Continuous blood pressure measurement device, system and method with data quality assessment functionality

Abstract

The application provides a continuous blood pressure measuring instrument, system and method with data quality evaluation function, wherein an electrocardio pulse wave acquisition module is used to collect electrocardio signals ECG and pulse wave signals PPG in real time; a data quality evaluation module is used to evaluate the quality of the electrocardio signals and the pulse wave signals, and to screen out signal data with qualified quality; a blood pressure prediction module is used to combine an Unet network and a self-attention mechanism network to form a deep learning model, and to use the signal data with qualified quality as the input of the deep learning model so as to output a blood pressure prediction value for display. The application combines the PPG features of deep learning with the ECG related features selected manually to improve the prediction accuracy; the new deep learning network combining the Unet and the self-attention mechanism can improve the prediction accuracy. The evaluation mechanism is used to evaluate the quality of the signals in the period, and the high-quality signals are selected to improve the prediction accuracy in the training and prediction process.

Description

Technical Field

[0001] This invention belongs to the field of medical device technology, and specifically relates to a continuous blood pressure measuring instrument, system and method with data quality assessment function. Background Technology

[0002] Existing blood pressure measurement methods can be divided into non-invasive and invasive methods. Common non-invasive methods include the Korotkoff sound method and the oscillometric method. The Korotkoff sound method requires a trained physician to perform the measurement, while the oscillometric method is the mainstream method commonly used in electronic blood pressure monitors. However, both methods require the cuff to be fixed to the patient's upper arm and inflated for measurement, making accurate long-term continuous blood pressure monitoring impossible. Current continuous blood pressure monitoring is an invasive, direct measurement method. Due to its invasiveness, the measurement is inconvenient and mainly used in intensive care units, limiting its application scenarios.

[0003] There is currently a considerable body of research on non-invasive continuous blood pressure measurement. Measurement methods mainly fall into two categories: pneumatic cuff inflation and pulse wave (PPG) measurement. Algorithms primarily include traditional feature fitting and deep learning methods. For example, Chinese patent CN108742574A discloses a non-invasive continuous blood pressure measuring device. It collects brachial artery blood pressure data via a cuff module and finger blood pressure data via a continuous blood pressure module. The brachial artery blood pressure data, converted from the finger blood pressure data, is then adjusted using the brachial artery blood pressure data, and the adjusted continuous blood pressure data is displayed on a display module. This allows for continuous blood pressure data and enables long-term blood pressure measurement. Essentially, this patent still uses a pneumatic cuff inflation method for measurement.

[0004] For example, Chinese patent CN113456043A mainly includes the following steps: obtaining the original data of the target object; preprocessing the original data to obtain the current data; extracting feature parameters from the current data; constructing a continuous blood pressure prediction model based on several feature parameters; and obtaining a multi-parameter fusion blood pressure prediction model for continuous blood pressure monitoring. The prediction model of this invention has the technical advantages of high prediction accuracy and small error, thus effectively preventing the occurrence of hypertension and its complications. Essentially, this patent is a method for fitting PPG waveform features.

[0005] For example, Chinese patent CN115349836A includes the following steps: First, the input signal for continuous blood pressure prediction based on PPG is segmented and sampled; second, the parameterization of the input signal for continuous blood pressure prediction based on PPG is evaluated, and unsuitable signals are removed; then, the cleaned PPG signal data is used as input to the ResNet model; subsequently, the model predicts systolic blood pressure (SBP) and diastolic blood pressure (DBP); finally, the trained model is fine-tuned using 20% ​​of the test data, thereby further improving the accuracy of the model's blood pressure prediction. Essentially, this patent uses deep learning methods to analyze PPG signals and predict blood pressure.

[0006] Existing measurement instruments mostly rely on traditional manual feature extraction for regression analysis or deep learning to build pulse wave models. The quality of extracted features is affected by the wearing posture of the pulse wave (PPG), further reducing measurement accuracy. Furthermore, existing instruments use basic filtering processes in their preprocessing, but signal quality varies at different times, leading to significant differences in measurement results and impacting accuracy. Summary of the Invention

[0007] To address the aforementioned problems in the prior art, this invention provides a continuous blood pressure monitor, system, and method with data quality assessment capabilities. The technical problem to be solved by this invention is achieved through the following technical solution:

[0008] In a first aspect, the present invention provides a continuous blood pressure measuring instrument with data quality assessment function, comprising: an electrocardiogram pulse wave acquisition module, a data quality assessment module, and a blood pressure prediction module;

[0009] The ECG and PPG acquisition module is used to acquire ECG and PPG signals in real time.

