Blood pressure detection method, device, apparatus and medium

Abstract

The application provides a blood pressure detection method, device, equipment and medium, and belongs to the technical field of medical detection. The blood pressure detection method compares the blood pressure data of the same detection position of the detection object at different times, and compares the blood pressure data of at least two different detection positions of the detection object at the same time, to determine whether the blood pressure of the detection object at the current time is abnormal. The blood pressure detection method is based on the comparison of the blood pressure of the same position of the detection object at different times, and the comparison of the blood pressure of different positions at the same time, to obtain the blood pressure abnormality judgment result of the detection object, without relying on the preset standard, and the detection result is accurate and is not prone to blood pressure abnormality missed detection or false detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical detection, and in particular to a blood pressure detection method, device, equipment and medium. BACKGROUND

[0002] In modern society, due to the combined effects of unreasonable dietary structure and work-rest time, insufficient exercise, smoking and drinking and other risk factors, the incidence of chronic cardiovascular diseases, hypertension and other blood pressure-related diseases continues to rise, and the age of patients is gradually decreasing. Blood pressure-related diseases pose an increasingly greater threat to human physical health, and their prevention and treatment have always been a concern. Therefore, accurately detecting and timely discovering abnormal changes in blood pressure is crucial for the prevention and treatment of diseases.

[0003] The existing blood pressure detection method usually sets the judgment standard of normal blood pressure in advance, then collects the blood pressure data of the detection object, and judges whether the detection object has blood pressure abnormalities according to the pre-set judgment condition. For example, an existing specific implementation method of blood pressure detection includes: using a pre-trained time series model to predict the blood pressure trend after the current time according to the blood pressure data within a period of time before the current time, to confirm whether the detection object will have blood pressure abnormalities. The pre-trained time series model is obtained by comparing the classification result of the classifier or the blood pressure threshold. The class of the classifier and the blood pressure threshold are both pre-set judgment conditions, however, different people may have different physical conditions, medical histories and other factors, and the corresponding blood pressure characteristics are also different. Therefore, the existing blood pressure detection method is prone to miss detection or false detection of blood pressure abnormalities, i.e., the accuracy of blood pressure abnormality detection is low, which is not conducive to the prevention of blood pressure-related diseases. SUMMARY

[0004] To solve the existing technical problems, the present application provides a blood pressure detection method, device, equipment and computer readable storage medium with high abnormality detection accuracy.

[0005] According to a first aspect of an embodiment of the present application, a blood pressure detection method is provided, comprising:

[0006] comparing the current blood pressure data corresponding to the current time and the previous blood pressure data corresponding to the previous time of the same detection site of the detection object, determining the first similarity between the current blood pressure data and the previous blood pressure data of the same detection site; and

[0007] comparing the current blood pressure data corresponding to the current time of different detection sites of the detection object, determining the second similarity between the current blood pressure data of different detection sites;

[0008] determine whether the detection object has blood pressure abnormality at the current time according to the first similarity and the second similarity.

[0009] According to a second aspect of the embodiment of the present application, a blood pressure detection device is provided, comprising:

[0010] The first determining module is configured to compare current blood pressure data corresponding to the current time and previous blood pressure data corresponding to a previous time of a same detection site of a detection object, and determine a first similarity between the current blood pressure data and the previous blood pressure data of the same detection site.

[0011] The second determining module is configured to compare current blood pressure data corresponding to the current time of different detection sites of the detection object, and determine a second similarity between the current blood pressure data of the different detection sites.

[0012] The third determining module is configured to determine whether the detection object has blood pressure abnormality at the current time according to the first similarity and the second similarity.

[0013] According to a third aspect of the embodiment of the present application, a blood pressure detection device is provided, comprising a memory and a processor, the memory stores a computer readable program, and the processor implements the blood pressure detection method when executing the computer readable program.

[0014] According to a fourth aspect of the embodiment of the present application, a computer readable storage medium is provided, the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the blood pressure detection method.

[0015] As can be seen from the above, the blood pressure detection method provided by the embodiment of the present application compares the blood pressure data corresponding to different times of a same detection site in multiple detection sites of a detection object to obtain a first similarity representing the blood pressure change of the same detection site at different times, and compares the blood pressure data of at least two different detection sites in the multiple detection sites of the detection object at the same time to obtain a second similarity representing the blood pressure change of the different detection sites at the same time, so as to determine whether the detection object has blood pressure abnormality at the current time through the first similarity and the second similarity. Therefore, the blood pressure detection method provided by the embodiment of the present application always compares the blood pressure of a same site of a detection object at different times and compares the blood pressure of different sites at the same time to obtain the blood pressure abnormality judgment result of the detection object, without relying on the preset standard, so that the detection result is accurate and the blood pressure abnormality is not easy to be missed or misjudged. BRIEF DESCRIPTION OF DRAWINGS

[0016] The accompanying drawings are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification. The drawings are not intended to be restrictive in any way. In the drawings:

[0017] Figure 1 A flowchart of a blood pressure detection method according to some embodiments of the present application;

[0018] Figure 2 A flowchart of determining a first similarity in a blood pressure detection method according to some embodiments of the present application;

[0019] Figure 3 A flowchart of determining a second similarity in a blood pressure detection method according to some embodiments of the present application;

[0020] Figure 4 A schematic diagram of a calculation manner of acquiring blood pressure data of each time of the first detection site and the second detection site according to some embodiments of the present application;

[0021] Figure 5 A schematic diagram of blood pressure time series data corresponding to each detection site in a blood pressure detection method according to some embodiments of the present application;

[0022] Figure 6 A block diagram of a blood pressure detection device according to some embodiments of the present application;

[0023] Figure 7 A schematic diagram of a specific implementation structure of a blood pressure detection device according to some embodiments of the present application;

[0024] Figure 8 A block diagram of a blood pressure detection device according to some embodiments of the present application;

[0025] Figure 9 A schematic diagram of an appearance of a blood pressure detection device according to some embodiments of the present application. DETAILED DESCRIPTION

[0026] The technical solutions of the present application will be further described in detail below in combination with the accompanying drawings and specific embodiments.

