Cuff-free continuous blood pressure signal reconstruction method and system

Abstract

The invention provides a cuff-free continuous blood pressure signal reconstruction method and system, and relates to the technical field of blood pressure monitoring, and the method comprises the steps: obtaining a PPG signal and an ECG signal which are synchronized; respectively extracting a PPG feature vector and an ECG feature vector from the PPG signal and the ECG signal by using the trained attention sequential network, carrying out multi-modal fusion on the PPG feature vector and the ECG feature vector, and reconstructing a cuff-free continuous blood pressure signal based on the fused feature vectors; according to feature vector extraction, an information screening machine and a time sequence learning machine are additionally arranged between an encoder and a decoder and are respectively used for learning a channel probability distribution vector and a time sequence weight vector, guiding the decoder to decode high-level semantic features and decoding the sequential relation between PPG and ECG; the attention sequential network is provided on the basis of multi-modal fusion, and continuous blood pressure signal reconstruction is achieved by analyzing photoelectric volume pulse wave signals and electrocardiosignals.

Description

Technical Field

[0001] The present disclosure relates to the technical field of blood pressure monitoring, and in particular to a cuffless continuous blood pressure signal reconstruction method and system. Background Art

[0002] Hypertension is an important risk factor for stroke, cardiovascular and cerebrovascular diseases, and chronic kidney disease; due to the influence of the 'white coat' effect and behavioral factors, the deviation between discrete blood pressure (BP) measurements and the true value ranges from -24 mmHg to 33 mmHg. Most medical associations recommend the use of cuffless continuous BP measurement technology.

[0003] According to the photoplethysmographic signal (PPG) and the electrocardiogram (ECG), the pulse transmission time (PTT), pulse arrival time (PAT) and pulse wave velocity (PWV) are the common cuffless BP detection methods. The essence of these methods is to perform mathematical physical modeling (MPM) on the propagation process of blood between two given points to estimate instantaneous BP, which is affected by the distance between the test points, the thickness of the blood vessel wall and the blood density. Although the study of the pre-ejection period can assist PAT, the pre-ejection period can change suddenly due to external factors and aging.

[0004] Machine learning (ML) based on feature engineering is applied in the estimation of instantaneous BP values; the relationship between PPG and BP is nonlinear, so it is a common method to input multimodal features such as morphology, time domain, frequency domain and nonlinear domain into a nonlinear regressor; however, the parameter adjustment of ML's nonlinear regressor is difficult to overcome the specific differences among subjects, and the nonlinear analysis capability of long-range signals needs to be further improved.

[0005] Data-driven deep learning (DL) has more powerful nonlinear modeling capabilities. It has the advantage of being convenient to estimate BP by analyzing unimodal signals. However, unimodal signals are greatly interfered by physiological and environmental factors. DL requires diverse data as training sets and the mapping relationship between BP signals and other physiological signals is not clear. There is a risk of missing information when estimating BP using unimodal signals.

[0006] In order to solve the defects of unimodal signals in BP estimation, researchers use signals of different modes as the source domain of DL or adopt different neural networks to perform multimodal analysis on unimodal signals. Although the technology based on multimodal fusion can reduce the error of BP estimation, the above studies perform simple linear addition and subtraction or cascade operations on different features, and fail to effectively explore and utilize the nonlinear relationship of multimodality.

[0007] In summary, MPM and ML are affected by the variability of various physiological parameters, making it difficult to model long-range BP signals and analyze patient specificity; although multimodal fusion and DL technology have shown better performance than other methods, the above methods fail to reconstruct continuous BP signals, and face the game problem of insufficient complementary learning of individual information and difficulty in eliminating the redundancy of common information when processing multimodal data. Summary of the invention

[0008] In order to solve the above problems, the present invention proposes a cuffless continuous blood pressure signal reconstruction method and system, and proposes an attention timing network (AT-Net) based on multimodal fusion, which realizes the reconstruction of continuous BP signal by analyzing photoplethysmogram (PPG) signal and electrocardiogram (ECG) signal.