[0010] The data quality assessment module is used to assess the quality of the electrocardiogram signal and pulse wave signal, and to filter out signal data that meets the quality requirements.

[0011] The blood pressure prediction module is used to combine a Unet network and a self-attention mechanism network to form a deep learning model, and to use qualified signal data as input to the deep learning model to output a blood pressure prediction value; and to display the blood pressure prediction value.

[0012] Optionally, the ECG pulse acquisition module includes a three-lead electrode patch, a photoelectric sensor, a signal amplifier, an analog-to-digital converter, a microcontroller MCU, and a Bluetooth module;

[0013] The three-lead electrode patch collects human electrocardiogram (ECG) signals in real time, and the photoelectric sensor collects pulse wave signals in real time. The ECG signals and pulse wave signals are amplified by a signal amplifier, converted from analog to digital by an analog-to-digital converter, and then sent to the microcontroller MCU.

[0014] The microcontroller MCU transmits the processed signal to the data quality assessment module via Bluetooth.

[0015] Optionally, both the data quality assessment module and the blood pressure prediction module are located on a mobile device, and the data quality assessment module includes a filter and a data quality assessment unit.

[0016] The filter is used to perform comprehensive filtering on the input signal data. The comprehensive filtering includes EMD filtering, Chebyshev low-pass filtering, and notch band-stop filtering.

[0017] The data quality assessment unit is used to perform frequency domain quality analysis and time domain quality analysis on the input PPG and ECG signals in sequence to filter out signal data with qualified quality.

[0018] Optionally, the data quality assessment unit is used for:

[0019] The signal spectrum is divided into three regions;

[0020] Perform spectrum analysis on the input PPG and ECG signals respectively to confirm whether the signal spectrum of the PCG and ECG signals is mainly located in the mid-frequency band. If the peak of the signal spectrum is not located in the mid-frequency band or there are multiple peaks in other frequency bands, it is determined that the signal data does not meet the quality standard.

[0021] The signal data is divided into single-cycle data;

[0022] Using the dynamic time warping algorithm, the similarity analysis is performed between the single-cycle data and the template signal to measure the similarity between the signal segments of the single-cycle data and the template signal, and the matching cost is obtained.

[0023] Signal data whose matching cost satisfies the 3σ criterion within the numerical range are identified as qualified signal data.

[0024] Unqualified signal data is fed back to the user through the blood pressure prediction module to remind the user to make adjustments.

[0025] The matching cost is:

[0026] d ij =|x i -y j |

[0027] D(i,j)=d ij+min{D(i,j-1),(i-1,j-1),D(i-1,j)}

[0028] D = ∑D(i,j)

[0029] Where, x i For the actual signal sampling points, y j Let D(i,j) be the template signal sampling point, D(i,j) be the matching distance between the current actual signal sampling point and the template signal sampling point, and D be the total cost of the current matching path. Dynamic programming can be used to calculate the optimal matching path D. min It is used to evaluate the similarity between the actual signal and the template signal.

[0030] Optionally, the data quality assessment unit is further configured to:

[0031] Save the start time of single-cycle PPG and ECG signals, as well as the peak value of the ECG R peak;

[0032] PPT on calculating the pulse transit time of ECG signals.

[0033] Optionally, the blood pressure prediction module includes: a blood pressure correction unit and a prediction network unit; the prediction network unit includes a Unet network and a self-attention network, and the blood pressure correction unit is used to correct the Unet network and the self-attention network to improve the accuracy of prediction;

[0034] The Unet network has a completely symmetrical structure, consisting of a contraction path and an expansion path. The expansion path amplifies the features through upsampling and fuses them with the features in the contraction path through cross-layer connections, thereby fully combining shallow and deep features. The input of the Unet network is a single-cycle PPG signal, and the output is the features extracted by the network from the PPG.

