Blood pressure monitoring method, device and electronic equipment
By preprocessing and feature extraction of array PPG signals, combined with graph convolution and LSTM network, the problem of insufficient extraction of convolutional neural networks in array PPG signals is solved, and the accuracy of blood pressure monitoring and interval prediction effect is improved.
Patent Information
- Application Number
- CN202211554662.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-12-05
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2042-12-05
AI Technical Summary
Under the irregular correlation of array PPG signals, existing convolutional neural networks cannot accurately extract feature information with high distinction, resulting in low accuracy of blood pressure monitoring.
By preprocessing the array PPG signal, an adjacency matrix of the graph structure is constructed, a graph convolutional neural network and long and short-term memory neural network are combined to extract spatial and temporal features, and feature fusion is performed. The blood pressure value is predicted using linear regression algorithm to determine contraction. The interval between pressure and diastolic blood pressure.
It improves the prediction accuracy of blood pressure value and blood pressure interval, enhances the feature extraction ability of array PPG signals, and improves the accuracy of monitoring.
Smart Images

Figure CN115956889B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of blood pressure monitoring, and in particular to a blood pressure monitoring method, device and electronic equipment. Background Art
[0002] At present, in blood pressure monitoring technology, CNN (Convolutional Neural Networks) can automatically extract discriminative features and regress the extracted features for systolic and diastolic pressure, thus realizing end-to-end blood pressure measurement based on regression prediction. However, when the array PPG (Photoplethysmographic) signal is used to monitor the systolic and diastolic pressure of blood pressure in real time, the correlation between the PPG signals in the array is irregular and cannot be accurately described by the traditional Euclidean distance. The use of the CNN model not only fails to fully extract the discriminative feature information, but also leads to the problem of low accuracy of blood pressure monitoring.
[0003] Therefore, it is necessary to propose an algorithm for real-time monitoring of array PPG signals at different systolic and diastolic pressures, which can improve the problem of low accuracy of traditional deep learning methods in real-time monitoring of array PPG signals at different systolic and diastolic pressure values. Summary of the invention
[0004] The embodiments of the present application provide a blood pressure monitoring method, device and electronic device, which can improve the accuracy of blood pressure monitoring.
[0005] The technical solution of the embodiment of the present application is implemented as follows:
[0006] In a first aspect, an embodiment of the present application provides a blood pressure monitoring method, comprising:
[0007] Preprocessing the array PPG signal to obtain a standardized array PPG signal and an adjacency matrix of a graph structure corresponding to the array PPG signal;
[0008] Determining spatial features and temporal features corresponding to the array PPG signal based on the standardized array PPG signal and the adjacency matrix;
[0009] Fusing the spatial feature and the temporal feature to obtain a space-time feature;
[0010] Determining a predicted blood pressure value based on the space-time characteristics;
[0011] Based on the blood pressure prediction value, a blood pressure interval of systolic pressure and a blood pressure interval of diastolic pressure corresponding to the array PPG signal are determined.
[0012] In the above scheme, the preprocessing of the array PPG signal to obtain the standardized array PPG signal and the adjacency matrix of the graph structure corresponding to the array PPG signal includes:
[0013] The array PPG signal is standardized to obtain the standardized array PPG signal.
[0014] In the above scheme, the preprocessing of the array PPG signal to obtain the standardized array PPG signal and the adjacency matrix of the graph structure corresponding to the array PPG signal includes:
[0015] Selecting an array PPG signal of a first length based on a sliding window;
[0016] For each selected array PPG signal of the first length, the adjacency matrix is constructed based on the average value of the sum of the maximum mutual information coefficients between each PPG signal and the adjacent PPG signals.
[0017] In the above solution, the determining of the spatial features and temporal features corresponding to the array PPG signal based on the standardized array PPG signal and the adjacency matrix includes:
[0018] Obtaining spatial features corresponding to the array PPG signal based on a graph convolutional neural network;
[0019] The time characteristics corresponding to the array PPG signal are obtained based on the long short-term memory neural network.
[0020] In the above solution, the step of fusing the spatial feature and the temporal feature to obtain the space-time feature includes:
[0021] The spatial feature and the temporal feature are fused through a fully connected layer to obtain the space-time feature.
[0022] In the above solution, the step of determining the predicted blood pressure value based on the space-time feature includes:
[0023] The space-time characteristics are predicted based on a linear regression algorithm to obtain a predicted value of systolic pressure and a predicted value of diastolic pressure.
[0024] In the above solution, determining the systolic blood pressure interval and the diastolic blood pressure interval corresponding to the array PPG signal based on the blood pressure prediction value includes:
[0025] Determining a first upper limit value and a first lower limit value corresponding to the systolic pressure based on the predicted value of the systolic pressure and the predicted value of the diastolic pressure;
[0026] Determining the blood pressure interval of the systolic blood pressure based on the first upper limit value and the first lower limit value;
[0027] Determining a second upper limit value and a second lower limit value corresponding to the diastolic pressure based on the predicted value of the systolic pressure and the predicted value of the diastolic pressure;
[0028] Based on the second upper limit value and the second lower limit value, a blood pressure interval of the diastolic pressure is determined.
[0029] In a second aspect, an embodiment of the present application provides a training method for a blood pressure monitoring model, characterized in that the method comprises:
[0030] Preprocessing the array PPG signal in the sample data set to obtain a standardized array PPG signal and an adjacency matrix of a graph structure corresponding to the array PPG signal, wherein the sample data set includes the array PPG signal and the labeled values of the systolic pressure and the diastolic pressure corresponding to the array PPG signal;
[0031] Using the standardized array PPG signal and the adjacency matrix as inputs of a first sub-model in a blood pressure monitoring model to obtain a blood pressure prediction value, wherein the blood pressure prediction value includes a systolic pressure prediction value and a diastolic pressure prediction value;
[0032] The predicted value of the systolic pressure and the predicted value of the diastolic pressure are used as inputs of the second sub-model to obtain a blood pressure interval range of the systolic pressure and a blood pressure interval range of the diastolic pressure.