[0009] According to some embodiments, the present disclosure adopts the following technical solutions:

[0010] A cuff-free continuous blood pressure signal reconstruction method, comprising:

[0011] Acquire synchronized PPG and ECG signals and perform standardization on them;

[0012] Using the trained attention time series network, PPG feature vectors and ECG feature vectors are extracted from PPG signals and ECG signals respectively, and multimodal fusion is performed on the PPG feature vectors and ECG feature vectors. Based on the fused feature vectors, the cuff-free continuous blood pressure signal is reconstructed.

[0013] Among them, the feature vector is extracted by adding an information screening machine and a timing learning machine between the encoder and the decoder, which are used to learn the channel probability distribution vector and the timing weight vector respectively, and guide the decoder to decode the high-level semantic features and the timing relationship between PPG and ECG.

[0014] According to some embodiments, the present disclosure adopts the following technical solutions:

[0015] A cuffless continuous blood pressure signal reconstruction system includes an acquisition module and a reconstruction module:

[0016] The acquisition module is configured to: acquire synchronized PPG signals and ECG signals, and perform standardization processing on them;

[0017] The reconstruction module is configured to: use the trained attention time series network to extract PPG feature vectors and ECG feature vectors from PPG signals and ECG signals respectively, perform multimodal fusion on the PPG feature vectors and ECG feature vectors, and reconstruct a cuff-free continuous blood pressure signal based on the fused feature vectors;

[0018] Among them, the feature vector is extracted by adding an information screening machine and a timing learning machine between the encoder and the decoder, which are used to learn the channel probability distribution vector and the timing weight vector respectively, and guide the decoder to decode the high-level semantic features and the timing relationship between PPG and ECG.

[0019] According to some embodiments, the present disclosure adopts the following technical solutions:

[0020] A computer program product comprises a computer program, wherein when the computer program is executed by a processor, the cuffless continuous blood pressure signal reconstruction method is implemented.

[0021] According to some embodiments, the present disclosure adopts the following technical solutions:

[0022] A non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, a cuffless continuous blood pressure signal reconstruction method is implemented.

[0023] According to some embodiments, the present disclosure adopts the following technical solutions:

[0024] An electronic device comprises: a processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device implements the cuffless continuous blood pressure signal reconstruction method.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] Combining DL with multimodal fusion technology, the present invention designs an attention timing network (AT-Net) with information screening and timing feature mining capabilities, and balances the redundancy and complementarity of PPG and ECG in multimodal fusion, thereby achieving high-precision continuous BP signal reconstruction, specifically:

[0027] (1) Multimodal fusion: The multimodal fusion method is used to learn the autocorrelation information and cross-correlation information of PPG and ECG respectively.

[0028] (2) Information filtering machine: Integrates multiple sets of parallel attention mechanisms to focus on the individual information and common information of features of different scales during ventricular contraction and relaxation to eliminate the redundancy of the two.

[0029] (3) Temporal learning machine: The memory function based on GRU captures the dynamic change trend of PPG and ECG to improve AT-Net’s ability to represent the temporal features of the ventricular ejection period. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The accompanying drawings constituting a part of the present disclosure are used to provide a further understanding of the present disclosure. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure.

[0031] Figure 1 The present invention is a method flow chart of an embodiment of the present invention.

[0032] Figure 2 A network structure diagram of an embodiment of the present disclosure.

[0033] Figure 3 The figure is a flow chart of the information screening machine according to an embodiment of the present disclosure.

[0034] Figure 4 The flowchart of a memory unit in the time series learning machine of the embodiment of the present disclosure.

[0035] Figure 5 This is an example diagram of a continuous arterial BP waveform reconstructed according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0036] The present disclosure is further described below in conjunction with the accompanying drawings and embodiments.

[0037] It should be noted that the following detailed descriptions are all illustrative and are intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present disclosure belongs.