[0035] The self-attention network fuses the features output by the Unet network with the PPT features, adds position encoding to the fused feature vector to reflect the information of the relative and absolute time step sequence, and obtains the corresponding continuous blood pressure prediction value through an attention module composed of a single-head self-attention module and a feedforward network, and finally obtains the corresponding continuous blood pressure prediction value through dense interpolation.

[0036] Secondly, the present invention provides a continuous blood pressure measurement system with data quality assessment function, including an electrocardiogram pulse wave acquisition module and a mobile terminal. The mobile terminal is equipped with a data quality assessment module and a blood pressure prediction module. The electrocardiogram pulse wave acquisition module and the mobile terminal communicate wirelessly or via wired mode.

[0037] The ECG and PPG acquisition module is used to acquire ECG and PPG signals in real time.

[0038] The data quality assessment module is used to assess the quality of the electrocardiogram signal and pulse wave signal, and to filter out signal data that meets the quality requirements.

[0039] The blood pressure prediction module is used to combine a Unet network and a self-attention mechanism network to form a deep learning model, and to use qualified signal data as input to the deep learning model to output a blood pressure prediction value; and to display the blood pressure prediction value.

[0040] Thirdly, the present invention provides a continuous blood pressure measurement method with data quality assessment function, which is applied to a second continuous blood pressure measurement system with data quality assessment function, and the measurement is completed according to the measurement process of each module of the first aspect.

[0041] Fourthly, the present invention provides a mobile terminal that performs the measurement process of the mobile terminal in the second aspect.

[0042] This invention provides a continuous blood pressure measurement device, system, and method with data quality assessment functionality. The device acquires ECG and PPG signals in real time via an ECG and PPG acquisition module. A data quality assessment module evaluates the quality of the ECG and PPG signals, selecting qualified signal data. A blood pressure prediction module combines a Unet network and a self-attention mechanism network to form a deep learning model, using the qualified signal data as input to output a predicted blood pressure value for display. This invention improves prediction accuracy by combining PPG features from deep learning with manually selected ECG-related features; the use of a novel deep learning network combining Unet and a self-attention mechanism further enhances prediction accuracy. An evaluation mechanism assesses the quality of signals over a given time period, selecting high-quality signals for use in training and prediction processes to improve prediction accuracy.

[0043] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of a continuous blood pressure measuring instrument with data quality assessment function provided by the present invention;

[0045] Figure 2 This is a structural block diagram of the electrocardiogram pulse wave acquisition module provided by the present invention;

[0046] Figure 3 This is a schematic diagram of the evaluation process of the data quality assessment module provided by the present invention;

[0047] Figure 4 This is a schematic diagram of the deep learning model provided by the present invention. Detailed Implementation

[0048] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0049] refer to Figure 1 The present invention provides a continuous blood pressure measuring device with data quality assessment function, including: an electrocardiogram pulse wave acquisition module, a data quality assessment module, and a blood pressure prediction module; the data quality assessment module and the blood pressure prediction module are both located on a mobile terminal, and the electrocardiogram pulse wave acquisition module can communicate with the mobile terminal via Bluetooth, or via wired or other wireless means.

[0050] The ECG and PPG acquisition module is used to acquire ECG and PPG signals in real time.

[0051] The data quality assessment module is used to assess the quality of the electrocardiogram signal and pulse wave signal, and to filter out signal data that meets the quality requirements.

[0052] The blood pressure prediction module is used to combine a Unet network and a self-attention mechanism network to form a deep learning model, and to use qualified signal data as input to the deep learning model to output a blood pressure prediction value; and to display the blood pressure prediction value.

[0053] refer to Figure 2 As shown, in one specific embodiment, the ECG pulse acquisition module includes a three-lead electrode patch, a photoelectric sensor, a signal amplifier, an analog-to-digital converter, a microcontroller MCU, and a Bluetooth module;

[0054] The three-lead electrode patch collects human electrocardiogram (ECG) signals in real time, and the photoelectric sensor collects pulse wave signals in real time. The ECG signals and pulse wave signals are amplified by a signal amplifier, converted from analog to digital by an analog-to-digital converter, and then sent to the microcontroller MCU.

[0055] The microcontroller MCU transmits the processed signal to the data quality assessment module via Bluetooth.