[0027] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the present application. As used in this description, the terms "may" and "could" include any one of a wide and varied array of combinations of the associated listed items in any and all of its senses, all of which mean to include a or a possibility of there being a combination of any one of the items listed.

[0028] Reference will now be made to Figure 1This is a flowchart illustrating a blood pressure detection method provided according to some embodiments of this application. In some embodiments, the blood pressure detection method provided by this application specifically includes S02, S04, and S06, and the specific description of each step is as follows.

[0029] S02: Compare the current blood pressure data of the same detection site of the test subject at the current moment with the previous blood pressure data at a previous moment, and determine the first similarity between the current blood pressure data of the same detection site and the previous blood pressure data.

[0030] The test subject refers to the object on which blood pressure is measured using the blood pressure measuring device implemented according to the blood pressure measuring method provided in the embodiments of this application. This object can be a person or other biological organism. The testing site is the part of the test subject's body used to collect blood pressure data. For example, when the blood pressure measuring device is a wristband, the corresponding testing site can be any part of the wristband that comes into contact with the wrist.

[0031] Blood pressure data refers to data used to characterize the magnitude of a person's blood pressure. That is, the blood pressure data value can be the direct blood pressure value of the person being tested, or it can be a value determined by its relationship to the person's blood pressure value. Based on the blood pressure data value, the person's blood pressure value can be determined. For example, blood pressure data can be directly collected by a blood pressure sensor, or it can be data obtained by processing and transforming the data collected by the blood pressure sensor.

[0032] The blood pressure detection method provided in this application detects the blood pressure of the subject based on blood pressure data collected by a blood pressure sensor. At a certain detection frequency (different from the sensor's acquisition frequency), the method determines whether the subject has abnormal blood pressure at regular intervals, based on blood pressure data collected at the current time and / or before the current time. The detection frequency is the frequency at which abnormal blood pressure is determined. The current time and all previous times are the dividing points of the detection frequency; that is, the time in this application refers to the dividing point of the detection frequency. The time between two adjacent times is the detection cycle corresponding to the detection frequency (the reciprocal of the detection frequency).

[0033] Current blood pressure data refers to the blood pressure data at the current moment, while previous blood pressure data refers to the blood pressure data at a previous moment before the current moment. A previous moment can be the moment before the current moment that is adjacent to the current moment, or it can be multiple moments before the current moment, or it can be all moments before the current moment.

[0034] The first similarity is used to characterize the degree of similarity or similarity category between current blood pressure data and prior blood pressure data at the same detection site. The first similarity can be obtained by directly comparing the current blood pressure data and prior blood pressure data, or by processing the current blood pressure data and prior blood pressure data at the same detection site separately and then comparing the processed data. For example, the first similarity can be obtained by comparing the feature-extracted data of the current blood pressure data and the feature-extracted data of the prior blood pressure data. Furthermore, in other embodiments, comparing the current blood pressure data and prior blood pressure data at the same detection site can also involve performing a first similarity comparison between the current blood pressure data corresponding to the same detection site and multiple blood pressure data from before the current time, respectively, to obtain the first similarity. In other words, the first similarity can be determined based on the comparison results of multiple first similarity comparisons.

[0035] The higher the degree of similarity represented by the first similarity, the lower the degree of change in the current blood pressure at the same detection site compared to the previous blood pressure, and the more it indicates that the current blood pressure of the test subject is normal, and vice versa.

[0036] The similarity category for the first similarity score can include, but is not limited to, similarity and dissimilarity. If similar, it means that the current blood pressure at the same testing site has not changed much compared to the previous blood pressure, indicating that the current blood pressure of the test subject may be stable compared to the previous blood pressure, i.e., it may be normal. If dissimilarity, it means that the current blood pressure at the same testing site has changed significantly compared to the previous blood pressure, indicating that the current blood pressure of the test subject may be unstable compared to the previous blood pressure, i.e., it may be abnormal.

[0037] S04: Compare the current blood pressure data of different detection sites of the test subject at the current moment to determine the second similarity between the current blood pressure data of different detection sites.

[0038] Different detection sites include two or more detection sites. These different detection sites can include the same detection site from step S02, or at least two other detection sites besides the same detection site. Therefore, based on the blood pressure detection method provided in this application embodiment, the detection object corresponds to two or more detection sites, i.e., it has multiple detection sites. One of the multiple detection sites is used to compare blood pressure changes at different times, and at least two of the multiple detection sites are used to compare blood pressure changes at different detection sites.

[0039] The second similarity is used to characterize the degree of similarity or similarity category between current blood pressure data from different detection sites. The second similarity can be obtained by directly comparing the current blood pressure data from different detection sites, or by processing the current blood pressure data from different detection sites separately and then comparing the processed data. For example, the second similarity can be obtained by extracting features from the current blood pressure data from different detection sites and comparing the results. Furthermore, in other embodiments, comparing the current blood pressure data from different detection sites can also involve performing pairwise second similarity comparisons between the current blood pressure data from three or more detection sites to obtain the second similarity; that is, the second similarity can be determined based on the comparison results of multiple second similarity comparisons.

[0040] The higher the degree of similarity represented by the second similarity, the lower the degree of change between the current blood pressure at the different detection sites, and the more it indicates that the current blood pressure of the test subject is normal, and vice versa.

