is used to acquire pulse wave signals of multiple consecutive cardiac cycles during the pressure holding phase of each pressure step in the process of non-invasive blood pressure measurement by gradually reducing blood pressure according to a preset pressure gradient. The filtering module is used to filter out outlier samples from all pulse wave signals collected within the same pressure step based on statistical outlier criteria using multi-dimensional morphological-temporal features, thereby obtaining the effective pulse wave set corresponding to that pressure step. The determination module is used to perform similarity clustering based on the waveform features of each pulse wave in the effective pulse wave set, determine the cluster with the largest number of samples as the dominant pulse wave set of the pressure step, and map the cluster center of the dominant pulse wave set in the clustering feature space to the amplitude dimension, and use the amplitude dimension as the dominant amplitude of the pressure step. The sorting module is used to arrange the dominant amplitudes corresponding to the pressure steps of all the acquired effective pulse waves in descending order of pressure value to form an envelope skeleton sequence. The verification module is used to verify the trend consistency of the envelope skeleton sequence based on the prior law of physiological envelope in oscillometric blood pressure measurement. The prior law of physiological envelope is: during step-low blood pressure reduction, the pulse wave amplitude changes with the cuff pressure from high to low, showing an overall spindle-shaped curve characteristic of first rising and then falling, with a unique main peak. The amplitude on the left side of the main peak shows an upward trend, and the amplitude on the right side of the main peak shows a downward trend. Abnormal skeleton points in the envelope skeleton sequence that violate the trend of spindle-shaped curve are removed or corrected to obtain the verified envelope skeleton sequence. The reconstruction module is used to perform interpolation, smoothing, and fitting reconstruction processing on the verified envelope skeleton sequence to obtain a continuous and stable pulse wave envelope for blood pressure identification.

[0007] Thirdly, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described pulse wave envelope reconstruction method.

[0008] Fourthly, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described pulse wave envelope reconstruction method.

[0009] The beneficial effects of this application are: This application acquires pulse wave signals from multiple consecutive cardiac cycles within the same pressure gradient and filters out outliers using statistical outlier criteria based on multi-dimensional morphological and temporal characteristics. This effectively identifies and eliminates abnormal pulse waves caused by motion artifacts, arrhythmias, or sensor noise, avoiding interference from abnormal waveforms to subsequent envelope reconstruction and improving the quality of the original data.

[0010] This application performs similarity clustering on the effective pulse wave set and maps the cluster center of the cluster with the largest number of samples in the cluster feature space to the amplitude dimension as the dominant amplitude of the pressure step. Compared with the traditional method of directly taking the mean or median amplitude, it can more accurately extract the most representative pulse wave amplitude under the pressure step, effectively reduce the deviation of atypical waveforms in the estimation of pressure step amplitude, and enhance the accuracy of envelope reconstruction.

[0011] This application utilizes the inherent physiological law that the pulse wave amplitude changes in a spindle-shaped curve with cuff pressure in oscillometric blood pressure measurement to perform trend consistency verification on the envelope skeleton sequence. Abnormal skeleton points that violate the single-peak change trend of first rising and then falling are removed or corrected. This verification step ensures the quality of the skeleton sequence from a physiological rationality perspective, ensuring that all data points participating in the final fitting and reconstruction conform to the basic physiological characteristics of oscillometric blood pressure measurement, thus guaranteeing the physiological reliability of the envelope morphology from the source.

[0012] This application organically combines outlier filtering, cluster center mapping to determine dominant amplitude, physiological envelope trend consistency verification, and interpolation smoothing fitting reconstruction to form a three-level quality control system from the signal layer to the feature layer and then to the physiological law layer. These steps work synergistically and progressively, ensuring that the reconstructed pulse wave envelope maintains high stability and accuracy even under strong interference conditions such as atrial fibrillation and limb movement. This improves the success rate and accuracy of oscillometric blood pressure measurement in complex clinical scenarios. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the 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.

[0014] in: Figure 1 This is a flowchart of a pulse wave envelope reconstruction method in one embodiment; Figure 2 This is a schematic diagram of multiple continuous cardiac cycle pulse waves at each step in one embodiment; Figure 3 This is a schematic diagram of the extraction of the dominant pulse wave within the stepped pressure point in one embodiment; Figure 4 This is a schematic diagram of the original rough envelope of single beat amplitude and the smooth reconstruction of the envelope in one embodiment; Figure 5 This is a schematic diagram of pressure interpolation reconstruction and blood pressure identification feature points in one embodiment; Figure 6 This is a structural block diagram of a pulse wave envelope reconstruction device in one embodiment; Figure 7 This is a schematic diagram of a computer storage medium in one embodiment; Figure 8 This is a structural block diagram of an electronic device in one embodiment. Detailed Implementation

[0015] 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, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0016] Please see Figure 1 As shown, Figure 1 A schematic flowchart of the pulse wave envelope reconstruction method provided in this application embodiment includes the following steps: S1. During the stepwise blood pressure reduction process of non-invasive blood pressure measurement, pulse wave signals of multiple consecutive cardiac cycles are collected sequentially for each pressure step during the pressure holding phase of that pressure step.

[0017] The cuff is controlled to perform a step-down blood pressure reduction operation, acquiring pulse wave signals during the pressure holding phase of each pressure step. Step-down blood pressure reduction refers to the cuff pressure being reduced step by step at a fixed increment, maintaining a stable pressure after each reduction, and starting signal acquisition once the cuff pressure has stabilized. Multiple complete cardiac cycles of pulse wave signals are continuously acquired during each pressure holding phase, and four core characteristic parameters of each pulse wave are extracted for subsequent processing. Optionally, during step-down blood pressure reduction, the pressure increment of each pressure step is 8 mmHg; pulse wave signals of 3–8 consecutive cardiac cycles are acquired during the pressure holding phase of each pressure step. The four core characteristic parameters are pulse wave amplitude, pulse wave interval, pulse wave area, and rising slope. The formula for calculating pulse wave amplitude is Amp = P. peak -P valley , where P peak P represents the peak value of the pulse wave signal. valley This represents the trough of the pulse wave signal. The formula for calculating the pulse wave interval is T = t0. n+1,peak -t n,peak , where t n,peak This represents the peak time of the nth pulse wave. The pulse wave area is the integral value of a single pulse wave from the start of the rising edge to the end of the falling edge, calculated using the trapezoidal integration method, with the integration step size consistent with the system sampling period. The rising edge slope is the linear fitting slope of the pulse wave rising edge segment from the 10% peak value to the 90% peak value, calculated using the formula k = (0.9P) / (0.9P). peak -0.1P peak ) / (t 90% -t 10% ), where t 10% The moment when the rising edge reaches a 10% peak value, t 90% This is the moment when the rising edge reaches 90% of its peak value.

[0018] For example, the control cuff is initially inflated to 160 mmHg. After confirming sufficient initial pressure, a step-down pressure reduction process is initiated, with each pressure step differing by 8 mmHg. Once the cuff pressure drops to 120 mmHg and remains stable, pulse wave signals for five complete cardiac cycles are continuously acquired. For each pressure step, the amplitude, interval, area, and rise slope of the acquired pulse wave are calculated, and these characteristic parameters are stored along with the corresponding pressure value. Subsequently, the control cuff is further depressurized to 112 mmHg, and the above acquisition and feature extraction operations are repeated until the cuff pressure drops to a level where no obvious pulse wave signal can be detected.

[0019] See Figure 2 , Figure 2 It is a schematic diagram of multiple continuous cardiac cycles of pulse waves at each step.

[0020] Figure 2The horizontal axis represents the measurement time in milliseconds, and the vertical axis represents the pressure value in millimeters of mercury. The graph contains two curves: the blue broken line is the stepped pressure curve, and the orange oscillating waveform is the continuous cardiac cycle pulse wave curve.