[0056] It is worth noting that the principle of ECG signal acquisition is as follows:

[0057] Electrocardiogram (ECG) signals are important physiological signals in the human body. They are generated by the potential difference on the skin surface due to the heartbeat and exhibit a periodic variation trend. Essentially, an ECG signal is a point signal on the body surface. A three-lead electrode patch can be attached to the skin surface, and the ECG signal can be acquired through analog-to-digital conversion. The right upper lead is located in the first intercostal space along the midclavicular line on the right side of the sternum; the left lower lead is located at the level of the xiphoid process on the left side of the midclavicular line; and the left upper lead is located in the first intercostal space along the midclavicular line on the left side of the sternum. The obtained ECG signal has a relatively large amplitude and strong anti-interference properties.

[0058] It is worth noting that the principle of pulse wave signal acquisition is as follows:

[0059] A pulse wave is formed when the heart rhythmically ejects blood into the arteries during each cardiac cycle, causing periodic changes in arterial pressure and consequently, periodic changes in the degree of vascular filling. These periodic changes propagate to peripheral blood vessels, thus forming a pulse wave.

[0060] In practical applications, pulse waves can be detected using photoplethysmography (PPG), a technique based on Beer-Lambert's law that collects pulse waves through photoelectric conversion. Common detection sites are at the extremities of the body, such as the fingertips. When light of a specific wavelength shines on skin tissue and is reflected back to the photodetector, the intensity of the light is attenuated. When light shines on the human body, non-blood components such as bones, muscles, veins, and other connecting tissues absorb light at a constant rate. However, for arteries, which contain flowing blood, the degree of light absorption varies. Therefore, we denote the constant signal absorbed by other tissues and fed back to the photodetector as a direct current (DC) signal, and the changing signal absorbed by arterial blood and fed back as an alternating current (AC) signal. This AC signal effectively reflects the characteristics of blood flow; this technique is called photoplethysmography.

[0061] After the three-lead electrode patch and photoelectric sensor collect the signal, they are filtered, amplified, and converted from analog to digital before being transmitted to the MCU processor. The MCU transmits the ECG pulse wave signal to the mobile device via Bluetooth in the data transmission unit. The mobile device processes the ECG pulse wave signal data accordingly and estimates the user's blood pressure using a deployed deep learning network.

[0062] In one specific embodiment, the data quality assessment module includes a filter and a data quality assessment unit;

[0063] The filter is used to perform comprehensive filtering on the input signal data. The comprehensive filtering includes EMD filtering, Chebyshev low-pass filtering, and notch band-stop filtering.

[0064] It's worth noting that after the mobile device receives ECG and pulse wave data, the ECG and PPG data are first filtered due to various factors affecting their acquisition. This filtering removes common power frequency interference and baseline drift. Commonly used filters include EMD filters, Chebyshev low-pass filters, and notch band-stop filters. EMD filters are mainly used to remove baseline drift from the original signal; low-pass filters smooth the signal profile and suppress high-frequency noise; and notch band-stop filters are typically used to eliminate power frequency interference around 50Hz.

[0065] The data quality assessment unit is used to perform frequency domain quality analysis and time domain quality analysis on the input PPG and ECG signals in sequence to filter out signal data with qualified quality.

[0066] Deep learning network prediction models primarily rely on the quality of PPG and ECG signals. To ensure the accuracy of training and prediction models and improve their prediction precision, quality checks are performed on ECG and pulse wave data after preprocessing. If the data quality fails to meet standards for an extended period, the user will be prompted via the mobile device display to adjust the measurement method. Once the data quality meets the standards, subsequent blood pressure calibration or prediction will be performed.

[0067] In one specific embodiment, reference is made to... Figure 3 The data quality assessment unit is used for:

[0068] The signal spectrum is divided into three regions;

[0069] Perform spectrum analysis on the input PPG and ECG signals respectively to confirm whether the signal spectrum of the PCG and ECG signals is mainly located in the mid-frequency band. If the peak of the signal spectrum is not located in the mid-frequency band or there are multiple peaks in other frequency bands, it is determined that the signal data does not meet the quality standard.

