CN116919363A

CN116919363A - Personalized blood pressure prediction method and device based on big data characteristics

Info

Publication number: CN116919363A
Application number: CN202210379756.6A
Authority: CN
Inventors: 吴泽剑; 王思翰; 张碧莹; 张洪盼; 李瑞莱; 曹君
Original assignee: Lepu Medical Technology Beijing Co Ltd
Current assignee: Lepu Medical Technology Beijing Co Ltd
Priority date: 2022-04-12
Filing date: 2022-04-12
Publication date: 2023-10-24
Also published as: WO2023197425A1

Abstract

The embodiment of the invention relates to a personalized blood pressure prediction method and device based on big data characteristics, wherein the method comprises the following steps: receiving PPG signals of a plurality of data sources and corresponding blood pressure information to form a first large database; receiving a plurality of PPG signals from a designated data source and corresponding blood pressure information to form a first personality database; training a big data blood pressure prediction model based on the first big database; calibrating, recording and confirming the first personality database based on the big data blood pressure prediction model; training a personality blood pressure prediction model based on the big data blood pressure prediction model, the first personality database and the first calibrated personality data record; and based on the big data blood pressure prediction model, the personalized blood pressure prediction model and the first calibrated personalized data record corresponding to the appointed data source, personalized blood pressure prediction processing is carried out on the PPG signal received from the appointed data source at any moment. The invention solves the personalized prediction problem and improves the prediction precision.

Description

Personalized blood pressure prediction method and device based on big data characteristics

Technical Field

The invention relates to the technical field of data processing, in particular to a personalized blood pressure prediction method and device based on big data characteristics.

Background

The heart is the center of the blood circulation of the human body, and the heart generates blood pressure through regular pulsation so as to supply blood to the whole body to complete metabolism of the human body, and the blood pressure is one of important physiological signals of the human body. Human blood pressure contains two important values: systolic and diastolic pressures. Most of the traditional blood pressure measurement modes adopt invasive interventional measurement or external pressure meter measurement, so that the operation is complex, discomfort of a measured person is easily caused, and the blood pressure meter can not be used for multiple times to achieve the purpose of continuous monitoring.

Photoplethysmography (PPG) signals are a set of signals acquired by a photo sensor from the change in light intensity of a particular light source transmitted through the skin of an observer. The heart beat causes a periodic change in the blood flow in the intravascular unit space, and the corresponding blood volume in the intravascular unit space changes accordingly, which in turn causes a periodic change in the PPG signal reflecting the amount of absorbed light. With the development of artificial intelligence technology, in recent years, a technical scheme for predicting blood pressure according to PPG signals by using an artificial intelligence model is also gradually mature and widely applied. However, the mature model of conventional application needs massive non-personalized PPG data to participate in training to be perfected, which brings about a problem: in the prediction result of the model, the personalized difference features cannot be represented.

Disclosure of Invention

The invention aims at overcoming the defects of the prior art and provides a personalized blood pressure prediction method, a device, electronic equipment and a computer readable storage medium based on big data characteristics, wherein a big data model, namely a big data blood pressure prediction model is firstly constructed based on the structure of a convolutional neural network (Convolutional Neural Networks, CNN) +an artificial neural network (Artificial Neural Network, namely ANN), and the big data is used for training the big data model; then, constructing a personalized model, namely a personalized blood pressure prediction model, by referring to a model structure of a support vector machine model, a random forest model or a decision tree model, and training the personalized model by using appointed personalized data; in practical application, the appointed personalized PPG data and the corresponding PPG calibration data are sent to a big data blood pressure prediction model for big data feature extraction, and then the obtained feature data, the corresponding personalized systolic pressure calibration data and the corresponding personalized diastolic pressure calibration data are sent to a personalized blood pressure prediction model for personalized blood pressure prediction. According to the invention, the characteristic information with higher accuracy can be obtained by utilizing the big data model, and a personalized model suitable for a specified object can be customized so as to predict more matched blood pressure data (systolic pressure and diastolic pressure), so that the problem that the mature model in the current market cannot be subjected to personalized prediction is solved, and the blood pressure prediction precision is improved.

To achieve the above object, a first aspect of the present invention provides a personalized blood pressure prediction method based on big data features, the method including:

receiving photoplethysmography (PPG) signals from a plurality of data sources and corresponding blood pressure information to form a first large database;

receiving a plurality of PPG signals from a designated data source and corresponding blood pressure information to form a first personality database;

training a big data blood pressure prediction model based on the first big database;

performing calibration record confirmation processing on the first personality database based on the trained big data blood pressure prediction model to obtain a corresponding first calibration personality data record;

training a personality blood pressure prediction model based on the trained big data blood pressure prediction model, the first personality database and the first calibrated personality data record;

based on the big data blood pressure prediction model, the personalized blood pressure prediction model and the first calibrated personalized data record corresponding to the appointed data source, personalized blood pressure prediction processing is carried out on PPG signals received from the appointed data source at any moment to generate corresponding first predicted systolic pressure data and first predicted diastolic pressure data.

Preferably, the first big database comprises a plurality of first big data records; the first big data record comprises one or more groups of first acquisition data groups and first calibration data groups; the first acquisition data set includes first PPG data, first systolic data, and first diastolic data; the first calibration data set comprises first calibration PPG data, first calibration systolic pressure data and first calibration diastolic pressure data;

the first personality database includes a plurality of first personality data records; the first personality data record includes first personality PPG data, first personality systolic pressure data, and first personality diastolic pressure data;

the big data blood pressure prediction model consists of a convolutional neural network and an artificial neural network;

the model structure of the personalized blood pressure prediction model adopts a model structure of a support vector machine model, a random forest model or a decision tree model.

Preferably, the receiving the photoplethysmography PPG signals and the corresponding blood pressure information from a plurality of data sources forms a first large database, specifically including:

forming a first received data set from the PPG signals of the single data source received at the present time and corresponding blood pressure information; the first received data set includes a plurality of first received data records; the first received data record comprises a first PPG signal and corresponding first acquired systolic pressure data and first acquired diastolic pressure data;

Performing signal sampling processing on each first PPG signal according to a specified signal sampling frequency to generate corresponding first PPG sampling data;

selecting one first calibration PPG data which meets the standard from a plurality of obtained first PPG sampling data according to a preset calibration data selection standard and is used as corresponding first calibration PPG data; and taking the other remaining first PPG sampled data as the corresponding first PPG data;

taking the first acquired systolic pressure data and the first acquired diastolic pressure data corresponding to the first calibration PPG data as corresponding first calibration systolic pressure data and first calibration diastolic pressure data, and forming a corresponding first calibration data set by the first calibration PPG data, the first calibration systolic pressure data and the first calibration diastolic pressure data; the first acquired systolic pressure data and the first acquired diastolic pressure data corresponding to the first PPG data are used as the corresponding first systolic pressure data and the first diastolic pressure data, and the first PPG data and the corresponding first systolic pressure data and the first diastolic pressure data form the corresponding first acquired data set; and one or more groups of obtained first acquired data sets and one group of first calibration data sets form corresponding first big data records; and adding the first big data record to the first big database.