[0021] In the initial measurement phase, the cuff is rapidly inflated, causing the step pressure curve to rise quickly. During this pressure rise, before effective arterial pulsation is established, the corresponding pulse wave curve remains at the baseline without significant fluctuations or oscillations. Once the overall cuff pressure drops to a level where arterial pulsation can be transmitted, the pulse wave curve begins to exhibit periodic up-and-down oscillations, indicating that a valid cardiac cycle pulse wave signal has been acquired.

[0022] During subsequent measurements, the cuff pressure was gradually decreased according to a fixed pressure step, resulting in a step-like decrease in the corresponding pressure curve. After each pressure reduction, the cuff pressure was maintained at the current pressure step, forming a horizontal stable pressure range. Within each stable pressure range, pulse waves corresponding to multiple consecutive complete cardiac cycles were continuously collected. The corresponding pulse wave curve showed multiple sets of independent oscillating waveforms densely arranged within a single pressure step period, visually demonstrating the implementation of simultaneous multi-pulse acquisition under a single pressure step.

[0023] The waveform morphology reveals irregular heart rhythms in the tested subjects. Significant variations in peak amplitude and beat intervals exist among different pulse waves within the same pressure step. These irregular beats represent typical statistically discrete samples under atrial fibrillation rhythms and are the targeted samples for non-characteristic sample filtering, cluster extraction of dominant amplitudes, and envelope reconstruction in this approach. As the cuff pressure gradually decreases, the overall pulse wave oscillation amplitude first gradually increases and then slowly decreases, fully demonstrating the overall trend of pulse wave amplitude changing with cuff pressure. This provides complete raw sampling data support for subsequent extraction of dominant amplitudes at each step, construction of envelope skeleton sequences, and fitting to generate continuous and stable pulse wave envelopes.

[0024] S2. Based on the statistical outlier criterion of multi-dimensional morphological-temporal characteristics, non-featured sample sequential filtering is performed on all pulse wave signals collected within the same pressure step to obtain the effective pulse wave set corresponding to the pressure step.

[0025] After acquiring and storing all pulse wave signals and corresponding characteristic parameters for a single pressure step, the statistical outlier sample cleaning process is initiated. Given the significant physiological random variability in the amplitude and interval of each beat under atrial fibrillation rhythm, this step does not consider beats deviating from the mean as "pathological abnormalities," but rather defines them as non-dominant characteristic samples that are not representative of the current pressure step. The overall cleaning process adopts a hierarchical sequential filtering architecture: First layer: Morphological domain primary screening (based on amplitude and area integrals).

[0026] Amplitude Robustness Outlier Criterion: Calculate the median and median absolute deviation (MAD) of all pulse wave amplitudes within the current step. Use MAD instead of standard deviation as the dispersion estimate to enhance the robustness against extreme values. Set an upper threshold of Median + 3 * MAD and a lower threshold of Median - 3 * MAD (equivalent to ±130% and ±50% of the mean after conversion). Pulses exceeding these ranges are considered statistical amplitude outliers and are filtered out.

[0027] Low-value cutoff for area energy: Calculate the mean of the area integral and set a lower limit of 50% of the mean. Pulses below this threshold usually correspond to weak conduction or fused waveforms, and their energy characteristics are insufficient to characterize the true hemodynamic response under this pressure, so they are excluded.

[0028] Second layer: Temporal-morphological domain joint secondary screening (based on interval and rising slope).

[0029] Outlier criterion for interval variability: Utilizing the random distribution characteristics of RR intervals in atrial fibrillation, the absolute deviation of the interval mean is calculated. Samples with a deviation exceeding ±30% of the mean are considered outliers in the time series characteristics. This threshold is selected based on the Poisson distribution model to ensure that only extreme interval samples falling within the long-tailed distribution range are excluded.

[0030] Outlier criterion for slope morphology: Extract the fitted slope from the 10% to 90% segment of the rising edge, calculate the mean slope within this step, and filter out samples with deviations exceeding the mean ±40%. This step aims to exclude morphologically distorted samples caused by sudden changes in ventricular filling time, resulting in excessively steep or gentle rising edges.

[0031] Robustness assurance mechanisms for the cleaning process: To ensure that effective samples are not excessively sacrificed in highly discrete atrial fibrillation signals, each screening layer is subject to hard constraints: the filtering ratio for a single dimension cannot exceed 15% of the initial total number of samples in the current step, and the number of remaining samples after filtering across all dimensions must not be less than 3. If a dimension triggers the upper limit of the quantity constraint, the system automatically switches the criterion for that dimension to a mild outlier identification based on the interquartile range (IQR) and skips that round of filtering, maintaining the integrity of the effective set with a minimum sample priority strategy. The final set of remaining samples is defined as the effective pulse wave set for that pressure step, used for subsequent cluster analysis, thereby mathematically separating the interference of atrial fibrillation random variability on feature extraction, rather than a clinically "abnormal" diagnosis.

[0032] S3. Based on the waveform characteristics of each pulse wave in the effective pulse wave set, similarity clustering is performed. The cluster with the largest number of samples is determined as the dominant pulse wave set of the pressure step. The cluster center of the dominant pulse wave set in the cluster feature space is mapped to the amplitude dimension, and the amplitude dimension is used as the dominant amplitude of the pressure step.

[0033] Similarity clustering is performed on each pulse wave in the effective pulse wave set obtained in step S2. The input feature for clustering is the waveform feature vector of each pulse wave, which contains at least two dimensions: peak-to-peak amplitude and waveform area integral. The processor uses a preset clustering algorithm, such as K-means clustering or hierarchical clustering, to divide each pulse wave in the effective pulse wave set into several clusters based on the similarity of their waveform features. Pulse waves within the same cluster are similar in waveform features, while pulse waves in different clusters show significant differences in waveform features. After clustering, the processor counts the number of samples in each cluster and determines the cluster with the largest number of samples as the dominant pulse wave set for that pressure step. This dominant pulse wave set represents the most representative pulse wave morphology under that pressure step. Subsequently, the processor maps the cluster centers of all pulse waves in the dominant pulse wave set back to the amplitude dimension in the clustering feature space, using this amplitude dimension as the dominant amplitude of that pressure step. The cluster center is the geometric center of the cluster in the feature space, and its coordinates in each dimension are the mean of all samples in that cluster along that dimension. Extracting the component corresponding to the amplitude dimension from the coordinates of the cluster center yields the dominant amplitude of the pressure step. This dominant amplitude serves as the sole representative amplitude of the pressure step and is used for subsequent envelope skeleton sequence construction.

[0034] See Figure 3 As shown, Figure 3 This is a schematic diagram of the extraction of the dominant pulse wave within the stepped pressure point.

[0035] The horizontal axis of this diagram represents measurement time in milliseconds, and the vertical axis represents pressure values ​​in millimeters of mercury. There are three curves in the diagram: the blue curve is the stepped pressure curve, the orange oscillating waveform is the continuous cardiac cycle pulse wave curve obtained from the acquisition, and the green waveform is the dominant pulse wave extracted after filtering out non-featured samples and similarity clustering.

[0036] In the initial measurement phase, the cuff is inflated and pressurized, causing a rapid rise in the step-pressure curve. During this phase, the cuff pressure is high, and arterial pulsations cannot be effectively transmitted to the cuff. The pulse wave curve remains close to the baseline, without forming a valid pulsation waveform. Once the overall pressure drops to a range where arterial pulsations can be transmitted, the pulse wave curve begins to exhibit a continuous oscillating pattern, subsequently entering a step-pressure reduction mode. The step-pressure curve decreases gradually, and after each reduction, the current pressure is stabilized. Multiple sets of raw pulse waves corresponding to consecutive cardiac cycles are collected within a single pressure stabilization range.