[0070] The signal data is divided into single-cycle data;

[0071] Using the dynamic time warping algorithm, the similarity analysis is performed between the single-cycle data and the template signal to measure the similarity between the signal segments of the single-cycle data and the template signal, and the matching cost is obtained.

[0072] Signal data whose matching cost satisfies the 3σ criterion within the numerical range are identified as qualified signal data.

[0073] Unqualified signal data is fed back to the user through the blood pressure prediction module to remind the user to make adjustments.

[0074] Save the start time of single-cycle PPG and ECG signals, as well as the peak value of the ECG R peak;

[0075] PPT on calculating the pulse transit time of ECG signals.

[0076] It is worth noting that: the data quality assessment flowchart is as follows Figure 3 As shown, the frequency domain quality analysis of the PPG signal is as follows:

[0077] PPG signal quality is primarily evaluated in both the frequency and time domains. First, a spectral analysis of the input signal is performed, dividing the signal spectrum into three main regions: a low-frequency band (below 0.5 Hz), mainly containing random baseline drift and segments influenced by low-frequency respiratory signals; a mid-frequency band (0.5-3 Hz), primarily containing the PPG signal; and a high-frequency band (above 3 Hz), mainly containing unfiltered high-frequency noise. After spectral analysis, quality is assessed by analyzing whether the signal frequency distribution is primarily concentrated in the mid-frequency band. If spectral peaks are not present in the mid-frequency band, or if multiple peaks exist in other frequency bands, the current PPG signal quality is considered substandard, and the user is prompted to make adjustments.

[0078] The time-domain quality analysis of the PPG signal is as follows:

[0079] After performing spectral quality testing on the PPG signal, the PPG signal is analyzed in the time domain. The first step is to periodically segment the entire PPG signal, requiring the detection of the starting point of each single-cycle PPG signal. This invention uses an adaptive threshold detection algorithm based on PPG waveform characteristics, mainly comprising three parts: threshold setting, nonlinear transformation, and decision rules. The threshold setting mainly includes the average value of the current signal amplitude, the starting amplitude of the single-cycle PPG signal, the period length of the single-cycle PPG signal, and the absolute values ​​of the slopes at the starting and ending points of the single-cycle PPG signal. The nonlinear transformation first performs median filtering to smooth the signal, then applies nonlinear processing such as squaring the signal amplitude to make the signal starting point characteristics more obvious. The decision rules first use a difference algorithm to calculate the minimum value of the waveform, and then determine whether the current minimum value is the starting point of the PPG signal according to the rules in the threshold setting section. If it is the starting point, the threshold is updated to achieve an adaptive effect.

[0080] After segmenting the entire signal into single-period PPG signals, since the lengths of the single-period PPG signals are not entirely the same, a Dynamic Time Warping (DTW) algorithm is used to perform similarity analysis with a template signal. The template is generated by manually selecting high-quality single-period signals. The similarity between the signal segment and the template signal is measured by calculating the matching cost between the current signal and the template signal. The matching cost is:

[0081] d ij =|x i -yj|

[0082] D(i,j)=d ij +min{D(i,j-1),D(i-1,j-1),D(i-1,j)}

[0083] D = ∑D(i,j)

[0084] Where, x iFor the actual signal sampling points, y j Let D(m,j) be the template signal sampling point, D(m,j) be the matching distance between the current actual signal sampling point and the template signal sampling point, and D be the total cost of the current matching path. Dynamic programming can be used to calculate the optimal matching path D. min It is used to evaluate the similarity between the actual signal and the template signal.

[0085] ECG exhibits similar characteristics within a single cycle, and the above evaluation method is also applicable to ECG signals. In the time-domain analysis process, the single-cycle PPG, ECG start time, and ECG R peak-to-peak time are saved. The pulse transit time (PTT) is calculated based on the corresponding characteristics for subsequent deep network feature fusion.

[0086] In one specific embodiment, reference is made to... Figure 4 The blood pressure prediction module includes a blood pressure correction unit and a prediction network unit; the prediction network unit includes a Unet network and a self-attention network, and the blood pressure correction unit is used to correct the Unet network and the self-attention network to improve the accuracy of prediction;

[0087] The Unet network has a completely symmetrical structure, consisting of a contraction path and an expansion path. The expansion path amplifies the features through upsampling and fuses them with the features in the contraction path through cross-layer connections, thereby fully combining shallow and deep features. The input of the Unet network is a single-cycle PPG signal, and the output is the features extracted by the network from the PPG.

