Sphygmomanometer intelligent management method and system and readable storage medium

Abstract

The invention provides a sphygmomanometer intelligent management method and system and a readable storage medium. The sphygmomanometer intelligent management method comprises the steps that blood pressure reference deviation, environment temperature and humidity deviation and blood pressure measurement channel hardware deviation are obtained; analyzing and comparing the blood pressure reference with the calibration result of the previous blood pressure reference to obtain a first analysis and comparison result; analyzing and comparing the environment temperature and humidity with the calibration result of the last environment temperature and humidity to obtain a second analysis and comparison result; analyzing and comparing the calibration result of the blood pressure measurement channel hardware with the calibration result of the previous blood pressure measurement channel hardware to obtain a third analysis and comparison result; collecting and processing user blood pressure data to obtain blood pressure data in a measurement stage; based on a preset analysis model, calculating an optimal estimated value of the blood pressure data in the measurement stage by adopting an adaptive Kalman filtering algorithm; comparing the optimal estimation value with a preset standard value, and calculating to obtain a comparison deviation; compensating the blood pressure data in the measurement stage based on the comparison deviation to obtain target blood pressure data.

Description

Technical Field

[0001] The present invention belongs to the technical field of sphygmomanometers, and in particular, relates to a sphygmomanometer intelligent management method, system and readable storage medium. Background Art

[0002] With the increasing prevalence of hypertension, patients have a growing demand for accurate, convenient and intelligent management of blood pressure monitoring. Although traditional sphygmomanometers can measure blood pressure values, they have obvious deficiencies in automatic calibration stability, algorithm accuracy, data depth analysis, abnormal trend warning and personalized health guidance, making it difficult to meet the requirements of modern medical care for refined management of hypertension.

[0003] Technical solution of the prior art 1: Ordinary electronic blood pressure monitors usually use fixed pressure sensor calibration parameters, and are calibrated once using a standard pressure source during the production process. The calibration data is stored inside the device, and during measurement, the electrical signal collected by the pressure sensor is converted into a blood pressure value according to a preset conversion formula, and the result is displayed on a local display screen.

[0004] Technical solution of the second existing technology: Some blood pressure monitors have data transmission function, which can transmit the measurement data to mobile phone APP or computer software through Bluetooth or USB interface, but most of these APP or software only provide simple data recording and chart display functions, and their blood pressure calculation algorithms are still based on traditional empirical formulas, and have not combined big data and intelligent algorithms to conduct in-depth data mining.

[0005] Disadvantages of the first prior art: Due to factors such as aging of electronic components, changes in ambient temperature and humidity, fixed calibration parameters will gradually deviate from actual values, causing measurement errors to increase over time. Moreover, this method cannot track and adjust the calibration process in real time, affecting the accuracy of long-term measurements.

[0006] Disadvantages of the second existing technology: The applicability of traditional algorithms to different individuals is limited, because the blood pressure fluctuation patterns of different populations are affected by multiple factors such as age, gender, living habits, and underlying diseases. At the same time, it is impossible to conduct real-time intelligent analysis of blood pressure data, making it difficult to detect potential abnormal blood pressure trends in advance, and it is also impossible to provide users with personalized blood pressure control plans and health management suggestions. Summary of the invention

[0007] The present invention provides a blood pressure monitor intelligent management method, system and readable storage medium, which can at least solve the technical problems that the blood pressure measurement system provided in the related art has inaccurate measurement data, the accuracy of the measurement data is easily affected by multiple influencing factors, and it is impossible to provide personalized blood pressure control plans and health management suggestions according to the actual needs and actual conditions of the group.

[0008] On the one hand, a method for intelligent management of a blood pressure monitor is provided, comprising: Obtain blood pressure baseline deviation, ambient temperature and humidity deviation, and blood pressure measurement channel hardware deviation; Analyze and compare the blood pressure pressure reference with the calibration result of the last blood pressure pressure reference to obtain a first analysis and comparison result; analyze and compare the ambient temperature and humidity with the calibration result of the last ambient temperature and humidity to obtain a second analysis and comparison result; analyze and compare the blood pressure measurement channel hardware with the calibration result of the last blood pressure measurement channel hardware to obtain a third analysis and comparison result; When the first analysis and comparison result, the second analysis and comparison result, and the third analysis and comparison result are all normal, the user's blood pressure data is collected and processed to obtain the blood pressure data of the measurement phase; Calculating an optimal estimate of the blood pressure data of the measurement phase using an adaptive Kalman filter algorithm based on a preset analysis model; Compare and calculate the optimal estimated value with a preset standard value to obtain a comparison deviation; The blood pressure data of the measurement phase is compensated based on the comparison deviation to obtain target blood pressure data.

[0009] In an optional implementation, the method further includes filtering the digital signal converted by ADC using a digital filter to remove noise interference.

[0010] In an optional implementation, the method further includes performing a linear or nonlinear correction on the pressure signal in the collected user blood pressure data according to the sensor characteristic curve to ensure the accuracy of the user blood pressure data.