[0037] The orange raw pulse wave clearly shows that the subject has atrial fibrillation. Within the same pressure step, there are obvious discrete fluctuations in the peak amplitude and beat interval of different pulse waves. Some beats are statistically atypical beats caused by rhythm disorder. If all raw pulse amplitude values ​​are used directly, it will cause distortion of the subsequent envelope trend.

[0038] Non-characteristic samples were filtered out from all raw pulse waves within a single pressure step to remove outliers with significant statistical deviations in morphology and rhythm. Then, similarity clustering was performed based on pulse wave amplitude and area characteristics. The number of samples within each cluster was counted, and the cluster with the largest number of samples was identified as the dominant pulse wave set corresponding to that step. The extracted results are represented by the green dominant pulse waves in the figure. A representative set of dominant pulse waves was extracted for each pressure step to determine the dominant amplitude corresponding to that step. This process filters out the chaotic non-characteristic pulsation interference from atrial fibrillation and accurately represents the true average pulsation level of the pulse wave under the current pressure. The dominant amplitudes of all steps were summarized to construct an envelope skeleton sequence, providing regular and reliable basic data for subsequent smooth reconstruction of the pulse wave envelope.

[0039] S4. Arrange the dominant amplitudes corresponding to the pressure steps of all the collected effective pulse waves in descending order of pressure value to form an envelope skeleton sequence.

[0040] The dominant amplitudes corresponding to all pressure steps of valid pulse waves are collected and arranged in descending order of pressure value to form an envelope skeleton sequence. Pressure steps for which no valid pulse waves were obtained after filtering out non-characteristic samples are excluded from the envelope skeleton sequence. The envelope skeleton sequence consists of a series of discrete pressure-amplitude data pairs, each corresponding to a pressure step and its dominant amplitude. The envelope skeleton sequence reflects the basic trend of pulse wave amplitude variation with cuff pressure and forms the basis for subsequent envelope reconstruction.

[0041] For example, during the step-down process, there were six pressure steps: 120 mmHg, 112 mmHg, 104 mmHg, 96 mmHg, 88 mmHg, and 80 mmHg. The 96 mmHg pressure step, after filtering out non-characteristic samples, did not yield sufficient effective pulse waves. The other five pressure steps all yielded effective pulse waves, with dominant amplitudes of 12 mmHg, 19 mmHg, 25 mmHg, 21 mmHg, and 14 mmHg, respectively. Arranging these five pressure-amplitude pairs in descending order of pressure as (120 mmHg, 12 mmHg), (112 mmHg, 19 mmHg), (104 mmHg, 25 mmHg), (88 mmHg, 21 mmHg), and (80 mmHg, 14 mmHg), this sequence is the envelope skeleton sequence.

[0042] See Figure 4 As shown, Figure 4 This is a schematic diagram of the original single beat amplitude rough envelope and the envelope smooth reconstruction.

[0043] Figure 4 The horizontal axis represents the measurement time in milliseconds, and the vertical axis represents the pressure value in millimeters of mercury. The figure shows three curves: the blue stepped line is the stepped pressure curve corresponding to the entire measurement; the red line with dots is the rough envelope directly constructed using the original single-stroke amplitude; and the blue smooth curve with dots is the reconstructed envelope obtained after interpolation and weighted smoothing in this scheme.

[0044] The stepped pressure curve fully reproduces the pressure change trend throughout the measurement process. Initially, the cuff rapidly inflates to the preset starting pressure, then gradually decreases pressure according to a fixed differential, stabilizing and collecting data at each pressure level, resulting in an overall stepped decrease. In atrial fibrillation, the amplitude of each pulse wave within a single pressure step is highly dispersed. If any single pulse wave amplitude is directly selected and connected to the corresponding pressure step, a rough envelope of the original single-beat amplitude is formed in the figure. This curve has abrupt inflection points, significant local distortion, and a shift in peak position, resulting in a chaotic overall trend. Directly locating blood pressure characteristic points based on this rough envelope easily leads to excessive discrepancies in the identification of systolic, diastolic, and mean blood pressure, failing to meet the accuracy requirements for blood pressure measurement in atrial fibrillation patients.

[0045] This method first extracts the dominant amplitude of each pressure step after clustering and filtering, and arranges them from high to low pressure to form a discrete envelope skeleton sequence. Then, linear interpolation is performed on the envelope skeleton sequence to supplement intermediate data points. After that, a five-point weighted smoothing filter operation is used to eliminate random fluctuations in the curve, and finally a smooth interpolation reconstructed envelope is generated. Comparing the two envelope curves, it can be seen intuitively that the smooth interpolation reconstructed envelope eliminates the abnormal bulges and steep inflection points caused by the original single-beat amplitude. The curve transition is smooth and continuous, and the true trend of pulse amplitude change with cuff pressure is completely restored. The peak position is accurately located, which can provide a reliable curve basis for subsequent accurate location of characteristic pressure points corresponding to mean pressure, systolic pressure, and diastolic pressure. It can improve the problem of blood pressure measurement error caused by atrial fibrillation pulse wave envelope distortion.

[0046] S5. Perform trend consistency verification on the envelope skeleton sequence based on the prior law of physiological envelope in oscillometric blood pressure measurement. The prior law of physiological envelope is: during step-lowering blood pressure, the pulse wave amplitude changes with the cuff pressure from high to low, and the overall curve is characterized by a spindle shape that first rises and then falls, with a unique main peak. The amplitude on the left side of the main peak shows an upward trend, and the amplitude on the right side of the main peak shows a downward trend. Abnormal skeleton points in the envelope skeleton sequence that violate the trend of spindle shape are removed or corrected to obtain the verified envelope skeleton sequence.

[0047] After forming the envelope skeleton sequence, the processor performs a trend consistency check on the sequence based on the physiological envelope prior law. According to the basic principle of oscillometric blood pressure measurement, as the cuff pressure gradually decreases from a high value higher than the systolic pressure to a low value lower than the diastolic pressure, the pulse wave amplitude exhibits a pattern of first gradually increasing and then gradually decreasing, with an overall shape of a spindle-shaped curve that first rises and then falls. This physiological law originates from the mechanical response characteristics of arterial blood vessels under different applied pressures: when the cuff pressure is much higher than the systolic pressure, the artery is completely occluded, and the pulse wave amplitude is minimal or zero; as the cuff pressure decreases to near the systolic pressure, the artery begins to periodically open, and the pulse wave amplitude gradually increases; when the cuff pressure approaches the mean arterial pressure, the arterial wall is in a state of maximum compliance, and the pulse wave amplitude reaches its peak; as the cuff pressure continues to decrease to below the diastolic pressure, the artery remains open throughout the entire cardiac cycle, and the pulse wave amplitude gradually decreases. This spindle-shaped curve characteristic is an inherent physiological law of oscillometric blood pressure measurement and is applicable to all subjects. The processor uses this prior rule to validate the envelope skeleton sequence: it checks whether each data point in the skeleton sequence satisfies the trend requirement of having a unique main peak with an increasing amplitude to the left of the main peak and a decreasing amplitude to the right of the main peak. For abnormal skeleton points that violate this increasing or decreasing trend, such as a decrease in amplitude to the left of the main peak or an increase in amplitude to the right of the main peak, the processor removes the abnormal skeleton point or corrects it by interpolation based on the values ​​of adjacent data points. The skeleton sequence after validation eliminates abnormal data points that are physiologically impossible to exist, thereby improving the data quality of subsequent fitting and reconstruction.