[0033] In the above scheme, the standardized array PPG signal and the adjacency matrix are used as inputs of the first sub-model in the blood pressure monitoring model to obtain the blood pressure prediction value, including:
[0034] Obtaining spatial features corresponding to the array PPG signal based on a graph convolutional neural network;
[0035] Obtaining the time characteristics corresponding to the array PPG signal based on a long short-term memory neural network;
[0036] The spatial features and the temporal features are fused through a fully connected layer, and the fused spatial and temporal features are predicted based on a linear regression algorithm to obtain a predicted value of systolic pressure and a predicted value of diastolic pressure;
[0037] The first sub-model adjusts the parameters of the first sub-model and performs model training based on the difference between the labeled value of the systolic pressure and the predicted value of the systolic pressure, and the difference between the labeled value of the diastolic pressure and the predicted value of the diastolic pressure.
[0038] In the above solution, the predicted value of systolic pressure and the predicted value of diastolic pressure are used as inputs of the second sub-model to obtain the blood pressure interval range of systolic pressure and the blood pressure interval range of diastolic pressure, including:
[0039] The second sub-model performs Gaussian regression processing based on the predicted value of the systolic blood pressure to determine a first upper boundary value and a first lower boundary value corresponding to the systolic blood pressure, and determines the blood pressure interval of the systolic blood pressure based on the first upper boundary value and the first lower boundary value;
[0040] The second sub-model performs Gaussian regression processing based on the predicted value of the diastolic pressure to determine a second upper boundary value and a second lower boundary value corresponding to the diastolic pressure, and determines the blood pressure range of the diastolic pressure based on the second upper boundary value and the second lower boundary value.
[0041] In a third aspect, an embodiment of the present application provides a blood pressure monitoring device, the blood pressure monitoring device comprising:
[0042] A preprocessing module, used to preprocess the array PPG signal to obtain a standardized array PPG signal and an adjacency matrix of a graph structure corresponding to the array PPG signal;
[0043] a blood pressure value determination module, configured to obtain the spatial features and the temporal features corresponding to the array PPG signal based on the standardized array PPG signal and the adjacency matrix; perform feature fusion on the spatial features and the temporal features, and determine the predicted value of the systolic pressure and the predicted value of the diastolic pressure based on the space-time features obtained after the fusion;
[0044] The blood pressure interval determination module is used to determine the blood pressure interval of the systolic pressure and the blood pressure interval of the diastolic pressure corresponding to the array PPG signal based on the predicted value of the systolic pressure and the predicted value of the diastolic pressure.
[0045] In a fourth aspect, an embodiment of the present application provides a blood pressure monitoring model training device, the device comprising:
[0046] A preprocessing module, used to preprocess the array PPG signal in the sample data set to obtain a standardized array PPG signal and an adjacency matrix of a graph structure corresponding to the array PPG signal, wherein the sample data set includes the array PPG signal and the labeled values of the systolic pressure and the diastolic pressure corresponding to the array PPG signal;
[0047] a blood pressure value determination module, which uses the standardized array PPG signal and the adjacency matrix as inputs of a first sub-model in a blood pressure monitoring model to obtain a blood pressure prediction value, wherein the blood pressure prediction value includes a systolic pressure prediction value and a diastolic pressure prediction value;
[0048] The blood pressure interval determination module uses the predicted value of systolic pressure and the predicted value of diastolic pressure as inputs of the second sub-model to obtain the blood pressure interval range of systolic pressure and the blood pressure interval range of diastolic pressure.
[0049] In a fifth aspect, an embodiment of the present application provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the blood pressure monitoring method provided in the embodiment of the present application or the blood pressure monitoring model training method provided in the embodiment of the present application.
[0050] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, which includes a set of computer-executable instructions, which, when executed, are used to execute the blood pressure monitoring method or blood pressure monitoring model training method provided in an embodiment of the present application.
[0051] The blood pressure monitoring method provided in the embodiment of the present application preprocesses the array PPG signal to obtain the corresponding standardized array PPG signal and the adjacency matrix of the graph structure corresponding to the array PPG signal; based on the standardized array PPG signal and the adjacency matrix, the spatial characteristics and time characteristics corresponding to the array PPG signal are determined; the spatial characteristics and the time characteristics are fused to obtain the space-time characteristics; the blood pressure prediction value is determined based on the space-time characteristics; based on the blood pressure prediction value, the systolic blood pressure interval and the diastolic blood pressure interval corresponding to the array PPG signal are determined. The blood pressure monitoring method of the present application improves the accuracy of blood pressure value prediction and the accuracy of blood pressure interval prediction by extracting and fusing spatial characteristics and time characteristics. BRIEF DESCRIPTION OF THE DRAWINGS
[0052] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present application.