[0041] The second similarity category can include, but is not limited to, similarity and dissimilarity. If similar, it means that the current blood pressure at different test sites does not change much, indicating that the current blood pressure at different sites of the test subject is stable, which is likely normal. If dissimilarity, it means that the current blood pressure at different test sites changes significantly, indicating that the current blood pressure of the test subject may be abnormal.

[0042] S06: Based on the first similarity and the second similarity, determine whether the subject has abnormal blood pressure at the current moment.

[0043] Based on the foregoing analysis, we know that the first similarity score characterizes the degree of blood pressure change at the same location of the test subject at different times, while the second similarity score characterizes the degree of blood pressure change at different locations of the test subject at the same time. Therefore, based on the first and second similarity scores corresponding to the current time, we can determine the blood pressure status of the test subject at the current time, such as whether the test subject's blood pressure is abnormal at the current time.

[0044] The blood pressure detection method provided in this application compares the similarity of blood pressure data from the same detection site at different times across multiple detection sites of the test subject to obtain a first similarity, representing the blood pressure changes at the same detection site at different times. It also compares the similarity of blood pressure data from at least two different detection sites at the same time across multiple detection sites of the test subject to obtain a second similarity, representing the blood pressure changes at different detection sites at the same time. Based on the first and second similarities, the method determines whether the test subject has abnormal blood pressure at the current time. Therefore, the blood pressure detection method provided in this application always compares the blood pressure at the same site of the test subject at different times, and compares the blood pressure at different sites at the same time to obtain the result of judging abnormal blood pressure. It does not rely on preset standards, has high accuracy, and is less prone to missed or false detections of abnormal blood pressure.

[0045] Please see Figure 2 The diagram shown illustrates the process of determining the first similarity in a blood pressure detection method according to some embodiments of this application. Specifically, in some embodiments, S02 specifically includes S022 and S024, and the steps are described below.

[0046] S022: Input the first current blood pressure time series data corresponding to the same detection site at the current time into the first feature extraction network model for feature extraction to obtain the first feature data and the second feature data.

[0047] The first current blood pressure time series data is the time series data corresponding to the current blood pressure data of the same detection site as the last data. It is a one-dimensional data composed of the current blood pressure data corresponding to the same detection site and the previous blood pressure data corresponding to multiple times before the current time arranged in time.

[0048] The first feature extraction network model is a neural network model trained on a blood pressure dataset to learn the feature representation of blood pressure data. The trained neural network model can extract features from the input blood pressure data to obtain blood pressure feature data.

[0049] Specifically, in some embodiments, the first feature extraction network model can be a Long Short-Term Memory (LSTM) network model trained on a blood pressure dataset. The LSTM network model has feature extraction layers corresponding to the number of blood pressure time-series data points it receives as input. Each feature extraction layer is an LSTM feature extraction unit layer. The output of the previous LSTM feature extraction unit layer serves as the input to the next LSTM feature extraction unit layer. Simultaneously, each LSTM feature extraction unit layer is also used to input the blood pressure data at the corresponding time point from the blood pressure time-series data. For example, one input to the last LSTM feature extraction unit layer is the output of the penultimate LSTM feature extraction unit layer, and the other input is the current blood pressure data from the first current blood pressure time-series data. The LSTM network model can record the output of the previous time point as the input to the current time point. Combined with the blood pressure data input at the current time point, the current time point of the LSTM network model includes two inputs. This allows the model to utilize historical data, possess temporal sequence characteristics, and accurately extract effective blood pressure feature data from the blood pressure time-series data, thereby avoiding the adverse effects of noise on the accuracy of blood pressure detection.

[0050] The first feature data is the feature data output by the first feature extraction network model at the previous time step, and the second feature data is the feature data output by the first feature extraction network model at the current time step. Taking a Long Short-Term Memory (LSTM) network model as an example, the first feature data is the feature data output by the penultimate LSTM feature extraction unit layer, and the second feature data is the feature data output by the last LSTM feature extraction unit layer. The feature data output by each LSTM feature extraction unit layer contains information about the blood pressure data from the time step before that point.

[0051] In other embodiments, the first feature extraction network model may also be a recurrent network model such as GRU (Gated Recurrent Unit) that can extract features from blood pressure time series data.

[0052] S024: Input the first feature data and the second feature data into the first Siamese network model to obtain the first similarity output by the first Siamese network model.

[0053] The first feature data is used to characterize the blood pressure at the same detection site before the current time, and the second feature data is used to characterize the blood pressure at the same detection site at the current time. The first similarity between the first feature data and the second feature data is used to characterize the change in blood pressure at the same detection site at the current time relative to the blood pressure before the current time.

[0054] In this embodiment, the first comparison network model used to compare the first feature data and the second feature data is a first Siamese network model. In other embodiments, the first comparison network model can also be any other neural network model that compares the similarity between two input data. The first Siamese network model is used to compare the Euclidean distance between the first feature data and the second feature data, thereby outputting a first similarity value representing the similarity between the first feature data and the second feature data. The first similarity value is either 1 or 0. A first similarity value of 1 indicates that the first feature data and the second feature data are similar, meaning that the blood pressure at the same detection site at the current time has not changed much compared to the blood pressure before the current time, i.e., the blood pressure at different times is relatively stable. A first similarity value of 0 indicates that the first feature data and the second feature data are dissimilar, meaning that the blood pressure at the same detection site at the current time has changed significantly compared to the blood pressure before the current time, i.e., the blood pressure is relatively unstable.

[0055] Please see Figure 3 The diagram shown illustrates the process of determining a second similarity in a blood pressure detection method according to some embodiments of this application. Specifically, in some embodiments, S04 specifically includes steps S042 and S044, described below.

[0056] S042: Input the current blood pressure data corresponding to the two different detection sites at the current time into the second feature extraction network model for feature extraction to obtain the third feature data and the fourth feature data.