[0048] In one example, the processor performs trend consistency checks on a skeleton sequence consisting of 18 data points. In this sequence, starting at 0.625 volts at 174 mmHg, the amplitude gradually increases to a peak of 1.200 volts at 120 mmHg, then gradually decreases to 0.250 volts at 72 mmHg. The overall shape conforms to a spindle-shaped curve. However, the processor detects that after 1.180 volts at 114 mmHg, the data point at 108 mmHg is missing. While the 0.950 volt point at 102 mmHg continues the downward trend, the decrease is too large compared to the preceding and following data points. The processor determines that this data point may be subject to abnormal interference, removes it from the skeleton sequence, and corrects the 102 mmHg position using interpolated data points at 0.800 volts at 96 mmHg and 108 mmHg. Simultaneously, the processor detects missing data at 162 mmHg and 102 mmHg. After verification and correction, the processor obtained a complete verified envelope skeleton sequence that satisfies the spindle curve characteristic trend, containing 20 data points and covering the entire pressure step.

[0049] S6. Perform smooth fitting and reconstruction on the verified envelope skeleton sequence to obtain a continuous and stable pulse wave envelope for blood pressure identification.

[0050] After reconstructing the continuous and stable pulse wave envelope, the three types of blood pressure values ​​are identified and calculated based on the complete and smooth pressure amplitude correspondence curve. Mean pressure corresponds to the cuff pressure value at the peak of the envelope amplitude; this point represents the highest pulse wave intensity and is the core feature point for determining mean pressure using the oscillometric method. Systolic pressure corresponds to the pressure value at a preset percentage of the peak amplitude during the rising phase of the envelope curve; this position indicates that the cuff pressure has just decreased to a critical level sufficient to transmit arterial pulsations. Diastolic pressure corresponds to the pressure value at another preset percentage of the peak amplitude during the falling phase of the envelope curve; this position indicates that the cuff pressure is too low, and the transmission effect of arterial pulsations is significantly weakened. Traversing the reconstructed complete envelope curve along the pressure axis, the pressure values ​​corresponding to the three feature points are located and output as mean pressure, systolic pressure, and diastolic pressure, respectively. The reconstructed envelope curve has no discrete abrupt changes or irregular spikes, and the location of feature points is not affected by random pulsations caused by atrial fibrillation, resulting in stable value readings.

[0051] For example, after reconstructing a complete, continuous, and stable pulse wave envelope, the peak amplitude of the curve corresponds to a pressure value of 98 mmHg, which is directly determined as the mean pressure. The device's preset systolic pressure determination ratio is 50% of the peak amplitude. The point where the amplitude equals half the peak value is found along the rising segment of the envelope, corresponding to a pressure value of 132 mmHg, which is the systolic pressure. The device's preset diastolic pressure determination ratio is 70% of the peak amplitude. The point where the amplitude equals 70% of the peak value is found along the falling segment of the envelope, corresponding to a pressure value of 76 mmHg, which is the diastolic pressure. These three sets of values ​​are then processed to complete the blood pressure result output.

[0052] See Figure 5 , Figure 5 This is a schematic diagram of pressure interpolation reconstruction and blood pressure identification feature points.

[0053] Figure 5 The horizontal axis represents measurement time in milliseconds, and the vertical axis represents pressure values ​​in millimeters of mercury. The graph contains four curves and three marked feature points. The blue stepped line represents the stepped pressure curve throughout the measurement process, the orange diagonal line represents the original layout trend line of the envelope skeleton, the blue curve with dots represents the smoothed interpolation reconstructed envelope after linear interpolation and weighted smoothing, the pink pentagram marks the systolic pressure feature point, the red pentagram marks the mean pressure feature point, and the green pentagram marks the diastolic pressure feature point.

[0054] The stepped pressure curve fully presents the entire pressure change process of this blood pressure measurement. In the early stage of measurement, the cuff was rapidly inflated to raise the pressure. After reaching an initial pressure higher than the body's systolic blood pressure, the cuff was gradually depressurized. After each depressurization, the current pressure was maintained to remain stable, forming an overall stepped descent pattern. The orange line visually represents the overall trend of the discrete envelope framework formed by the dominant amplitudes of each pressure step. This framework only contains discrete step node data and cannot directly and accurately locate the corresponding characteristic points of blood pressure.

[0055] Linear interpolation and five-point weighted smoothing filtering are performed on the envelope skeleton sequence to generate a smooth interpolated reconstructed envelope with a continuous and gentle trend. This curve completely restores the continuous change of pulse wave amplitude with cuff pressure. Based on this reconstructed envelope, three types of blood pressure characteristic points are located. The red pentagram indicates the pressure position corresponding to the peak value of the reconstructed envelope amplitude, and this position is identified as the mean pressure characteristic point. Along the rising segment of the reconstructed envelope pressure increase, the corresponding point whose amplitude equals a preset proportion of the peak amplitude is found. This point is marked with a pink pentagram and is identified as the systolic pressure characteristic point. Along the falling segment of the reconstructed envelope pressure decrease, the corresponding point whose amplitude equals another preset proportion of the peak amplitude is found. This point is marked with a green pentagram and is identified as the diastolic pressure characteristic point.

[0056] By relying on the smooth reconstruction envelope after distortion correction to locate blood pressure feature points, the problem of local abrupt changes in the envelope curve caused by the disordered fluctuations of the pulse in atrial fibrillation can be avoided, the misjudgment and deviation of feature points can be prevented, the accuracy of systolic blood pressure, diastolic blood pressure and mean blood pressure values ​​can be improved, and finally stable and reliable non-invasive blood pressure measurement results can be output.

[0057] In one possible embodiment, the non-feature sample filtering process in step S2 is performed in the following order: First, filter out outliers by amplitude and area, then filter out outliers by interval and slope. Furthermore, the number of pulse waves filtered out under each criterion of amplitude outlier, area outlier, interval outlier, and slope outlier does not exceed the preset upper limit of the total number of pulse waves initially collected within the pressure step, and the number of valid pulse waves remaining after removal is not less than the preset minimum valid number.

[0058] The non-feature sample filtering process follows a fixed sequence, consisting of two rounds of screening. The first round identifies and filters out amplitude and area outliers. The second round processes the pulse wave samples remaining from the first round, simultaneously identifying and filtering out interval and slope outliers. Priority is given to waveform distortion issues related to amplitude and area, as these outliers have significant defects in their complete pulse wave shape, causing greater interference to subsequent calculations. Early removal reduces the computational overhead of invalid data in later steps. Interval and slope outliers primarily reflect pulsation rhythm and local waveform changes, with relatively lower interference levels, and are processed in the later round. The two rounds of screening are executed independently; the remaining samples from the first round serve as input data for the second round, preventing reverse cross-judgment.

[0059] A uniform quantity constraint standard is set for each outlier filtering operation. For the four categories of outliers—amplitude, area, interval, and slope—the number of pulse waves removed under any single category must not exceed the preset upper limit corresponding to the total number of pulse waves initially collected for the current pressure ladder. A minimum effective sample control logic is simultaneously set. After all outlier filtering operations are completed, the total number of pulse waves remaining must reach the preset minimum effective number. If the remaining samples after filtering out all categories of outliers do not reach this value, the filtering operation for that category is stopped, and pulse waves with slight statistical deviations are retained for subsequent calculations. The preset upper limit is fixed at 15%, a value calibrated based on a large number of multi-cycle pulse acquisition trials of atrial fibrillation patients, to avoid excessive removal of samples from a single category. The preset minimum effective number is fixed at 3, which is the minimum sample size to ensure stable and reliable results for subsequent cluster analysis.

[0060] In one possible embodiment, step S3, which involves similarity clustering based on waveform features, specifically includes: S31. Similarity clustering is performed based on the amplitude and area characteristics of pulse waves, wherein the ratio of the amplitude difference between any two pulse waves within the same cluster to the average amplitude of the cluster does not exceed a preset cluster density ratio threshold.