[0011] In an optional embodiment, the method further includes obtaining a change in the ambient pressure during the user's blood pressure measurement, and controlling the inflation pressure during the blood pressure measurement based on the change in the ambient pressure.

[0012] In an optional embodiment, the method further includes obtaining physiological characteristics of the user, and controlling the deflation speed during blood pressure measurement based on the physiological characteristics of the user.

[0013] In an optional embodiment, it also includes real-time acquisition of pressure change data in the cuff, filtering the digital signal converted by the ADC through a digital filter, and calculating the systolic pressure and diastolic pressure using the oscillation method principle.

[0014] In an optional embodiment, the target blood pressure data is also transmitted back to a cloud processing center for the cloud processing center to train the preset analysis model.

[0015] In an optional implementation, it also includes performing an alarm process when the target blood pressure data exceeds a preset blood pressure value.

[0016] On the other hand, a sphygmomanometer intelligent management system is provided, the system comprising: A deviation acquisition unit, used to acquire a blood pressure pressure reference deviation, an ambient temperature and humidity deviation, and a blood pressure measurement channel hardware deviation; A comparison and analysis unit, configured to analyze and compare the blood pressure pressure reference with the calibration result of the last blood pressure pressure reference to obtain a first analysis and comparison result; analyze and compare the ambient temperature and humidity with the calibration result of the last ambient temperature and humidity to obtain a second analysis and comparison result; analyze and compare the blood pressure measurement channel hardware with the calibration result of the last blood pressure measurement channel hardware to obtain a third analysis and comparison result; A blood pressure data acquisition unit, used to collect and process the user's blood pressure data when the first analysis and comparison result, the second analysis and comparison result, and the third analysis and comparison result are all normal, to obtain the blood pressure data of the measurement stage; The target blood pressure data acquisition unit is used to calculate the optimal estimate value of the blood pressure data of the measurement phase based on a preset analysis model using an adaptive Kalman filter algorithm; compare the optimal estimate value with a preset standard value to obtain a comparison deviation; and compensate the blood pressure data of the measurement phase based on the comparison deviation to obtain target blood pressure data.

[0017] In yet another aspect, a readable storage medium is provided, on which any of the above-described methods is stored.

[0018] The sphygmomanometer intelligent management method provided by the embodiment of the present invention has at least the following beneficial effects: The intelligent management method for a sphygmomanometer provided in an embodiment of the present invention calculates an optimal estimate of the blood pressure data in the measurement phase based on a preset analysis model using an adaptive Kalman filtering algorithm; the optimal estimate is compared with a preset standard value to obtain a comparison deviation; based on the comparison deviation, the blood pressure data in the measurement phase is compensated to obtain target blood pressure data, thereby achieving dynamic correction and high-precision calculation of blood pressure measurement errors, breaking through the calibration limitations of traditional sphygmomanometers, and improving long-term measurement accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and other objects, features and advantages of the present disclosure will become more apparent through a more detailed description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, wherein like reference numerals generally represent like components throughout the exemplary embodiments of the present disclosure.

[0020] Figure 1 A schematic flow chart of a blood pressure monitor intelligent management method provided in an embodiment of the present invention.

[0021] Figure 2A schematic diagram of the automatic calibration process after initialization of the blood pressure monitor intelligent management system provided by an embodiment of the present invention.

[0022] Figure 3 A schematic diagram of the MCU and peripheral circuits of the blood pressure monitor intelligent management system provided in an embodiment of the present invention.

[0023] Figure 4 This is a schematic diagram of the pressure sensor interface circuit of the blood pressure monitor intelligent management system provided in an embodiment of the present invention.

[0024] Figure 5 This is a schematic diagram of the LCD display interface circuit of the blood pressure monitor intelligent management system provided by an embodiment of the present invention.

[0025] Figure 6 A schematic diagram of a power management circuit for a blood pressure monitor intelligent management system provided in an embodiment of the present invention.

[0026] Figure 7 A schematic diagram of the wireless communication circuit of the blood pressure monitor intelligent management system provided in an embodiment of the present invention.

[0027] Figure 8 Schematic diagram of the environmental sensor acquisition circuit of the blood pressure monitor intelligent management system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0028] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0029] As used herein, the term "including" and its variations mean open inclusion, i.e., "including but not limited to". Unless otherwise stated, the term "or" means "and / or". The term "based on" means "based at least in part on". The terms "an example embodiment" and "an embodiment" mean "at least one example embodiment". The term "another embodiment" means "at least one additional embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0030] On the one hand, an embodiment of the present invention provides a method for intelligent management of a blood pressure monitor, including: step S101, step S102 and step S103.

