KR20240098407A - Apparatus for measuring blood pressure using recalibration - Google Patents

Abstract

A wearable device for measuring blood pressure according to the present invention includes an ECG sensor unit that senses a user's electrocardiogram (ECG) signal and outputs ECG data; A PPG sensor unit that senses the user's pulse wave (Photoplethysmogram, PPG) signal and outputs PPG data; And after performing preprocessing on the ECG data and the PPG data, estimate the user's blood pressure by applying it to a predetermined deep learning model learned in advance, wherein the predetermined deep learning model is a CNN (Convolutional Neural Network) model, It is a model learned by binding a Bi-GRU (Bidirectional Gated Recurrent Unit) model and an attention model, and the processor inputs the preprocessed ECG and PPG data into a CNN model to output CNN weight parameters, and the output CNN weights. The parameter is applied as input data to the Bi-GRU model to output Bi-GRU weight parameters, and the Bi-GRU weight parameters are applied to the Attention model as input data and then the blood pressure value is estimated.

Description

Blood pressure measuring device using recalibration {APPARATUS FOR MEASURING BLOOD PRESSURE USING RECALIBRATION}

The present invention relates to a device for measuring blood pressure, and more specifically, to a device for measuring blood pressure using recalibration and remeasurement.

Biosignals contain various information that can be used to determine health status. Therefore, it is possible to measure biosignals and predict health status using the measured biosignals. Blood pressure is a biological signal that is mainly used to determine health status in real life.

Additionally, as each year passes, the number of patients with various heart diseases such as high blood pressure and arrhythmia is increasing. Therefore, real-time blood pressure measurement is necessary to detect these heart diseases early. Currently, most people use pressurized blood pressure measuring machines to measure blood pressure. A new method is needed rather than the existing cumbersome pressurization method of measuring blood pressure.

There is a cuff-based method of measuring blood pressure that is measured by wrapping it around the forearm. The existing cuff-based method has the disadvantage of not being able to measure blood pressure anytime, anywhere, and even if a machine is available, it is cumbersome. Recently, a cuff-less method of measuring blood pressure without using a cuff is being studied. Cuff-less blood pressure monitors measure blood pressure indirectly using biometric information that is highly correlated with blood pressure.

Cuff-less blood pressure monitors indirectly measure blood pressure using biometric information that is highly correlated with blood pressure, but the accuracy of blood pressure estimates is very low and irregular, and the computational complexity is high, making it very difficult to install them in wearable devices.

Accordingly, the present invention proposes a blood pressure measurement method with low computational complexity and very high accuracy in a wearable device.

The technical problem to be achieved by the present invention is to provide a wearable device for measuring blood pressure.

Another technical problem to be achieved by the present invention is to provide a method for a wearable device to measure blood pressure.

Another technical problem to be achieved by the present invention is to provide a computer-readable recording medium recording a computer program that performs the method described in any one of claims 9 to 14 on a computer.

The technical problems to be achieved in the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the description below. You will be able to.

In order to achieve the above technical problem, a wearable device for measuring blood pressure according to the present invention includes an ECG sensor unit that senses a user's electrocardiogram (ECG) signal and outputs ECG data; A PPG sensor unit that senses the user's pulse wave (Photoplethysmogram, PPG) signal and outputs PPG data; And after performing preprocessing on the ECG data and the PPG data, estimate the user's blood pressure by applying it to a predetermined deep learning model learned in advance, wherein the predetermined deep learning model is a CNN (Convolutional Neural Network) model, It is a model learned by binding a Bi-GRU (Bidirectional Gated Recurrent Unit) model and an attention model, and the processor inputs the preprocessed ECG and PPG data into a CNN model to output CNN weight parameters, and the output CNN weights. The parameters can be applied as input data to the Bi-GRU model to output Bi-GRU weight parameters, and the blood pressure value can be estimated after applying the Bi-GRU weight parameters to the attention model as input data.