[0053] Figure 1 This is a schematic diagram of an optional processing flow of the blood pressure monitoring method provided in an embodiment of the present application;
[0054] Figure 2 It is a schematic diagram of a process of constructing graph data based on an array PPG signal provided in an embodiment of the present application;
[0055] Figure 3 is a schematic diagram of the spatial feature extraction model structure provided in an embodiment of the present application;
[0056] Figure 4 It is a schematic diagram of the structure of the time feature extraction model provided in the embodiment of the present application;
[0057] Figure 5 It is a schematic diagram of a process for obtaining a predicted blood pressure value provided in an embodiment of the present application;
[0058] Figure 6This is another optional processing flow diagram of the blood pressure monitoring method provided in the embodiment of the present application;
[0059] Figure 7 is a schematic diagram of an optional structure of a blood pressure monitoring device provided in an embodiment of the present application;
[0060] Figure 8 This is a schematic diagram of an optional processing flow of the blood pressure monitoring model training method provided in an embodiment of the present application;
[0061] Fig. 9 It is another optional processing flow diagram of the blood pressure monitoring model training method provided in the embodiment of the present application;
[0062] Fig.10 is a schematic diagram of the prediction results of the blood pressure interval provided in the embodiment of the present application;
[0063] Fig.11 This is an optional structural diagram of the blood pressure monitoring model training device provided in the embodiment of the present application;
[0064] Fig.12 It is a schematic block diagram of an optional electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0065] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.
[0066] In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it can be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.
[0067] In the following description, the terms "first\second" involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It can be understood that "first\second" can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.
[0068] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.
[0069] A blood pressure monitoring method provided by an embodiment of the present application is described below. Figure 1 , Figure 1 This is an optional processing flow diagram of the blood pressure monitoring method provided in the embodiment of the present application. Figure 1 Steps S101-S105 are shown for explanation.
[0070] Step S101: preprocess the array PPG signal to obtain a standardized array PPG signal and an adjacency matrix of a graph structure corresponding to the array PPG signal.
[0071] In some embodiments, the array PPG signal includes two dimensions, one dimension is time, and the other dimension is the number of arrays, such as a 3*3 array PPG signal with a duration of 2s. Figure 2 Shown are PPG signals with an array of 3*3, including v1, v2, v3, v4, v5, v6, v7, v8 and v9, a total of 9 PPG signals.
[0072] In some embodiments, it is assumed that the sequence of the original array PPG signal is: s1, ..., s n , where the average value of the signal value of all array PPG signals is μ, and the standard deviation of the signal value of all array PPG signals is σ, then the standardized array PPG signals are: Among them, the standardized array PPG signal obeys the normal distribution.
[0073] In some embodiments, Figure 2 For example, the array PPG signal is a 3*3 PPG signal with 9 channels. The sliding window method can be used to select the original array PPG signal collected in real time according to a fixed first length. For each selected array PPG signal of the first length, the adjacency matrix of the graph corresponding to the array PPG signal is constructed based on the average value of the sum of the maximum mutual information coefficients between each PPG signal and the adjacent PPG signal. The following formula (1) shows the calculation formula for the maximum mutual information coefficient between two adjacent PPG signals in the array PPG signal.
[0074]
[0075] In formula (1), V i and V j represents two adjacent PPG signals in the array PPG signal, M IC (V i, V j ) indicates V i and V j The maximum mutual information coefficient between i and V jTwo PPG signals, x represents V i Any PPG signal value in; y represents V j For any PPG signal value in , R(n) represents the maximum value of the product between x and y, and I(D,x,y) represents V i and V j The maximum value of the relevant mutual information.
[0076] After the maximum mutual information coefficient of each PPG signal in the array PPG signal is obtained based on formula (1), the average value of the sum of the maximum mutual information coefficients between each PPG signal and the adjacent PPG signal is obtained as shown in formula (2) to obtain the final constructed adjacency matrix.
[0077]
[0078] In formula (2), For example, due to Figure 2 In , the adjacent PPG signals of v1 are v2 and v4, therefore, the average of the sum of the maximum mutual information number between v1 and v2 and the maximum mutual information number between v1 and v4 is determined as the vertex value corresponding to v1 in the adjacency matrix.
[0079] Step S102: determining the spatial features and temporal features corresponding to the array PPG signal based on the standardized array PPG signal and the adjacency matrix.
[0080] In some embodiments, the standardized array PPG signal and adjacency matrix can be used to obtain spatial features through a graph convolutional neural network (GCN) model, and the temporal features can be obtained through a long short-term memory neural network (LSTM) model.
[0081] In this case, the normalized array PPG signal and the adjacency matrix are input into Figure 3 The spatial feature extraction model shown can obtain the spatial features corresponding to the array PPG signal. Figure 3 In , the spatial feature extraction model includes two graph convolutional layers and one fully connected layer.
[0082] By inputting the normalized array PPG signal and adjacency matrix into Figure 4 The time feature extraction model shown can obtain the time features corresponding to the array PPG signal. Figure 4 In , the temporal feature extraction model includes two LSTM layers and one fully connected layer.
[0083] Step S103: Fusing the spatial feature and the temporal feature to obtain a space-time feature.
[0084] In some embodiments, the spatial features and the temporal features may be fused through a fully connected layer to obtain the spatial-temporal features of the array PPG signal.
[0085] Since the space-time features are obtained by further linear mapping of space features and time features, the problem of insufficient extraction of discriminative features due to different amplitudes of different features is eliminated, and the extracted features are more accurate.
[0086] Step S104: Determine a predicted blood pressure value based on the space-time characteristics.
[0087] In some embodiments, the fused space-time features can be predicted based on a linear regression algorithm to determine a blood pressure prediction value, wherein the blood pressure prediction value includes a systolic pressure prediction value and a diastolic pressure prediction value. Figure 5 As shown, Figure 5 A flow chart of obtaining a predicted blood pressure value is shown. Figure 5 The structure includes Figure 3 The spatial feature extraction model and Figure 4 The time feature extraction model extracts spatial features based on the spatial feature extraction model and temporal features based on the temporal feature extraction model, obtains space-time features based on the spatial features and the temporal features through the fully connected layer, and performs linear regression calculation on the space-time features based on the regression layer to determine the predicted values of systolic pressure and diastolic pressure.