[0031] Step S101, obtaining a blood pressure reference deviation, an ambient temperature and humidity deviation, and a blood pressure measurement channel hardware deviation; Step S102: Analyze and compare the blood pressure pressure reference with the calibration result of the last blood pressure pressure reference to obtain a first analysis and comparison result; analyze and compare the ambient temperature and humidity with the calibration result of the last ambient temperature and humidity to obtain a second analysis and comparison result; analyze and compare the blood pressure measurement channel hardware with the calibration result of the last blood pressure measurement channel hardware to obtain a third analysis and comparison result; Step S103 collects and processes the user's blood pressure data when there are no abnormalities in the first analysis comparison results, the second analysis comparison results, and the third analysis comparison results to obtain the blood pressure data of the measurement phase; calculates the optimal estimate value of the blood pressure data of the measurement phase using an adaptive Kalman filter algorithm based on a preset analysis model; compares the optimal estimate value with the preset standard value to obtain a comparison deviation; and compensates the blood pressure data of the measurement phase based on the comparison deviation to obtain the target blood pressure data.

[0032] The intelligent management method for a sphygmomanometer provided in an embodiment of the present invention calculates an optimal estimate of the blood pressure data in the measurement phase based on a preset analysis model using an adaptive Kalman filtering algorithm; the optimal estimate is compared with a preset standard value to obtain a comparison deviation; based on the comparison deviation, the blood pressure data in the measurement phase is compensated to obtain target blood pressure data, thereby achieving dynamic correction and high-precision calculation of blood pressure measurement errors, breaking through the calibration limitations of traditional sphygmomanometers, and improving long-term measurement accuracy.

[0033] See also Figure 2 After the system starts, it is initialized and automatically calibrates the deviation, that is, the blood pressure pressure reference deviation, the ambient temperature and humidity deviation, and the blood pressure measurement channel hardware deviation provided in the embodiment of the present invention are obtained, and the blood pressure pressure reference is analyzed and compared with the calibration result of the previous blood pressure pressure reference to obtain the first analysis and comparison result; the ambient temperature and humidity are analyzed and compared with the calibration result of the previous ambient temperature and humidity to obtain the second analysis and comparison result; the blood pressure measurement channel hardware is analyzed and compared with the calibration result of the previous blood pressure measurement channel hardware to obtain the third analysis and comparison result. It is judged whether the first analysis and comparison result, the second analysis and comparison result, and the third analysis and comparison result are abnormal. If there is an abnormality, the abnormality processing is performed, that is, the automatic calibration deviation is performed, and the judgment is performed again after calibration until there is no abnormality and blood pressure data is measured.

[0034] In an optional implementation, the method further includes filtering the digital signal converted by ADC using a digital filter to remove noise interference.

[0035] In an optional implementation, the pressure signal in the collected user blood pressure data is linearly or nonlinearly corrected according to the sensor characteristic curve to ensure the accuracy of the user blood pressure data. For example, a voltage-blood pressure mapping table can be directly established to calculate the actual blood pressure by interpolation (linear / spline). Or deep learning (such as BP neural network) can be used to fit complex nonlinear relationships.

[0036] In an optional embodiment, the method further includes obtaining changes in the ambient pressure during the user's blood pressure measurement, and controlling the inflation pressure during the blood pressure measurement based on the changes in the ambient pressure.

[0037] Blood pressure measurement is usually based on inflating a cuff to block blood flow, then slowly deflate it, and determine systolic and diastolic pressures by detecting Korotkoff sounds or oscillation waves. Changes in environmental pressure (such as lower air pressure at high altitudes) can affect the absolute value of the cuff inflation pressure, leading to measurement errors.

[0038] Furthermore, the embodiment of the present invention can perform system initialization: measure the current ambient pressure Penv. Set the basic inflation pressure Pbase according to the user's historical data or the default value. Inflation control: calculate the target inflation pressure Ptarget according to the formula Ptarget=Pbase+ΔP+k·ΔPenv. Control the inflation pump to inflate at a constant rate until the cuff pressure reaches Ptarget. Deflation and measurement: Deflate at a constant rate while detecting Korotkoff sounds or oscillation waves. Determine the systolic and diastolic pressures based on the detection results. Real-time adjustment: During the measurement process, continuously monitor the changes in ambient pressure. If the ambient pressure changes exceed the threshold, recalculate Ptarget and adjust the inflation pressure.

[0039] In an optional embodiment, the method further includes obtaining physiological characteristics of the user, and controlling the deflation speed during blood pressure measurement based on the physiological characteristics of the user.

[0040] Furthermore, basic physiological parameters may include: static data: age, gender, height, weight, medical history (such as hypertension, arteriosclerosis). Dynamic data: current heart rate (HR) and heart rate variability (HRV); vascular elasticity index (which can be estimated by pulse wave velocity or waveform analysis); recent blood pressure history values ​​(such as systolic blood pressure SBP, diastolic blood pressure DBP). Measurement environment data, ambient temperature, user position (sitting / lying), whether after exercise (determined by acceleration sensor).

[0041] Exemplarily, based on the baseline speed of user classification, user types are divided according to physiological characteristics, and a preset baseline deflation rate (for example, 2~3 mmHg / second) is provided: for healthy adults, for example, it may be 3 mmHg / second; for the elderly / those with vascular sclerosis, for example, it may be 1.5~2 mmHg / second, etc.