If the average absolute error between the blood pressure value estimated from the predetermined deep learning model and the reference value is greater than a predetermined value, the processor may perform relearning by recalibrating the estimated blood pressure value. The predetermined value may be 5.

The processor may perform preprocessing on the ECG data by applying a bandpass filter of 0.5 to 35 hertz (Hz).

The processor may perform preprocessing on the PPG data by applying a bandpass filter of 0.5 to 15 hertz (Hz).

The processor may divide the preprocessed ECG and PPG data into 5-second increments and input them into the CNN model.

The wearable device may include a smart watch.

The reference value may be a blood pressure value measured based on a cuff.

In order to achieve the above other technical problems, a method for measuring blood pressure by a wearable device according to the present invention includes the steps of: sensing a user's electrocardiogram (ECG) signal and outputting ECG data; Sensing the user's pulse wave (Photoplethysmogram, PPG) signal and outputting PPG data; And performing preprocessing on the ECG data and the PPG data and then applying them to a predetermined deep learning model to estimate the user's blood pressure,

The predetermined deep learning model is a model learned by binding a CNN (Convolutional Neural Network) model, a Bi-GRU (Bidirectional Gated Recurrent Unit) model, and an attention model,

The step of estimating the blood pressure includes inputting the preprocessed ECG and PPG data into a CNN model and outputting CNN weight parameters; Applying the output CNN weight parameters as input data of the Bi-GRU model to output Bi-GRU weight parameters; And it may include estimating a blood pressure value after applying the Bi-GRU weight parameters as input data to an attention model.

The method may further include the step of re-learning the size of the average absolute error between the blood pressure value estimated from the predetermined deep learning model and the reference value by recalibrating the estimated blood pressure value. The predetermined value may be 5.

In the method, the step of estimating the blood pressure may include dividing the preprocessed ECG and PPG data into 5-second increments and inputting them into the CNN model.

In the method, estimating the blood pressure may include performing preprocessing on the ECG data by applying a bandpass filter of 0.5 to 35 hertz (Hz).

In the method, estimating the blood pressure may include performing preprocessing on the PPG data by applying a bandpass filter of 0.5 to 15 hertz (Hz).

In the case where a deep learning model according to the present invention was remeasured/recalibrated and relearned, the average absolute error size became much smaller, and as a result, it can be seen that the accuracy was greatly improved compared to the actual blood pressure value. .

The size of the mean absolute error of blood pressure values predicted through a certain deep learning model (a model that binds three models) proposed in the present invention is much larger than the size of the mean absolute error of blood pressure values measured with a PTT-based deep learning model. The simulation results are small, so the blood pressure value measured through a certain deep learning model proposed in the present invention is very accurate, and unnecessary calculations can be reduced, so it has the advantage of being optimal for wearable devices with low computational complexity.

The effects that can be obtained from the present invention are not limited to the effects mentioned above, and other effects not mentioned can be clearly understood by those skilled in the art from the description below. will be.

The accompanying drawings, which are included as part of the detailed description to aid understanding of the present invention, provide embodiments of the present invention, and together with the detailed description, explain the technical idea of the present invention.
Figure 1 is a diagram illustrating the layer structure of an artificial neural network.
Figure 2 is a diagram showing an example of a deep neural network.
Figure 3 is a diagram showing an example of PTT (Pulse Transit Time), a feature calculated using ECG, PPG, and BCG.
Figure 4 is a block diagram for explaining the components of the blood pressure measuring device 400 using the user's biological signals according to the present invention.
FIG. 5 is a diagram illustrating a process for the wearable device 400 to estimate blood pressure.
Figure 6 is a diagram illustrating input data of the deep learning model proposed in the present invention.
FIG. 7 is a diagram illustrating a method of estimating a user's blood pressure in the wearable device 400 according to the present invention.
Figure 8 is a diagram showing the average absolute error between the estimated value of the user's blood pressure and the reference value through a deep learning model proposed in the present invention.

Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the attached drawings. The detailed description set forth below in conjunction with the accompanying drawings is intended to illustrate exemplary embodiments of the invention and is not intended to represent the only embodiments in which the invention may be practiced. The following detailed description includes specific details to provide a thorough understanding of the invention. However, one skilled in the art will appreciate that the present invention may be practiced without these specific details.

In some cases, in order to avoid ambiguity of the concept of the present invention, well-known structures and devices may be omitted or may be shown in block diagram form focusing on the core functions of each structure and device. In addition, the same components are described using the same reference numerals throughout this specification.

Before explaining the present invention, artificial intelligence (AI), machine learning, and deep learning will be explained. The easiest way to understand the relationship between these three concepts is to imagine three concentric circles. Artificial intelligence is the largest circle, followed by machine learning, and deep learning, which leads the current artificial intelligence boom, is the smallest circle.

The concept of artificial intelligence first appeared at the Dartmouth Conference held by Professor John McCarthy at Dartmouth College in 1956, and has been growing explosively in recent years. In particular, it has been accelerating since 2015 with the introduction of GPUs, which provide fast and powerful parallel processing performance. The advent of the big data era, in which storage capacity is increasing explosively and data in all areas, including images, text, and mapping data, are flooding in, also had a significant impact on this growth.

Artificial Intelligence - Implementing human intelligence into machines

The dream of artificial intelligence pioneers in 1956 was to ultimately create a complex computer with characteristics similar to human intelligence. In this way, artificial intelligence that thinks like a human while possessing human senses and thinking skills is called 'general AI', but artificial intelligence that can be created at the current level of technological development is called 'narrow AI'. Included. Narrow AI is characterized by its ability to perform certain tasks with better than human capabilities, such as social media image classification services or facial recognition functions.

Machine Learning - A Specific Approach to Implementing Artificial Intelligence

Machine learning automatically filters spam from your mailbox. Meanwhile, machine learning basically analyzes data using algorithms, learns through analysis, and makes judgments or predictions based on what has been learned. Therefore, the ultimate goal is to learn how to perform tasks by 'learning' the computer itself through large amounts of data and algorithms, rather than directly coding specific instructions for decision-making criteria into the software. Machine learning is derived from concepts directly proposed by early artificial intelligence researchers, and algorithmic methods include decision tree learning, inductive logic programming, clustering, reinforcement learning, and Bayesian networks. However, none of these have achieved general AI, which is the ultimate goal, and it is true that in many cases it was difficult to complete even narrow AI with early machine learning approaches.

Currently, machine learning is making great achievements in fields such as computer vision, but it has encountered a limitation in that a certain amount of coding work is involved throughout the process of implementing artificial intelligence, even if it is not a specific guideline. For example, when recognizing an image of a stop sign based on a machine learning system, developers use a boundary detection filter to programmatically identify the start and end of the object, shape detection to identify the side of the object, and letters such as 'S-T-O-P'. A classifier that recognizes must be produced through direct coding. In this way, machine learning works by recognizing images from a 'coded' classifier and 'learning' stop signs through an algorithm.

The image recognition rate of machine learning achieves sufficient performance for commercialization, but the image recognition rate may drop in certain situations where signs are difficult to see due to fog or trees. The reason computer vision and image recognition have not been able to reach human level until recently is because of these recognition rate issues and frequent errors.

Deep Learning - The technology that makes complete machine learning possible

Another algorithm created by early machine learning researchers, the artificial neural network, was inspired by the biological properties of the human brain, particularly the neuronal connection structure. However, unlike the brain, where any neuron in physical proximity can be interconnected, artificial neural networks have constant layer connections and data propagation directions.

For example, if an image is cut into a number of tiles and fed into the first layer of a neural network, the neurons pass the data to the next layer and repeat the process until the final output is produced in the last layer. Each neuron is then assigned a weight that represents the accuracy of the input based on the task it performs, and then the weights are all added together to determine the final output. In the case of a stop sign, characteristics of that image - octagonal shape, red color, sign letters, size, whether it's moving or not - are chopped up and 'examined' by neurons, and the neural network's job is to identify whether this is a stop sign or not. Here, a 'probability vector' is used, which predicts the result according to weights based on sufficient data.