[0061] After obtaining the set of all valid pulse waves corresponding to the current pressure step, the amplitude and area features of each valid pulse wave are extracted as the basis for clustering calculation, and all similarity division is completed using only these two types of features. First, the first valid pulse wave in the set is designated as the initial sample of the first cluster. Then, each remaining valid pulse wave is read sequentially, and the pulse wave to be judged is matched and verified against all existing clusters. During the verification process, the amplitude dimension is used alone to determine the density. First, the average amplitude of all pulse waves within the target cluster is calculated. Then, the difference between the amplitude of the pulse wave to be judged and the average amplitude of the cluster is calculated. This difference is divided by the average amplitude of the cluster to obtain the corresponding ratio. If this ratio is less than or equal to a preset cluster density ratio threshold, the pulse wave to be judged is assigned to the corresponding cluster. If the matching ratio of the pulse wave to be judged with all existing clusters exceeds the cluster density ratio threshold, a new independent cluster is created based on this pulse wave. The preset clustering tightness threshold is set to 15%. This value was determined through a comparative amplitude distribution test of the dominant beat and scattered outliers in atrial fibrillation patients, clearly distinguishing stable beats from discrete, non-characteristic beats within the same rhythm. Area features are used to help differentiate pulse waves with similar amplitudes but significantly different waveform integrity. If two pulse waves meet the tightness requirement but their areas differ too much, they will not be assigned to the same cluster. The entire clustering process is executed cyclically until all valid pulse waves within the set have been clustered.

[0062] For example, a set of effective pulse waves for a certain pressure gradient contains four effective pulse waves with amplitudes of 18 mmHg, 19 mmHg, 21 mmHg, and 25 mmHg, respectively. The cluster density threshold is set to 15%. The first cluster is established using the pulse wave with an amplitude of 18 mmHg, and the initial average amplitude of this cluster is 18 mmHg. The second pulse wave with an amplitude of 19 mmHg is used for verification; the amplitude difference is 1 mmHg, corresponding to a ratio of approximately 5.6%, which is less than 15%. Furthermore, the area characteristics of the two pulse waves are similar, so they are directly assigned to the first cluster, updating the average amplitude of this cluster to 18.5 mmHg. The third pulse wave with an amplitude of 21 mmHg is used for verification; the amplitude difference is 2.5 mmHg, corresponding to a ratio of approximately 13.5%, which does not exceed the 15% threshold. The area characteristics match, so it is assigned to the first cluster, updating the average amplitude of this cluster to 19.3 mmHg. The pulse wave with an amplitude of 25 mmHg differs from the average amplitude of the first cluster by 5.7 mmHg, corresponding to a ratio of approximately 29.5%, which exceeds the 15% cluster density threshold. Therefore, a second cluster is created to store this pulse wave. Ultimately, two independent clusters are formed: the first cluster contains 3 pulse waves, and the second cluster contains only 1 pulse wave.

[0063] This embodiment combines amplitude and area waveform features for clustering. By relying on the amplitude ratio to constrain the fluctuation range of samples within the cluster, it accurately distinguishes between stable and repetitive effective beats and scattered distorted beats within the same pressure step under atrial fibrillation. This provides a reliable classification basis for the subsequent screening of the dominant pulse wave set with the largest number of samples, reduces the interference of irregular heartbeats on the amplitude values ​​of the step representative, and improves the overall stability of the subsequent envelope skeleton sequence.

[0064] In one possible embodiment, the smooth fitting reconstruction process in step S5 specifically includes: S51. Using the dominant amplitude corresponding to each pressure step as the interpolation node, linear interpolation is performed using a preset pressure step size to obtain a discrete interpolation sequence in the pressure dimension.

[0065] After constructing the envelope skeleton sequence, each pressure value and its corresponding dominant amplitude are extracted as interpolation nodes, with a fixed pressure difference between adjacent interpolation nodes. A fixed value is selected as the preset pressure step size, and several new pressure and amplitude data are added between adjacent interpolation nodes using a linear interpolation algorithm. During the linear interpolation calculation, the slope of the amplitude change between two adjacent interpolation nodes is first calculated. Then, the pressure interval between the two nodes is evenly divided according to the preset pressure step size, and the amplitude value matching each newly added pressure point is calculated point by point based on the slope. All original interpolation nodes and newly added points obtained from the interpolation calculation are arranged uniformly in descending order of pressure, forming a discrete interpolation sequence that completely covers the entire pressure interval. The preset pressure step size is 2 mmHg, a value determined through multiple sets of stepped pressure reduction tests, which balances curve smoothness and computational cost.

[0066] S52. Perform weighted smoothing filtering on the discrete interpolation sequence with a preset window size to obtain a continuous and stable pulse wave envelope.

[0067] After obtaining the complete discrete interpolation sequence, weighted smoothing filtering is performed. The filtering uses a fixed-size sliding window to traverse all amplitude data points within the sequence. The sliding window simultaneously covers the current calculation point and two sets of data before and after that point. The total window length is 5, corresponding to a preset window size of 5 points. Fixed weight values ​​are assigned to the five points within the window, with weights of 1, 1, 2, 1, and 1 respectively. The current point at the center of the window is assigned the highest weight to ensure that the original effective amplitude data is not excessively weakened. The pressure-corresponding amplitude values ​​of all points within the window are sequentially taken, and a weighted sum is calculated based on the corresponding weights. This sum is then divided by the total weight value to obtain the smoothed amplitude value. The original discrete interpolation amplitudes are replaced according to the original pressure order of the sequence. After the filtering calculations for all points are completed, a continuous and stable pulse wave envelope is generated.

[0068] In one possible embodiment, an automatic heart rhythm determination step is included before step S1: S01. Acquire the initial pulse wave signal of 5 consecutive cardiac cycles before the start of stepwise blood pressure reduction.

[0069] After initial cuff inflation and confirmation of stable cuff pressure, initiate a stepped deflation process. When a pulse signal is detected for the first time within a pressure range and the pressure stabilizes, acquire initial pulse wave signals for consecutive cardiac cycles. The number of cardiac cycles acquired is controlled to be 5. During acquisition, the pulse interval corresponding to each pulse wave is extracted simultaneously, and all initial pulse wave interval data are completely stored for subsequent heart rhythm determination. Stepped deflation operations are paused during signal acquisition, and the cuff is maintained at its current stable pressure to prevent pressure fluctuations from compromising the integrity of the initial pulse wave signal. After acquisition, all initial pulse wave interval data are stored in a temporary storage area, awaiting subsequent computation.

[0070] S02. Calculate the coefficient of variation of the pulse interval based on the initial pulse wave continuous cardiac signal to assess the degree of irregularity of the pulse wave rhythm.

[0071] The pulse interval data of all initial pulse wave signals stored in step S01 are retrieved, and the coefficient of variation of the pulse interval is calculated based on multiple sets of interval values. The coefficient of variation of the pulse interval is used to quantify the degree of fluctuation of the pulse interval between different cardiac cycles; the larger the coefficient of variation, the higher the degree of irregularity of the pulse rhythm. First, the average value of all pulse intervals is calculated, then the standard deviation of multiple sets of interval values ​​is calculated, and the coefficient of variation of the pulse interval is obtained by dividing the standard deviation by the average value of the intervals. This coefficient can intuitively reflect the regularity of the current pulse rhythm of the subject and is the core criterion for distinguishing between atrial fibrillation rhythm and sinus rhythm.

[0072] S03. When the coefficient of variation of the pulse interval exceeds the preset atrial fibrillation threshold, the current heart rhythm is determined to be atrial fibrillation rhythm, and steps S1 to S5 are initiated.

[0073] The system retrieves the preset atrial fibrillation threshold from within the device and compares the coefficient of variation of pulse intervals calculated in step S02 with this threshold. If the calculated coefficient of variation of pulse intervals is greater than the preset atrial fibrillation threshold, the subject's current heart rhythm is determined to be atrial fibrillation. Under atrial fibrillation rhythms, the fluctuation in pulse wave amplitude and interval is significantly increased, making conventional oscillometric blood pressure identification insufficient. The system automatically executes the entire set of steps (S1 to S5) for dominant pulse extraction and envelope reconstruction under step pressure, processing irregular pulse wave signals using a dedicated algorithm. The preset atrial fibrillation threshold is calibrated through a large-scale comparative study of pulse interval samples from sinus rhythm and atrial fibrillation patients, stably distinguishing the rhythmic fluctuation characteristics of the two types of heart rhythms.