[0042] In an optional embodiment, it also includes real-time collection of pressure change data in the cuff, filtering the digital signal converted by ADC through a digital filter, and calculating the systolic and diastolic pressures using the oscillation method. The pressure in the cuff is converted into an analog electrical signal by a high-precision pressure sensor, and sampled as a digital signal by ADC (analog-to-digital converter), and the sampling rate is usually ≥ 100 Hz (to ensure the capture of high-frequency oscillation waves). Digital filtering is used for denoising. Then the oscillation wave is extracted and the characteristics are analyzed. Further, oscillation wave separation is performed: the oscillation wave component is separated from the filtered cuff pressure signal: Posc(t)=Praw(t)-Ptrend(t)Posc(t)=Praw(t)-Ptrend(t) Wherein, PrawPraw: original pressure signal; PtrendPtrend: slowly changing baseline obtained by moving average or polynomial fitting.

[0043] Then perform amplitude calculation: calculate the oscillation wave amplitude (peak-to-peak value) in each cardiac cycle, usually expressed as the difference between adjacent peaks and troughs.

[0044] Identify key feature points: locate the maximum oscillation wave amplitude (AmaxAmax). During deflation, the oscillation wave amplitude shows a trend of "increasing first and then decreasing" as the cuff pressure decreases. Record the cuff pressure corresponding to the maximum amplitude point (PmaxPmax). Empirical association: PmaxPmax is usually close to the mean arterial pressure (MAP).

[0045] Perform amplitude envelope fitting: sort the amplitudes of each oscillation wave by cuff pressure and fit them into an envelope curve (usually using cubic spline interpolation or Gaussian fitting). The envelope peaks correspond to AmaxAmax and PmaxPmax.

[0046] Finally, the systolic blood pressure (SBP) and diastolic blood pressure (DBP) are calculated: first determine the mean arterial pressure (MAP): MAP≈Pmax (directly take the pressure value at the maximum amplitude point). Use the empirical proportional method to calculate SBP and DBP. Among them, ASBPASBP and ADBPADBP represent: the oscillation wave amplitude corresponding to systolic blood pressure and diastolic blood pressure, respectively, usually taking ASBP≈0.5AmaxASBP≈0.5Amax, ADBP≈0.8AmaxADBP≈0.8Amax. k1,k2k1,k2: proportionality coefficient (can be 0.3~0.8).

[0047] In an optional embodiment, the target blood pressure data is also transmitted back to a cloud processing center for the cloud processing center to train a preset analysis model.

[0048] It should be noted that the preset analysis model provided in the embodiment of the present invention can adopt the existing AI large model, and can also be trained using the model training method provided in the embodiment of the present invention.

[0049] Among them, the training method of the preset analysis model includes: Step 1: Build the initial algorithm architecture.

[0050] The embodiment of the present invention establishes a PTT-BP mapping equation based on the PTT benchmark model: SBP = α / (PTT) 2 + β. Among them, the benchmark model of PTT (PTT (Peripheral Pulse Transit Time) refers to the time required for blood to be ejected from the heart in the arteries to the peripheral blood vessels (such as fingers or toes). PTT is often used to assess the health of the cardiovascular system, especially in the diagnosis and management of diseases such as hypertension and diabetes. The benchmark model based on PTT is mainly used to analyze and interpret PTT data in order to better understand its physiological significance.) SBP: Systolic Blood Pressure, which indicates the highest pressure in the aorta when the heart contracts. α: A coefficient in the model, representing a constant term, used to adjust the fit of the model. This parameter is usually determined by data fitting. PTT (Peripheral Pulse Transit Time): Peripheral pulse wave transit time, that is, the time required for blood to be ejected from the heart to the peripheral blood vessels (such as fingers or toes). β: Another coefficient in the model, representing another constant term, also used to adjust the fit of the model. This parameter is also determined by data fitting.

[0051] Step 2: Use piecewise linear regression to calibrate individual differences. Step 2 includes the following steps: Step A: Pulse wave feature extraction.

[0052] Step B: Extract 12-dimensional time domain features (main wave amplitude, dicrotic trough depth, etc.).

[0053] Step C: Extract 24-dimensional frequency domain features (0.5-10Hz wavelet energy distribution). And Step D: Calculate waveform characteristic index (K value, stiffness index, etc.).

[0054] It should be noted that pulse wave feature extraction is an important part of signal processing and biomedical engineering, which aims to extract useful information from pulse wave signals for health assessment of the cardiovascular system, disease diagnosis, etc.

[0055] Step B: Extract 12-dimensional time domain features.