Deep learning is a form of artificial intelligence that evolved from artificial neural networks and learns data using information input and output layers similar to neurons in the brain. However, because even basic neural networks require a huge amount of computation, the commercialization of deep learning faced difficulties from the beginning. Nevertheless, the researchers' research continued, and they succeeded in parallelizing an algorithm that proved the concept of deep learning based on a supercomputer. And the emergence of GPUs optimized for parallel computation dramatically accelerated the computation speed of neural networks, ushering in the emergence of true deep learning-based artificial intelligence.

Neural networks are likely to produce numerous wrong answers during the 'learning' process. Going back to the stop sign example, we might need to learn hundreds, thousands, or even millions of images to adjust the weights of neuron inputs precisely enough to always produce the correct answer, regardless of weather conditions or changes between day and night. Only when this level of accuracy is reached can the neural network be considered to have properly learned stop signs. In 2012, Google and Stanford University professor Andrew Ng implemented a 'Deep Neural Network' consisting of more than 1 billion neural networks with 16,000 computers. Through this, they extracted and analyzed 10 million images from YouTube and succeeded in having the computer classify pictures of people and cats. The computer learned the process of recognizing and judging the shape and appearance of the cat shown in the video.

The image recognition capabilities of systems trained with deep learning already surpass those of humans. Other areas of deep learning include the ability to identify cancer cells in the blood and tumors in MRI scans. Google's AlphaGo learned the basics of Go and further strengthened its neural network by repeatedly playing against AI like itself. With the advent of deep learning, the practicality of machine learning has been strengthened and the scope of artificial intelligence has expanded. Deep learning breaks down tasks in any way that can be supported by a computer system. Deep learning-based technologies, such as driverless cars, better preventive medicine, and more accurate movie recommendations, are already being used in our daily lives or are about to be put into practical use. Deep learning is evaluated as the present and future of artificial intelligence with the potential to realize general AI that appeared in science fiction.

Below we look at deep learning in more detail.

Deep learning is a type of artificial neural network (ANN) that uses human neural network theory. It is composed of a layer structure and has one input layer and one output layer. It is a set of machine learning models or algorithms that refer to a deep neural network (DNN) with more than one hidden layer (hereinafter referred to as the middle layer). Simply put, Deep Learning can be said to be an artificial neural network with deep layers.

It is estimated that the human brain consists of 25 billion nerve cells. The brain is made up of nerve cells, and each nerve cell (neuron) refers to one nerve cell that makes up a neural network. A nerve cell contains one cell body, one axon or nurite, which is a projection of the cell body, and usually several dendrites (dendrite or protoplasmic process). Information is exchanged between these nerve cells through junctions between nerve cells called synapses. If you look at a single nerve cell in isolation, it is very simple, but when these nerve cells come together, it can have human intelligence. The dendrites are the part that receives signals from other nerve cells (Input), and the axon is the part that extends very long from the cell body and is the part that transmits signals to other nerve cells (Output). There is a connection called a synapse that connects the axon and dendrites, which transmit signals between nerve cells. Rather than transmitting signals from nerve cells unconditionally, signals are only transmitted when the signal strength exceeds a certain value (threshold). It is done. In other words, not only does each synapse have a different connection strength, but it also determines whether or not to transmit a signal.

Artificial neural network (ANN), a field of artificial intelligence, is a mathematical model modeled by imitating the brain structure (neural network) of biology (usually human). In other words, artificial neural networks are implemented by imitating the information processing and transmission processes of biological nerve cells. Neural networks, which are implemented similarly to the way the human brain solves problems, have excellent parallelism because each nerve cell operates independently. In addition, because information is distributed across many connection lines, even if a problem occurs in a few nerve cells, it does not have a significant impact on the overall system, so it is resistant to a certain level of error and has the ability to learn about a given environment.

