Multi-index based oscillographic blood pressure measurement system for automatic evaluation of pulse wave quality

Abstract

The application provides a kind of automatic evaluation system of pulse wave quality in oscillographic blood pressure measurement based on multi-index, it is related to wearable medical health monitoring technical field.The system includes data preprocessing module for filtering processing to pulse wave signal;For the preprocessed pulse wave signal, the signal quality once evaluation module is removed to filter out the signal of poor quality evaluation;Signal interception module extracts the effective signal area required for calculating blood pressure;According to the morphological feature, the signal feature index extraction module obtains the feature index of the pulse wave signal without quality evaluation;According to the feature index of pulse wave signal, the signal quality secondary evaluation module further divides the quality of pulse wave signal;Signal quality correction module uses airbag pressure data to correct the quality of general signal.The system overcomes the limitations of manual threshold setting in traditional quality evaluation method, and the evaluation algorithm is more efficient and has higher adaptability.

Description

Technical Field

[0001] This invention relates to the field of wearable medical health monitoring technology, and in particular to an automatic pulse wave quality assessment system based on multi-index oscillometric blood pressure measurement. Background Technology

[0002] Hypertension is a common cardiovascular disease, consistently ranking first in cardiovascular mortality in my country. Meanwhile, the prevalence of hypertension among Chinese adults has been rising annually. Timely and accurate blood pressure measurement is fundamental to the diagnosis and treatment of hypertension, as well as the prevention and treatment of other common cardiovascular diseases. In routine blood pressure monitoring, electronic blood pressure monitors are a common type of automatic measurement device, mostly based on the oscillometric method. This process requires calculation based on the collected pulse wave. However, during the data collection process, environmental factors, equipment limitations, and operator error can all affect the signal, leading to noise, interference, and signal instability. These issues can negatively impact the accuracy and reliability of blood pressure prediction, resulting in diagnostic errors.

[0003] Therefore, assessing the quality of pulse waves is particularly important and necessary. Research on blood pressure measurement can effectively improve the accuracy and reliability of blood pressure data, thereby increasing the efficiency of hospital diagnosis and enabling patients to receive timely treatment. Simultaneously, for home monitoring, it can effectively reduce errors during the measurement process. Only when the accuracy of measurement data is guaranteed can the prevention and treatment of hypertension be achieved, safeguarding people's lives and health.

[0004] The invention with patent number CN115530783A provides a method, device, and electronic blood pressure monitor for measuring blood pressure. This invention utilizes the pressure sensor of the electronic blood pressure monitor to acquire pulse wave signals, extract quality parameters of the pulse wave signals, and determine the quality of the pulse wave signals based on these parameters. However, the quality parameters are determined based on several manually defined empirical thresholds, making the evaluation results susceptible to human influence. Furthermore, this method only performs a binary classification of pulse wave signal quality, which is not detailed enough and can easily lead to the discarding of some usable processed data, thus limiting the ability for further signal processing and improvement.

[0005] The invention with patent number CN113440114A provides a method and apparatus for pulse wave optimization based on a thin-film pressure sensor. It acquires pulse wave signals from multiple channels by covering the thin-film pressure sensor at the radial artery in the wrist. A similarity algorithm is used to determine the quality of the pulse wave signals from each channel, selecting the optimal channel and its pulse wave signal, and discarding channels with poor quality. For a specific channel's pulse wave signal, the optimal pulse wave signal is determined through eigenvalue analysis; otherwise, it is discarded. However, this method does not assess signal quality based on the overall envelope shape of the pressure pulse wave, making it difficult to use for evaluating the pulse wave envelope quality required for oscillometric blood pressure calculation. Furthermore, the selection of eigenvalues ​​and thresholds is subjective.

[0006] The invention with patent number CN113558584A provides a pulse wave preprocessing method based on signal quality assessment. This method can adjust the placement or parameters of the pulse wave sensor in a timely manner based on the waveform characteristics, amplitude characteristics, and period difference information of the acquired pulse wave signal, thereby reducing the probability of acquiring invalid pulse wave signals. It solves the problem in existing technologies that treat all pulse wave signals acquired by the pulse wave sensor as usable pulse wave signals for generating pulse diagnosis results, ignoring some morphologically distorted pulse wave signals. However, this method does not consider the pulse wave signal during the inflation or deflation process, making it difficult to assess the overall quality of the pulse wave signal.

[0007] The invention with patent number CN112183354A provides a method for quality assessment of single-cycle pulse wave signals based on SVM. Feature selection is performed based on the physiological characteristics of the pulse wave, and the quality assessment results can well reflect the physiological characteristics of the pulse wave. Even for sample types that are less common in the dataset, key features can be used for identification, eliminating interference from non-key features and compensating for the deficiencies of the classifier caused by an incomplete dataset, resulting in high classification accuracy. However, this method only assesses the quality of single-cycle pulse wave signals, and the classifier is a binary classifier, without providing more detailed quality assessment results, which can easily lead to data loss.