[0056] Time domain features are features extracted from the time series data of the pulse wave signal. These features directly reflect the temporal characteristics of the pulse wave. Common time domain features include but are not limited to the following categories: Amplitude of Main Wave (AMW): The maximum amplitude of the main wave (usually P wave or Q wave). The main wave reflects the strength of the heart's contractile force. Depth of Dicrotic Notch (DDN): Definition: The depth of a small valley that appears during the decline of the main wave. The depth of the dicrotic notch reflects the elasticity of the large arteries and the resistance to blood flow. Period (T): The time interval between two adjacent points of the same phase. The period reflects the heart rate and cycle frequency. Area Under Curve (AUC): The area under the pulse wave curve. The integrated area reflects the total blood flow. Maximum Slope (MS): The maximum slope of the pulse wave curve. The maximum slope reflects the heart's ejection capacity and arterial compliance. Minimum Slope (mS): The minimum slope of the pulse wave curve. Significance: Reflects the blood return rate and arterial resistance. Peak-to-Peak Distance (PPd): The distance between two adjacent main wave peaks. The peak-to-peak distance reflects the heart rate and circulatory stability. Root Mean Square (RMS): The root mean square value of the pulse wave signal. The root mean square value reflects the overall signal strength. Standard Deviation (SD): The standard deviation of the pulse wave signal. The standard deviation reflects the degree of signal fluctuation. Skewness (SKEW): The asymmetry of the probability distribution of the pulse wave signal. The skewness reflects the degree of asymmetry of the signal shape.

[0057] Step C: Extract 24-dimensional frequency domain features.

[0058] Frequency domain features are features extracted from the frequency components of the pulse wave signal. These features reflect the energy distribution of the pulse wave in different frequency ranges. Commonly used frequency domain features include but are not limited to the following categories: 0.5-10Hz wavelet energy distribution: Use wavelet transform to decompose the pulse wave signal into wavelet coefficients of different frequencies, and calculate the energy of each frequency interval. Significance: Different frequency intervals correspond to different physiological mechanisms. For example, low-frequency bands may reflect long-term changes, and high-frequency bands may reflect short-term changes. Specifically, a variety of wavelet basis functions (such as Daubechies wavelet, Morlet wavelet, etc.) can be used to perform multi-resolution analysis on the pulse wave signal, and calculate the wavelet energy of each frequency interval. Usually, the frequency range of 0.5Hz to 10Hz is selected and subdivided into multiple sub-intervals (for example, one interval every 0.2Hz), so as to obtain the energy distribution of 24 frequency intervals.

[0059] Step D: Calculate the waveform characteristic index.

[0060] The waveform characteristic index is a comprehensive index extracted from the pulse wave morphology, which is used to more comprehensively describe the characteristics of the pulse wave. Common waveform characteristic indices include but are not limited to the following categories: K Index: The ratio of the distance from a certain point of the pulse wave curve to the baseline to the amplitude of the main wave. Significance: Reflects the shape of the pulse wave and the elasticity of the large arteries. Stiffness Index: A quantitative index of large artery stiffness calculated by analyzing various characteristic parameters of the pulse wave (such as rise time and fall time). Significance: Reflects the elasticity and hardness of the blood vessel wall and is often used to assess the degree of arteriosclerosis. Augmentation Index (AIx): The ratio of the amplitude of the reflected wave to the total amplitude. Significance: Reflects the degree of large artery reflected waves and is often used as an indicator for assessing arteriosclerosis. Pulse Wave Velocity (PWV): The speed at which the pulse wave propagates along the artery. Significance: Reflects the elastic state of the artery and is often used to assess the degree of arteriosclerosis. Central Aortic Pressure (CAP): Approximates the pressure at the root of the aorta and is obtained by inversion of the peripheral pulse wave. Significance: Closer to the actual blood pressure level and used to assess cardiovascular risk.

[0061] Step 3: Build a machine learning optimization module. Step 3 includes the following steps: Step a: Construct feature engineering.

[0062] Step b: Use the mutual information method to screen the TOP30 significant features.

[0063] Step c: Perform high-dimensional feature visualization analysis through t-SNE.

[0064] Step d: Construct time series features (statistics of 10s windows).

[0065] Step a: In the sphygmomanometer intelligent management method, feature engineering is a key step to improve data quality and model performance. It combines the elements of blood pressure measurement and the management of blood pressure measurement data.

[0066] Furthermore, in step b, the mutual information method can be used to screen the TOP30 significant features.

[0067] Blood pressure measurement involves multiple physiological parameters and environmental factors, such as heart rate, blood oxygen saturation, body temperature, ambient temperature and humidity, etc. These parameters may have an impact on blood pressure values. Through feature selection, the most influential features can be identified, thereby improving the accuracy and efficiency of the model.

[0068] The collected blood pressure measurement data include but are not limited to: heart rate (HR), blood pressure value (SBP, DBP), ambient temperature and humidity, blood oxygen saturation (SpO2), etc.

[0069] Calculate mutual information value: For each feature and target variable (such as systolic blood pressure SBP or diastolic blood pressure DBP), calculate the mutual information value between them. The mutual information formula is as follows: Where I(X;Y) is the mutual information value, p(x, y) is the joint probability distribution, and p(x) and p(y) are the marginal probability distributions.