Preferably, the receiving the multiple PPG signals from the specified data source and the corresponding blood pressure information form a first personality database, which specifically includes:

forming a second received data set by a plurality of PPG signals of appointed data sources received at the present time and corresponding blood pressure information; the second received data set includes a plurality of second received data records; the second received data record comprises a first personalized PPG signal and corresponding first personalized systolic pressure data and first diastolic pressure data;

performing signal sampling processing on each first personalized PPG signal according to a specified signal sampling frequency to generate corresponding first personalized PPG data; the first personality data record is composed of the first personality PPG data, the first personality systolic pressure data and the first personality diastolic pressure data;

counting the total number of first personality data records in the first personality database to generate a corresponding first total number of records; identifying whether the first total number of records is smaller than a preset total number of records threshold value or not; if the total number of the first records is smaller than the total number of records threshold, adding the first personality data record obtained currently into the first personality database; and if the total number of the first records is equal to the threshold value of the total number of the records, replacing the first personality data record with the earliest adding time in the first personality database by using the first personality data record obtained currently.

Preferably, the training the big data blood pressure prediction model based on the first big database specifically includes:

extracting the first big data records from the first big database one by one to serve as current training records; extracting the first calibration PPG data, the first calibration systolic pressure data and the first calibration diastolic pressure data of the current training record as corresponding current calibration PPG data, current calibration systolic pressure data and current calibration diastolic pressure data; extracting the first PPG data, the first systolic pressure data and the first diastolic pressure data of the current training record one by one to serve as corresponding current training PPG data, current comparison systolic pressure data and current comparison diastolic pressure data; and the current training PPG data and the current calibration PPG data form corresponding current model input data;

based on the convolutional neural network of the big data blood pressure prediction model, carrying out characteristic data extraction processing on the input data of the current model to generate corresponding first characteristic data; the current calibration systolic pressure data and the current calibration diastolic pressure data are used as calibration characteristics to perform characteristic addition processing on the first characteristic data to obtain corresponding second characteristic data; based on the artificial neural network of the big data blood pressure prediction model, performing full-connection calculation on the second characteristic data to obtain corresponding current predicted systolic pressure data and current predicted diastolic pressure data;

Based on the current comparison systolic pressure data and the current comparison diastolic pressure data, carrying out prediction error analysis on the current prediction systolic pressure data and the current prediction diastolic pressure data to obtain a corresponding first analysis result; if the first analysis result is that the model is converged, the big data blood pressure prediction model is trained to be mature; and if the first analysis result is that the model is not converged, modulating network parameters of the convolutional neural network and the artificial neural network of the big data blood pressure prediction model, and continuously selecting the next group of first acquisition data groups and the corresponding first calibration data groups from the first big database to train the modulated model until the first analysis result is that the model is converged.

Preferably, the training maturation-based big data blood pressure prediction model performs calibration record confirmation processing on the first personality database to obtain a corresponding first calibration personality data record, and specifically includes:

extracting any one of the first personality data record as a first record in the first personality database, and taking the first personality PPG data, the first personality systolic pressure data, and the first personality diastolic pressure data of the first record as corresponding first selected PPG data, first selected systolic pressure data, and first selected diastolic pressure data; recording the first personality data records except the first record as corresponding second records, and recording the first personality PPG data, the first personality systolic pressure data and the first personality diastolic pressure data of each second record as corresponding first measured PPG data, first measured systolic pressure data and first measured diastolic pressure data;

Polling each of the second records; when polling, recording the second record currently polled as a current record; and composing current model input data from the first measured PPG data and the first selected PPG data of the current record; performing feature data extraction on the current model input data based on the convolutional neural network of the big data blood pressure prediction model which is trained and mature to generate corresponding current first feature data; performing feature addition processing on the first selected systolic pressure data and the first selected diastolic pressure data serving as calibration features to obtain corresponding current second feature data; based on the artificial neural network of the big data blood pressure prediction model which is trained and mature, carrying out full-connection calculation on the current second characteristic data to obtain corresponding current predicted systolic pressure data and current predicted diastolic pressure data; calculating the absolute difference between the first measured systolic pressure data and the current predicted systolic pressure data recorded currently to generate a corresponding first differential pressure, calculating the absolute difference between the first measured diastolic pressure data and the current predicted diastolic pressure data recorded currently to generate a corresponding second differential pressure, and summing the first differential pressure and the second differential pressure to obtain a corresponding first differential pressure sum;

When the polling of each second record is finished, calculating the sum of the obtained plurality of first differential pressure sums, and taking the result of the sum calculation as a first sum corresponding to the current first record;

and confirming the first personality data record corresponding to the first record with the first sum of the first personality data record being the minimum value in the first personality database as a calibration record and taking the calibration record as the corresponding first calibration personality data record.

Preferably, the training of the personality blood pressure prediction model based on the big data blood pressure prediction model, the first personality database, and the first calibrated personality data record specifically includes:

the first personality PPG data, the first personality systolic pressure data and the first personality diastolic pressure data recorded by the first calibration personality data are used as corresponding first calibration PPG data, first calibration systolic pressure data and first calibration diastolic pressure data;

extracting all the first personality data records except the first calibration personality data record in the first personality database to serve as a first training record set; the first training record set comprises a plurality of first training records, and each first training record corresponds to one first personality data record except the first calibration personality data record;

Training the personalized blood pressure prediction model based on the mature big data blood pressure prediction model and the first training record set, specifically: extracting a first training record from the first training record set as a current training record; extracting the first personality PPG data of the current training record and the first calibration PPG data of the first calibration personality data record to form corresponding current model input data; based on the convolutional neural network of the big data blood pressure prediction model which is trained and mature, carrying out characteristic data extraction processing on the input data of the current model to generate corresponding third characteristic data; carrying out characteristic addition processing on a preset personal characteristic data group corresponding to the appointed data source to third characteristic data to obtain corresponding fourth characteristic data; taking the fourth characteristic data as independent variables, and taking the first calibration systolic pressure data and the first calibration diastolic pressure data recorded by the first calibration individual data as dependent variables, carrying out primary model parameter modulation on the individual blood pressure prediction model; after the parameter modulation of the current model is finished, extracting the next first training record from the first training record set as a new current training record, and continuing training the personalized blood pressure prediction model after the current modulation until the last first training record finishes training; the personal characteristic data set may be empty, and may also include age, gender, height, weight, heart rate difference value and heart rate variability difference value.