[0057] The second feature extraction network model may include, but is not limited to, two convolutional neural network (CNN) models, each trained on a blood pressure dataset to learn feature representations of blood pressure data. These two CNN models are used to extract features from the current blood pressure data corresponding to two different detection sites, respectively, to output third and fourth feature data. The current blood pressure data from the first detection site is input into the first CNN model of the second feature extraction network model, causing the first CNN model to output the third feature data. The current blood pressure data from the second detection site is input into the second CNN model of the second feature extraction network model, causing the second CNN model to output the fourth feature data. In other embodiments, the second feature extraction network model may also be other neural network models capable of extracting spatial features from the input data.

[0058] S044: Input the third feature data and the fourth feature data into the second Siamese network model to obtain the second similarity output by the second Siamese network model.

[0059] The third feature data is used to characterize the blood pressure at the first detection site before the current time, and the fourth feature data is used to characterize the blood pressure at the second detection site at the current time. The second similarity between the third feature data and the fourth feature data is used to characterize the change between the current blood pressure at the first detection site and the current blood pressure at the second detection site.

[0060] In this embodiment, the second comparison network model used to compare the third feature data and the fourth feature data is a second Siamese network model. In other embodiments, the second comparison network model can also be any other neural network model that compares the similarity between two input data. The second Siamese network model is used to compare the Euclidean distance between the third feature data and the fourth feature data, thereby outputting a second similarity value that characterizes the similarity between the third feature data and the fourth feature data. The second similarity value is either 1 or 0. A second similarity value of 1 indicates that the third feature data and the fourth feature data are similar, meaning that the current blood pressure at the first detection site does not change much compared to the current blood pressure at the second detection site, indicating that the blood pressure at different detection sites is relatively stable. A second similarity value of 0 indicates that the third feature data and the fourth feature data are dissimilar, meaning that the current blood pressure at the first detection site changes significantly compared to the current blood pressure at the second detection site, indicating that the blood pressure at different detection sites is relatively unstable.

[0061] In some embodiments, before steps S02 and S04, the blood pressure detection method provided in this application further includes: acquiring blood pressure data collected in real time at each detection site according to a preset acquisition frequency; calculating blood pressure data corresponding to each detection site at each time point based on the blood pressure acquisition data to obtain the current blood pressure time series data corresponding to each detection site; the blood pressure data corresponding to each time point is calculated based on the blood pressure acquisition data within the corresponding calculation time period, and the calculation time periods corresponding to two adjacent times points have a preset overlap rate. It should be noted here that, unless otherwise specified, the time in this application refers to the calculation time of the blood pressure data, that is, the time corresponding to the detection frequency.

[0062] The blood pressure detection method provided in some embodiments of this application uses two blood pressure sensors to synchronously collect blood pressure data at a first detection site and a second detection site at the same acquisition frequency, respectively obtaining blood pressure data corresponding to the first and second detection sites. For details, please refer to... Figure 4The diagram illustrates the calculation method for obtaining blood pressure data at various times for the first and second detection sites according to an embodiment of this application. The blood pressure data acquired at each collection time point for the first detection site includes a0, a1, a2, a3, ..., and the blood pressure data acquired at each collection time point for the second detection site includes b0, b1, b2, b3, ... Taking a collection frequency of 1 second and a calculation time period of 6 seconds for each time point as an example, one blood pressure data point can be obtained for each detection site every 1 second. After acquiring valid blood pressure data, the calculation of the blood pressure data for the first time point is performed after 6 seconds. Taking the first detection site as an example, the blood pressure data P11 corresponding to the first time point is calculated as follows: 6 seconds after acquiring valid blood pressure data a0 is the first calculation time period, and the blood pressure data within the first calculation time period are a0, a1, a2, ..., a6. Based on the blood pressure data a0, a1, a2, ..., a6, the corresponding calculation is performed to obtain the blood pressure data P11 corresponding to the first detection site at the first time point.

[0063] In this embodiment, to improve the temporal logic of the first feature extraction network model and / or the second feature extraction network model, so as to improve the accuracy of blood pressure detection without causing the current computing power to be insufficient to meet the implementation of the blood pressure detection method, the calculation time intervals corresponding to two adjacent moments of the same detection site are set to have a certain overlap rate. For example, in some embodiments, the above overlap rate can be 25%. If the time when valid blood pressure data a0 is collected is defined as the start time, then the second calculation time interval corresponding to the second moment of the first detection site should be the period from 4.5s to 10.5s. The blood pressure data collected within the second calculation time interval are a5, a6, ..., a10. Based on the blood pressure data a5, a6, ..., a10, corresponding calculations are performed to obtain the blood pressure data P12 corresponding to the second moment of the first detection site.

[0064] Similarly, following the above calculation method, blood pressure data P13 at the third moment, P14 at the fourth moment, ... at the first detection site can be obtained, as well as blood pressure data P21 at the first moment, P22 at the second moment, P23 at the third moment, P24 at the fourth moment, ... at the second detection site. Specifically, in some embodiments, calculating the blood pressure data at the corresponding moment based on the blood pressure data collected within each calculation time period can be achieved by averaging the blood pressure data collected within each calculation time period, using the average value of the blood pressure data collected within each calculation time period as the blood pressure data at the corresponding moment.