[0070] Sort features: Sort all features in descending order based on their mutual information values.

[0071] In step c, select TOP30 features: select the top 30 features with the highest mutual information values ​​as significant features. Perform high-dimensional feature visualization analysis through t-SNE.

[0072] When dealing with high-dimensional features, it is very difficult to directly understand and explain the relationship between these features. t-SNE can help map high-dimensional data into two-dimensional or three-dimensional space so that the distribution and clustering of the data can be observed intuitively.

[0073] For example, data preparation is required: using the TOP30 feature data set after mutual information screening.

[0074] Standardize the features to ensure that different features are on the same scale. Use the t-SNE algorithm to map high-dimensional features to two-dimensional or three-dimensional space. Draw the t-SNE result graph to observe the distribution and clustering of data points.

[0075] In step d, construct the time series features (statistics of 10s window): Blood pressure measurement is a dynamic process, and time series features can capture the changing trends and patterns of blood pressure over time. Constructing time series features helps improve the model's understanding of time dependency.

[0076] Furthermore, you can first define the window size: select a fixed time window (e.g. 10 seconds) to extract statistics. Sliding window: slide this window across the entire data set to extract features segment by segment. Statistics calculation: calculate multiple statistics for the data in each window, such as mean, standard deviation, maximum, minimum, slope, etc. Generate a new feature set: use the statistics of each window as a new feature to form a new feature set.

[0077] Through the above steps, the most useful information can be effectively extracted from complex blood pressure measurement data, improving the performance of subsequent data analysis and prediction models.

[0078] Step 4: Build a hybrid model architecture, including the following structure: Base layer: XGBoost regressor (learning rate 0.05, depth 6).

[0079] Enhancement layer: 1D-CNN processes the original waveform (3 layers of convolution + MaxPooling).

[0080] Fusion layer: attention mechanism weights multiple model outputs.

[0081] In the sphygmomanometer intelligent management method, the construction of a hybrid model architecture aims to combine the advantages of multiple machine learning and deep learning technologies to improve the accuracy and robustness of blood pressure prediction. The following is a specific explanation of the hybrid model architecture steps provided in an embodiment of the present invention, and a detailed explanation is given in combination with the elements of blood pressure measurement and the management of blood pressure measurement data.

[0082] The base layer of the hybrid model architecture: XGBoost regressor. XGBoost: is an efficient gradient boosting tree algorithm widely used in regression and classification tasks. Parameter settings: Learning Rate: set to 0.05, a smaller learning rate helps to adjust the model weights more finely. Max Depth: set to 6 to control the maximum depth of the tree to prevent overfitting. Feature input: Use the TOP30 significant features filtered by the mutual information method as input. For example, these features may include heart rate, ambient temperature and humidity, blood oxygen saturation, etc.

[0083] Training process: Input feature data into XGBoost regressor for training. The target variable is the target blood pressure value (such as systolic blood pressure SBP or diastolic blood pressure DBP). Output: XGBoost regressor outputs a preliminary blood pressure prediction result. Enhancement layer: 1D-CNN processes the original waveform.

[0084] 1D-CNN: One-dimensional convolutional neural network is suitable for time series data and signal processing. It can automatically extract local features and is insensitive to noise.

[0085] The specific steps include: using the original pulse waveform data as input. Data preprocessing includes normalization and window length division (e.g. 10-second window). Model structure: three layers of convolutional layers (Convolutional Layers): extract features at different levels. Pooling layer (MaxPooling Layers): reduce the spatial dimension of the feature map and retain important features. Fully connected layer (Fully Connected Layers): output the final blood pressure prediction result. Training process: input the original waveform data, and perform feature extraction and prediction through CNN. The target variable is the target blood pressure value (e.g. systolic blood pressure SBP or diastolic blood pressure DBP). Output: 1D-CNN outputs another preliminary blood pressure prediction result. Fusion layer: attention mechanism weighted multi-model output. Among them, the attention mechanism refers to the importance weight used to assign different model outputs, combining the advantages of multiple models. It can improve the overall performance of the model and enhance the generalization ability.

[0086] Through the above steps, the embodiment of the present invention constructs a hybrid model architecture that combines the advantages of XGBoost regressor and 1D-CNN: that is, the base layer (XGBoost): uses the TOP30 significant features screened by the mutual information method for training. Provides preliminary predictions based on statistical features. Enhancement layer (1D-CNN): processes the original pulse waveform data. Automatically extracts time and space features to improve the robustness of the model. Fusion layer (attention mechanism): combines the prediction results of XGBoost and 1D-CNN. Use the attention mechanism to dynamically assign weights to optimize the final prediction effect. This hybrid model architecture can make full use of different types of data and features to improve the accuracy and reliability of blood pressure prediction.

[0087] Step 5: Build a dynamic optimization system. This includes the following steps: Step f: Establish an online learning mechanism.

[0088] Step g: Design a sliding window (1000 samples) for incremental updates.