[0038] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0039] Example 1

[0040] In one embodiment of the present disclosure, a cuff-free continuous blood pressure signal reconstruction method is provided. Figure 1 As shown, including:

[0041] Step 1: Obtain synchronized PPG and ECG signals and perform standardization on them;

[0042] Step 2: Use the trained attention time series network to extract PPG feature vectors and ECG feature vectors from PPG signals and ECG signals respectively, perform multimodal fusion on the PPG feature vectors and ECG feature vectors, and reconstruct the cuff-free continuous blood pressure signal based on the fused feature vectors;

[0043] Among them, the feature vector is extracted by adding an information screening machine and a timing learning machine between the encoder and the decoder, which are used to learn the channel probability distribution vector and the timing weight vector respectively, and guide the decoder to decode the high-level semantic features and the timing relationship between PPG and ECG.

[0044] As an embodiment, a cuffless continuous blood pressure signal reconstruction method disclosed in the present invention proposes an attention timing network (AT-Net) based on multimodal fusion, and realizes the reconstruction of continuous BP signal by analyzing the photoplethysmogram signal (PPG) and the electrocardiogram signal (ECG). The specific implementation process is as follows:

[0045] Step S1: Signal acquisition

[0046] The training data used in this embodiment comes from the Multiparameter Intelligent Monitoring in Intensive Care (MIMIC) dataset, and the sampling rate of all signals is 125 Hz. This embodiment selects data with a recording time of 8 minutes or more and no obvious noise from the sorted dataset, and a total of 1,400 subjects are obtained.

[0047] In clinical practice, some patients may have extremely low and high BP values. The BP range of this embodiment is set to 50 mmHg to 200 mmHg. In addition, the synchronized PPG and ECG are both Z-score standardized and segmented into 1-second segments; all data are divided into training set: validation set: test set according to 8:1:1. According to the Office blood pressure threshold, 1,400 subjects were divided into 4 groups. The division criteria and subject proportions are shown in Table 1, where mean pressure (MAP), diastolic pressure (DBP) and systolic pressure (SBP) are the average, minimum and maximum values ​​of the BP segments, respectively.

[0048] Table 1. Classification criteria and subject proportions

[0049]

[0050] The above data is used for network training. For the persons to be tested in actual applications, their PPG signals and ECG signals should be collected synchronously and standardized. After processing, they should be input into the trained network to reconstruct the cuffless continuous blood pressure signal.

[0051] Step S2: Reconstructing continuous blood pressure signals using attention temporal networks

[0052] The attention temporal network (AT-Net) proposed in this embodiment based on multimodal fusion adopts the U-Net framework. The overall structure of the network is as follows: Figure 2 As shown, it includes an encoder, an information screening machine, a time series learning machine, a decoder, and a multimodal fusion module connected in sequence.

[0053] The encoder, consisting of one-dimensional convolutions of three scales of 1×7, 1×5 and 1×3, is used to extract implicit information related to blood pressure in PPG and ECG signals; the amplitudes and gradients of different bands in multimodal signals are analyzed with the help of a multi-scale mechanism to learn the complementarity between blood flow and cardiac electrical activity.

[0054] The information screening machine, based on the attention mechanism, performs two excitation operations on the output of the encoder, learns the importance of each feature channel in the information mining process, and obtains the channel probability distribution vector to guide the decoder to decode high-level semantic features.

[0055] The temporal learning machine, built on the basis of a GRU with 128 memory units, captures the temporal transformation trend of long-term signals from the output of the encoder and obtains a temporal weight vector, which is used to guide the decoder to decode the temporal relationship between PPG and ECG.

[0056] The decoder, which consists of three scales of deconvolution, 1×3, 1×5, and 1×7, is used to perform weighted decoding on the implicit information extracted by the encoder using the channel probability distribution vector and the temporal weight vector to obtain the feature vectors related to BP in the PPG and ECG signals.

[0057] It can be seen that the difference from U-Net is that: in this embodiment, the output of the c∈(1≤c≤5)th layer of convolution is converted into a channel probability distribution vector S2 and a timing weight vector G after two independent encodings (i.e., an information screening machine and a timing learning machine), and the vectors S2 and G are used to weight the output of the 6-cth layer of deconvolution and serve as the input of the 6-c+1th layer of deconvolution.