Deep neural network can be considered a descendant of artificial neural network, and is the latest version of artificial neural network, surpassing existing limitations and achieving success in areas where many artificial intelligence technologies had failed in the past. Looking at the modeling of an artificial neural network by imitating a biological neural network, in terms of processing units, biological neurons are used as nodes, and in terms of connectivity, synapses are used as weights. It was modeled as shown in Table 1.

biological neural network artificial neural network cell body node dendrites input Axon output synapse weight

Figure 1 is a diagram illustrating the layer structure of an artificial neural network.

Just as human biological neurons connect not one but many, but many, meaningful tasks, in the case of artificial neural networks, individual neurons are connected to each other through synapses, so multiple layers are connected to each other, and the connection strength between each layer is It is possible to edit (update) the weight. In this way, it is used in the field of learning and cognition due to its multi-layered structure and connection strength.

Each node is connected by weighted links, and the entire model learns by repeatedly adjusting the weights. Weights are the basic means for long-term memory and express the importance of each node. Simply put, an artificial neural network trains the entire model by initializing these weights and updating and adjusting the weights with the data set to be trained. After training is complete, when new input values come in, an appropriate output value is inferred. The learning principle of artificial neural networks can be viewed as a process in which intelligence is formed from the generalization of experience, and is done in a bottom-up manner. In Figure 1, when there are two or more middle layers (i.e. 5 to 10), the layers are considered to be deeper and are called a deep neural network, and the learning and inference model created through such a deep neural network can be referred to as deep learning. there is.

An artificial neural network can perform a certain role even if it has one middle layer (usually referred to as a 'hidden layer') excluding the input and output, but as the complexity of the problem increases, the number of nodes or the number of layers increases. The number must be increased. Among these, it is effective to use a multi-layer structure model by increasing the number of layers, but its scope of use is limited due to the inability to learn efficiently and the large amount of calculations to learn the network.

However, by overcoming the existing limitations as described above, artificial neural networks have been able to achieve deep structures. This makes it possible to build complex and highly expressive models, and groundbreaking results are being announced in various fields such as voice recognition, face recognition, object recognition, and text recognition.

Figure 2 is a diagram showing an example of a deep neural network.

A deep neural network (DNN) is an artificial neural network (ANN) made up of several hidden layers between an input layer and an output layer. A machine learning (Machine Learning) model or algorithm that refers to a Deep Neural Network (DNN) that has one or more hidden layers between the input layer and the output layer. is a set of The connection of a neural network is from the input layer to the hidden layer and from the hidden layer to the output layer.

Deep neural networks, like general artificial neural networks, can model complex non-linear relationships. For example, in a deep neural network structure for an object identification model, each object can be expressed as a hierarchical composition of basic elements of the image. At this time, additional layers can gradually integrate the characteristics of the gathered lower layers. This feature of deep neural networks allows complex data to be modeled with fewer units (nodes) than similarly performed artificial neural networks.

Previous deep neural networks have usually been designed as forward-feeding neural networks, but recent studies have successfully applied deep learning structures to recurrent neural networks (RNNs). For example, there is a case of applying a deep neural network structure to the field of language modeling. In the case of Convolutional Neural Network (CNN), not only has it been well applied in the field of computer vision, but each successful application case is also well documented. More recently, convolutional neural networks have been applied to the field of acoustic modeling for Automatic Speech Recognition (ASR), and are evaluated to be more successful than existing models. Deep neural networks can be trained with the standard error backpropagation algorithm. At this time, the weights can be updated through stochastic gradient descent using the equation below.