[0089] Step h: Adopt elastic weight consolidation (EWC) to prevent catastrophic forgetting.

[0090] Step k: Set up the model drift detection module (KL divergence monitoring).

[0091] In the sphygmomanometer intelligent management method, the dynamic optimization system aims to ensure the continuous learning ability of the model, prevent the forgetting of old knowledge, detect model drift, and improve the efficiency and accuracy of the model through a series of mechanisms. The following is a specific explanation of the steps of the dynamic optimization system provided, and a detailed explanation is given in combination with the elements of blood pressure measurement and the management of blood pressure measurement data.

[0092] Online learning: The model can be updated in real time when new data is received, without retraining the entire model. The online learning model includes: Incremental update: Design a sliding window (1000 samples) for incremental update. Whenever new data arrives, update the data in the sliding window and use this new data to fine-tune the model. Data management: Collect blood pressure measurement data (such as heart rate, blood pressure value, ambient temperature and humidity, etc.) in real time. Use a sliding window mechanism to ensure that the model is always trained based on the latest data.

[0093] Use Elastic Weight Consolidation (EWC) to prevent catastrophic forgetting. EWC (Elastic WeightConsolidation): A regularization technique used to prevent the model from forgetting previously learned knowledge when learning a new task. It can keep the model's performance stable on multiple tasks. The specific steps include: 1. Calculate the importance matrix: Calculate the importance of each parameter, that is, the impact of the parameter on the loss of the previous task. 2. Add regularization terms: Add regularization terms to the loss function to limit the change of parameters to protect important parameters. 3. Model update: Consider both new data and previous regularization terms each time the model is updated.

[0094] Set up the model drift detection module (KL divergence monitoring). Monitor the changes in model performance over time and detect changes in data distribution in a timely manner. KL divergence: used to measure the difference between two probability distributions. It can detect model performance degradation early and take measures to adjust the model. The specific steps include: 1. Calculate KL divergence: compare the probability distribution predicted by the current model with the probability distribution predicted by the historical model. 2. Threshold setting: set a threshold, and when the KL divergence exceeds the threshold, an alarm is triggered. 3. Model adjustment: retrain or adjust the model based on the alarm results.

[0095] Through the above steps, the embodiment of the present invention constructs a dynamic optimization system to ensure the continuous improvement and stability of the blood pressure monitor intelligent management system: through the online learning mechanism: use the sliding window for incremental updates to ensure that the model is always trained based on the latest data. Improve the adaptability and flexibility of the model. Elastic Weight Consolidation (EWC): Prevent the model from forgetting old knowledge when learning new data. Keep the performance of the model stable on multiple tasks. Through the model drift detection module (KL divergence monitoring): the changes in model performance over time can be monitored, and changes in data distribution can be discovered in time. Model performance degradation can be discovered early, and measures can be taken to adjust the model. This dynamic optimization system can ensure that the blood pressure monitor intelligent management system maintains high performance and reliability in a constantly changing environment.

[0096] In an optional implementation, it also includes performing an alarm process when the target blood pressure data exceeds a preset blood pressure value.

[0097] On the other hand, a sphygmomanometer intelligent management system is provided, the system comprising: A deviation acquisition unit, used to acquire a blood pressure pressure reference deviation, an ambient temperature and humidity deviation, and a blood pressure measurement channel hardware deviation; A comparison and analysis unit is used to analyze and compare the blood pressure pressure reference with the calibration result of the last blood pressure pressure reference to obtain a first analysis and comparison result; analyze and compare the ambient temperature and humidity with the calibration result of the last ambient temperature and humidity to obtain a second analysis and comparison result; analyze and compare the blood pressure measurement channel hardware with the calibration result of the last blood pressure measurement channel hardware to obtain a third analysis and comparison result; A blood pressure data acquisition unit, used to collect and process the user's blood pressure data when the first analysis and comparison result, the second analysis and comparison result, and the third analysis and comparison result are all normal, to obtain the blood pressure data of the measurement stage; The target blood pressure data acquisition unit is used to calculate the optimal estimate value of the blood pressure data in the measurement phase based on the preset analysis model using an adaptive Kalman filter algorithm; compare the optimal estimate value with the preset standard value to obtain a comparison deviation; and compensate the blood pressure data in the measurement phase based on the comparison deviation to obtain the target blood pressure data.

[0098] The blood pressure monitor intelligent management system device of the present invention starts the automatic calibration deviation program after being turned on, including collecting the pressure reference deviation, ambient temperature and humidity deviation and blood pressure measurement channel hardware deviation through the deviation acquisition unit, and comparing and analyzing these deviation data with the last calibration results based on the comparison and analysis unit, analyzing and optimizing them, storing them in the instrument main control MCU flash, and participating in subsequent blood pressure measurement calculations.