[0058] Assume that the output of the cth convolution layer in the encoder is U c = {U c1 ,U c2 ,U c3 ,…,U cn}, after learning the information screening machine, the channel probability distribution vector S2 = {s 21 ,s 22 ,s 23 ,…,s 2n}, after learning by the time series learning machine, the time series weight vector G = {g1, g2, g3, …, g n}. If the output of the 6-c layer deconvolution is U cT = {UcT1 ,U cT2 ,U cT3 ,…,U cTn}, then the input U of the 6-c+1th layer deconvolution ct =(S2⊕G)⊙U cT = {U ct1 ,U ct2 ,U ct3 ,…,U ctn}, where U cti =(s 2i +g i )×U cTi The feature vector after weighting by vectors S2 and G can reduce the redundancy of the encoder's multimodal information and preserve the timing of PPG and ECG.

[0059] It is an existing technology to use an encoder and a decoder to extract features. On this basis, this embodiment adds an information screening machine and a timing learning machine to respectively learn the channel probability distribution vector and the timing weight vector, and guide the decoder to decode high-level semantic features and the timing relationship between PPG and ECG, so as to solve the problem of redundancy and complementarity game between PPG and ECG.

[0060] The information screening machine, time series learning machine, and multimodal fusion module are described in detail below.

[0061] 1. Information screening machine

[0062] The implicit information extracted from different bands of PPG and ECG after multi-scale encoding has information redundancy. Indiscriminately adding it to the decoder will lead to a decrease in the performance of the model. In order to eliminate the redundancy of individual information and common information of different scales during ventricular contraction and relaxation, this embodiment proposes an information screening machine based on the attention mechanism. The principle is shown in Algorithm 1 in Table 2:

[0063]

[0064]

[0065] Assume that is an arbitrary set of real numbers, and the input signal of AT-Net is X undergoes a nonlinear transformation in the encoder After that, we get multi-dimensional features

[0066] Information filtering machines, such as Figure 3 As shown, first, using the compression operation F sq (·) will U cConverted into a vector with global representation characteristics To obtain the channel importance under the wide receptive field; then, use the weight vector And Relu function performs the first excitation operation on Z Get the feature vector To increase the sparsity of AT-Net; secondly, in order to learn the importance of each feature channel in the information mining process and convert it into a weight vector, the weight vector Perform a second excitation operation on the obtained feature vector S1 And through the Softmax activation function, S1 is converted into a channel probability distribution vector with a probability sum of 1 Wherein, m, n, H and L are positive integers.

[0067] This embodiment uses the channel probability distribution vector S2 as a weight to guide the decoder to decode high-level semantic features.

[0068] 2. Sequential Learning Machine

[0069] PPG, ECG and BP are all time series signals. When using neural networks to mine deep abstract features, retaining the time series relationship of the signals helps decode the blood flow and cardiac electrical activity during the ventricular ejection period; convolutional neural networks have strong spatial information learning capabilities, but their time series information learning capabilities are relatively weak. This embodiment designs a time series learning machine based on GRU to mine the time transformation relationship between PPG and ECG. It is constructed based on GRU with 128 memory units, captures the time transformation trend of long-term signals from the output of the encoder, and obtains a time series weight vector to guide the decoder to decode the time series relationship between PPG and ECG. The principle is shown in Algorithm 2 in Table 3:

[0070]

[0071]

[0072] Assume that, given any period of time [τ1,τ n ], the input signal After c nonlinear transformations in the encoder The extracted feature vector is

[0073] For a certain memory unit, such as Figure 4 As shown, the specific processing flow is:

[0074] Current input The corresponding feature is U tt , the hidden state at the previous moment is h t-1, first, the weight matrix W r For the matrix [h t-1 ,U tt ] Perform multi-dimensional linear transformation To learn h t-1 with U tt The temporal relationship between them. The weight matrix W z For the matrix [h t-1 ,U tt ] to perform linear transformations in high-dimensional space To obtain U tt Secondly, Sigmoid transforms the temporal relationship and long-term memory of PPG and ECG into weight value r through nonlinear transformation. t and weight value z t , r t With h t-1 Perform the Hadamard product to get h t-1 The remaining information. Then, the weight W has an effect on r t *h t-1 with U tt Nonlinear weighting is performed, and the candidate hidden state is obtained after a secondary nonlinear transformation using the Tanh function Finally, after selective forgetting (1-z t )*h t-1 and selective memory After that, get the hidden state h at the next moment t and the output g of the current memory unit t , τ1, τ t and τ n is a positive real number, and a, b, c, d and t are positive integers.