[0065] Please see Figure 5As shown, this is a schematic diagram of blood pressure time-series data corresponding to various detection sites in the blood pressure detection method provided according to the embodiments of this application. To facilitate observation of the temporal relationship between blood pressure data at two adjacent moments, one of the two adjacent blood pressure data points is intentionally set as a solid-line box, and the other as a dashed-line box. Furthermore, Figure 5 The diagram illustrates the blood pressure change curves at the first and second detection sites. The horizontal extension width of the blood pressure data represents its corresponding calculation time period. Clearly, for blood pressure data from two adjacent moments, since their calculation time periods partially overlap, the blood pressure data from the later moment includes a portion of the blood pressure data from the previous moment, which helps improve detection accuracy. Starting from the first moment, blood pressure data is calculated every certain period of time (4.5 seconds in the example above) according to the detection frequency. If this period of time is a first preset time, and the duration of the calculation time period corresponding to each moment is a second preset time, the first preset time is shorter than the second preset time. As an optional implementation, the first preset time is 75% of the second preset time.

[0066] In some embodiments, the blood pressure data at each time point are arranged chronologically to obtain the blood pressure time-series data for each undetected site. Specifically, the blood pressure time-series data for the first detection site at the current time is the first current blood pressure time-series data P1, and the blood pressure time-series data for the second detection site at the current time is the second current blood pressure time-series data P2. The last data point in the blood pressure time-series data for each detection site at each time point is the blood pressure data for that detection site at that corresponding time point.

[0067] If the first detection site is the same as the above-mentioned detection site, then the input data of the first feature extraction network model at the current time is the first current blood pressure time series data P1, and the input data of the second feature extraction network model at the current time is the last blood pressure data in the first current blood pressure time series data P1 and the second current blood pressure time series data P2, that is, the current blood pressure data of the first detection site and the current blood pressure data of the second detection site.

[0068] Furthermore, in some embodiments, S042 specifically includes: converting the current blood pressure data corresponding to two different detection sites at the current time into image data, and inputting the two image data into the corresponding convolutional neural network models for feature extraction.

[0069] In the blood pressure detection method provided in the embodiments of this application, the convolutional neural network model used in the second feature extraction network model is a one-dimensional convolutional neural network model. However, the ability of a one-dimensional convolutional neural network model to extract spatial features is limited, and it is more suitable for extracting features from image data. In order to extract the third and fourth feature data more accurately, before inputting the current blood pressure data of the first and second detection sites into the first and second convolutional neural network models in the second feature extraction network model, respectively, the current blood pressure data corresponding to the first and second detection sites are first converted into image data, and then the image data is input into the corresponding convolutional neural network models for image feature extraction. In this embodiment, the first and second convolutional neural network models are models trained on the blood pressure image dataset converted from the blood pressure dataset.

[0070] In some embodiments, an optional implementation of converting the blood pressure data corresponding to each detection site at each time point, i.e., the time-series blood pressure data corresponding to each site, into image data is as follows: convert each blood pressure data into a Gram angle field map. The converted Gram angle field map is then input into the corresponding convolutional neural network model to obtain the corresponding feature data.

[0071] Specifically, the main steps for converting blood pressure time-series data from various testing sites into Gram field plots include the following four steps.

[0072] Step 1: Normalize the blood pressure time series data to scale each data point in the blood pressure time series data to the range of [-1, 1].

[0073] Step 2: Convert the normalized data into polar coordinate data. Specifically, calculate the corresponding polar coordinates from the sequence values ​​of the normalized data to obtain polar coordinate data.

[0074] Step 3: Convert the polar coordinate data into matrix data. The matrix data can be GASF or GADF matrix data.

[0075] Step 4: Convert each element in the matrix data into a pixel value within the corresponding grayscale range to obtain the corresponding Gram area map. The corresponding grayscale range can be, but is not limited to, a grayscale value range of 0-256. Specifically, a heatmap method can be used to convert the matrix data into an image, where the values ​​at different positions in the matrix data represent the colors in the image.

[0076] In some embodiments, S06 specifically includes: determining whether both the first similarity and the second similarity satisfy the corresponding preset conditions; if not, determining that the detected object has abnormal blood pressure at the current moment.

[0077] Whether the first similarity and the second similarity meet the corresponding preset conditions can be specifically defined as whether the values ​​of the first similarity and the second similarity are both 1. If yes, it means that both the first similarity and the second similarity meet the corresponding preset conditions, indicating that the blood pressure changes at the same detection site of the test subject are relatively stable at different times, and the blood pressure at different detection sites is also relatively stable, and the blood pressure of the test subject is normal at the current time. If no, it means that at least one of the first similarity and the second similarity does not meet the corresponding preset conditions, indicating that the test subject has abnormal blood pressure at the current time.

[0078] In some embodiments, after S06, the method further includes: issuing an early warning when it is determined that the subject has an abnormal blood pressure at the current moment.

[0079] The warning system can display corresponding images and / or text, or provide sound and / or light alerts. At any given moment, once an abnormal blood pressure is detected in the subject, a warning is issued, promptly reminding the subject to pay attention to the abnormality and take appropriate measures, effectively improving the subject's health index.

[0080] In some other embodiments, after S06, the method further includes: when it is determined that the test subject has abnormal blood pressure at the current moment, if the number of times the test subject has abnormal blood pressure in the preset number of times before the current moment meets the preset number, then a warning is issued.

[0081] The blood pressure detection method provided in this application always determines whether blood pressure is abnormal based on the blood pressure changes of the subject at different times and at different detection locations. It determines that the subject has an abnormal blood pressure at the current moment when either the blood pressure change at the same detection site at different times is relatively large, or the blood pressure change at different detection sites at the same time is relatively large. If an alert is issued every time an abnormal blood pressure is detected, frequent alerts may occur, potentially causing anxiety for the subject. Therefore, the blood pressure detection method provided in this application provides an abnormal blood pressure alert method that does not directly issue an alert when an abnormal blood pressure is detected at the current moment. Instead, it first determines whether the number of times the subject has experienced abnormal blood pressure within the most recent preset number of moments before the current moment reaches a certain threshold, or whether the number of times the subject has experienced abnormal blood pressure within the most recent preset number of moments, including the current moment, reaches a certain threshold before issuing an alert. This avoids frequent alerts. For example, if an abnormal blood pressure is detected at the current moment, and an abnormal blood pressure is detected at 6 out of the last 10 moments, including the current moment, then an alert is issued.