[0099] When the user starts the blood pressure measurement operation, the cuff starts to deflate after it is inflated to the set pressure. The blood pressure collection channel provides the blood pressure data acquisition unit to collect pressure data in real time during the deflation process, and sends the collection result calculation value to the cloud AI processing center (target blood pressure data acquisition unit) through the wireless communication module within a unit time. The cloud AI processing center calculates the accurate blood pressure value based on the established AI model algorithm and the received data, and transmits it back to the blood pressure monitor for display, and stores the data in the cloud database.

[0100] The hardware circuit mainly consists of U1 main control MCU and peripheral circuits (such as Figure 3 )、Pressure sensor interface circuit (such as Figure 4 )、LCD liquid crystal display interface circuit (such as Figure 5 ), power management circuits (such as Figure 6 ), wireless communication circuits (such as Figure 7 )、Environmental sensor acquisition circuit (such as Figure 8 ) is composed of MCU software flow (such as Figure 2 ) covers system initialization, automatic calibration deviation, blood pressure measurement data collection and transmission, cloud data reception and display, etc., to ensure stable system operation and accurate data processing.

[0101] In yet another aspect, a readable storage medium is provided, on which the above-mentioned method is stored.

[0102] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for intelligent management of a blood pressure monitor, characterized in that: include: Obtain blood pressure baseline deviation, ambient temperature and humidity deviation, and blood pressure measurement channel hardware deviation; Analyze and compare the blood pressure pressure reference with the calibration result of the last blood pressure pressure reference to obtain a first analysis and comparison result; analyze and compare the ambient temperature and humidity with the calibration result of the last ambient temperature and humidity to obtain a second analysis and comparison result; analyze and compare the blood pressure measurement channel hardware with the calibration result of the last blood pressure measurement channel hardware to obtain a third analysis and comparison result; When the first analysis and comparison result, the second analysis and comparison result, and the third analysis and comparison result are all normal, the user's blood pressure data is collected and processed to obtain the blood pressure data of the measurement phase; Calculating an optimal estimate of the blood pressure data of the measurement phase using an adaptive Kalman filter algorithm based on a preset analysis model; Compare and calculate the optimal estimated value with a preset standard value to obtain a comparison deviation; The blood pressure data of the measurement phase is compensated based on the comparison deviation to obtain target blood pressure data.

2. The method for intelligent management of blood pressure monitor according to claim 1, characterized in that: The method also includes filtering the digital signal converted by the ADC using a digital filter to remove noise interference.

3. The method for intelligent management of blood pressure monitor according to claim 1, characterized in that: It also includes performing linear or nonlinear correction on the pressure signal in the collected user blood pressure data according to the sensor characteristic curve to ensure the accuracy of the user blood pressure data.

4. The method for intelligent management of blood pressure monitor according to claim 1, characterized in that: It also includes obtaining a change in the ambient pressure during the blood pressure measurement of the user, and controlling the inflation pressure during the blood pressure measurement based on the ambient pressure change.

5. The method for intelligent management of blood pressure monitor according to claim 1, characterized in that: The method also includes obtaining physiological characteristics of the user, and controlling the deflation speed during blood pressure measurement based on the physiological characteristics of the user.

6. The method for intelligent management of blood pressure monitor according to claim 1, characterized in that: It also includes real-time collection of pressure change data in the cuff, filtering the digital signal converted by the ADC through a digital filter, and calculating the systolic and diastolic pressures using the oscillation method principle.

7. The method for intelligent management of blood pressure monitor according to claim 1, characterized in that: It also includes transmitting the target blood pressure data back to a cloud processing center for the cloud processing center to train the preset analysis model.

8. The method for intelligent management of blood pressure monitor according to claim 1, characterized in that: It also includes alarm processing when the target blood pressure data exceeds a preset blood pressure value.

9. A blood pressure monitor intelligent management system, characterized in that: The system comprises: A deviation acquisition unit, used to acquire a blood pressure pressure reference deviation, an ambient temperature and humidity deviation, and a blood pressure measurement channel hardware deviation; A comparison and analysis unit, configured to analyze and compare the blood pressure pressure reference with the calibration result of the last blood pressure pressure reference to obtain a first analysis and comparison result; analyze and compare the ambient temperature and humidity with the calibration result of the last ambient temperature and humidity to obtain a second analysis and comparison result; analyze and compare the blood pressure measurement channel hardware with the calibration result of the last blood pressure measurement channel hardware to obtain a third analysis and comparison result; A blood pressure data acquisition unit, used to collect and process the user's blood pressure data when the first analysis and comparison result, the second analysis and comparison result, and the third analysis and comparison result are all normal, to obtain the blood pressure data of the measurement stage; The target blood pressure data acquisition unit is used to calculate the optimal estimate value of the blood pressure data of the measurement phase based on a preset analysis model using an adaptive Kalman filter algorithm; compare the optimal estimate value with a preset standard value to obtain a comparison deviation; and compensate the blood pressure data of the measurement phase based on the comparison deviation to obtain target blood pressure data.

10. A readable storage medium, characterized in that: The readable storage medium stores the method according to any one of claims 1 to 7.

Images