[0008] The invention with patent number CN115886746A provides a method for classifying pulse wave waveform quality levels and extracting stable waveforms. This method can preferentially eliminate abnormal single-cycle pulse waves, avoiding the influence of abnormal data and providing more accurate results than using the averaging method. Simultaneously, it offers a more refined classification of waveform quality levels, ensuring the reliability of subsequent analysis results. However, this method only focuses on the morphology of a single pulse wave cycle, neglecting the overall quality of the entire pulse wave signal. Furthermore, its stable waveform extraction method is not applicable to pressure pulse waves during pressurization, leading to a loss of blood pressure calculation information.

[0009] The invention with patent number CN112587104A provides a method for filtering out invalid pulse waveforms. By calculating the sum of the absolute values ​​of the Euclidean distances from all points in each waveform to their corresponding points in the average waveform, a threshold is set to filter out invalid waveforms, providing a new approach to filtering out invalid or poor waveforms. However, the method of comparing with the average waveform for screening may struggle to capture all the key features of more complex waveforms, potentially overlooking local features and subtle changes within the waveform.

[0010] The invention with patent number CN102670182B provides a real-time pulse wave acquisition quality analysis device. It includes a data input module, a data buffer and preprocessing module, a data analysis module, a weight setting module, and an evaluation module. It can perform real-time analysis and processing of the input pulse wave signal and automatically identify the pulse wave signal quality. However, it relies on user-set weight coefficients, which may lead to inconsistencies in the results.

[0011] Liu Jian's master's thesis at Southeast University, titled "Evaluation and Quantification of Pulse Waveform Quality in Oscillometric Blood Pressure Measurement," proposes a quantitative evaluation algorithm for pulse waveforms in oscillometric blood pressure measurement by analyzing data from a hardware acquisition system. The study also quantifies the impact of five noise conditions on blood pressure values ​​and their significant differences from standard conditions, further validating the reliability of the four-class classification boundary. However, the algorithm still suffers from poor generalization ability and does not fully consider the individual information of the subjects.

[0012] Most of the methods described above focus only on the quality of a single-cycle pulse wave signal, neglecting the overall quality of the pulse wave signal. Some methods assess the overall quality of the pulse wave but fail to extract the effective region of the signal, while noise information contained in the ineffective region interferes with the quality assessment. Furthermore, current methods can only classify pulse wave quality based on some obvious morphological features, and often use manually defined empirical thresholds as quality assessment standards, resulting in poor generalization ability. In addition, some patents only differentiate between "good" and "bad" quality levels, failing to provide accurate and detailed quality assessment results and limiting further signal processing and improvement capabilities. Summary of the Invention

[0013] The technical problem to be solved by the present invention is to address the shortcomings of the prior art by providing an automatic evaluation system for pulse wave quality in oscillometric blood pressure measurement based on multiple indicators, thereby realizing the automatic evaluation of pulse wave quality in oscillometric blood pressure measurement.

[0014] To solve the above-mentioned technical problems, the technical solution adopted by this invention is: an automatic pulse wave quality assessment system based on multi-index oscillometric blood pressure measurement, including a data preprocessing module, a primary signal quality assessment module, a signal interception module, a signal feature index extraction module, a secondary signal quality assessment module, and a signal quality correction module; the data preprocessing module is used to filter the acquired pulse wave signal to remove common noise; the primary signal quality assessment module receives the pulse wave signal preprocessed by the data preprocessing module, filters out a portion of signals with poor quality assessment, and removes them; the signal interception module receives the other pulse wave signals after filtering and extracts the effective signal region required for blood pressure calculation; the signal feature index extraction module receives the effective region of the pulse wave signal and extracts feature indicators based on morphological characteristics to obtain the feature indicators of the pulse wave signal with unassessed quality; the secondary signal quality assessment module receives the feature indicators of the pulse wave signal and further divides the pulse wave signal into three categories: good quality, average quality, and poor quality; the signal quality correction module receives the signals with average quality, uses synchronously acquired airbag pressure data for quality correction, obtains the quality-corrected pulse wave signal, and completes the quality assessment of the pulse wave signal.

[0015] Preferably, the data preprocessing module first uses a low-pass filter to remove high-frequency noise from the original pulse wave signal; then it performs a high-pass filter to remove the influence of baseline drift on the pulse wave.

[0016] Preferably, the signal quality primary evaluation module performs a primary evaluation of signal quality by extracting the amplitude and morphological features of the preprocessed pulse wave signal, and simultaneously performs a primary evaluation of signal quality using airbag pressure data; the amplitude feature is the overall amplitude of the pulse wave signal envelope; the morphological feature consists of the number of troughs, the number of peaks, the integrity of troughs, the integrity of peaks, the rising slope, the falling slope of the pulse wave signal, and the amplitude, rising slope, and falling slope of the airbag pressure data.