[0082] Please see Figure 6The diagram shows a structural block diagram of a blood pressure detection device provided according to some embodiments of this application. In some embodiments, the blood pressure detection device provided by this application includes a first determining module 100, a second determining module 200, and a third determining module 300. Specifically, the first determining module 100 is used to compare the current blood pressure data corresponding to the same detection site of the test subject at the current time with the previous blood pressure data corresponding to a previous time, and determine a first similarity between the current blood pressure data of the same detection site and the previous blood pressure data. The second determining module 200 is used to compare the current blood pressure data corresponding to different detection sites of the test subject at the current time, and determine a second similarity between the current blood pressure data of different detection sites. The third determining module 300 is used to determine whether the test subject has an abnormal blood pressure at the current time based on the first similarity and the second similarity. The blood pressure detection device and the blood pressure detection method provided by the embodiments of this application can achieve the same detection effect, and will not be described again here.

[0083] Please see Figure 7 The diagram illustrates a specific implementation structure of a blood pressure detection device according to some embodiments of this application. In some embodiments, the first determining module 100 specifically includes a first feature extraction module 101 and a first comparison module 102. The first feature extraction module 101 is used to input the first current blood pressure time series data P1 corresponding to the same detection site at the current time into a first feature extraction network model for feature extraction, obtaining first feature data and second feature data. The first feature extraction module 101 includes a first feature extraction network model, which may be, but is not limited to, a Long Short-Term Memory (LSTM) network model trained based on a blood pressure dataset. The first feature data and second feature data have been described above and will not be repeated here. The first comparison module 102 is used to input the first feature data and second feature data into a first Siamese network model to obtain a first similarity output by the first Siamese network model. That is, the first comparison module 102 specifically includes a first Siamese network model.

[0084] Please continue reading. Figure 7As shown, in some embodiments, the second determining module 200 further includes a second feature extraction module 201 and a second comparison module 202. The second feature extraction module 201 is used to input the current blood pressure data corresponding to two different detection sites at the current time into the second feature extraction network model for feature extraction to obtain third feature data and fourth feature data. Specifically, the two different detection sites here include the same detection site mentioned above, and the blood pressure data corresponding to the two different detection sites at the current time are the last data in the first current blood pressure time series data P1 and the second current blood pressure time series data P2, that is, the current blood pressure data corresponding to the two different detection sites at the current time. Therefore, in this embodiment, the second feature extraction module 201 is specifically used to input the last data in the first current blood pressure time series data P1 and the second current blood pressure time series data P2 into the corresponding convolutional neural network (CNN) model in the second feature extraction network model for feature extraction to obtain the third feature data and the fourth feature data respectively.

[0085] The second comparison module 202 is used to input the third feature data and the fourth feature data into the second Siamese network model to obtain the second similarity output by the second Siamese network model. That is, the second comparison module 202 specifically includes the second Siamese network model.

[0086] In some embodiments, the blood pressure detection device provided in this application further includes an acquisition module. The acquisition module is used to: acquire blood pressure data collected in real time at each detection site according to a preset acquisition frequency; calculate the blood pressure data corresponding to each detection site at each time based on the blood pressure data, and obtain the current blood pressure time series data corresponding to each detection site; the blood pressure data corresponding to each time is calculated based on the blood pressure data collected within the corresponding calculation time period, and the calculation time periods corresponding to two adjacent times have a preset overlap rate.

[0087] Furthermore, the aforementioned different detection sites include a first detection site (the same detection site mentioned above) and a second detection site. The acquisition module is specifically used to acquire blood pressure data collected in real time from the first and second detection sites at a certain acquisition frequency. Then, according to the aforementioned detection frequency, it calculates the currently acquired blood pressure data at each moment to obtain the blood pressure data corresponding to each moment. The blood pressure data corresponding to each moment are arranged in chronological order to obtain the blood pressure time-series data corresponding to the first and second detection sites. Specifically, the blood pressure time-series data corresponding to the first detection site at the current moment is the first current blood pressure time-series data P1, and the blood pressure time-series data corresponding to the second detection site at the current moment is the second current blood pressure time-series data P2.

[0088] Furthermore, in some embodiments, the second feature extraction module 201 further includes a data conversion module, which is used to convert the current blood pressure corresponding to the two different detection sites at the current time into image data, and input the two image data into the corresponding convolutional neural network models for feature extraction.

[0089] In some embodiments, the third determining module 300 is specifically used to determine whether both the first similarity and the second similarity meet the corresponding preset conditions, and when it is determined that both the first similarity and the second similarity meet the corresponding preset conditions, it determines that the detection object is normal at the current moment; otherwise, it determines that the detection object has abnormal blood pressure at the current moment. Further details can be found in the following sections. Figure 7 As shown, in some embodiments, the third determining module 300 includes an XOR calculation determining module 301. The XOR calculation module 301 is used to calculate the XOR value of the first similarity output by the first comparison module 102 and the second similarity output by the second comparison module 202, and determine whether the test subject has abnormal blood pressure at the current moment based on the XOR value. If the values ​​of the first similarity and the second similarity are both 1, and the XOR value obtained by the XOR calculation determining module 301 is 1, then it is determined that the blood pressure of the test subject is normal at the current moment; if one of the values ​​of the first similarity and the second similarity is 0, and the XOR value obtained by the XOR calculation determining module 301 is 0, then it is determined that the test subject has abnormal blood pressure at the current moment.