Preferably, the performing, based on the trained big data blood pressure prediction model, the personalized blood pressure prediction model and the first calibrated personalized data record corresponding to the specified data source, personalized blood pressure prediction processing on the PPG signal received from the specified data source at any moment to generate corresponding first predicted systolic pressure data and first predicted diastolic pressure data specifically includes:

recording the PPG signal received from the specified data source at any instant as the current PPG signal; and performing signal sampling processing on the current PPG signal according to the appointed signal sampling frequency to generate corresponding current PPG sampling data;

the first personality PPG data, the first personality systolic pressure data and the first personality diastolic pressure data recorded by the first calibration personality data are used as corresponding second calibration PPG data, second calibration systolic pressure data and second calibration diastolic pressure data;

the current PPG sampling data and the second calibration PPG data form corresponding current model input data;

based on the convolutional neural network of the big data blood pressure prediction model which is trained and mature, carrying out characteristic data extraction processing on the input data of the current model to generate corresponding fifth characteristic data;

Performing characteristic addition processing on a preset personal characteristic data set corresponding to the appointed data source, the second calibration systolic pressure data and the second calibration diastolic pressure data to fifth characteristic data to obtain corresponding sixth characteristic data; the personal characteristic data set may be empty, and may also include age, gender, height, weight, heart rate difference, and heart rate variability difference;

and carrying out blood pressure prediction processing on the sixth characteristic data based on the trained and mature personalized blood pressure prediction model to obtain corresponding first predicted systolic pressure data and first predicted diastolic pressure data.

A second aspect of an embodiment of the present invention provides an apparatus for implementing the method described in the first aspect, including: the model prediction system comprises a receiving module, a model training processing module and a model prediction processing module;

the receiving module is used for receiving the photoplethysmography (PPG) signals from a plurality of data sources and corresponding blood pressure information to form a first large database; receiving a plurality of PPG signals from a designated data source and corresponding blood pressure information to form a first personality database;

the model training processing module is used for training a big data blood pressure prediction model based on the first big database; performing calibration record confirmation processing on the first personality database based on the trained big data blood pressure prediction model to obtain a corresponding first calibration personality data record; training the personality blood pressure prediction model based on the trained big data blood pressure prediction model, the first personality database and the first calibrated personality data record;

The model prediction processing module is used for performing personalized blood pressure prediction processing on the PPG signal received from the appointed data source at any moment to generate corresponding first predicted systolic pressure data and first predicted diastolic pressure data based on the trained and mature big data blood pressure prediction model, the personalized blood pressure prediction model and the first calibrated personalized data record corresponding to the appointed data source.

A third aspect of an embodiment of the present invention provides an electronic device, including: memory, processor, and transceiver;

the processor is configured to couple to the memory, and read and execute the instructions in the memory, so as to implement the method steps described in the first aspect;

the transceiver is coupled to the processor and is controlled by the processor to transmit and receive messages.

A fourth aspect of the embodiments of the present invention provides a computer-readable storage medium storing computer instructions that, when executed by a computer, cause the computer to perform the instructions of the method of the first aspect.

The embodiment of the invention provides a personalized blood pressure prediction method, a personalized blood pressure prediction device, electronic equipment and a computer readable storage medium based on big data characteristics, wherein a big data model, namely a big data blood pressure prediction model, is firstly constructed based on a CNN+ANN neural network structure, and is trained by utilizing big data; then, constructing a personalized model, namely a personalized blood pressure prediction model, by referring to a model structure of a support vector machine model, a random forest model or a decision tree model, and training the personalized model by using appointed personalized data; establishing a personalized database for each appointed object to update personalized calibration data (personalized PPG calibration data, personalized systolic pressure calibration data and personalized diastolic pressure calibration data); in practical application, the individual data of the appointed object is sent to the big data blood pressure prediction model to carry out big data feature extraction, and then the obtained feature data and the corresponding individual systolic pressure and diastolic pressure calibration data are sent to the individual blood pressure prediction model to carry out individual blood pressure prediction. According to the invention, the characteristic information with higher accuracy is obtained through the big data model, and the personalized model adapting to the specific object is customized for each specific object, so that the blood pressure data (systolic pressure and diastolic pressure) matched with the specific object can be predicted for each specific object, the problem that the mature model in the current market cannot be subjected to personalized prediction is solved, and the blood pressure prediction precision is improved.

Drawings

Fig. 1 is a schematic diagram of a personalized blood pressure prediction method based on big data features according to a first embodiment of the present invention;

fig. 2 is a block diagram of a personalized blood pressure prediction device based on big data features according to a second embodiment of the present invention;

fig. 3 is a schematic structural diagram of an electronic device according to a third embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail below with reference to the accompanying drawings, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

The personalized blood pressure prediction method based on big data features provided in the first embodiment of the present invention, as shown in fig. 1, is a schematic diagram of the personalized blood pressure prediction method based on big data features provided in the first embodiment of the present invention, and mainly includes the following steps:

step 1, receiving photoplethysmography (PPG) signals from a plurality of data sources and corresponding blood pressure information to form a first large database;

Wherein the first big database comprises a plurality of first big data records; the first big data record comprises one or more groups of first acquisition data groups and first calibration data groups; the first acquisition data set comprises first PPG data, first systolic data and first diastolic data; the first calibration data set comprises first calibration PPG data, first calibration systolic pressure data and first calibration diastolic pressure data;

here, each data source actually corresponds to a specific individual object, and in the first embodiment of the present invention, a first large database is used to store a huge amount of individual object collected data;

the method specifically comprises the following steps: step 11, forming a first received data set from the PPG signals of the single data source received at the present time and the corresponding blood pressure information;

wherein the first received data set comprises a plurality of first received data records; the first received data record comprises a first PPG signal and corresponding first acquired systolic data and first acquired diastolic data;

here, each received first received data set corresponds to a batch of multiple acquired data of the individual subject, and each first received data record corresponds to a single PPG signal+blood pressure acquired data in the batch of acquired data; in the process of acquisition, in order to ensure that the accuracy of acquired data is not continuously acquired, a time interval, namely an appointed acquisition time interval, is provided for multiple acquisitions, and a PPG signal with fixed duration, namely appointed acquisition duration, is acquired as a current first PPG signal during each acquisition, wherein the appointed acquisition duration is actually related to the input data length of a subsequently used big data blood pressure prediction model; when PPG signal acquisition is carried out, the fingertip position of an appointed finger of an individual object is selected as an appointed PPG signal acquisition part by default, a appointed light source (an infrared light source, a red light source or a green light source) is used for carrying out light source irradiation on one side of the fingertip of the appointed finger, and a light sensor is used for carrying out light intensity signal acquisition on the other side of the fingertip of the appointed finger so as to obtain a corresponding first PPG signal; in addition, at the previous time or the next time of PPG signal acquisition, a conventional blood pressure detection mode is also used for carrying out real-time blood pressure detection on the current individual object so as to obtain a group of blood pressure data, namely first acquired systolic pressure data and first acquired diastolic pressure data, and the first acquired systolic pressure data and the first acquired diastolic pressure data which are obtained at the time before/after the current PPG signal acquisition and the first PPG data which are acquired at the current time can form a corresponding relation because the real-time property of blood pressure change is not strong;