[0017] Preferably, the signal quality primary evaluation module extracts the upper envelope amplitude A of the pulse wave signal from the filtered pulse wave signal data. up and lower envelope amplitude A down Calculate the overall envelope amplitude A, i.e., A = A up -A downThe system determines whether the pulse wave signal quality exceeds a preset threshold. If the overall amplitude A of the envelope exceeds the preset threshold, the signal is retained; otherwise, it is discarded, and the signal quality is assessed as poor. The system extracts the initial trough, peak, and final trough of the pulse wave signal envelope by setting thresholds, obtaining the rising and falling edges of the envelope, ultimately dividing the pulse wave signal into five segments. Multiple indicators are extracted, including the number of trough points, the number of peak points, trough integrity, peak integrity, rising edge slope, and falling edge slope, serving as standards for pulse wave signal quality assessment. Simultaneously, the system acquires and calculates the amplitude, rising edge slope, and falling edge slope of the airbag pressure data, using these as standards for pulse wave signal quality assessment. The system calculates the indicators contained in the morphological features and compares them with preset thresholds to determine if the pulse wave signal quality meets the requirements. If all indicators meet the preset threshold requirements, the signal is retained; otherwise, the signal quality is assessed as poor, and the pulse wave signal is discarded.

[0018] Preferably, the signal interception module includes a front-end noise removal unit and a back-end noise removal unit, which intercepts the pulse wave signal by removing front-end and back-end noise; the front-end noise removal unit is used to remove unstable signals generated in the early stage of data acquisition; and the upper envelope amplitude A of the pulse wave signal is extracted. up Set the interception threshold A front The system first extracts the signal portion with an initial amplitude less than the extraction threshold, removes front-end noise, and obtains the effective front-end region. The back-end noise removal unit is used to remove the reverse signal generated at the end of the data acquisition period. First, the coordinates of the peak and valley points of each cycle pulse wave are extracted. When the sum of the ordinate of the valley point and the ordinate of the peak point is less than 0, it indicates that a reverse waveform has appeared. Second, the identified reverse waveform is removed to obtain the effective back-end region. The effective front-end region and the effective back-end region are then spliced ​​together to obtain the effective region of the pulse wave signal.

[0019] Preferably, the signal feature extraction module is used to extract abrupt change feature indicators, reverse wave feature indicators, waveform integrity feature indicators, and multi-peak feature indicators, and combines these feature indicators into a feature matrix, thereby integrating the key feature information of the signal; wherein,

[0020] Abrupt change point feature indicators are used to extract abrupt changes in the effective region of a signal, i.e., the locations where the signal peak point changes significantly. By analyzing the changes in the slope and amplitude of the signal, abrupt changes in the signal are identified and used as one of the feature indicators.

[0021] The reverse wave characteristic index is used to analyze the reverse waveform in a signal; the reverse wave refers to the waveform in the signal that is opposite in direction to the main waveform; by detecting and analyzing the occurrence position and amplitude of the reverse wave, the reverse wave characteristic index is extracted.

[0022] Waveform integrity characteristic indicators are used to evaluate the integrity of signal waveforms; by analyzing the shape, smoothness, and duration of signal waveforms, the waveform integrity of the signal is quantified and used as one of the characteristic indicators;

[0023] Multi-peak feature indicators are used to detect multiple envelope peaks in a signal. By analyzing the number, location, and amplitude of envelope peaks in the signal, multi-peak feature indicators are extracted to describe the envelope peak situation in the signal.

[0024] Preferably, the secondary signal quality assessment module employs a fusion model, combining random forest and SVM methods to further assess the quality of the pulse wave signal and add new features to the original feature matrix for training and classification by the SVM model. First, a random forest classifier is used to classify the pulse wave signal. The classification results obtained from the random forest classifier are then used as new features and added to the feature matrix extracted by the signal feature index extraction module to construct an expanded new feature matrix. Finally, an SVM is used as the classifier, and the expanded new feature matrix is ​​used to train the model, constructing a classification model that divides the pulse wave signal samples after the primary signal quality assessment into three categories: good quality, average quality, and poor quality. Combining the results of the primary signal quality assessment, the quality assessment of the pulse wave signal is completed.

[0025] Preferably, the signal quality correction module uses the airbag pressure data corresponding to the pulse wave to correct the quality of the pulse wave signal. First, it identifies abnormal regions in the pulse wave signal. Then, it determines the corresponding pressure range based on the location of the abnormal regions. Next, it compares the abnormal regions with the airbag pressure data detected by the airbag pressure sensor. Finally, it calculates the amplitude coefficient between the airbag pressure data and the pulse wave, adjusts the amplitude of the airbag pressure data, and replaces the extracted airbag pressure data with the abnormal pulse wave data to complete the quality correction of the signal with poor quality.