[0088] Step S105: Determine a blood pressure interval range of systolic pressure and a blood pressure interval range of diastolic pressure based on the predicted blood pressure value.
[0089] In some embodiments, a first upper limit value and a first lower limit value corresponding to the systolic pressure may be determined based on the predicted value of the systolic pressure and the predicted value of the diastolic pressure, and a blood pressure interval of the systolic pressure may be determined based on the first upper limit value and the first lower limit value;
[0090] Based on the predicted value of the diastolic pressure and the predicted value of the diastolic pressure, a second upper limit value and a second lower limit value corresponding to the diastolic pressure can be determined, and based on the second upper limit value and the second lower limit value, a blood pressure range of the diastolic pressure can be determined.
[0091] Another processing flow of the blood pressure monitoring method provided in the embodiment of the present application is described below. Figure 6 , Figure 6 This is another optional processing flow diagram of the blood pressure monitoring method provided in an embodiment of the present application.
[0092] In some embodiments, the blood pressure monitoring method can be implemented based on the graph convolutional neural network spatial-temporal feature fusion deep learning model GT-net model proposed in this application.
[0093] Step 1: Input the original array PPG signal into the preprocessing unit to obtain the standardized array PPG signal and the corresponding adjacency matrix.
[0094] Step 2: The standardized array PPG signal and the adjacency matrix are used to extract the corresponding spatial features through the spatial feature extraction unit, and at the same time, the corresponding temporal features are extracted through the temporal feature extraction unit.
[0095] Step 3: The spatial features and temporal features are fused through a feature fusion unit to obtain space-time features.
[0096] Step 4: The space-time features are output through a point output unit to output the blood pressure value, which includes the predicted value of the systolic pressure and the predicted value of the diastolic pressure.
[0097] Step 5: Pass the blood pressure value through a Gaussian process regression unit to obtain a first upper limit value and a first lower limit value of the systolic pressure, and a second upper limit value and a second lower limit value of the diastolic pressure.
[0098] Step 6: Output the blood pressure value interval through the interval output unit, using the first upper limit value and the first lower limit value of the systolic pressure, and the second upper limit value and the second lower limit value of the diastolic pressure. The blood pressure value interval includes the blood pressure range of the systolic pressure and the blood pressure range of the diastolic pressure.
[0099] The following is an introduction to the blood pressure monitoring device provided in the embodiments of the present application. Figure 7 This is a schematic diagram of an optional device structure of a blood pressure monitoring device provided in an embodiment of the present application. The blood pressure monitoring device 700 includes a preprocessing module 701, a blood pressure value determination module 702, and a blood pressure interval determination module 703. Among them,
[0100] A preprocessing module 701 is used to preprocess the array photoelectric pulse wave PPG signal to obtain a standardized array PPG signal corresponding to the array PPG signal and an adjacency matrix of a graph structure corresponding to the array PPG signal;
[0101] The blood pressure value determination module 702 is used to obtain the spatial features and time features corresponding to the array PPG signal based on the standardized array PPG signal and the adjacency matrix; perform feature fusion on the spatial features and the time features to obtain space-time features, and determine the blood pressure prediction value based on the space-time features;
[0102] The blood pressure interval determination module 703 is used to determine the blood pressure interval of the systolic pressure and the blood pressure interval of the diastolic pressure corresponding to the array PPG signal based on the blood pressure prediction value.
[0103] In some embodiments, the preprocessing module 701 is used to: perform standardization processing on the array PPG signal to obtain the standardized array PPG signal.
[0104] In some embodiments, the preprocessing module 701 is further used to: select an array PPG signal of a first length based on a sliding window; and construct the adjacency matrix based on the average value of the sum of the maximum mutual information coefficients between each PPG signal and an adjacent PPG signal for each selected array PPG signal of the first length.
[0105] In some embodiments, the blood pressure value determination module 702 is used to: obtain the spatial features corresponding to the array PPG signal based on a graph convolutional neural network; and obtain the temporal features corresponding to the array PPG signal based on a long short-term memory neural network.
[0106] In some embodiments, the blood pressure value determination module 702 is further used to: perform feature fusion on the spatial feature and the temporal feature through a fully connected layer to obtain the space-time feature.
[0107] In some embodiments, the blood pressure value determination module 702 is further used to: predict the space-time characteristics based on a linear regression algorithm to obtain a predicted value of systolic pressure and a predicted value of diastolic pressure.
[0108] In some embodiments, the blood pressure interval determination module 703 is used to determine a first upper limit value and a first lower limit value corresponding to the systolic pressure based on the predicted value of the systolic pressure and the predicted value of the diastolic pressure; determine the blood pressure interval of the systolic pressure based on the first upper limit value and the first lower limit value; determine the second upper limit value and the second lower limit value corresponding to the diastolic pressure based on the predicted value of the systolic pressure and the predicted value of the diastolic pressure; determine the blood pressure interval of the diastolic pressure based on the second upper limit value and the second lower limit value.
[0109] It should be noted that the blood pressure monitoring device of the embodiment of the present application is similar to the description of the blood pressure monitoring method embodiment described above, and has similar beneficial effects as the method embodiment, so it will not be described in detail. Figures 1 to 6 The present invention can be understood by referring to the description of any one of the accompanying drawings.
[0110] The following is an introduction to a training method for a blood pressure monitoring model provided in an embodiment of the present application. The blood pressure monitoring model proposed in the present application is a GT-GPR model. Figure 8-Figure 9 Provide explanation.