Step 12, performing signal sampling processing on each first PPG signal according to the designated signal sampling frequency to generate corresponding first PPG sampling data;

here, the designated signal sampling frequency is a preset system parameter, the higher the frequency is, the longer the data length of the first PPG sampling data is, and the designated signal sampling frequency is actually related to the input data length of the big data blood pressure prediction model used later and the designated acquisition time length in the foregoing;

step 13, selecting one first calibration PPG data which meets the standard from a plurality of obtained first PPG sampling data according to a preset calibration data selection standard and is used as corresponding first calibration PPG data; and taking the other remaining first PPG sampling data as corresponding first PPG data;

here, the big data blood pressure prediction model to be used later performs feature extraction and blood pressure prediction on the real-time acquired data based on a set of calibration data, so before training the big data blood pressure prediction model, a set of calibration data (calibration PPG data, calibration systolic pressure data, calibration diastolic pressure data) needs to be selected, and because the systolic pressure/diastolic pressure corresponds to the PPG data, one calibration PPG data is determined from all the obtained PPG sampling data;

It should be noted that, the calibration data selection criterion in the first embodiment of the present invention is a screening criterion for selecting calibration PPG data from all PPG sampling data, and may be implemented in various manners; one specific implementation manner is as follows, and one first calibration PPG data which meets the standard and is corresponding to the standard is selected from a plurality of obtained first PPG sampling data according to a preset calibration data selection standard, which specifically comprises the following steps:

respectively carrying out signal-to-noise ratio statistics on an effective signal and a noise signal on the obtained multiple first PPG sampling data to obtain corresponding first signal-to-noise ratios; performing heart rate conversion processing according to each first PPG sampling data to obtain a corresponding first heart rate; performing blood oxygen saturation conversion processing according to each first PPG sampling data to obtain corresponding first blood oxygen saturation;

acquiring preset calibration data selection standard data; the calibration data selection criterion data comprises a state data set and a threshold data set; the state data set comprises a signal-to-noise ratio check state word, a heart rate check state word and an oxygen saturation check state word, wherein the signal-to-noise ratio check state word, the heart rate check state word and the oxygen saturation check state word comprise an on state and an off state; the threshold data set includes a signal-to-noise ratio check threshold, a heart rate check threshold range, and a blood oxygen saturation check threshold range; the signal-to-noise ratio check state word corresponds to a signal-to-noise ratio check threshold, the heart rate check state word corresponds to a heart rate check threshold range, and the blood oxygen saturation check state word corresponds to a blood oxygen saturation check threshold range;

Identifying whether the signal-to-noise ratio checking state word is in an on state or not, if so, extracting first PPG sampling data with the first signal-to-noise ratio exceeding a signal-to-noise ratio checking threshold value to be classified as a first data set, and if not, classifying all the first PPG sampling data as the first data set; identifying whether the heart rate check status word is in an on state or not, if so, extracting first PPG sampling data with the first heart rate meeting the heart rate check threshold range from the first data set to be classified into a second data set, and if not, taking the first data set as the second data set; identifying whether the blood oxygen saturation checking state word is in an on state or not, if so, extracting first PPG sampling data of which the first blood oxygen saturation in the first data set meets the blood oxygen saturation checking threshold range to be classified into a third data set, and if not, taking the second data set as the third data set; sequencing the first PPG sampling data in the third data set according to the sequence from the high to the low of the corresponding first signal-to-noise ratio, and taking the first PPG sampling data in the first sequencing bit as first calibration PPG data meeting the standard;

step 14, taking the first acquired systolic pressure data and the first acquired diastolic pressure data corresponding to the first calibration PPG data as corresponding first calibration systolic pressure data and first calibration diastolic pressure data, and forming a corresponding first calibration data set by the first calibration PPG data, the first calibration systolic pressure data and the first calibration diastolic pressure data; the first acquired systolic pressure data and the first acquired diastolic pressure data corresponding to each first PPG data are used as corresponding first systolic pressure data and first diastolic pressure data, and each first PPG data and the corresponding first systolic pressure data and first diastolic pressure data form a corresponding first acquired data set; and one or more groups of obtained first acquisition data groups and one group of first calibration data groups form corresponding first big data records; and adding the first big data record to the first big database.

Step 2, receiving a plurality of PPG signals from a designated data source and corresponding blood pressure information to form a first personality database;

the first personality database comprises a plurality of first personality data records; the first personality data record includes first personality PPG data, first personality systolic pressure data, and first personality diastolic pressure data;

here, the designated data source actually corresponds to a designated individual object, and in the first embodiment of the present invention, a first personality database is configured for each first personality database to store personalized acquisition data corresponding to the designated object;

the method specifically comprises the following steps: step 21, forming a second received data set from the multiple PPG signals of the designated data source received at the present time and the corresponding blood pressure information;

wherein the second received data set comprises a plurality of second received data records; the second received data record comprises a first personalized PPG signal, corresponding first personalized systolic pressure data and first personalized diastolic pressure data;

here, each received specified data source, that is, a second received data set of the specified object, is actually a batch of multiple acquired data of the specified object, and each second received data record corresponds to a single PPG signal+blood pressure acquired data in the batch of acquired data; as for the first personalized PPG signal and the corresponding acquisition mode of the first personalized systolic pressure data and the first personalized diastolic pressure data, default is consistent with the acquisition mode mentioned in the step 11, and further description is omitted herein;

Step 22, performing signal sampling processing on each first personalized PPG signal according to the designated signal sampling frequency to generate corresponding first personalized PPG data; and forming a corresponding first personality data record by each first personality PPG data and corresponding first personality systolic pressure data and first personality diastolic pressure data;

here, the signal sampling is similar to step 12, and after the first personality PPG data is obtained, the first personality PPG data and the first personality systolic pressure data and the first personality diastolic pressure data obtained in step 21 form a first personality data record;

step 23, counting the total number of the first personality data records in the first personality database to generate a corresponding first total number of records; identifying whether the first total number of records is smaller than a preset total number of records threshold value or not; if the total number of the first records is smaller than the total number of records threshold, adding the first personality data record obtained currently into a first personality database; if the total number of the first records is equal to the threshold value of the total number of the records, replacing the first personality data record with the earliest adding time in the first personality database by using the first personality data record obtained currently.