[0075] Using the output G of the memory unit t The timing weight vector G composed of guides the decoder to perform deep decoding on the timing relationship between PPG and ECG.

[0076] 3. Multimodal Fusion

[0077] The mutual assistance of other physiological signals and PPG can improve the accuracy of BP estimation. Among various physiological signals, ECG can accurately reflect the changes in skin potential when blood flows. Algorithms based on PTT, PAT and PWV all use ECG as an auxiliary signal, which shows that there are features in ECG that are closely related to BP.

[0078] This embodiment uses an encoder, an information filter, a time series learning machine, and a decoder to extract features, which actually maps the source domains of PPG and ECG to their respective feature domains, specifically:

[0079] Using synchronized PPG and ECG as input to AT-Net, we implement end-to-end modeling of continuous BP signals according to different source domains. In the PPG segment, and the ECG segment is Using the Encoder Source Domain Mapping to high-dimensional space In this paper, we obtain two sets of high-dimensional feature vectors and Then use the decoder High dimensional space Projection to feature domain and feature domain In this paper, we obtain two sets of final feature vectors and

[0080] After obtaining two sets of final feature vectors, the PPG feature vector y ppg and ECG feature vector y ecg There is a corresponding complementary relationship, but there is also a certain degree of information redundancy. In order to achieve information coupling between the two sets of feature vectors, this embodiment performs a PPG feature vector y ppg and ECG feature vector y ecg Perform secondary modeling to transform the feature domain and Unified mapping to BP target domain middle.

[0081] Assume that The BP in (i.e., the real blood pressure signal) is BP True ={BP True1 ,BP True2 ,BP True3 ,…,BP Truen}, middle y ppg The modeled BP (i.e., the output signal obtained by taking PPG as input) is BP PPg ={BP PPG1 ,BP PPG2 ,BP PPG3 ,…,BP PPGn}, middle y ecg The modeled BP (output signal obtained by taking ECG as input) is BP ECG ={BP ECG1 ,BP ECG2 ,BP ECG3 ,…,BP ECGn}, any (BP PPGi +BP ECGi) / 2 and BP Truei The error relationship between them is shown in Table 4:

[0082] Table 4BP PPGi , BP ECGi and BP Truei The error situation

[0083]

[0084]

[0085] Clinical data is unlabeled data, so it is impossible to obtain PPGi With BP ECGi The estimated value with the smallest error is selected as the clinical prediction result. PPGi With BP ECGi If we simply add the values ​​and take the average, there is a 2 / 3 chance that the estimated value will deviate from the true value.

[0086] To avoid deviation The present embodiment uses re-learning to and Fusion is performed to obtain the transition domain Then Map to The process is shown in formula (1-2):

[0087]

[0088]

[0089] Specifically, assuming that the feature domain With the target domain The error between ppg , feature domain With the target domain The error between ecg , as shown in formula (3-4):

[0090]

[0091]

[0092] In this embodiment, BP PPG With BP ECG Dynamically adjust and use nonlinear transformation to make |γ ppg +γ ecg | reaches the minimum value, the process of nonlinear transformation is shown in formula (5~8),

[0093]

[0094]

[0095]

[0096] BP PRE =U{BP PRE1 ,BP PRE2 ,BP PRE3 ,…,BP PREn} (8)

[0097] First, using the weight matrix and bias BP 22G With BP ECG Each BP value in is weighted to achieve This process is based on the back propagation of the loss function to achieve BP PPGi With BP ECGi Dynamic adjustment of , that is, using low weights to suppress estimates with large errors. Then, using the weight matrix and bias right BP vector BP pre Perform regression calculation to obtain BP True Closest BP PRE , as the reconstructed cuff-free continuous blood pressure signal; among them, and are all 1×n vectors.

[0098] In order to evaluate the effect of this method, the mean absolute error (MAE) and root mean square error (RMSE) are used as evaluation indicators. The calculation process of MAE and RMSE is shown in formulas (9) and (10).