[0074] For example, with an internally preset atrial fibrillation threshold of 0.25, the calculated coefficient of variation for the pulse interval is 0.38, which is greater than 0.25. The system determines that the current subject has atrial fibrillation and automatically executes a complete processing flow, including step-down multi-cycle pulse acquisition, filtering of non-characteristic samples, clustering to extract dominant amplitudes, constructing the envelope framework, and smoothing the envelope reconstruction, without invoking conventional blood pressure measurement logic.

[0075] S04. When the coefficient of variation of the pulse interval does not exceed the preset atrial fibrillation threshold, the current heart rhythm is determined to be sinus rhythm, and the process is switched to the conventional oscillometric blood pressure measurement procedure.

[0076] After comparing the pulse interval coefficient of variation with the preset atrial fibrillation threshold, if the calculated coefficient of variation is less than or equal to the preset atrial fibrillation threshold, the subject's current heart rhythm is determined to be sinus rhythm. Sinus rhythm corresponds to small pulse interval fluctuations and stable pulse wave morphology, allowing for accurate blood pressure identification using conventional oscillometric methods. Stop initiating the envelope reconstruction steps from S1 to S5, and switch to the device's built-in conventional oscillometric blood pressure measurement process. Complete the blood pressure measurement using a common acquisition method of uniform deflation or conventional step-down, reducing device power consumption and measurement time caused by complex clustering and envelope reconstruction calculations.

[0077] This application can automatically distinguish the heart rhythm type of the person being measured before the formal blood pressure measurement. For atrial fibrillation, it automatically uses a high-precision envelope reconstruction algorithm to make up for the accuracy defects of conventional oscillometric methods. For sinus rhythm switching, it streamlines the conventional measurement process. While ensuring the accuracy of blood pressure measurement for people with atrial fibrillation, it shortens the measurement time for people with sinus rhythm and reduces the power consumption of the device.

[0078] In one possible embodiment, a multimodal triggering control step is included before step S1: A1. Receives atrial fibrillation detection signals or premature beat detection signals from the ECG detection module.

[0079] Before initiating the stepped blood pressure reduction acquisition process, the device continuously interacts with the ECG monitoring module, receiving real-time screening results for atrial fibrillation and premature beats from the ECG module. The ECG monitoring module identifies the presence of atrial fibrillation or premature beats based on surface ECG signals. If atrial fibrillation or premature beats are detected, a signal is sent to the blood pressure module. The system continuously monitors the signal transmission channel. If the ECG monitoring module outputs an atrial fibrillation or premature beat detection signal, the system fully stores the corresponding rhythm determination result and simultaneously acquires the rhythm determination result of the previously acquired initial pulse wave signal, forming a two-dimensional judgment basis for subsequent trigger condition verification. The system also retains the pulse interval variation coefficient calculation result as a pulse wave side judgment basis. The calculation method follows the aforementioned process: the average and standard deviation of multiple pulse interval values ​​are calculated, and then the standard deviation is divided by the average to obtain the pulse interval variation coefficient, which is used to determine whether atrial fibrillation or premature beats exist at the pulse level.

[0080] For example, the device is equipped with an ECG detection module, which performs initial cuff compression and collects initial pulse wave signals for five consecutive cardiac cycles, while simultaneously receiving data transmitted in real time from the ECG detection module. When the ECG detection module detects a continuous irregular ECG waveform in the subject and sends an atrial fibrillation detection signal, the device caches the ECG determination result and simultaneously reads the initial pulse wave signal to calculate the pulse interval variation coefficient, which is 0.37. Both sets of determination information are stored for subsequent trigger logic judgments.

[0081] A2. When any of the following triggering conditions are met, steps S1 to S5 will be initiated: (1) Single-mode triggering: Only the ECG detection module outputs atrial fibrillation detection signal or premature beat detection signal, or only the initial pulse wave signal is used to determine the presence of atrial fibrillation or premature beat; (2) Dual-modal verification trigger: The ECG detection module outputs an atrial fibrillation detection signal or a premature beat detection signal, and determines the presence of atrial fibrillation or premature beat based on the initial pulse wave signal.

[0082] The heart rhythm signal output from the ECG detection module is integrated with the initial pulse wave rhythm determination result. The judgment is completed by comparing it with two types of preset trigger conditions. Once either condition is met, the entire envelope reconstruction process from step S1 to S5 can be started.

[0083] The first type is single-modal triggering, which has two independent scenarios. In the first scenario, the determination relies solely on the ECG monitoring module. As long as the ECG monitoring module outputs an atrial fibrillation or premature beat detection signal, regardless of whether the pulse wave rhythm determination result is normal, the stepwise blood pressure reduction and envelope reconstruction algorithms are directly initiated. In the second scenario, the determination relies solely on the initial pulse wave. The ECG monitoring module does not output any non-sinus rhythm signals; only the calculated pulse interval variation coefficient exceeds the preset atrial fibrillation threshold, indicating the presence of atrial fibrillation or premature beats at the pulse level, and the entire processing procedure is initiated accordingly.

[0084] The second type is dual-modal verification triggering. This triggering mode sets dual verification standards for the judgment results. It requires the ECG detection module to output an atrial fibrillation detection signal or a premature beat detection signal, and at the same time, the coefficient of variation of the pulse interval calculated based on the initial pulse wave exceeds the preset atrial fibrillation threshold. Only when both judgment results point to non-sinus rhythm will steps S1 to S5 be initiated.

[0085] The two trigger modes can be pre-configured by the device according to the clinical use scenario. The judgment logic is executed according to the selected mode. If the corresponding trigger condition is not met, it will directly switch to the conventional oscillometric blood pressure measurement process.

[0086] For example, if the device is configured in single-modal trigger mode, in the first scenario, the ECG detection module pushes a premature beat detection signal. The calculated pulse interval variation coefficient is 0.21, which does not exceed the atrial fibrillation threshold of 0.25, thus still meeting the single-modal trigger condition. The entire process of stepwise blood pressure reduction, non-featured sample filtering, cluster extraction of dominant amplitude, envelope skeleton construction, and smooth envelope reconstruction is then directly initiated. In the second scenario, the ECG detection module has no abnormal signal output, and the calculated pulse interval variation coefficient is 0.32, exceeding the threshold. The entire algorithm is triggered solely based on the pulse wave determination result.

[0087] The device was reconfigured to dual-modal verification trigger mode. The ECG detection module output an atrial fibrillation detection signal, but the pulse interval variation coefficient of 0.22 did not reach the threshold. The two sets of judgment results could not meet the trigger criteria at the same time. The envelope reconstruction process was not started, and the conventional oscillometric method was directly run to complete the blood pressure measurement.

[0088] In one possible embodiment, the method further includes: B1. During the step-down process, if the number of effective pulse waves obtained after filtering out non-characteristic samples in the current pressure step does not reach the preset minimum effective number, the pressure holding time of the pressure step is extended to the preset minimum holding time.

[0089] After completing the non-featured sample filtering operation at each pressure step, the total number of valid pulse waves remaining at the current step is counted and compared with the preset minimum valid number. When the total number of valid pulse waves is less than the preset minimum valid number, it is determined that the usable samples collected at the current step are insufficient to support subsequent cluster analysis. The solenoid valve is then controlled to maintain the cuff pressure at the current step, extending the pressure holding time to the preset minimum holding time. During the extended pressure holding phase, new cardiac cycle pulse wave signals are continuously collected, and the non-featured sample filtering process is executed again to supplement valid samples. The preset minimum valid number is set to 3, which is the minimum sample size to ensure stable clustering results. The preset minimum holding time is set to 4 seconds, which is calculated based on the average human cardiac cycle and can likely collect a sufficient number of complete beats. If the number of valid pulse waves still cannot meet the target after extending the minimum holding time, the time is not extended further, and the existing small number of valid pulse waves are used directly in subsequent calculations.