[0088] The self-attention network fuses the features output by the Unet network with the PPT features, adds position encoding to the fused feature vector to reflect the information of the relative and absolute time step sequence, and obtains the corresponding continuous blood pressure prediction value through an attention module composed of a single-head self-attention module and a feedforward network, and finally obtains the corresponding continuous blood pressure prediction value through dense interpolation.

[0089] If it's the first measurement, the blood pressure calibration unit calibrates the model to improve prediction accuracy. If it has already been calibrated, the data is fed into the deep learning model for calculation, and the results are displayed on the mobile screen for user reference. The specific blood pressure calibration process is as follows: During the first measurement, the network learning rate is reduced. Simultaneously, the user can input the blood pressure value measured using an external blood pressure monitor during the first measurement to adjust the network parameters, further improving accuracy.

[0090] The measuring instrument of this invention uses a deep learning model combining Unet and a self-attention mechanism. It also incorporates ECG-related features such as PTT (Power Tolerance) into the network for information fusion to improve prediction accuracy. The Unet network structure, as shown in the figure, is a completely symmetrical U-shaped structure consisting of a contraction path and an expansion path. The expansion path amplifies features through upsampling and fuses them with features from the contraction path through cross-layer connections, fully combining shallow and deep features. The Unet network input is a single-cycle PPG signal, and the output is the features extracted from the PPG signal.

[0091] The output of the Unet network is fused with ECG-related features such as PTT (Patient Time Tolerance). The combined features are then input to a self-attention mechanism network, with positional encoding added to the input vector to reflect information about the relative and absolute time step sequences. This is followed by an attention module consisting of a single-head self-attention module and a feedforward network. Finally, dense interpolation and remaining connections are used to obtain the corresponding continuous blood pressure prediction values. After training, the model is deployed to a mobile device, and the predicted values ​​are displayed in real-time on the interface during measurement.

[0092] This invention provides a continuous blood pressure measurement system with data quality assessment function, including an electrocardiogram pulse wave acquisition module and a mobile terminal. The mobile terminal is equipped with a data quality assessment module and a blood pressure prediction module. The electrocardiogram pulse wave acquisition module and the mobile terminal communicate wirelessly or via wired mode.

[0093] The ECG and PPG acquisition module is used to acquire ECG and PPG signals in real time.

[0094] The data quality assessment module is used to assess the quality of the electrocardiogram signal and pulse wave signal, and to filter out signal data that meets the quality requirements.

[0095] The blood pressure prediction module is used to combine a Unet network and a self-attention mechanism network to form a deep learning model, and to use qualified signal data as input to the deep learning model to output a blood pressure prediction value; and to display the blood pressure prediction value.

[0096] This invention provides a continuous blood pressure measurement method with data quality assessment function, which is applied to a continuous blood pressure measurement system with data quality assessment function. The method is characterized in that the measurement process of each module completes the measurement.

[0097] This invention provides a mobile device that executes the measurement process of a continuous blood pressure measurement system with data quality assessment function.

[0098] This invention provides a continuous blood pressure measurement device, system, and method with data quality assessment function. The specific process is as follows: Electrocardiogram (ECG) signals are acquired in real time via a three-lead electrode patch, and pulse wave signals are acquired in real time via a photoelectric sensor. After processing, the ECG and pulse wave signals are transmitted to a mobile application via a Bluetooth signal transmission unit. In the mobile application, the data quality is first assessed. If the data for the current time period meets the quality standards, the data is passed to subsequent processes for prediction. If the data does not meet the standards for an extended period, the user is prompted via the mobile device to make adjustments to meet the standards. If the user is using the system for blood pressure measurement for the first time, calibration can be performed through the blood pressure model calibration unit, using transfer learning on a pre-set deep learning model to improve prediction accuracy. If calibration has already been performed, after quality detection of the ECG and pulse wave signal input data, signals meeting the standards are passed to the blood pressure model prediction unit. Real-time blood pressure prediction is then performed using a deployed deep learning model. The deep learning model combines PPG and ECG features to improve prediction performance, and the results are displayed in real time on the mobile device interface, achieving non-invasive continuous blood pressure monitoring.