[0099]

[0100]

[0101] This example also uses the standards of the British Hypertension Society (BHS) and the Association for the Advancement of Medical Instrumentation (AAMI) to evaluate the clinical application value of this method.

[0102] This embodiment reconstructs the BP signal based on continuous PPG and ECG. In order to observe the long-range tracking ability of the algorithm, a record is randomly selected from each type of subject as a visualization sample. The reconstruction effect of the BP signal is shown in Figure 2. Figure 5As shown, the blue solid line represents the real BP signal, and the red dotted line represents the reconstructed BP signal. The results show that the AT-Net of this embodiment can accurately reconstruct the BP signal based on PPG and ECG.

[0103] During the ventricular ejection phase, the accelerated ejection velocity causes the slope and amplitude of the ascending branch to increase rapidly in a short period of time, while the slowed ejection velocity causes the slopes of the anterior and posterior branches of the descending branch to decrease slowly. The method of this embodiment has a good fitting effect for both the rapid increase and the slow decrease of the amplitude and slope; especially for the dicrotic wave (RW) in the BP signal with small jitter and easy to form an inflection point, the method of this embodiment has a good fitting effect, and the inflection point formed by RW can also be accurately reconstructed. The reconstruction effect of RW is as follows: Figure 5 As shown, the method of this embodiment can better fit RW, and can also accurately reconstruct the inflection point formed by RW.

[0104] This embodiment collects some existing technologies for BP estimation between 2019 and 2024 and compares them with the results of this embodiment. The comparison results are shown in Table 5. In this embodiment, the MAE and RMSE of the mean pressure (MAP) are 1.24 mmHg and 2.19 mmHg, the MAE and RMSE of the systolic blood pressure (SBP) are 2.94 mmHg and 4.49 mmHg, and the MAE and RMSE of the diastolic blood pressure (DBP) are 1.65 mmHg and 2.77 mmHg; by comparison, it is found that some indicators are second only to the solutions proposed by Hajj and Zhang. However, the solution proposed by Hajj only selected 500 sets of waveforms, and the solution proposed by Zhang only tested 25 healthy subjects.

[0105] In contrast, this embodiment uses 1400 subjects, and the BP interval is [50mmHg, 200mmHg], that is, the test data of this embodiment is more diversified, and the test results have higher reliability. In addition, the task of this embodiment is not a simple SBP and DBP detection, but the reconstruction of continuous BP signals, that is, this embodiment is more in line with the requirements of continuous, long-term and dynamic monitoring.

[0106] Table 5 Method comparison

[0107]

[0108]

[0109] In summary, this embodiment adopts a multimodal fusion strategy and designs a new AT-Net to realize the reconstruction of continuous BP signals; the encoder of AT-Net extracts implicit information related to BP in PPG and ECG, the information screening machine solves the information redundancy problem in the feature extraction process, the timing learning machine captures the time transformation trend of long-term signals, and the decoder accurately decodes the semantic vectors related to BP in PPG and ECG; compared with the existing algorithms, this embodiment realizes the task conversion from estimating discrete BP values ​​to reconstructing continuous BP signals, and has lower error and stronger reliability.

[0110] Example 2

[0111] In one embodiment of the present disclosure, a cuffless continuous blood pressure signal reconstruction system is provided, including an acquisition module and a reconstruction module:

[0112] The acquisition module is configured to: acquire synchronized PPG signals and ECG signals, and perform standardization processing on them;

[0113] The reconstruction module is configured to: use the trained attention time series network to extract PPG feature vectors and ECG feature vectors from PPG signals and ECG signals respectively, perform multimodal fusion on the PPG feature vectors and ECG feature vectors, and reconstruct a cuff-free continuous blood pressure signal based on the fused feature vectors;

[0114] Among them, the feature vector is extracted by adding an information screening machine and a timing learning machine between the encoder and the decoder, which are used to learn the channel probability distribution vector and the timing weight vector respectively, and guide the decoder to decode the high-level semantic features and the timing relationship between PPG and ECG.