[0090] B2. After initially pressurizing to the preset initial pressure, if a pulse wave signal is detected, it is determined that the cuff pressure is insufficient. The cuff pressure is automatically and gradually increased until no pulse wave signal is detected, and then the step-by-step depressurization process begins.

[0091] Before officially initiating the stepped blood pressure reduction process, the cuff is first pressurized to a preset initial pressure of 160 mmHg using the inflation pump. Once the pressure stabilizes, dynamic pressure signals within the cuff are continuously collected to extract the pulse wave component. The presence of a identifiable pulse wave signal at the current pressure is determined. If a clear pulse wave signal is detected, it indicates that the current cuff pressure is lower than the body's systolic blood pressure, and the cuff pressure is deemed insufficient. The cuff pressure is then gradually increased in small increments. After each pressurization, the pressure is allowed to stabilize before the pulse wave signal is detected again. This process of pressurization and signal detection is repeated until no pulse wave signal can be detected at the current cuff pressure. Once no pulse wave signal is confirmed, inflation is stopped, and the stepped blood pressure reduction process officially begins to collect pulse waves. This step-by-step pressurization method avoids the intense discomfort caused by a single large increase in pressure, while ensuring that the initial pressure is higher than the body's systolic blood pressure, thus fully covering the entire pulse wave variation range corresponding to blood pressure.

[0092] Please see Figure 6 As shown, in one embodiment, a pulse wave envelope reconstruction device is provided, the device comprising: The acquisition module 301 is used to acquire pulse wave signals of multiple consecutive cardiac cycles during the process of non-invasive blood pressure measurement by gradually reducing blood pressure according to a preset pressure gradient, for each pressure step, during the pressure holding phase of that pressure step. The filtering module 302 is used to filter out outlier samples of all pulse wave signals collected within the same pressure step based on statistical outlier criteria of multi-dimensional morphological-temporal features, so as to obtain the effective pulse wave set corresponding to the pressure step. The determination module 303 is used to perform similarity clustering based on the waveform characteristics of each pulse wave in the effective pulse wave set, determine the cluster with the largest number of samples as the dominant pulse wave set of the pressure step, and map the cluster center of the dominant pulse wave set in the clustering feature space to the amplitude dimension, and use the amplitude dimension as the dominant amplitude of the pressure step. The sorting module 304 is used to arrange the dominant amplitudes corresponding to the pressure steps of all the collected effective pulse waves in descending order of pressure value to form an envelope skeleton sequence. The verification module 305 is used to perform trend consistency verification on the envelope skeleton sequence based on the physiological envelope prior law of oscillometric blood pressure measurement. The physiological envelope prior law is as follows: during the step-down blood pressure process, the pulse wave amplitude changes with the cuff pressure from high to low, and the overall curve has the characteristics of first rising and then falling, with a unique main peak. The amplitude on the left side of the main peak shows an upward trend, and the amplitude on the right side of the main peak shows a downward trend. Abnormal skeleton points in the envelope skeleton sequence that violate the trend of the spindle curve are removed or corrected to obtain the verified envelope skeleton sequence. The reconstruction module 306 is used to perform interpolation, smoothing, and fitting reconstruction processing on the verified envelope skeleton sequence to obtain a continuous and stable pulse wave envelope for blood pressure identification.

[0093] For other details regarding the implementation of the above technical solution by each module in the above pulse wave envelope reconstruction device, please refer to the description in the pulse wave envelope reconstruction method provided in the above application embodiments, which will not be repeated here.

[0094] It should be noted that the device 3 provided in the above embodiments is only illustrated by the division of the above functional modules when performing the pulse wave envelope reconstruction method. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the above functions. In addition, the pulse wave envelope reconstruction device and the pulse wave envelope reconstruction method embodiments provided in the above embodiments belong to the same concept, and the implementation process is detailed in the method embodiments, which will not be repeated here.

[0095] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0096] like Figure 7 The diagram shown is a schematic of a computer storage medium provided in an embodiment of this application. The computer storage medium can store multiple instructions (i.e., ... Figure 7 The computer program shown above), the instructions are adapted to be loaded and executed by a processor as described above. Figure 1 The method steps of the illustrated embodiment can be found in the following documentation for detailed execution. Figure 1The specific details of the illustrated embodiments will not be elaborated here.

[0097] This application also provides a computer program product that stores at least one instruction, which is loaded and executed by a processor to implement the pulse wave envelope reconstruction method as described in the above embodiments.

[0098] Please see Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 8 As shown, the electronic device 500 may include: at least one processor 501, at least one network interface 504, user interface 503, memory 505, and at least one communication bus 502.

[0099] The communication bus 502 is used to enable communication between these components.

[0100] The user interface 503 may include input units such as a mouse and keyboard.

[0101] The network interface 504 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0102] The processor 501 may include one or more processing cores. The processor 501 connects to various parts within the electronic device 500 using various interfaces and lines, and performs various functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 505, and by calling data stored in the memory 505. Optionally, the processor 501 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 501 may integrate one or more of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content to be displayed on the screen; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 501 and may be implemented as a separate chip.

[0103] The memory 505 may include random access memory (RAM) or read-only memory. Optionally, the memory 505 may include a non-transitory computer-readable storage medium. The memory 505 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 505 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 505 may also be at least one storage device located remotely from the aforementioned processor 501. Figure 8 As shown, the memory 505, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and application programs.

[0104] exist Figure 8 In the illustrated electronic device 500, the user interface 503 is mainly used to provide an input interface for the user and to acquire user input data; while the processor 501 can be used to call the application program stored in the memory 505 and specifically execute, for example... Figure 1 The method shown can be referred to for details. Figure 1 As shown, it will not be elaborated further here.

[0105] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a disk, optical disk, read-only memory, or random access memory, etc.