[0026] The beneficial effects of adopting the above technical solution are as follows: The pulse wave quality assessment system based on multi-index oscillometric blood pressure measurement provided by this invention overcomes the limitations of manually setting thresholds in traditional quality assessment methods, making the assessment algorithm more efficient and more adaptable; the signal quality primary assessment module evaluates the overall morphological characteristics of the pulse wave signal, screening the signal quality as a whole; the signal interception module reduces the interference of noise in the invalid region of the pulse wave signal on the quality assessment, simplifying the calculation and analysis process; the signal feature index extraction module can better analyze the local features and subtle changes in the overall pulse wave signal, improving the accuracy and sensitivity of the assessment algorithm; the use of a fusion model for pulse wave signal quality assessment integrates the advantages of multiple models, improves the algorithm's generalization ability, and better adapts to complex feature correlations; the proposed three-classification model for signal quality makes the assessment results more refined, enhancing the ability for further signal processing and improvement; the use of a signal quality correction module can better retain the effective information of the data, improving data accuracy. Attached Figure Description

[0027] Figure 1 A structural block diagram of a pulse wave quality assessment system based on multi-index oscillometric blood pressure measurement provided in an embodiment of the present invention;

[0028] Figure 2 This is a block diagram illustrating the implementation of the data preprocessing module provided in an embodiment of the present invention.

[0029] Figure 3 A block diagram illustrating the implementation of a primary signal quality assessment module provided in an embodiment of the present invention;

[0030] Figure 4 This is a block diagram illustrating the implementation of the signal interception module provided in an embodiment of the present invention.

[0031] Figure 5 The image shows the result of signal interception provided in the embodiment of the present invention, where (a) is the original signal and (b) is the signal of the effective region after signal interception.

[0032] Figure 6 This is a block diagram illustrating the implementation of the signal feature index extraction module provided in an embodiment of the present invention.

[0033] Figure 7 This is a block diagram illustrating the implementation of the signal quality secondary evaluation module provided in an embodiment of the present invention.

[0034] Figure 8 This is an example of a signal evaluation that is considered good in the embodiments of the present invention;

[0035] Figure 9The quality assessment provided in this embodiment of the invention is a general example of signal evaluation, wherein (a) is a signal incompletely extracted from the effective region of the pulse wave, and (b) is a signal with a large abrupt change point in the pulse wave;

[0036] Figure 10 This is an example of a signal evaluation with poor quality assessment provided in an embodiment of the present invention;

[0037] Figure 11 This is a block diagram illustrating the implementation of the signal quality correction module provided in an embodiment of the present invention. Detailed Implementation

[0038] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0039] In this embodiment, an automatic pulse wave quality assessment system based on multi-index oscillometric blood pressure measurement, such as... Figure 1 As shown, the system includes a data preprocessing module, a primary signal quality assessment module, a signal interception module, a signal feature extraction module, a secondary signal quality assessment module, and a signal quality correction module. The data preprocessing module filters the acquired pulse wave signal to remove common noise such as baseline and power frequency interference. The primary signal quality assessment module receives the preprocessed pulse wave signal from the data preprocessing module, filters out signals with poor quality assessment, and removes them. The signal interception module receives the remaining filtered pulse wave signals and extracts the effective signal region needed for blood pressure calculation. The signal feature extraction module receives the effective region of the pulse wave signal and extracts feature indicators based on morphological characteristics to obtain the feature indicators of the unassessed pulse wave signal. The secondary signal quality assessment module receives the feature indicators of the pulse wave signal and further classifies the pulse wave signal into three categories: good quality, average quality, and poor quality. The signal quality correction module receives signals with average quality and uses synchronously acquired airbag pressure data for quality correction to obtain a quality-corrected pulse wave signal, thus completing the pulse wave signal quality assessment.

[0040] In this embodiment, the data preprocessing module, such as Figure 2 As shown, a low-pass filter is first used to remove high-frequency noise such as power frequency interference from the original pulse wave signal; then a high-pass filter is performed to remove the influence of baseline drift on the pulse wave. In this embodiment, the data preprocessing module uses a 5th-order Butterworth filter with a cutoff frequency of 20Hz for noise reduction. Meanwhile, due to the patient's own breathing or other reasons, the pulse wave may experience baseline drift, and excessive baseline drift can affect the judgment of pulse wave morphological indicators, thus affecting the accuracy of blood pressure prediction. Therefore, a Butterworth filter is chosen for high-pass filtering, which can effectively remove the influence of baseline drift.