[0111] Step 801: preprocess the array PPG signal in the sample data set to obtain a standardized array PPG signal corresponding to the array PPG signal and an adjacency matrix of a graph structure corresponding to the array PPG signal, wherein the sample data set includes the array PPG signal and the labeled values of the systolic pressure and the diastolic pressure corresponding to the array PPG signal.
[0112] In some embodiments, array PPG signals of multiple subjects may be collected, and the true systolic pressure value and diastolic pressure value corresponding to the array PPG signal of each subject may be obtained;
[0113] In some embodiments, all array PPG signals are normalized to obtain standardized array PPG signals.
[0114] The process of standardization is as follows: Assume that the signal sequences of all array PPG signals are s1,…,s n , where the mean and standard deviation of the PPG signal are μ and σ respectively, then the standardized array PPG signal sequence is: The standardized array PPG signal obeys the normal distribution.
[0115] In some embodiments, all array PPG signals are selected based on a sliding window method, and a corresponding adjacency matrix is obtained for each selected array PPG signal of fixed length.
[0116] Among them, for each selected array PPG signal of fixed length, an adjacency matrix of the graph corresponding to the array PPG signal can be constructed based on the average value of the sum of the maximum mutual information coefficients between each PPG signal and the adjacent PPG signals.
[0117] In some embodiments, the array PPG signal obtained after preprocessing is used as sample data, and the systolic pressure value and diastolic pressure value corresponding to each array PPG signal are used as sample labels. The sample data is divided into a test set and a training set. The test set is used for training the blood pressure monitoring model, and the test set is used for testing the trained blood pressure monitoring model.
[0118] Step 802: Use the standardized array PPG signal and the adjacency matrix as inputs of a first sub-model in a blood pressure monitoring model to obtain a blood pressure prediction value, wherein the blood pressure prediction value includes a systolic pressure prediction value and a diastolic pressure prediction value.
[0119] In some embodiments, the first sub-model is Fig. 9 The GT-net model in is a GCN-LSTM neural network model consisting of a spatial feature extraction network, a temporal feature extraction network, a feature fusion network, and a blood pressure prediction value output.
[0120] Among them, the training process of the GT-net model is:
[0121] The spatial feature extraction network consisting of two layers of GCN and one fully connected layer obtains the spatial features corresponding to the array PPG signal; at the same time, the temporal feature extraction network consisting of two layers of LSTM and one fully connected layer obtains the temporal features corresponding to the array PPG signal. The spatial features and temporal features are fused through the feature fusion network consisting of the fully connected layer, and the blood pressure value is predicted based on the fused spatial and temporal features based on the linear regression algorithm to obtain the predicted value of blood pressure, where the predicted value of blood pressure includes the predicted value of systolic pressure and the predicted value of diastolic pressure. Then, according to the difference between the labeled value of systolic pressure and the predicted value of systolic pressure, and the difference between the labeled value of diastolic pressure and the predicted value of diastolic pressure, the parameters of the GT-net model are updated until the training termination condition is met and the model training is completed.
[0122] Step 803: Use the predicted value of the systolic pressure and the predicted value of the diastolic pressure as inputs of the second sub-model to obtain a blood pressure interval range of the systolic pressure and a blood pressure interval range of the diastolic pressure.
[0123] In some embodiments, the second sub-model is Fig. 9 In the GPR model, the second sub-model can be composed of a GPR model for predicting the blood pressure interval range of systolic blood pressure and a GPR model for predicting the blood pressure interval range of diastolic blood pressure.
[0124] Among them, the GPR model for predicting the blood pressure range of systolic blood pressure performs Gaussian regression processing based on the predicted value of systolic blood pressure to determine the first upper boundary value and the first lower boundary value corresponding to the systolic blood pressure, and based on the first upper boundary value and the first lower boundary value, determines that the blood pressure range of systolic blood pressure is the blood pressure range between the first upper boundary value and the first lower boundary value, including the first upper boundary value and the first lower boundary value.
[0125] A GPR model for predicting the blood pressure range of diastolic pressure performs Gaussian regression processing based on the predicted value of diastolic pressure to determine the second upper boundary value and the second lower boundary value corresponding to the diastolic pressure, and based on the second upper boundary value and the second lower boundary value, determines that the blood pressure range of diastolic pressure is the blood pressure range between the second upper boundary value and the second lower boundary value including the second upper boundary value and the second lower boundary value.
[0126] In summary, the first sub-model and the second sub-model constitute the GT-GPR model proposed in this application, and the data of the test set can be used to perform model verification on the trained GT-GPR model.
[0127] As shown in the table below, the GT-GPR model proposed in this application uses the public dataset MIMIC-III developed by the MIT Computational Physiology Laboratory to verify SBP (Systolic Blood Pressure). The performance indicators MAE (Mean Absolute Error), ME (Mean Error) and STD are used for point prediction; CP is used for interval prediction. α (Coverage Probability, coverage rate), MWP α (Mean Width Percentage, the average percentage of interval width to observations) and custom MC α Among them, MC in this application α Defined as
[0128] MC α The definition of is due to the fact that the ideal prediction interval should have a high CP α and low MWP α , where α is the confidence interval, and this application uses a 95% confidence interval. Compared with the CNN-GPR and CLSTM-GPR models with similar structures, the GT-GPR model has a significant performance improvement, and the point prediction value is within the AAMI standard range. The ME±STD of the AAMI standard is 5±8.