[0106] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. A method of reconstructing a pulse wave envelope, characterized by, include: Acquire initial pulse wave signals from at least two consecutive cardiac cycles before the start of stepwise blood pressure reduction; calculate the pulse interval variation coefficient based on the initial pulse wave signals to assess the degree of pulse wave rhythm irregularity; when the pulse interval variation coefficient exceeds a preset atrial fibrillation threshold, determine that the current heart rhythm is atrial fibrillation rhythm and initiate the pulse wave envelope reconstruction method; when the pulse interval variation coefficient does not exceed the preset atrial fibrillation threshold, determine that the current heart rhythm is sinus rhythm and switch to the conventional oscillometric blood pressure measurement procedure; During the non-invasive blood pressure measurement process of gradually reducing blood pressure according to a preset pressure gradient, pulse wave signals of multiple consecutive cardiac cycles are collected sequentially for each pressure step during the pressure holding phase of that pressure step. Based on a statistical outlier criterion using multi-dimensional morphological-temporal characteristics, outlier samples are filtered out from all pulse wave signals collected within the same pressure step using a statistical distribution-based method to obtain the effective pulse wave set corresponding to that pressure step. The outlier filtering based on statistical distribution specifically includes one or more of the following removal criteria set based on robust statistics: Amplitude outlier detection: Pulse waves whose amplitude is higher than the first upper limit of the average pulse wave amplitude within the current pressure ladder, or lower than the first lower limit of the average pulse wave amplitude, are removed. The thresholds are determined based on quantile fitting of the amplitude distribution in the atrial fibrillation population. Area energy outlier detection: Pulse waves whose area integral is lower than the second proportional threshold of the mean area integral within the current pressure step are removed to filter out invalid samples with insufficient energy or dominated by noise. Outlier detection for inter-pulse variation: Pulse waves whose absolute difference between the pulse wave interval and the mean interval within the current pressure ladder exceeds the third proportional threshold, which is set with reference to the clinical statistical cutoff value of ventricular rhythm variation; Outlier detection based on rising slope: Pulse waves whose absolute difference between the rising slope and the mean slope within the current pressure step exceeds the fourth proportional threshold are removed to exclude samples with significant morphological distortion caused by local rhythm disturbances. The non-feature sample filtering process is performed in a hierarchical sequential filtering order: first, primary morphological domain screening is performed in the amplitude and area dimensions, followed by secondary time-morphological domain joint screening in the interval and slope dimensions. The number of samples filtered under each criterion does not exceed the preset proportional upper limit of the total number of pulse waves initially collected for that pressure step, and the number of remaining valid samples after filtering is not less than the preset minimum effective capacity to ensure the statistical significance of subsequent cluster analysis. Based on the waveform characteristics of each pulse wave in the effective pulse wave set, similarity clustering is performed. The cluster with the largest number of samples is determined as the dominant pulse wave set of the pressure step. The cluster center of the dominant pulse wave set in the clustering feature space is mapped to the amplitude dimension, and the amplitude dimension is used as the dominant amplitude of the pressure step. The dominant amplitudes corresponding to the pressure steps of all the collected effective pulse waves are arranged in descending order of pressure value to form an envelope skeleton sequence. The envelope skeleton sequence is subjected to trend consistency verification based on the prior law of physiological envelope in oscillometric blood pressure measurement. The prior law of physiological envelope is as follows: during step-low blood pressure reduction, the pulse wave amplitude changes with the cuff pressure from high to low, showing an overall spindle-shaped curve characteristic of first rising and then falling, with a unique main peak. The amplitude on the left side of the main peak shows an upward trend, and the amplitude on the right side of the main peak shows a downward trend. Abnormal skeleton points in the envelope skeleton sequence that violate the trend of spindle-shaped curve are removed or corrected to obtain the verified envelope skeleton sequence. The verified envelope skeleton sequence is subjected to interpolation, smoothing, and fitting reconstruction to obtain a continuous and stable pulse wave envelope for blood pressure identification.

2. The method of claim 1, wherein, The similarity clustering based on the waveform features of each pulse wave in the effective pulse wave set includes: Similarity clustering is performed based on the amplitude and area characteristics of pulse waves, wherein the ratio of the amplitude difference between any two pulse waves within the same cluster to the average amplitude of the cluster does not exceed a preset cluster density ratio threshold.

3. The method of claim 1, wherein, The process of interpolating, smoothing, and fitting the verified envelope skeleton sequence to obtain a continuous and stable pulse wave envelope for blood pressure identification includes: Using the dominant amplitude corresponding to each pressure step as the interpolation node, linear interpolation is performed using a preset pressure step size to obtain a discrete interpolation sequence in the pressure dimension. The discrete interpolation sequence is subjected to a weighted smoothing filter with a preset window size to obtain the continuous and stable pulse wave envelope.

4. The method of claim 1, wherein, The method further includes: During the step-down pressure reduction process, if the number of effective pulse waves obtained after filtering out non-characteristic samples within the current pressure step does not reach the preset minimum effective number, the pressure holding time of that pressure step is extended to the preset minimum holding time; and After initially pressurizing to the preset initial pressure, if a pulse wave signal is detected, it is determined that the cuff pressure is insufficient. The cuff pressure is automatically and gradually increased until no pulse wave signal is detected, and then the step-by-step depressurization process begins.

5. A pulse wave envelope reconstruction apparatus characterized by comprising: include: The evaluation module is used to acquire initial pulse wave signals from at least two consecutive cardiac cycles before the start of stepwise blood pressure reduction; calculate the pulse interval variation coefficient based on the initial pulse wave signals to assess the degree of pulse wave rhythm irregularity; when the pulse interval variation coefficient exceeds a preset atrial fibrillation threshold, the current heart rhythm is determined to be atrial fibrillation rhythm, and the pulse wave envelope reconstruction method is initiated; when the pulse interval variation coefficient does not exceed the preset atrial fibrillation threshold, the current heart rhythm is determined to be sinus rhythm, and the system switches to the conventional oscillometric blood pressure measurement procedure. The acquisition module is used to acquire pulse wave signals of multiple consecutive cardiac cycles during the pressure holding phase of each pressure step in the process of non-invasive blood pressure measurement by gradually reducing blood pressure according to a preset pressure gradient. The filtering module is used to filter out outliers from all pulse wave signals collected within the same pressure step based on statistical outlier criteria using multi-dimensional morphological-temporal features, thereby obtaining the set of valid pulse waves corresponding to that pressure step. The outlier filtering based on statistical distribution specifically includes one or more of the following removal criteria set based on robust statistics: Amplitude outlier detection: Pulse waves whose amplitude is higher than the first upper limit of the average pulse wave amplitude within the current pressure ladder, or lower than the first lower limit of the average pulse wave amplitude, are removed. The thresholds are determined based on quantile fitting of the amplitude distribution in the atrial fibrillation population. Area energy outlier detection: Pulse waves whose area integral is lower than the second proportional threshold of the mean area integral within the current pressure step are removed to filter out invalid samples with insufficient energy or dominated by noise. Outlier detection for inter-pulse variation: Pulse waves whose absolute difference between the pulse wave interval and the mean interval within the current pressure ladder exceeds the third proportional threshold, which is set with reference to the clinical statistical cutoff value of ventricular rhythm variation; Outlier detection based on rising slope: Pulse waves whose absolute difference between the rising slope and the mean slope within the current pressure step exceeds the fourth proportional threshold are removed to exclude samples with significant morphological distortion caused by local rhythm disturbances. The non-feature sample filtering process is performed in a hierarchical sequential filtering order: first, primary morphological domain screening is performed in the amplitude and area dimensions, followed by secondary time-morphological domain joint screening in the interval and slope dimensions. The number of samples filtered under each criterion does not exceed the preset proportional upper limit of the total number of pulse waves initially collected for that pressure step, and the number of remaining valid samples after filtering is not less than the preset minimum effective capacity to ensure the statistical significance of subsequent cluster analysis. The determination module is used to perform similarity clustering based on the waveform features of each pulse wave in the effective pulse wave set, determine the cluster with the largest number of samples as the dominant pulse wave set of the pressure step, and map the cluster center of the dominant pulse wave set in the clustering feature space to the amplitude dimension, and use the amplitude dimension as the dominant amplitude of the pressure step. The sorting module is used to arrange the dominant amplitudes corresponding to the pressure steps of all the acquired effective pulse waves in descending order of pressure value to form an envelope skeleton sequence. The verification module is used to verify the trend consistency of the envelope skeleton sequence based on the prior law of physiological envelope in oscillometric blood pressure measurement. The prior law of physiological envelope is: during step-low blood pressure reduction, the pulse wave amplitude changes with the cuff pressure from high to low, showing an overall spindle-shaped curve characteristic of first rising and then falling, with a unique main peak. The amplitude on the left side of the main peak shows an upward trend, and the amplitude on the right side of the main peak shows a downward trend. Abnormal skeleton points in the envelope skeleton sequence that violate the trend of spindle-shaped curve are removed or corrected to obtain the verified envelope skeleton sequence. The reconstruction module is used to perform interpolation, smoothing, and fitting reconstruction processing on the verified envelope skeleton sequence to obtain a continuous and stable pulse wave envelope for blood pressure identification.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the pulse wave envelope reconstruction method as described in any one of claims 1 to 4.

7. An electronic device, comprising: It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the pulse wave envelope reconstruction method as described in any one of claims 1 to 4.

Images