[0090] The blood pressure detection device provided in this application further includes an early warning module. In some embodiments, the early warning module is used to issue an early warning when it is determined that the blood pressure of the subject is abnormal at the current moment. In other embodiments, the early warning module is further specifically used to: when it is determined that the blood pressure of the subject is abnormal at the current moment, if the number of times the subject's blood pressure is abnormal within a preset number of time periods prior to the current moment meets a preset number, then issue an early warning.

[0091] Please see Figure 8 The diagram shown is a structural block diagram of a blood pressure detection device provided according to some embodiments of this application. The blood pressure detection device provided in the embodiments of this application includes a memory 1 and a processor 2. The memory 1 stores a computer program that can be executed by the processor 2. When the computer program is executed by the processor 2, it implements the blood pressure detection method provided in any embodiment of this application and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0092] In some embodiments, the blood pressure detection method provided in this application further includes at least two sensors connected to the processor 2. The at least two sensors are used to sense blood pressure at different detection sites of the test subject, to collect blood pressure data corresponding to different detection sites, and to send the blood pressure data to the processor 2 for corresponding processing. Please continue reading.Figure 8 As shown, at least two sensors include a first sensor 3 and a second sensor 4. The first sensor 3 is used to collect blood pressure data from a first detection site of the test subject in real time at a first acquisition frequency, and sends the first blood pressure data to the processor 2. The second sensor 4 is used to collect blood pressure data from a second detection site of the test subject in real time at a second acquisition frequency, and sends the second blood pressure data to the processor 2. The first acquisition frequency and the second acquisition frequency can be the same. The first sensor 3 and the second sensor 4 synchronously collect blood pressure data from the corresponding detection sites at the same acquisition frequency, obtaining first blood pressure data from the first detection site and second blood pressure data from the second detection site. The same detection site in the blood pressure detection method can be the first detection site collected by the first sensor 3, and different detection sites in the blood pressure detection method include the first detection site and the second detection site collected by the second sensor 4. To improve detection accuracy, the distance between the first detection site and the second detection site meets a certain distance condition. For example, the distance between the first detection site and the second detection site is greater than a set distance.

[0093] In some embodiments, the blood pressure monitoring device further includes an early warning module connected to the processor 2. The early warning module is used to issue corresponding early warning alerts based on the early warning information sent by the processor 2. For details, please refer to [link / reference]. Figure 8 As shown, in some embodiments, the warning module specifically includes a display screen 5 and / or a speaker 6. The display screen 5 is used to display warning prompts in the form of images and / or text. The speaker 6 is used to provide audible warning prompts based on the warning prompt information. Furthermore, the display screen 5 is also used to display the blood pressure information of the monitored object. The blood pressure information of the monitored object includes its corresponding blood pressure collection data, blood pressure data at various times, and / or blood pressure change information, etc. In other embodiments, the warning module may also include indicator lights for illuminating the warning prompt information sent by the processor 2.

[0094] Please see Figure 9 The diagram shown is an external schematic of a blood pressure monitoring device provided according to some embodiments of this application. In some embodiments, the blood pressure monitoring device provided in this application is a blood pressure monitoring watch. The blood pressure monitoring wristband includes a strap 11 and a case 12. The case 12 is provided with a display screen 5, and the case 12 contains the aforementioned memory 1 and processor 2. Figure 9(Not shown in the image), one of the first sensor 3 and the second sensor 4 is located on the watch strap, and the other is located on the lower surface of the watch strap or the watch case. Specifically, the first sensor 3 is located on the side of the watch case 12 close to the skin of the subject, and is used to collect blood pressure on the back of the subject's wrist. The second sensor 4 is located on the watch strap at a position opposite to the first sensor 3, and is used to collect blood pressure on the front of the subject's wrist. The blood pressure monitoring watch provided in this application embodiment collects blood pressure data from multiple detection sites of the subject in real time based on at least two sensors, and processes the blood pressure data according to the processor 2, accurately detecting whether the subject's blood pressure is abnormal and providing corresponding early warning prompts, which is beneficial for the subject to effectively prevent blood pressure-related diseases.

[0095] This application also provides a computer-readable storage medium storing a computer program. When executed by a processor, this computer program implements the various processes of the above-described blood pressure detection method embodiments and achieves the same technical effects. To avoid repetition, it will not be described again here. The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0096] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for detecting blood pressure, characterized in that, include: The current blood pressure data of the same detection site of the test object at the current time is compared with the previous blood pressure data at a previous time to determine the first similarity between the current blood pressure data of the same detection site and the previous blood pressure data. as well as, The current blood pressure data corresponding to different detection sites of the detection object at the current time are compared to determine the second similarity between the current blood pressure data of the different detection sites; Based on the first similarity and the second similarity, it is determined whether the detected object exhibits abnormal blood pressure at the current moment; The step of comparing the current blood pressure data of the same detection site at the current moment with the previous blood pressure data at a previous moment, and determining the first similarity between the current blood pressure data and the previous blood pressure data of the same detection site, includes: The first current blood pressure time series data corresponding to the same detection site at the current time is input into the first feature extraction network model for feature extraction to obtain the first feature data and the second feature data. The first feature extraction network model is a long short-term memory network model trained on the blood pressure dataset to learn the feature representation of blood pressure data. The first feature data is the feature data output by the first feature extraction network model at the time before the current time, and the second feature data is the feature data output by the first feature extraction network model at the current time. The first feature data and the second feature data are input into the first Siamese network model to obtain the first similarity output by the first Siamese network model; The step of comparing the current blood pressure data corresponding to different detection sites of the detection object at the current time to determine the second similarity between the current blood pressure data of the different detection sites includes: The current blood pressure data corresponding to the two different detection sites at the current time are input into the second feature extraction network model for feature extraction to obtain the third feature data and the fourth feature data. The second feature extraction network model includes two convolutional neural network models that have been trained on the blood pressure dataset to learn the feature representation of the blood pressure data. The two convolutional neural network models are used to extract features from the current blood pressure data corresponding to the two different detection sites to output the third feature data and the fourth feature data respectively. The third and fourth feature data are input into the second Siamese network model to obtain the second similarity output by the second Siamese network model; The first current blood pressure time series data is the time series data corresponding to the current blood pressure data of the same detection site as the last data. The first current blood pressure time series data is a one-dimensional data composed of the current blood pressure data corresponding to the same detection site and the previous blood pressure data corresponding to multiple times before the current time arranged in time. The output of the previous LSTM feature extraction unit layer of the Long Short-Term Memory Network model is used as the input of the next LSTM feature extraction unit layer. At the same time, each LSTM feature extraction unit layer is also used to input the blood pressure data at the corresponding time in the blood pressure time series data. The first feature data is the feature data output by the penultimate LSTM feature extraction unit layer of the Long Short-Term Memory network model, and the second feature data is the feature data output by the last LSTM feature extraction unit layer of the Long Short-Term Memory network model.