[0099] Although this application has been described herein in conjunction with various embodiments, those skilled in the art will understand and implement other variations of the disclosed embodiments by reviewing the accompanying drawings, the disclosure, and the appended claims in carrying out the claimed application. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude a plurality.

[0100] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A continuous blood pressure measuring apparatus having a data quality evaluation function, characterized by comprising: The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device.

2. The continuous blood pressure measuring apparatus with data quality evaluation function according to claim 1, characterized in that, The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device.

3. The continuous blood pressure measuring apparatus with data quality evaluation function according to claim 1, wherein, The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device.

4. The continuous blood pressure measuring apparatus with data quality evaluation function according to claim 3, characterized in that, The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction method and device. The application relates to a blood pressure prediction The single-cycle data is subjected to similarity analysis calculation with the template signal by using a dynamic time warping algorithm, to measure the similarity between the signal segment of the single-cycle data and the template signal, and to obtain a matching cost; Signal data in a numerical range corresponding to the matching cost satisfying the 3σ criterion is determined as qualified signal data; The unqualified signal data is fed back to the user through the blood pressure prediction module to remind the user to adjust.

5. The continuous blood pressure measuring apparatus with data quality evaluation function according to claim 4, characterized in that, The matching cost is: wherein, is the actual signal sample point, is the template signal sample point, is the matching distance of the current actual signal sample point and the template signal sample point, D is the total cost of the current matching path, and the optimal matching path can be calculated using dynamic programming is used to evaluate the similarity of the actual signal and the template signal.

6. The continuous blood pressure measuring apparatus with data quality evaluation function according to claim 4, wherein, The data quality assessment unit is further configured to: Save the start time of the single-cycle PPG and ECG signals, and the R peak time of the ECG; Calculate the pulse transit time (PPT) of the ECG signal.

7. A continuous blood pressure measurement system with data quality assessment functionality, characterized in that The system comprises an ECG pulse wave acquisition module and a mobile terminal, wherein the mobile terminal is provided with a data quality assessment module and a blood pressure prediction module, and the ECG pulse wave acquisition module and the mobile terminal communicate through wireless or wired mode; The ECG pulse wave acquisition module is configured to acquire ECG and PPG signals in real time; The data quality assessment module is configured to assess the quality of the ECG and PPG signals and screen out qualified signal data; The blood pressure prediction module is configured to form a deep learning model by combining a Unet network and a self-attention mechanism network, input the qualified signal data into the deep learning model to output a blood pressure prediction value, and display the blood pressure prediction value. The blood pressure prediction module comprises a blood pressure correction unit and a prediction network unit, the prediction network unit comprises a Unet network and a self-attention network, and the blood pressure correction unit is configured to correct the Unet network and the self-attention network to improve the prediction accuracy. The Unet network is a complete symmetrical structure composed of a contraction path and an expansion path, the expansion path enlarges the features through up-sampling and fuses the features with those in the contraction path through cross-layer connection, so as to fully combine the shallow features and the deep features; the Unet network input is a single-cycle PPG signal, and the output is features extracted from the PPG by the network; The self-attention network fuses the features output by the Unet network with PPT features, adds position encoding to the fused feature vector to reflect the information of relative and absolute time step sequences, passes through an attention module composed of a single-head self-attention module and a feedforward network, and finally obtains the corresponding continuous blood pressure prediction value through dense interpolation.

8. A continuous blood pressure measurement method with data quality assessment function applied to the continuous blood pressure measurement system with data quality assessment function according to claim 7, characterized in that, The system comprises: acquiring ECG and PPG signals in real time; assessing the quality of the ECG and PPG signals and screening out qualified signal data; forming a deep learning model by combining a Unet network and a self-attention mechanism network, inputting the qualified signal data into the deep learning model to output a blood pressure prediction value, and displaying the blood pressure prediction value.

9. A mobile terminal, characterized in that The system performs the functions of the data quality assessment module and the blood pressure prediction module deployed in the mobile terminal of the continuous blood pressure measurement system with data quality assessment function according to claim 7.

Images