Electrocardiogram (ECG), photoplethysmogram (PPG), and ballistocardiogram (BCG) signals can be considered as biometric information highly correlated with blood pressure. Hereinafter, we will briefly look at the ECG sensor unit, PPG sensor unit, and BCG sensor unit.

signal Justice ECG Record the electrical activity of the heart (measures changes in the electrical excitement state caused by depolarization and repolarization of the myocardium). PPG Record changes in blood vessel volume using optical characteristics (LED green light is transmitted to the skin and partially absorbed and reflected by the skin, measuring the change.) BCG Record physical vibrations that occur when the heart contracts

An electrocardiogram is a curved record of the potential difference that occurs with the heartbeat as the heart contracts. The heart is unique compared to other muscles in the body in that it contracts automatically and rhythmically. The contraction of the heart muscle is like a generator that supplies electricity to living things. In other words, the driving force that causes contraction is a small electric current generated in the sinoatrial node of the atrium. As this weak current passes through the heart muscle, an electric current flows within the body, and this current can be recorded from the surface of the body. The device that records this is called an electrocardiography sensor (ECG sensor), and this recording is called an electrocardiogram (ECG).

Heart rate (Photoplethysmograph, PPG)

Next, a photoplethysmograph (PPG) sensor will be described as an example of a heart rate sensor. The PPG sensor is called an optical red pulse sensor. Photoplethysmograph (PPG) is a pulse wave measurement method that can determine heart rate activity or heart rate by measuring blood flow in blood vessels using the optical properties of biological tissue. A pulse wave is a pulsating waveform that appears as blood waves in the heart, and can be measured through changes in blood flow that occur due to the relaxation and contraction of the heart, that is, changes in the volume of blood vessels. The optical red pulse wave measurement method is a method of measuring pulse waves using light. Changes in optical characteristics such as reflection and absorption/transmission ratio of biological tissue that occur when volume changes are detected and measured by an optical sensor, making it possible to measure pulse. . This method is widely used because it enables non-invasive pulse measurement and has advantages such as miniaturization and ease of use, and can be used as a biosignal detection sensor in wearable devices.

Figure 3 is a diagram showing an example of PTT (Pulse Transit Time), a feature calculated using ECG, PPG, and BCG.

There is a method to estimate blood pressure based on statistical methods using PTT (Pulse Transit Time), a feature calculated using ECG, PPG, and BCG. However, PTT has the disadvantage of being influenced by the characteristics of the artery and requiring periodic correction due to changes in blood vessel elasticity over time. Additionally, when measuring biosignals, noise is generated depending on the user's actions, making it difficult to obtain features, which may reduce the estimation accuracy of the blood pressure estimation algorithm. Therefore, a deep learning-based method in which learning is performed through a deep neural network, rather than the traditional method of obtaining and estimating features, may be a good way to solve this problem.

The present invention proposes a device and method for estimating blood pressure using biosignals such as electrocardiogram and pulse wave using a deep learning model, and the deep learning model used in the present invention is a model learned by binding a plurality of models.

Figure 4 is a block diagram for explaining the components of the blood pressure measuring device 400 using the user's biological signals according to the present invention.

Referring to FIG. 4, the blood pressure measuring device 400 using the user's biological signals includes a processor 410, an ECG sensor unit 420, a PPG sensor unit 430, a memory 440, a display unit 450, and a communication unit. It may include (460). The ECG sensor unit 420 and the PPG sensor unit 430 described in FIG. 4 are shown as separate sensor units, but may be integrated into one sensor device. The memory 440 is electrically coupled to the processor 410 and may store information on estimated blood pressure values, learning data, deep learning models, etc. The blood pressure measurement device 400 using the user's biological signals is a wearable device and can be any type of device that can be worn by the user, including a wrist-type wearable device, a smart watch, etc. Hereinafter, it will be described under the name of a wearable device for measuring blood pressure.

FIG. 5 is a diagram illustrating a process for the wearable device 400 to estimate blood pressure.

Referring to FIGS. 4 and 5 , the ECG sensor unit 420 can detect the user's biosignals and extract or obtain data about the electrocardiogram signal (ECG signal) and output it. The PPG sensor unit 430 can sense the user's biological signals and extract or obtain data about the heart rate signal (PPG signal) and output it. The processor 410 may receive ECG data output from the ECG sensor unit 420, and the PPG sensor unit 430 may transmit data on the extracted or acquired heart rate signal to the processor 410.