[0041] Signal quality primary assessment module, such as Figure 3 As shown, the signal quality is evaluated by extracting the amplitude and morphological features of the preprocessed pulse wave signal, and simultaneously by using airbag pressure data. The amplitude feature is the overall amplitude of the pulse wave signal envelope. The morphological feature consists of the number of troughs, peaks, trough integrity, peak integrity, rising slope, falling slope, and the amplitude, rising slope, and falling slope of the airbag pressure data. For the filtered pulse wave signal data, the upper envelope amplitude A is extracted. up and lower envelope amplitude A down Calculate the overall envelope amplitude A, i.e., A = A up -A down The signal is evaluated as poor quality if the overall amplitude A of the pulse wave signal is greater than a preset threshold. If the overall amplitude A is greater than the preset threshold, the signal is retained; otherwise, it is filtered out. This indicates that the pulse wave signal has a small amplitude and may be affected by noise or other interference, making it unsuitable for further processing and analysis. The initial trough, peak, and final trough of the pulse wave signal envelope are extracted by setting thresholds, resulting in the rising and falling edges of the envelope. The pulse wave signal is then divided into five segments. Multiple indicators are extracted, including the number of trough points, the number of peak points, trough integrity, peak integrity, rising edge slope, and falling edge slope, serving as standards for pulse wave signal quality evaluation. Simultaneously, airbag pressure data is acquired, and its amplitude, rising edge slope, and falling edge slope are calculated, also serving as standards for pulse wave signal quality evaluation. The method for extracting characteristic indicators from airbag pressure data is as follows: extract the upper envelope amplitude PA from the airbag pressure data. up and lower envelope amplitude PA down Next, calculate the overall amplitude PA of the airbag pressure data, i.e., PA = PA up -PA down Simultaneously, the rising and falling edges of the airbag pressure data are determined based on threshold values, and the rising and falling edge slopes are calculated. By calculating the indicators contained in the morphological features and comparing them with preset thresholds, it is determined whether the quality of the pulse wave signal meets the requirements. If all indicators meet the preset threshold requirements, the signal is retained; otherwise, the signal quality is evaluated as poor, and the pulse wave signal is filtered out.

[0042] Signal interception module, such as Figure 4 As shown, it includes a front-end noise removal unit and a back-end noise removal unit. The pulse wave signal is truncated by removing front-end and back-end noise. The front-end noise removal unit is used to remove unstable signals caused by noise interference generated during the early stages of data acquisition. The upper envelope amplitude A of the pulse wave signal is extracted. up Set the interception threshold Afront The system first extracts the signal portion with an initial amplitude less than a threshold to remove front-end noise and obtain the effective front-end region. The back-end noise removal unit removes the reverse signal generated at the end of data acquisition. First, the peak and trough coordinates of each pulse wave cycle are extracted. When the sum of the ordinate of the trough and the ordinate of the peak is less than 0, it indicates the presence of a reverse waveform. Second, the identified reverse waveform is removed to obtain the effective back-end region. The effective front-end region and the effective back-end region are then concatenated to obtain the effective region of the pulse wave signal. In this embodiment, the front-end noise removal method specifically involves: first, extracting the upper envelope amplitude A of the pulse wave signal. up Since unstable signals usually occur in the early stages of a signal, the truncation threshold A is further calculated. front The calculation formula is as follows:

[0043] A front =A up *C

[0044] Where C is the truncation coefficient, and the preset truncation coefficient can be set according to the specific application requirements.

[0045] Finally, the initial amplitude of the pulse wave was selected when it was less than A. front The signal is removed to obtain the effective front-end region of the pulse wave signal.

[0046] The method for removing back-end noise is as follows: At the end of data acquisition, a reverse waveform may sometimes appear. When the amplitude of the reverse waveform is too large, it will interfere with the envelope fitting. Therefore, the peak point coordinates P[(X1, Y1), (X2, Y2), ..., (X...] of each pulse wave cycle are first extracted. i Y i The coordinates of the trough points of each pulse wave cycle are V[(M1, N1), (M2, N2), ..., (M...). i N i Next, calculate the sum of the ordinates of the peak and valley points for each cycle. When N i +Y i A value less than 0 indicates the presence of a reverse waveform. Subsequent reverse waveforms are truncated and removed to obtain the effective rear region of the pulse wave signal. The effective front region and the effective rear region are then concatenated to obtain the effective region of the pulse wave signal. In this embodiment, the original pulse wave signal and the pulse wave signal processed by the signal truncation module are as follows: Figure 5 As shown.

[0047] Signal feature extraction module, such as Figure 6As shown, features such as abrupt change point characteristics, reverse wave characteristics, waveform integrity characteristics, and multi-peak characteristics are extracted and combined into a feature matrix, thereby integrating the key feature information of the signal; among them,

[0048] Abrupt change point feature indicators are used to extract abrupt changes in the effective region of a signal, i.e., the locations where the signal peak point changes significantly. By analyzing the changes in the slope and amplitude of the signal, abrupt changes in the signal can be identified and used as one of the feature indicators.