Here, in the embodiment of the present invention, the total number of records in the first personality database is limited, the maximum total number of records is a total number of records threshold, if the records in the first personality database are not full, that is, the total number of the first records is smaller than the total number of records threshold, the newly added records are directly added into the database, and if the records in the first personality database are full, that is, the total number of the first records is equal to the total number of records threshold, the earliest record in the database is replaced by the newly added records.

Step 3, training a big data blood pressure prediction model based on a first big database;

the big data blood pressure prediction model consists of a convolutional neural network and an artificial neural network; the convolutional neural network consists of a plurality of convolutional pooling layers; each convolution pooling layer consists of one convolution layer and one pooling layer; the artificial neural network consists of an input layer, one or more layers of full-connection layers and an output layer; the input layer comprises a plurality of input layer neurons, and each input layer neuron corresponds to one data element of the input characteristic data; the full-connection layer comprises a plurality of full-connection layer neurons, the full-connection layer neurons of the first full-connection layer are connected with the plurality of input layer neurons, the full-connection layer neurons of the next full-connection layer are connected with the plurality of full-connection layer neurons of the last full-connection layer, and the full-connection layer neurons default to use a ReLU function as an activation function; the output layer comprises 2 output layer neurons, each output layer neuron is connected with a plurality of full-connection layer neurons of the last full-connection layer, the output layer neurons do not use an activation function, and the 2 output layer neurons are respectively used for outputting predicted systolic pressure data and predicted diastolic pressure data;

The method specifically comprises the following steps: step 31, extracting first big data records from the first big database one by one to serve as current training records; extracting first calibration PPG data, first calibration systolic pressure data and first calibration diastolic pressure data of the current training record as corresponding current calibration PPG data, current calibration systolic pressure data and current calibration diastolic pressure data; extracting first PPG data, first systolic pressure data and first diastolic pressure data of the current training record one by one as corresponding current training PPG data, current comparison systolic pressure data and current comparison diastolic pressure data; and corresponding current model input data is formed by each current training PPG data and current calibration PPG data;

step 32, based on a convolutional neural network of the big data blood pressure prediction model, carrying out characteristic data extraction processing on the input data of the current model to generate corresponding first characteristic data; the current calibration systolic pressure data and the current calibration diastolic pressure data are used as calibration characteristics to perform characteristic addition processing on the first characteristic data to obtain corresponding second characteristic data; based on an artificial neural network of the big data blood pressure prediction model, performing full-connection calculation on the second characteristic data to obtain corresponding current predicted systolic pressure data and current predicted diastolic pressure data;

The actual process of adding the current calibration systolic pressure data and the current calibration diastolic pressure data serving as calibration features to the first feature data to obtain corresponding second feature data is that the current calibration systolic pressure data and the current calibration diastolic pressure data are added into the first feature data; the data adding position can be realized in a plurality of modes by self-definition, and the current calibration systolic pressure data and the current calibration diastolic pressure data are added at the tail part of the first characteristic data under the conventional condition;

step 33, based on the current comparison systolic pressure data and the current comparison diastolic pressure data, performing prediction error analysis on the current prediction systolic pressure data and the current prediction diastolic pressure data to obtain a corresponding first analysis result; if the first analysis result is that the model is converged, the visual big data blood pressure prediction model is trained to be mature; and if the first analysis result is that the model is not converged, modulating network parameters of a convolutional neural network and an artificial neural network of the big data blood pressure prediction model, and continuously selecting a next group of first acquisition data groups and corresponding first calibration data groups from a first big database to train the modulated model until the first analysis result is that the model is converged.

And if the calculation result of the loss function is in the reasonable loss range, outputting a first analysis result to be that the model is converged, otherwise, outputting the first analysis result to be that the model is not converged. When the network parameters of the convolutional neural network and the artificial neural network of the big data blood pressure prediction model are modulated, a conventional reverse modulation method is adopted in the first embodiment of the invention, and further description is omitted here.

The model training is completed by utilizing the big data to the big data blood pressure prediction model through the step 3, and the model is characterized in that calibration PPG data, calibration systolic pressure data and calibration diastolic pressure data can be set through input data, the convolutional neural network is used for respectively carrying out feature extraction on the two input PPG data (the PPG data for prediction and the calibration PPG data) to obtain feature data containing two sections of PPG features, namely first feature data, the artificial neural network carries out full-connection calculation related to systolic pressure prediction based on the two sections of PPG features in the first feature data to obtain a systolic pressure prediction parameter, carries out full-connection calculation related to diastolic pressure prediction to obtain a diastolic pressure prediction parameter, and finally carries out calculation on 2 output layer neurons of an output layer respectively based on the calibration systolic pressure data and the preset relation of the systolic pressure prediction parameter to obtain final prediction systolic pressure data, and calculation is carried out calculation based on the preset relation of the calibration diastolic pressure data and the diastolic pressure prediction parameter to obtain final prediction diastolic pressure data. The model is obtained by training PPG data of a large number of different testers, so that the model has the advantages that the first characteristic data extracted by the convolutional neural network is rich in characteristics, and the model has the defects that the predicted data output by the artificial neural network is insufficient in individuation. Therefore, the embodiment of the invention improves the deficiency of the big data blood pressure prediction model by creating a personalized model, namely a personalized blood pressure prediction model, through the subsequent steps.

Step 4, performing calibration record confirmation processing on the first personality database based on the trained big data blood pressure prediction model to obtain a corresponding first calibration personality data record;

the method specifically comprises the following steps: step 41, extracting any first personality data record in the first personality database as a first record, and taking first personality PPG data, first personality systolic pressure data and first personality diastolic pressure data of the first record as corresponding first selected PPG data, first selected systolic pressure data and first selected diastolic pressure data; recording other first personality data records except the first record as corresponding second records, and recording first personality PPG data, first personality systolic pressure data and first personality diastolic pressure data of each second record as corresponding first measurement PPG data, first measurement systolic pressure data and first measurement diastolic pressure data;

for example, if the first personality database includes 9 first personality data records, when the first record is first personality data record 1, the first personality data records 2-9 are second records; when the first record is the first personality data record 2, the first personality data records 1, 3-9 are the second records; similarly, when the first record is the first personality data record 9, the first personality data records 1-8 are the second records;

Step 42, polling each second record; when polling, recording the second record currently polled as the current record; and forming current model input data from the first measured PPG data and the first selected PPG data currently recorded; performing characteristic data extraction on the input data of the current model based on a convolutional neural network of a trained big data blood pressure prediction model to generate corresponding current first characteristic data; the first selected systolic pressure data and the first selected diastolic pressure data are used as calibration characteristics to perform characteristic addition processing on the current first characteristic data to obtain corresponding current second characteristic data; based on an artificial neural network of a training mature big data blood pressure prediction model, performing full-connection calculation on the current second characteristic data to obtain corresponding current predicted systolic pressure data and current predicted diastolic pressure data; calculating the absolute difference value of the first measured systolic pressure data recorded currently and the predicted systolic pressure data recorded currently to generate a corresponding first differential pressure, calculating the absolute difference value of the first measured diastolic pressure data recorded currently and the predicted diastolic pressure data recorded currently to generate a corresponding second differential pressure, and summing the first differential pressure and the second differential pressure to obtain a corresponding first differential pressure sum;