The processor 410 may perform preprocessing on the ECG data (or ECG signal) received from the ECG sensor unit 420. Preprocessing may include removing noise by applying bandpass filtering to the ECG data received from the ECG sensor unit 420. Specifically, the processor 410 may perform preprocessing on the ECG data by applying a bandpass filter of 0.5 to 35 hertz (Hz). The processor 410 may perform preprocessing on the PPG data by applying a bandpass filter of 0.5 to 15 hertz (Hz).

The processor 410 may input the pre-processed ECG data and the pre-processed PPG data as input values for a pre-trained deep learning model proposed in the present invention.

The deep learning model proposed in the present invention is a model learned by binding a CNN (Convolutional Neural Network) model, a Bi-GRU (Bidirectional Gated Recurrent Unit) model, and an Attention model. The processor 410 may perform learning or test blood pressure estimation through a model bound to preprocessed ECG data and PPG data as a CNN (Convolutional Neural Network) model, Bi-GRU (Bidirectional Gated Recurrent Unit) model, and Attention model. there is.

A CNN model generally consists of a convolution and pooling layer that plays a role in feature extraction, and a fully connected layer that plays a classification role. The processor 410 inputs the preprocessed ECG and PPG data into the CNN model and outputs CNN weight parameters.

GRU is a variation of LSTM (Long Short-Term Memory) used for problems requiring long-term memory, and includes a Reset gate and an Update gate. The Reset gate determines how to combine the new input with the previous memory, and the Update gate determines how much of the previous memory information is maintained to calculate a new state. GRU has a simpler structure than LSTM. The Bi-GRU model is a reset gate that appropriately resets past information, an update gate that determines the update rate of past and present information, a candidate that calculates the current information candidate group, and finally, a combination of the update gate results and candidate results. It consists of calculating the current hidden layer. The processor 410 applies the CNN weight parameters output from the CNN model as input data of the Bi-GRU model and outputs the Bi-GRU weight parameters.

Attention model/mechanism refers to the fact that when predicting an output word, the entire input sentence is not referred to at the same rate, but rather the part of the input word that is related to the word to be predicted at that point is more focused (attention). It is a mechanism that allows you to see it. The attention model refers back to the encoder's input sequence every time it predicts an output word from the decoder. At this time, the word Attention is used because, rather than referring to the input sequence with equal weight, more emphasis is placed on input words that are related to the predicted word. The attention function calculates the similarity of all 'Keys' for a given 'Query'. And, this similarity is reflected in each 'Value' mapped to the key. Then, all the 'similarity reflected' values are added up and returned, and the attention value is returned. Attention values are obtained by weighting each attention weight of the encoder and the corresponding hidden state. Finally, the final attention value is obtained through the attention weight obtained above and each hidden state.

The processor 410 estimates the blood pressure value by applying the Bi-GRU weight parameters output from the Bi-GRU model to the attention model as input data. The blood pressure estimated here is systolic blood pressure (SBP) and/or diastolic blood pressure (DBP).

Figure 6 is a diagram illustrating input data of the deep learning model proposed in the present invention.

Referring to Figures 5 and 6, the processor 410 divides the pre-processed time-series ECG data and the pre-processed time-series PPG data into 5-second increments and inputs them into the CNN model, outputs weight values, and outputs weight values. It learns by updating the value through the Bi-GRU model and the attention model, and at this time, the blood pressure value is estimated as the output value. If the mean absolute error size in prediction through the predetermined deep learning model exceeds a predetermined value (e.g., 5), the processor 410 performs re-measurement/re-calibration. The accuracy of the estimated blood pressure value can be significantly improved by relearning a certain deep learning model. If the mean absolute error between the blood pressure value predicted by the predetermined deep learning model and the cuff-based measured blood pressure value exceeds a predetermined value (e.g., 5), the processor 410 uses a dictionary Relearning is performed by recalling the deep learning model learned and updating parameters using new data.