[0049] The reverse wave characteristic index is used to analyze the reverse waveform in a signal; the reverse wave refers to the waveform in the signal that is opposite in direction to the main waveform; by detecting and analyzing the occurrence position and amplitude of the reverse wave, the reverse wave characteristic index is extracted.

[0050] Waveform integrity features are used to evaluate the integrity of a signal waveform. By analyzing the shape, smoothness, and duration of the signal waveform, the integrity of the waveform can be quantified and used as one of the features.

[0051] Multi-peak feature indicators are used to detect multiple envelope peaks in a signal. By analyzing the number, location, and amplitude of envelope peaks in a signal, multi-peak feature indicators can be extracted to describe the envelope peak situation in the signal.

[0052] In this embodiment, the specific method for extracting the feature indicators of abrupt change points is as follows: Generally, connecting the peak points of a pulse wave signal constitutes the initial envelope of the oscillation wave. Common envelope fitting methods include polynomial fitting and Gaussian fitting. However, during signal acquisition, due to subject movement or external interference, the relative position between the sensor and the skin changes, causing abrupt changes in the oscillation waveform. Excessive or too many abrupt change points can affect the accuracy of the oscillation envelope fitting, thus affecting blood pressure estimation. The feature indicators of abrupt change points are defined as follows:

[0053] A uoutlier (i)=A upeak (i)-A filtupeak (i) i = 1, 2, 3…N1

[0054]

[0055]

[0056] A loutlier (j)=A lpeak (j)-A filtlpeak (j)j=1,2,3…N2

[0057]

[0058]

[0059] Where F1 and F2 are the mean difference and standard deviation of all peak point mutation values ​​in the upper envelope, respectively; F3 and F4 are the mean difference and standard deviation of all oscillatory trough point mutation values ​​in the lower envelope, respectively; A uoutier A loutlier These represent the amplitude differences before and after filtering at the peak point and the amplitude differences before and after filtering at the valley point, respectively; A upeak A represents the peak amplitude before filtering. filtupeak Peak amplitude after median filtering; A lpeak A represents the amplitude of the valley point before filtering. filtlpeak N1 represents the amplitude of the valley points after filtering. N2 represents the number of peak points and valley points, respectively.

[0060] The method for extracting the inverted wave feature index is as follows: During the early or late stages of inflation, a segment of the pulse wave signal exhibits an inverted waveform. If the amplitude of the inverted waveform is large, it will affect the shape of the fitted envelope. The inverted wave feature index is extracted and defined as follows:

[0061]

[0062] F6=|Loc u -Loc l |

[0063] Where F5 is the ratio of the upper and lower envelope amplitudes; F6 is the distance between the peak points of the upper and lower envelopes; Loc u Loc represents the location of the upper envelope peak. l This indicates the location of the lower envelope peak.

[0064] The method for extracting waveform integrity features is as follows: To remove interference signals in the early and late stages of pressurization, the effective region of the signal is truncated. However, due to excessively large reverse wave amplitudes or significant abrupt changes, the usable portion of some signals is difficult to extract, potentially resulting in waveform gaps after truncation and increasing the difficulty of envelope fitting. To determine waveform integrity, the positions of the upper envelope peak and lower envelope valley of the oscillating wave are extracted. The peaks are then used to determine whether they occur at the critical points of the truncated waveform, i.e., the beginning and end. The waveform integrity index is defined as follows:

[0065]

[0066]

[0067] Where F7 is the peak point at the critical point of the waveform; F8 is the valley point at the critical point of the waveform; Bp is the number of upper envelope peak points at the critical point; and Bt is the number of lower envelope peak points at the critical point.

[0068] The method for extracting multi-peak feature indicators is as follows: Generally, as external pressure gradually increases, the pulse wave amplitude initially increases and then decreases, meaning the pulse wave envelope exhibits a single-peak curve shape. However, in actual data acquisition, the pulse wave envelope shape is complex, potentially containing multiple peaks and troughs. The location and shape of these peaks and troughs depend on various factors, such as cardiac contractility, vascular stiffness, vascular resistance, and heart valve condition. The presence of multiple peaks and troughs can interfere with the fitting of the pulse wave envelope. The multi-peak feature indicators for pulse waves are defined as follows:

[0069] F9 = Loc upeak (i+1)-Loc upeak (i)

[0070] F 10 =|A upeak (i+1)-A upeak (i)|

[0071] F 11 =|A upeak (i)-A lpeak |

[0072] Where F9 is the distance between two adjacent peak points; F 10 F is the amplitude difference between two adjacent peak points; 11 Loc represents the amplitude difference between two adjacent peak points and valley points. upeak A represents the location of the envelope peak. upeak A represents the amplitude of the envelope peak value. lpeak This represents the amplitude of the envelope valley value.

[0073] By extracting the above feature indicators and combining them into a feature matrix, the key feature information of the signal can be integrated together.