For example, when the first record is the first personality data record 1, the first personality PPG data, the first personality systolic pressure data, and the first personality diastolic pressure data of the first personality data record 1 are used as the corresponding first selected PPG data, the first selected systolic pressure data, and the first selected diastolic pressure data; the first personality data record 2-10 is a second record;

the first personality PPG data, the first personality systolic blood pressure data, and the first personality diastolic blood pressure data of the 1 st second record (first personality data record 2) are recorded as corresponding first measured PPG data 1, first measured systolic blood pressure data 1, and first measured diastolic blood pressure data 1; the method comprises the steps of inputting first measured PPG data 1+first selected PPG data serving as input data of a current model into a convolutional neural network of a big data blood pressure prediction model to obtain first characteristic data 1, inputting the first characteristic data 1+first selected systolic pressure data+first selected diastolic pressure data to obtain second characteristic data 1, inputting the second characteristic data 1 into an artificial neural network of the big data blood pressure prediction model to obtain current predicted systolic pressure data 1 and current predicted diastolic pressure data 1, obtaining a first differential pressure by |first measured systolic pressure data 1-current predicted systolic pressure data 1|, obtaining a second differential pressure by |first measured diastolic pressure data 1-current predicted diastolic pressure data 1|, and obtaining a first differential pressure sum 11 of 1 st second record by the first differential pressure+the second differential pressure; by analogy, the first pressure difference and 12-17 corresponding to 7 second records (the first personality data record 3-9) can be obtained;

Similarly, when the first record is the first personality data record 2, a first differential pressure sum 21-28 corresponding to 8 second records (the first personality data record 1 and the first personality data records 3-9) can be obtained; similarly, when the first record is the first personality data record 10, a first pressure difference sum 91-99 of 8 second records (first personality data records 1-8) may also be obtained;

step 43, when the polling of each second record is finished, calculating the sum of the obtained first differential pressure sums, and taking the result of the sum calculation as a first sum corresponding to the current first record;

for example, when the first record is the first personality data record 1, the first sum 1= (first differential pressure sum 11+first differential pressure sum 12+ … +first differential pressure sum 19); when the first record is the first personality data record 2, first sum 2= (first differential pressure sum 21+first differential pressure sum 22+ … +first differential pressure sum 29); similarly, when the first record is the first personality data record 9, first sum 9= (first differential pressure sum 91+first differential pressure sum 92+ … +first differential pressure sum 99);

and step 44, confirming the first personality data record corresponding to the first record with the first sum of the first sum being the minimum value in the first personality database as the calibration record and taking the calibration record as the corresponding first calibration personality data record.

For example, if the value of the first sum 4 is the smallest in the first sums 1-9, the first personality data record 4 corresponding to the first record of the first sum 4 in the first personality database.

It should be noted that, in the embodiment of the present invention, each specific object actually has a corresponding first personality database and a corresponding first calibration personality data record. Once the first calibration personality data record is determined, the calibration data content (PPG data, systolic pressure data, and diastolic pressure data) of the first calibration personality data record is used as calibration data required for model operation each time a personalized blood pressure prediction is performed on the specified subject. In addition, after the first calibration personality data record corresponding to the specified object is determined for the first time, in the first embodiment of the present invention, the first calibration personality data record is updated periodically, and when the first calibration personality data record is updated periodically, the steps 41-44 are repeated, and calibration record confirmation processing is performed on the first personality database based on the trained mature big data blood pressure prediction model to obtain the latest first calibration personality data record. In addition, when the content of the first personality database corresponding to the specified object changes (such as new record is added or old record is rewritten), the first calibration personality data record is updated in real time, and when the first calibration personality data record is updated in real time, the steps 41-44 are repeated, and calibration record confirmation processing is performed on the first personality database based on the training mature big data blood pressure prediction model to obtain the latest first calibration personality data record.

Step 5, training the personalized blood pressure prediction model based on the trained and mature big data blood pressure prediction model, the first personalized database and the first calibrated personalized data record;

the model structure of the personalized blood pressure prediction model adopts a model structure of a support vector machine model, a random forest model or a decision tree model;

the method specifically comprises the following steps: step 51, using first personality PPG data, first personality systolic pressure data and first personality diastolic pressure data recorded by the first calibration personality data as corresponding first calibration PPG data, first calibration systolic pressure data and first calibration diastolic pressure data;

step 52, extracting all the first personality data records except the first calibration personality data record in the first personality database as a first training record set;

the first training record set comprises a plurality of first training records, and each first training record corresponds to a first personality data record except a first calibration personality data record;

step 53, training the personalized blood pressure prediction model based on the training mature big data blood pressure prediction model and the first training record set, specifically:

extracting a first training record from the first training record set as a current training record; extracting first personality PPG data of the current training record and first calibration PPG data of the first calibration personality data record to form corresponding current model input data; based on a convolutional neural network of a training mature big data blood pressure prediction model, carrying out characteristic data extraction processing on input data of the current model to generate corresponding third characteristic data; carrying out characteristic addition processing on a preset personal characteristic data group corresponding to the appointed data source to third characteristic data to obtain corresponding fourth characteristic data; taking the fourth characteristic data as independent variables, and taking the first calibration systolic pressure data and the first calibration diastolic pressure data recorded by the first calibration individual data as dependent variables, and performing primary model parameter modulation on the individual blood pressure prediction model; after the parameter modulation of the current model is finished, extracting the next first training record from the first training record set as a new current training record, and continuing training the personalized blood pressure prediction model after the current modulation until the last first training record finishes training;

The personal characteristic data set may be empty, and may also include age, gender, height, weight, heart rate difference value, and heart rate variability difference value.

Here, the input of the individual blood pressure prediction model in the above step 5 is the integrated feature data (fourth feature data) composed of PPG feature (third feature data) +individual feature data set output by the convolutional neural network of the big data blood pressure prediction model. The model structure of the personalized blood pressure prediction model adopts a model structure of a support vector machine model, a random forest model or a decision tree model, and the model structure can be realized by referring to the related disclosed technology and is not further described herein. It should be noted that, compared with the fully connected layer structure of the artificial neural network, the three model structures have the advantages that the branch judgment condition can be better set based on the personal characteristic data set, and the settlement result is refined, that is, after different personal characteristic data sets are added into the input characteristic data, different predicted blood pressure values can be calculated even if the input PPG characteristic (third characteristic data) is close, which is difficult to realize by the artificial neural network. In order to assign a corresponding individual blood pressure prediction model to each specified subject, the first calibration PPG data is configured for the big data blood pressure prediction model based on the first calibration individual data record corresponding to the specified subject during training, the input of the individual blood pressure prediction model, that is, the fourth feature data, is set based on the individual feature data set corresponding to the specified subject, and the model parameters of the individual blood pressure prediction model modulated based on such an individual data structure tend to be related to the individual features of the specified subject.