2. The blood pressure detection method according to claim 1, characterized in that, Before comparing the current blood pressure data of the same detection site of the detection object at the current moment with the previous blood pressure data at a previous moment, the method further includes: Obtain blood pressure data collected in real time at each of the aforementioned detection sites according to a preset acquisition frequency; Based on the blood pressure data collected, the blood pressure data corresponding to each detection site at each time point is calculated to obtain the current blood pressure time sequence data corresponding to each detection site; the blood pressure data corresponding to each time point is calculated based on the blood pressure data collected within the corresponding calculation time period, and the calculation time periods corresponding to two adjacent times points have a preset overlap rate.

3. The blood pressure detection method according to claim 1, characterized in that, The step of inputting the current blood pressure data corresponding to two different detection sites at the current time into the second feature extraction network model for feature extraction includes: The current blood pressure data corresponding to the two different detection sites at the current time are converted into image data, and the two image data are respectively input into the corresponding convolutional neural network model for feature extraction.

4. The blood pressure detection method according to any one of claims 1 to 3, characterized in that, The step of determining whether the detected object exhibits abnormal blood pressure at the current moment based on the first similarity and the second similarity includes: Determine whether the first similarity and the multiple second similarities both satisfy the corresponding preset conditions; If not, determine that the subject being tested has an abnormal blood pressure at the current moment.

5. The blood pressure detection method according to any one of claims 1 to 3, characterized in that, Also includes: When it is determined that the detected object has an abnormal blood pressure at the current moment, an early warning is issued; or, When it is determined that the detected object has an abnormal blood pressure at the current time, if the number of times the detected object has an abnormal blood pressure in the preset number of times in the preset number of times in the previous time meets the preset number, then an early warning will be issued.

6. A blood pressure detection device, characterized in that, include: The first determining module is used to compare the current blood pressure data corresponding to the same detection site of the detection object at the current time with the previous blood pressure data corresponding to a previous time, and determine the first similarity between the current blood pressure data of the same detection site and the previous blood pressure data. The second determining module is used to compare the current blood pressure data corresponding to different detection sites of the detection object at the current time, and determine the second similarity between the current blood pressure data of the different detection sites; The third determining module is used to determine whether the detected object has abnormal blood pressure at the current time based on the first similarity and the second similarity. The step of comparing the current blood pressure data of the same detection site at the current moment with the previous blood pressure data at a previous moment, and determining the first similarity between the current blood pressure data and the previous blood pressure data of the same detection site, includes: The first current blood pressure time series data corresponding to the same detection site at the current time is input into the first feature extraction network model for feature extraction to obtain the first feature data and the second feature data. The first feature extraction network model is a long short-term memory network model trained on the blood pressure dataset to learn the feature representation of blood pressure data. The first feature data is the feature data output by the first feature extraction network model at the time before the current time, and the second feature data is the feature data output by the first feature extraction network model at the current time. The first feature data and the second feature data are input into the first Siamese network model to obtain the first similarity output by the first Siamese network model; The step of comparing the current blood pressure data corresponding to different detection sites of the detection object at the current time to determine the second similarity between the current blood pressure data of the different detection sites includes: The current blood pressure data corresponding to the two different detection sites at the current time are input into the second feature extraction network model for feature extraction to obtain the third feature data and the fourth feature data. The second feature extraction network model includes two convolutional neural network models that have been trained on the blood pressure dataset to learn the feature representation of the blood pressure data. The two convolutional neural network models are used to extract features from the current blood pressure data corresponding to the two different detection sites to output the third feature data and the fourth feature data respectively. The third and fourth feature data are input into the second Siamese network model to obtain the second similarity output by the second Siamese network model; Wherein, the first current blood pressure time series data is the time series data corresponding to the current blood pressure data of the same detection site as the last data, and the first current blood pressure time series data is a one-dimensional data composed of the current blood pressure data corresponding to the same detection site and the blood pressure data corresponding to multiple times before the current time arranged in time series; the output of the previous LSTM feature extraction unit layer of the Long Short-Term Memory Network model is used as the input of the next LSTM feature extraction unit layer, and each LSTM feature extraction unit layer is also used to input the blood pressure data of the corresponding time in the blood pressure time series data; the first feature data is the feature data output by the second-to-last LSTM feature extraction unit layer of the Long Short-Term Memory Network model, and the second feature data is the feature data output by the last LSTM feature extraction unit layer of the Long Short-Term Memory Network model.

7. A blood pressure detection device, characterized in that, The device includes a memory and a processor, the memory storing a computer program, and the processor, when executing the computer program, implementing the blood pressure detection method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the blood pressure detection method as described in any one of claims 1 to 5.

Images