[0129]
[0130]
[0131] In summary, in order to make full use of the information in the space-time domain of the array PPG signal, the present application fuses the features extracted by the GCN model and the LSTM model, and proposes a deep learning model GT-net based on the graph convolutional neural network for space-time feature fusion; the model proposed in the present application fully utilizes the advantages of each network, uses GCN to extract the space-domain features of the array PPG signal, uses LSTM to extract the time-domain features of the array PPG signal, and realizes the fusion of space-time features, thereby realizing the full extraction of the features of the array PPG signal for blood pressure measurement; based on the high accuracy of the prediction of systolic and diastolic pressure by the GT-net proposed in the present application, the present application combines the advantages of the GPR model in interval prediction, and proposes the GT-GPR model used in the present application for uncertain measurement of blood pressure values, thereby improving the accuracy of blood pressure interval monitoring; the model proposed in the present application avoids the problem of insufficient expression of discriminative features existing in artificial features, and improves the generalization performance of the model. The model proposed in the present application is more suitable for the uncertainty monitoring of cuffless blood pressure values based on array PPG signals. Among them, the prediction results of the systolic blood pressure interval and the diastolic blood pressure interval of the test set are as follows: Fig.10 shown.
[0132] The following introduces the blood pressure monitoring model training device provided in the embodiment of the present application. Fig.11 This is a schematic diagram of an optional device structure of a blood pressure monitoring model training device provided in an embodiment of the present application. The blood pressure monitoring model training device 1100 includes a preprocessing module 1101, a blood pressure value determination module 1102, and a blood pressure interval determination module 1103. Among them,
[0133] A preprocessing module 1101 is used to preprocess the array photoelectric pulse wave PPG signal in the sample data set to obtain a standardized array PPG signal and an adjacency matrix of a graph structure corresponding to the array PPG signal, wherein the sample data set includes the array PPG signal and the labeled values of the systolic pressure and the diastolic pressure corresponding to the array PPG signal;
[0134] A blood pressure value determination module 1102, configured to use the standardized array PPG signal and the adjacency matrix as inputs of a first sub-model in a blood pressure monitoring model to obtain a blood pressure prediction value, wherein the blood pressure prediction value includes a systolic pressure prediction value and a diastolic pressure prediction value;
[0135] The blood pressure interval determination module 1103 is used to use the predicted value of systolic pressure and the predicted value of diastolic pressure as inputs of the second sub-model to obtain a blood pressure interval range of systolic pressure and a blood pressure interval range of diastolic pressure.
[0136] In some embodiments, the preprocessing module 1101 is used to: preprocess the array photoelectric pulse wave PPG signal in the sample data set to obtain a standardized array PPG signal and an adjacency matrix of a graph structure corresponding to the array PPG signal, wherein the sample data set includes the array PPG signal and the labeled values of the systolic pressure and the diastolic pressure corresponding to the array PPG signal; use the standardized array PPG signal and the adjacency matrix as inputs of a first sub-model in a blood pressure monitoring model to obtain a blood pressure prediction value, wherein the blood pressure prediction value includes a systolic pressure prediction value and a diastolic pressure prediction value; use the systolic pressure prediction value and the diastolic pressure prediction value as inputs of a second sub-model to obtain a blood pressure interval range of the systolic pressure and a blood pressure interval range of the diastolic pressure.
[0137] In some embodiments, the blood pressure value determination module 1102 is used to: obtain the spatial features corresponding to the array PPG signal based on a graph convolutional neural network; obtain the time features corresponding to the array PPG signal based on a long short-term memory neural network; fuse the spatial features and the time features through a fully connected layer, and predict the fused space-time features based on a linear regression algorithm to obtain a predicted value of systolic pressure and a predicted value of diastolic pressure; wherein the first sub-model adjusts the parameters of the first sub-model based on the difference between the labeled value of systolic pressure and the predicted value of systolic pressure, and the difference between the labeled value of diastolic pressure and the predicted value of diastolic pressure, and performs model training.
[0138] In some embodiments, the blood pressure interval determination module 1103 is used for: the second sub-model performs Gaussian regression processing based on the predicted value of the systolic pressure to determine the first upper boundary value and the first lower boundary value corresponding to the systolic pressure, and determines the blood pressure interval of the systolic pressure based on the first upper boundary value and the first lower boundary value; the second sub-model performs Gaussian regression processing based on the predicted value of the diastolic pressure to determine the second upper boundary value and the second lower boundary value corresponding to the diastolic pressure, and determines the blood pressure interval of the diastolic pressure based on the second upper boundary value and the second lower boundary value.
[0139] Fig.12A schematic block diagram of an example electronic device 1200 that can be used to implement an embodiment of the present disclosure is shown. The electronic device 1200 is used to implement the blood pressure monitoring method or the blood pressure monitoring model training method of the embodiment of the present disclosure. In some optional embodiments, the electronic device 1200 can implement the blood pressure monitoring method or the blood pressure monitoring model training method provided in the embodiment of the present application by running a computer program. For example, the computer program can be a software module in the operating system; it can be a native APP (Application), that is, a program that needs to be installed in the operating system to run; it can also be a small program, that is, a program that can be run only by downloading it to a browser environment; it can also be a small program that can be embedded in any APP. In short, the above-mentioned computer program can be an application, module or plug-in in any form.
[0140] In practical applications, the electronic device 1200 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms, where cloud technology refers to a hosting technology that unifies a series of resources such as hardware, software, and networks in a wide area network or a local area network to achieve data calculation, storage, processing, and sharing. The electronic device 1200 can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart TV, a smart watch, etc., but is not limited to this.
[0141] Electronic devices are intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, vehicle-mounted terminals, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or required herein.
[0142] like Fig.12As shown, the electronic device 1200 includes a computing unit 1201, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1202 or a computer program loaded from a storage unit 1208 into a random access memory (RAM) 1203. In the RAM 1203, various programs and data required for the operation of the electronic device 1200 can also be stored. The computing unit 1201, the ROM 1202, and the RAM 1203 are connected to each other via a bus 1204. An input / output (I / O) interface 1205 is also connected to the bus 1204.