Step 6, based on the trained mature big data blood pressure prediction model, the personalized blood pressure prediction model and the first calibrated personalized data record corresponding to the appointed data source, personalized blood pressure prediction processing is carried out on the PPG signal received from the appointed data source at any moment to generate corresponding first predicted systolic pressure data and first predicted diastolic pressure data;

the method specifically comprises the following steps: step 61, recording the PPG signal received from the appointed data source at any moment as the current PPG signal; and according to the appointed signal sampling frequency, carrying out signal sampling processing on the current PPG signal to generate corresponding current PPG sampling data;

here, the PPG signal of the specified data source received at any moment, that is, the current PPG signal, is actually a real-time PPG acquisition signal of the specified subject, and the acquisition mode of the signal is similar to the acquisition mode introduced in the foregoing step 11, and further description is omitted herein. The signal sampling processing manner is similar to the foregoing step 12, and further description is omitted herein;

step 62, using first personality PPG data, first personality systolic pressure data and first personality diastolic pressure data recorded by the first calibration personality data as corresponding second calibration PPG data, second calibration systolic pressure data and second calibration diastolic pressure data;

Here, the first calibration personality data record is a calibration reference data set of the specified object, including three calibration data required for the subsequent model: calibrating PPG data, systolic pressure data and diastolic pressure data;

step 63, composing corresponding current model input data by the current PPG sampling data and the second calibration PPG data;

step 64, based on a convolutional neural network of a training mature big data blood pressure prediction model, carrying out characteristic data extraction processing on input data of the current model to generate corresponding fifth characteristic data;

here, the fifth feature data is actually feature data including two sets of PPG features, which are the PPG feature of the current PPG sample data and the PPG feature of the second calibration PPG data, respectively;

step 65, performing feature addition processing on the preset personal feature data set, the second calibration systolic pressure data and the second calibration diastolic pressure data corresponding to the designated data source to obtain corresponding sixth feature data;

wherein the personal characteristic data set may be empty, and may also include age, gender, height, weight, heart rate difference, and heart rate variability difference;

here, there are various ways of adding the personal characteristic data set of the specified object and the second calibration systolic pressure data and the second calibration diastolic pressure data corresponding to the specified object to the fifth characteristic data, which may be added in a concentrated manner at the tail of the fifth characteristic data, or may be added in a concentrated manner or in a dispersed manner in one or more specified positions of the fifth characteristic data;

And step 66, performing blood pressure prediction processing on the sixth characteristic data based on the trained mature individual blood pressure prediction model to obtain corresponding first predicted systolic pressure data and first predicted diastolic pressure data.

Here, the personalized blood pressure prediction model may assign corresponding personalized systolic pressure weights and personalized diastolic pressure weights to the systolic pressure prediction and the diastolic pressure prediction based on the personal feature data set through a support vector machine model, a random forest model, or a decision tree model; performing prediction calculation related to the systolic pressure on the two sections of PPG features in the sixth feature data based on the personalized systolic pressure and diastolic pressure weight to obtain corresponding weighted systolic pressure prediction parameters, and performing prediction calculation related to the diastolic pressure to obtain corresponding weighted diastolic pressure prediction parameters; and calculating first predicted systolic pressure data based on a preset corresponding relation between the second calibrated systolic pressure data and the weighted systolic pressure prediction parameter in the sixth characteristic data, and calculating first predicted diastolic pressure data based on a preset corresponding relation between the second calibrated diastolic pressure data and the weighted diastolic pressure prediction parameter.

Fig. 2 is a block diagram of a personalized blood pressure prediction device based on big data features according to a second embodiment of the present invention, where the device may be a terminal device or a server for implementing a method according to an embodiment of the present invention, or may be a device for implementing a method according to an embodiment of the present invention, which is connected to the terminal device or the server, and for example, the device may be a device or a chip system of the terminal device or the server. As shown in fig. 2, the apparatus includes: a receiving module 201, a model training processing module 201 and a model prediction processing module 203.

The receiving module 201 is configured to receive photoplethysmography PPG signals from a plurality of data sources and corresponding blood pressure information to form a first large database; and receiving a plurality of PPG signals from the appointed data source and corresponding blood pressure information to form a first personality database.

The model training processing module 202 is used for training a big data blood pressure prediction model based on the first big database; performing calibration record confirmation processing on the first personality database based on the trained big data blood pressure prediction model to obtain a corresponding first calibration personality data record; and training the personality blood pressure prediction model based on the trained big data blood pressure prediction model, the first personality database and the first calibrated personality data record.

The model prediction processing module 203 is configured to perform personalized blood pressure prediction processing on PPG signals received from a specified data source at any moment to generate corresponding first predicted systolic pressure data and first predicted diastolic pressure data based on the trained mature big data blood pressure prediction model, the personalized blood pressure prediction model, and the first calibrated personalized data record corresponding to the specified data source.

The personalized blood pressure prediction device based on big data features provided by the embodiment of the invention can execute the method steps in the method embodiment, and the implementation principle and the technical effect are similar and are not repeated here.

It should be noted that, it should be understood that the division of the modules of the above apparatus is merely a division of a logic function, and may be fully or partially integrated into a physical entity or may be physically separated. And these modules may all be implemented in software in the form of calls by the processing element; or can be realized in hardware; the method can also be realized in a form of calling software by a processing element, and the method can be realized in a form of hardware by a part of modules. For example, the receiving module may be a processing element that is set up separately, may be implemented in a chip of the above apparatus, or may be stored in a memory of the above apparatus in the form of program codes, and may be called by a processing element of the above apparatus to execute the functions of the above determining module. The implementation of the other modules is similar. In addition, all or part of the modules can be integrated together or can be independently implemented. The processing element described herein may be an integrated circuit having signal processing capabilities. In implementation, each step of the above method or each module above may be implemented by an integrated logic circuit of hardware in a processor element or an instruction in a software form.

For example, the modules above may be one or more integrated circuits configured to implement the methods above, such as: one or more specific integrated circuits (Application Specific Integrated Circuit, ASIC), or one or more digital signal processors (Digital Signal Processor, DSP), or one or more field programmable gate arrays (Field Programmable Gate Array, FPGA), etc. For another example, when a module above is implemented in the form of a processing element scheduler code, the processing element may be a general purpose processor, such as a central processing unit (Central Processing Unit, CPU) or other processor that may invoke the program code. For another example, the modules may be integrated together and implemented in the form of a System-on-a-chip (SOC).