[0115] Example 3

[0116] In one embodiment of the present disclosure, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the cuffless continuous blood pressure signal reconstruction method is implemented.

[0117] Example 4

[0118] In one embodiment of the present disclosure, a non-transitory computer-readable storage medium is provided, wherein the non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the cuffless continuous blood pressure signal reconstruction method is implemented.

[0119] Example 5

[0120] In one embodiment of the present disclosure, an electronic device is provided, comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device executes the cuffless continuous blood pressure signal reconstruction method.

[0121] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0122] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0123] Although the above describes the specific implementation methods of the present disclosure in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present disclosure. Technical personnel in the relevant field should understand that on the basis of the technical solution of the present disclosure, various modifications or variations that can be made by those skilled in the art without creative work are still within the scope of protection of the present disclosure.

Claims

1. A cuffless continuous blood pressure signal reconstruction method, characterized in that: include: Acquire synchronized PPG and ECG signals and perform standardization on them; Using the trained attention time series network, PPG feature vectors and ECG feature vectors are extracted from PPG signals and ECG signals respectively, and multimodal fusion is performed on the PPG feature vectors and ECG feature vectors. Based on the fused feature vectors, the cuff-free continuous blood pressure signal is reconstructed. Among them, the feature vector is extracted by adding an information screening machine and a timing learning machine between the encoder and the decoder, which are used to learn the channel probability distribution vector and the timing weight vector respectively, and guide the decoder to decode the high-level semantic features and the timing relationship between PPG and ECG.

2. A cuffless continuous blood pressure signal reconstruction method as claimed in claim 1, characterized in that: The attention time series network includes an encoder, an information screening machine, a time series learning machine, a decoder, and a multimodal fusion module which are connected in sequence.

3. A cuffless continuous blood pressure signal reconstruction method as claimed in claim 2, characterized in that: The encoder, consisting of multi-scale one-dimensional convolutions, is used to extract implicit information related to blood pressure in PPG signals and ECG signals; The decoder is composed of multi-scale one-dimensional deconvolution, which is used to perform weighted decoding on the implicit information extracted by the encoder using the channel probability distribution vector and the timing weight vector to obtain the feature vectors related to BP in the PPG signal and the ECG signal.

4. A cuffless continuous blood pressure signal reconstruction method as claimed in claim 2, characterized in that: The information screening machine, based on the attention mechanism, performs two excitation operations on the output of the encoder, learns the importance of each feature channel in the information mining process, and obtains a channel probability distribution vector.

5. The cuffless continuous blood pressure signal reconstruction method according to claim 1, characterized in that: The time series learning machine, based on GRU, captures the time transformation trend of the long-term signal from the output of the encoder to obtain a time series weight vector.

6. A cuffless continuous blood pressure signal reconstruction method as claimed in claim 1, characterized in that: The multimodal fusion is learned with the goal of minimizing the error between the feature domain and the target domain, and the PPG feature vector and the ECG feature vector are uniformly mapped from their respective feature domains to the target domain of blood pressure using the learned weight matrix and bias to obtain a fused feature vector.

7. A cuffless continuous blood pressure signal reconstruction system, characterized in that: Including acquisition module and reconstruction module: The acquisition module is configured to: acquire synchronized PPG signals and ECG signals, and perform standardization processing on them; The reconstruction module is configured to: use the trained attention time series network to extract PPG feature vectors and ECG feature vectors from PPG signals and ECG signals respectively, perform multimodal fusion on the PPG feature vectors and ECG feature vectors, and reconstruct a cuff-free continuous blood pressure signal based on the fused feature vectors; Among them, the feature vector is extracted by adding an information screening machine and a timing learning machine between the encoder and the decoder, which are used to learn the channel probability distribution vector and the timing weight vector respectively, and guide the decoder to decode the high-level semantic features and the timing relationship between PPG and ECG.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for reconstructing a cuffless continuous blood pressure signal according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by the processor, a cuffless continuous blood pressure signal reconstruction method as described in any one of claims 1 to 6 is implemented.

10. An electronic device, characterized in that: include: A processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes a cuffless continuous blood pressure signal reconstruction method as described in any one of claims 1-6.

Images