[0143] Multiple components in the electronic device 1200 are connected to the I / O interface 1205, including: an input unit 1206, such as a keyboard, a mouse, etc.; an output unit 1207, such as various types of displays, speakers, etc.; a storage unit 1208, such as a disk, an optical disk, etc.; and a communication unit 1209, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1209 allows the electronic device 1200 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.
[0144] The computing unit 1201 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 1201 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 1201 performs the various methods and processes described above, such as a blood pressure monitoring method or a training method for a blood pressure monitoring model. For example, in some optional embodiments, the blood pressure monitoring method or the training method for a blood pressure monitoring model may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 1208. In some optional embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 1200 via the ROM 1202 and / or the communication unit 1209. When the computer program is loaded into the RAM 1203 and executed by the computing unit 1201, one or more steps of the blood pressure monitoring method or the training method for the blood pressure monitoring model described above may be performed. Alternatively, in other embodiments, the computing unit 1201 may be configured as a blood pressure monitoring method or a training method for a blood pressure monitoring model in any other appropriate manner (for example, by means of firmware).
[0145] An embodiment of the present application provides a computer-readable storage medium storing executable instructions, wherein executable instructions are stored. When the executable instructions are executed by a processor, the processor will execute the blood pressure monitoring method or the blood pressure monitoring model training method provided in the embodiment of the present application.
[0146] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface storage, optical disk, or CD-ROM; or it may be various devices including one or any combination of the above memories.
[0147] In some embodiments, executable instructions may be in the form of a program, software, software module, script or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a stand-alone program or as a module, component, subroutine or other unit suitable for use in a computing environment.
[0148] By way of example, executable instructions may be deployed to be executed on one computing device, or on multiple computing devices located at one site, or on multiple computing devices distributed across multiple sites and interconnected by a communication network.
[0149] The present application 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 application. 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.
[0150] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0151] It should be understood that in the various embodiments of the present application, the size of the serial number of each implementation process does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.
[0152] The above are only embodiments of the present application and are not intended to limit the protection scope of the present application. Any modifications, equivalent substitutions and improvements made within the spirit and scope of the present application are included in the protection scope of the present application.
Claims
1. A blood pressure monitoring method, characterized in that: The method comprises: Preprocessing the array photoelectric pulse wave PPG signal to obtain a standardized array PPG signal and an adjacency matrix of a graph structure corresponding to the PPG signal; Determining spatial features and temporal features corresponding to the array PPG signal based on the standardized array PPG signal and the adjacency matrix; Fusing the spatial feature and the temporal feature to obtain a space-time feature; Determining a predicted blood pressure value based on the space-time characteristics; Based on the blood pressure prediction value, determining a blood pressure interval of systolic pressure and a blood pressure interval of diastolic pressure corresponding to the array PPG signal; The preprocessing of the array photoelectric pulse wave PPG signal to obtain a standardized array PPG signal and an adjacency matrix of a graph structure corresponding to the PPG signal includes: Performing standardization processing on the array PPG signal to obtain the standardized array PPG signal; Selecting an array PPG signal of a first length based on a sliding window; For each selected array PPG signal of the first length, the adjacency matrix is constructed based on the average value of the sum of the maximum mutual information coefficients between each PPG signal and the adjacent PPG signals.
2. The method according to claim 1, characterized in that The determining, based on the standardized array PPG signal and the adjacency matrix, the spatial features and the temporal features corresponding to the array PPG signal comprises: Obtaining spatial features corresponding to the array PPG signal based on a graph convolutional neural network; The time characteristics corresponding to the array PPG signal are obtained based on the long short-term memory neural network.
3. The method according to claim 1, characterized in that The step of fusing the spatial feature and the temporal feature to obtain the space-time feature includes: The spatial feature and the temporal feature are fused through a fully connected layer to obtain the space-time feature.
4. The method according to claim 1, characterized in that: The determining of the blood pressure prediction value based on the space-time feature comprises: The space-time characteristics are predicted based on a linear regression algorithm to obtain a predicted value of systolic pressure and a predicted value of diastolic pressure.
5. The method according to claim 1, characterized in that The determining, based on the blood pressure prediction value, a blood pressure interval of systolic pressure and a blood pressure interval of diastolic pressure corresponding to the array PPG signal comprises: Determining a first upper limit value and a first lower limit value corresponding to the systolic pressure based on the predicted value of the systolic pressure and the predicted value of the diastolic pressure; Determining the blood pressure interval of the systolic blood pressure based on the first upper limit value and the first lower limit value; Determining a second upper limit value and a second lower limit value corresponding to the diastolic pressure based on the predicted value of the systolic pressure and the predicted value of the diastolic pressure; Based on the second upper limit value and the second lower limit value, a blood pressure interval of the diastolic pressure is determined.
6. A training method for a blood pressure monitoring model, characterized in that: The method comprises: Preprocessing the array PPG signal in the sample data set to obtain a standardized array PPG signal and an adjacency matrix of a graph structure corresponding to the array PPG signal, wherein the sample data set includes the array PPG signal and the labeled values of the systolic pressure and the diastolic pressure corresponding to the array PPG signal; Using the standardized array PPG signal and the adjacency matrix as inputs of a first sub-model in a blood pressure monitoring model to obtain a blood pressure prediction value, wherein the blood pressure prediction value includes a systolic pressure prediction value and a diastolic pressure prediction value; Using the predicted value of systolic pressure and the predicted value of diastolic pressure as inputs of the second sub-model to obtain a blood pressure interval range of systolic pressure and a blood pressure interval range of diastolic pressure; The preprocessing of the array PPG signal in the sample data set to obtain the standardized array PPG signal and the adjacency matrix of the graph structure corresponding to the array PPG signal includes: Performing standardization processing on the array PPG signal to obtain the standardized array PPG signal; Selecting an array PPG signal of a first length based on a sliding window; For each selected array PPG signal of the first length, the adjacency matrix is constructed based on the average value of the sum of the maximum mutual information coefficients between each PPG signal and the adjacent PPG signals.