In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, produces, in whole or in part, the processes or functions described in accordance with embodiments of the present invention. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, from one website, computer, server, or data center via a wired (e.g., coaxial cable, fiber optic, digital subscriber line (Digital Subscriber Line, DSL)) or wireless (e.g., infrared, wireless, bluetooth, microwave, etc.) means. The computer readable storage media may be any available media that can be accessed by a computer or a data storage device such as a server, data center, or the like that contains an integration of one or more available media. The usable medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a magnetic tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a Solid State Disk (SSD)), or the like.

Fig. 3 is a schematic structural diagram of an electronic device according to a third embodiment of the present invention. The electronic device may be the aforementioned terminal device or server, or may be a terminal device or server connected to the aforementioned terminal device or server for implementing the method of the embodiment of the present invention. As shown in fig. 3, the electronic device may include: a processor 301 (e.g., a CPU), a memory 302, a transceiver 303; the transceiver 303 is coupled to the processor 301, and the processor 301 controls the transceiving actions of the transceiver 303. The memory 302 may store various instructions for performing various processing functions and implementing the methods and processes provided in the above-described embodiments of the present invention. Preferably, the electronic device according to the embodiment of the present invention further includes: a power supply 304, a system bus 305, and a communication port 306. The system bus 305 is used to implement communication connections between the elements. The communication port 306 is used for connection communication between the electronic device and other peripheral devices.

The system bus referred to in fig. 3 may be a peripheral component interconnect standard (Peripheral Component Interconnect, PCI) bus, or an extended industry standard architecture (Extended Industry Standard Architecture, EISA) bus, or the like. The system bus may be classified into an address bus, a data bus, a control bus, and the like. For ease of illustration, the figures are shown with only one bold line, but not with only one bus or one type of bus. The communication interface is used to enable communication between the database access apparatus and other devices (e.g., clients, read-write libraries, and read-only libraries). The Memory may comprise random access Memory (Random Access Memory, RAM) and may also include Non-Volatile Memory (Non-Volatile Memory), such as at least one disk Memory.

The processor may be a general-purpose processor, including a Central Processing Unit (CPU), a network processor (Network Processor, NP), etc.; but may also be a digital signal processor DSP, an application specific integrated circuit ASIC, a field programmable gate array FPGA or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component.

It should be noted that the embodiments of the present invention also provide a computer readable storage medium having instructions stored therein, which when executed on a computer, cause the computer to perform the methods and processes provided in the above embodiments.

The embodiment of the invention also provides a chip for running the instructions, which is used for executing the method and the processing procedure provided in the embodiment.

The embodiment of the invention provides a personalized blood pressure prediction method, a personalized blood pressure prediction device, electronic equipment and a computer readable storage medium based on big data characteristics, wherein a big data model, namely a big data blood pressure prediction model, is firstly constructed based on a CNN+ANN neural network structure, and is trained by utilizing big data; then, constructing a personalized model, namely a personalized blood pressure prediction model, by referring to a model structure of a support vector machine model, a random forest model or a decision tree model, and training the personalized model by using personalized data of a designated object; establishing a personalized database for each appointed object to update personalized calibration data (personalized PPG calibration data, personalized systolic pressure calibration data and personalized diastolic pressure calibration data); in practical application, personalized data of a specified object, namely real-time PPG data and corresponding personalized PPG calibration data, are sent to a big data blood pressure prediction model for big data feature extraction, and then the obtained feature data and corresponding personalized systolic pressure and diastolic pressure calibration data are sent to the personalized blood pressure prediction model for personalized blood pressure prediction. According to the invention, the characteristic information with higher accuracy is obtained through the big data model, and the personalized model adapting to the specific object is customized for each specific object, so that the blood pressure data (systolic pressure and diastolic pressure) matched with the specific object can be predicted for each specific object, the problem that the mature model in the current market cannot be subjected to personalized prediction is solved, and the blood pressure prediction precision is improved.

Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative elements and steps are described above generally in terms of function in order to clearly illustrate the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied in hardware, in a software module executed by a processor, or in a combination of the two. The software modules may be disposed in Random Access Memory (RAM), memory, read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

The foregoing description of the embodiments has been provided for the purpose of illustrating the general principles of the invention, and is not meant to limit the scope of the invention, but to limit the invention to the particular embodiments, and any modifications, equivalents, improvements, etc. that fall within the spirit and principles of the invention are intended to be included within the scope of the invention.

Claims

1. A personalized blood pressure prediction method based on big data features, the method comprising:

2. The personalized blood pressure prediction method according to claim 1, wherein,

the first big database comprises a plurality of first big data records; the first big data record comprises one or more groups of first acquisition data groups and first calibration data groups; the first acquisition data set includes first PPG data, first systolic data, and first diastolic data; the first calibration data set comprises first calibration PPG data, first calibration systolic pressure data and first calibration diastolic pressure data;

3. The personalized blood pressure prediction method based on big data features according to claim 2, wherein the receiving of the photoplethysmographic PPG signals of multiple data sources and the corresponding blood pressure information form a first big database, specifically comprising:

4. The personalized blood pressure prediction method based on big data features according to claim 2, wherein the receiving the plurality of PPG signals of the specified data source and the corresponding blood pressure information form a first personality database, specifically comprising:

5. The personalized blood pressure prediction method based on big data features according to claim 2, wherein the training of the big data blood pressure prediction model based on the first big database specifically comprises:

6. The personalized blood pressure prediction method based on big data features according to claim 2, wherein the training-matured big data blood pressure prediction model performs calibration record confirmation processing on the first personality database to obtain a corresponding first calibration personality data record, and specifically comprises the following steps:

7. The personalized blood pressure prediction method based on big data features according to claim 2, wherein the training of the personalized blood pressure prediction model based on the training mature big data blood pressure prediction model, the first personality database and the first calibrated personality data record specifically comprises:

8. The personalized blood pressure prediction method based on big data features according to claim 2, wherein the personalized blood pressure prediction processing is performed on PPG signals received from the specified data source at any moment to generate corresponding first predicted systolic pressure data and first predicted diastolic pressure data based on the trained and mature big data blood pressure prediction model, the personalized blood pressure prediction model and the first calibrated personalized data record corresponding to the specified data source, specifically comprising:

9. An apparatus for implementing the personalized blood pressure prediction method steps based on big data features of any of claims 1-8, the apparatus comprising: the model prediction system comprises a receiving module, a model training processing module and a model prediction processing module;

10. An electronic device, comprising: memory, processor, and transceiver;

the processor being adapted to be coupled to the memory, read and execute the instructions in the memory to implement the method steps of any one of claims 1-8;

11. A computer readable storage medium storing computer instructions which, when executed by a computer, cause the computer to perform the instructions of the method of any one of claims 1-8.