[0074] Signal quality secondary evaluation module, such as Figure 7As shown, a fusion model combining random forest and SVM methods is employed. This module aims to further assess the signal quality and add new features to the original feature matrix for SVM model training and classification. First, a random forest classifier is used to classify the signal. Second, the classification results obtained by the random forest classifier are used as new features and added to the feature matrix extracted by the signal feature index extraction module to construct an expanded new feature matrix. Finally, SVM is used as the classifier, and the expanded new feature matrix is ​​used for model training to construct a classification model that divides the pulse wave signal samples after the first signal quality assessment into three categories: good quality, average quality, and poor quality. Combining the results of the first signal quality assessment, the quality assessment of the pulse wave signal is completed. In this embodiment, the original feature matrix F = [F1, F2, ..., F...] is first constructed using the feature indices extracted by the signal feature index extraction module. 11 Then, a random forest classifier is trained using the original feature matrix, and the classification result is used as the extracted new feature F. 12 Next, the features F extracted by the random forest are... 12 As new features, they are added to the original feature matrix to obtain the expanded new feature matrix F = [F1, F2, ..., F 12 Finally, an SVM classifier is trained using the expanded new feature matrix to further divide the data after the first evaluation into three categories: good quality, average quality, and poor quality. In this embodiment, a pulse wave signal with a good quality evaluation is as follows: Figure 8 As shown; the quality assessment is a general pulse wave signal, such as Figure 9 As shown; pulse wave signals with poor quality assessment are as follows: Figure 10 As shown.

[0075] Signal quality correction module, such as Figure 11 As shown, the quality of a pulse wave signal with average quality is corrected using airbag pressure data corresponding to the pulse wave. First, through threshold detection and waveform shape analysis, abnormal regions in the pulse wave signal can be identified. Then, based on the location of the abnormal region, its corresponding pressure range is determined. Next, by correlating it with the airbag pressure data detected by the airbag pressure sensor, the signal quality correction module extracts the airbag pressure data within the corresponding pressure range. Finally, the amplitude coefficient between the airbag pressure data and the pulse wave is calculated, the amplitude of the airbag pressure data is adjusted, and the extracted airbag pressure data is replaced with the abnormal pulse wave data, thus completing the quality correction of the average quality signal.

[0076] In this embodiment, the amplitude coefficient r between the airbag pressure data and the pulse wave is calculated using the following formula:

[0077]

[0078] Among them, Aup PA is the upper envelope amplitude of the pulse wave. up This represents the upper envelope amplitude of the airbag pressure data.

[0079] The amplitude of the airbag pressure data is adjusted using the amplitude coefficient r, and the calculation formula is as follows:

[0080] PA r =PA*r

[0081] Among them, PA r PA represents the amplitude of the airbag pressure data after amplitude adjustment, while PA represents the original amplitude of the airbag pressure data.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.

Claims

1. An automatic pulse wave quality assessment system for oscillometric blood pressure measurement based on multiple indicators, characterized in that: The system includes a data preprocessing module, a primary signal quality assessment module, a signal interception module, a signal feature extraction module, a secondary signal quality assessment module, and a signal quality correction module. The data preprocessing module filters the acquired pulse wave signal. The primary signal quality assessment module receives the preprocessed pulse wave signal from the data preprocessing module, filters out signals with poor quality assessment, and removes them. The signal interception module receives the remaining pulse wave signals after filtering and extracts the effective signal region needed for blood pressure calculation. The signal feature extraction module receives the effective region of the pulse wave signal and extracts feature indicators based on morphological characteristics to obtain the feature indicators of the pulse wave signal with unassessed quality. The secondary signal quality assessment module receives the feature indicators of the pulse wave signal and further classifies the pulse wave signal into three categories: good quality, average quality, and poor quality. The signal quality correction module receives signals with average quality and uses synchronously acquired airbag pressure data for quality correction to obtain a quality-corrected pulse wave signal, thus completing the pulse wave signal quality assessment. The signal interception module includes a front-end noise removal unit and a back-end noise removal unit. It intercepts the pulse wave signal by removing front-end and back-end noise. The front-end noise removal unit removes unstable signals generated during the early stages of data acquisition. The upper envelope amplitude A of the pulse wave signal is extracted. up Set the interception threshold A front The signal portion with an initial amplitude less than the threshold is extracted, and front-end noise is removed to obtain the effective front-end region. The back-end noise removal unit is used to remove the reverse signal generated at the end of the data acquisition period. First, the peak and valley coordinates of each cycle pulse wave are extracted. When the sum of the ordinate of the valley point and the ordinate of the peak point is less than 0, it indicates that a reverse waveform has appeared. Secondly, the identified reverse waveform is removed to obtain the effective region at the back end; the effective region at the front end and the effective region at the back end are spliced ​​together to obtain the effective region of the pulse wave signal. The secondary signal quality assessment module employs a fusion model, combining random forest and SVM methods, to further assess the quality of pulse wave signals and add new features to the original feature matrix for SVM model training and classification. First, a random forest classifier is used to classify the pulse wave signals. The classification results from the random forest classifier are then used as new features and added to the feature matrix extracted by the signal feature index extraction module to construct an expanded new feature matrix. Finally, SVM is used as the classifier, and the expanded new feature matrix is ​​used for model training to construct a classification model that divides the pulse wave signal samples after the primary signal quality assessment into three categories: good quality, average quality, and poor quality. Combining the results of the primary signal quality assessment, the overall quality assessment of the pulse wave signals is completed. The signal quality correction module uses the airbag pressure data corresponding to the pulse wave to correct the quality of the pulse wave signal that is of average quality; first, it identifies the abnormal areas in the pulse wave signal. Then, based on the location of the abnormal area, its corresponding pressure range is determined; then, it is correlated with the airbag pressure data detected by the airbag pressure sensor; finally, the amplitude coefficient between the airbag pressure data and the pulse wave is calculated, the amplitude of the airbag pressure data is adjusted, and the extracted airbag pressure data is replaced with the abnormal pulse wave data to complete the quality correction of the general signal.