7. The method according to claim 6, characterized in that The step of using the standardized array PPG signal and the adjacency matrix as inputs of a first sub-model in a blood pressure monitoring model to obtain a blood pressure prediction value includes: Obtaining spatial features corresponding to the array PPG signal based on a graph convolutional neural network; Obtaining the time characteristics corresponding to the array PPG signal based on a long short-term memory neural network; The spatial features and the temporal features are fused through a fully connected layer, and the fused spatial and temporal features are predicted based on a linear regression algorithm to obtain a predicted value of systolic pressure and a predicted value of diastolic pressure; The first sub-model adjusts the parameters of the first sub-model and performs model training based on the difference between the labeled value of the systolic pressure and the predicted value of the systolic pressure, and the difference between the labeled value of the diastolic pressure and the predicted value of the diastolic pressure.
8. The method according to claim 6, characterized in that The predicted value of the systolic pressure and the predicted value of the diastolic pressure are used as inputs of the second sub-model to obtain a blood pressure interval range of the systolic pressure and a blood pressure interval range of the diastolic pressure, including: The second sub-model performs Gaussian regression processing based on the predicted value of the systolic blood pressure to determine a first upper boundary value and a first lower boundary value corresponding to the systolic blood pressure, and determines the blood pressure interval of the systolic blood pressure based on the first upper boundary value and the first lower boundary value; The second sub-model performs Gaussian regression processing based on the predicted value of the diastolic pressure to determine a second upper boundary value and a second lower boundary value corresponding to the diastolic pressure, and determines the blood pressure range of the diastolic pressure based on the second upper boundary value and the second lower boundary value.
9. A blood pressure monitoring device, characterized in that: The device comprises: A preprocessing module, used to preprocess the array photoelectric pulse wave PPG signal to obtain a standardized array PPG signal and an adjacency matrix of a graph structure corresponding to the array PPG signal; A blood pressure value determination module, configured to obtain the spatial features and the temporal features corresponding to the array PPG signal based on the standardized array PPG signal and the adjacency matrix; perform feature fusion on the spatial features and the temporal features to obtain the space-time features, and determine the blood pressure prediction value based on the space-time features; a blood pressure interval determination module, configured to determine a blood pressure interval of systolic pressure and a blood pressure interval of diastolic pressure corresponding to the array PPG signal based on the blood pressure prediction value; The preprocessing of the array photoelectric pulse wave PPG signal to obtain a standardized array PPG signal and an adjacency matrix of a graph structure corresponding to the PPG signal includes: Performing standardization processing on the array PPG signal to obtain the standardized array PPG signal; Selecting an array PPG signal of a first length based on a sliding window; For each selected array PPG signal of the first length, the adjacency matrix is constructed based on the average value of the sum of the maximum mutual information coefficients between each PPG signal and the adjacent PPG signals.
10. A blood pressure monitoring model training device, characterized in that: The device comprises: A preprocessing module, used to preprocess the array PPG signal in the sample data set to obtain a standardized array PPG signal and an adjacency matrix of a graph structure corresponding to the array PPG signal, wherein the sample data set includes the array PPG signal and the labeled values of the systolic pressure and the diastolic pressure corresponding to the array PPG signal; a blood pressure value determination module, configured to use the standardized array PPG signal and the adjacency matrix as inputs of a first sub-model in a blood pressure monitoring model to obtain a blood pressure prediction value, wherein the blood pressure prediction value includes a systolic pressure prediction value and a diastolic pressure prediction value; A blood pressure interval determination module, used to use the predicted value of systolic pressure and the predicted value of diastolic pressure as inputs of a second sub-model to obtain a blood pressure interval range of systolic pressure and a blood pressure interval range of diastolic pressure; The preprocessing of the array PPG signal in the sample data set to obtain the standardized array PPG signal and the adjacency matrix of the graph structure corresponding to the array PPG signal includes: Performing standardization processing on the array PPG signal to obtain the standardized array PPG signal; Selecting an array PPG signal of a first length based on a sliding window; For each selected array PPG signal of the first length, the adjacency matrix is constructed based on the average value of the sum of the maximum mutual information coefficients between each PPG signal and the adjacent PPG signals.
11. An electronic device, characterized in that: The electronic device comprises: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor so that the at least one processor can execute the method of any one of claims 1-5 or 6-8.
12. A computer-readable storage medium, characterized in that: The storage medium includes a set of computer executable instructions, which, when executed, are used to execute the blood pressure monitoring method described in any one of claims 1 to 5 or the training method of the blood pressure monitoring model described in any one of claims 6 to 8.
Citation Information
Patent Citations
Noninvasive continuous blood pressure measuring device
CN105877723A
Continuous blood pressure measurement method based on photoelectric array enhanced signal multi-feature extraction
CN112998676A
Noninvasive blood glucose detection method, processor and device based on graph convolutional network
CN113288131A
Blood pressure interval measurement method based on residual space-time feature fusion and Gaussian process regression
CN114748047A
Wearable device for non-invasive administration of continuous blood pressure monitoring without cuffing
US20200121258A1