2. The automatic pulse wave quality assessment system for oscillometric blood pressure measurement based on multiple indicators as described in claim 1, characterized in that: The data preprocessing module first uses a low-pass filter to remove high-frequency noise from the original pulse wave signal; then it performs a high-pass filter to remove the influence of baseline drift on the pulse wave.

3. The automatic pulse wave quality assessment system for oscillometric blood pressure measurement based on multiple indicators as described in claim 1, characterized in that: The signal quality primary evaluation module performs a primary evaluation of signal quality by extracting the amplitude and morphological features of the preprocessed pulse wave signal, and simultaneously uses airbag pressure data for a primary evaluation of signal quality. The amplitude feature is the overall amplitude of the pulse wave signal envelope. The morphological feature consists of the number of troughs, the number of peaks, the integrity of troughs, the integrity of peaks, the rising slope, the falling slope, and the amplitude, rising slope, and falling slope of the airbag pressure data.

4. The automatic pulse wave quality assessment system for oscillometric blood pressure measurement based on multiple indicators as described in claim 2, characterized in that: The signal quality primary evaluation module extracts the upper envelope amplitude A of the pulse wave signal from the filtered pulse wave signal data. up and lower envelope amplitude A down Calculate the overall envelope amplitude A, i.e., A = A up -A down The signal is determined to be greater than a preset threshold. If the overall amplitude A of the envelope is greater than the preset threshold, the signal is retained. If it is less than the preset threshold, the signal is filtered out and the quality of the pulse wave signal is evaluated as poor. By setting thresholds, the initial trough, peak, and final trough of the pulse wave signal envelope are extracted, resulting in the rising and falling edges of the pulse wave signal envelope. The pulse wave signal is then divided into five segments. Multiple indicators are extracted, including the number of trough points, the number of peak points, trough integrity, peak integrity, rising edge slope, and falling edge slope, serving as standards for pulse wave signal quality assessment. Simultaneously, airbag pressure data is acquired, and its amplitude, rising edge slope, and falling edge slope are calculated, also serving as standards for pulse wave signal quality assessment. By calculating the indicators contained in the morphological features and comparing them with preset thresholds, it is determined whether the pulse wave signal quality meets the requirements. If all indicators meet the preset threshold requirements, the signal is retained; otherwise, the signal quality is assessed as poor, and the pulse wave signal is rejected.

5. The automatic pulse wave quality assessment system for oscillometric blood pressure measurement based on multiple indicators as described in claim 1, characterized in that: The signal feature extraction module is used to extract abrupt change feature indicators, reverse wave feature indicators, waveform integrity feature indicators, and multi-peak feature indicators, and combines these feature indicators into a feature matrix, thereby integrating the key feature information of the signal; wherein... Abrupt change point feature indicators are used to extract abrupt changes in the effective region of a signal, i.e., the locations where the signal peak point changes significantly. By analyzing the changes in the slope and amplitude of the signal, abrupt changes in the signal are identified and used as one of the feature indicators. The reverse wave characteristic index is used to analyze the reverse waveform in a signal; the reverse wave refers to the waveform in the signal that is opposite in direction to the main waveform; by detecting and analyzing the occurrence position and amplitude of the reverse wave, the reverse wave characteristic index is extracted. Waveform integrity characteristic indicators are used to evaluate the integrity of signal waveforms; by analyzing the shape, smoothness, and duration of signal waveforms, the waveform integrity of the signal is quantified and used as one of the characteristic indicators; Multi-peak feature indicators are used to detect multiple envelope peaks in a signal. By analyzing the number, location, and amplitude of envelope peaks in the signal, multi-peak feature indicators are extracted to describe the envelope peak situation in the signal.

Images