The processor 410 can control information about the estimated blood pressure value to be displayed on the display unit 450 of the wearable device 400 so that the user can check it. If the estimated blood pressure value is outside the range of a predetermined normal value, the processor 410 can transmit information on the estimated blood pressure value to the network through the communication unit 450.

FIG. 7 is a diagram illustrating a method of estimating a user's blood pressure in the wearable device 400 according to the present invention.

Referring to (a) of FIG. 7, the processor 410 can input preprocessed ECG data and preprocessed PPG data as input values of a pretrained deep learning model (Pretrained Model) proposed in the present invention. there is. The processor 410 estimates blood pressure values through a model that binds pre-processed ECG data and PPG data to a pre-trained CNN (Convolutional Neural Network) model, Bi-GRU (Bidirectional Gated Recurrent Unit) model, and Attention model.

Referring to (b) of FIG. 7, if the mean absolute error of the blood pressure value predicted by a pretrained deep learning model (Pretrained Model) exceeds a predetermined value (e.g., 5) The processor 410 re-learns a predetermined deep learning model through re-measurement/re-calibration. The processor 410 performs retraining by recalling a pre-trained deep learning model and updating parameters using new data. Such

Figure 8 is a diagram showing the average absolute error between the estimated value of the user's blood pressure and the reference value through a deep learning model proposed in the present invention.

Figure 8(a) shows the average absolute error size from the reference value in the case of no re-learning due to remeasurement/recalibration in a given deep learning model according to the present invention, and Figure 8(b) shows the size of the average absolute error according to the present invention. It shows the size of the average absolute error from the reference value when re-measuring/re-calibrating and re-learning from a given deep learning model. Here, the reference value may be a blood pressure value actually measured based on a cuff on the arm or wrist when performing ECG and PPG sensing.

Compared to the case without re-learning due to re-measurement/re-calibration in a predetermined deep learning model according to the present invention in (a) of FIG. 8, (b) of FIG. 8 shows re-measurement/re-calibration in a predetermined deep learning model according to the present invention. In the case of correction and re-learning, the average absolute error size became much smaller, and as a result, it can be seen that the accuracy was greatly improved compared to the actual blood pressure value.

In addition, as shown in (b) of FIG. 8, the size of the average absolute error of the blood pressure value measured with the PTT-based deep learning model is greater than that of the predetermined deep learning model (a model that binds three models) proposed in the present invention. The simulation results show that the average absolute error of the predicted blood pressure value is much smaller, so the blood pressure value measured through the deep learning model proposed in the present invention has the advantage of being very accurate.

The embodiments described above combine the components and features of the present invention in a predetermined form. Each component or feature should be considered optional unless explicitly stated otherwise. Each component or feature may be implemented in a form that is not combined with other components or features. Additionally, it is also possible to configure an embodiment of the present invention by combining some components and/or features. The order of operations described in embodiments of the present invention may be changed. Some features or features of one embodiment may be included in other embodiments or may be replaced with corresponding features or features of other embodiments. It is obvious that claims that do not have an explicit reference relationship in the patent claims can be combined to form an embodiment or included as a new claim through amendment after filing.

The processor 410 may also be called a controller, microcontroller, microprocessor, microcomputer, etc. Meanwhile, the processor 320 may be implemented by hardware, firmware, software, or a combination thereof. When implementing embodiments of the present invention using hardware, application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), and programmable logic devices (PLDs) configured to perform the present invention. , FPGAs (field programmable gate arrays), etc. may be provided in the processor 410.

It is obvious to those skilled in the art that the present invention can be embodied in other specific forms without departing from the essential features of the present invention. Accordingly, the above detailed description should not be construed as restrictive in all respects and should be considered illustrative. The scope of the present invention should be determined by reasonable interpretation of the appended claims, and all changes within the equivalent scope of the present invention are included in the scope of the present invention.

Claims (15)

In a wearable device for measuring blood pressure,
An ECG sensor unit that senses the user's electrocardiogram (ECG) signal and outputs ECG data;
A PPG sensor unit that senses the user's pulse wave (Photoplethysmogram, PPG) signal and outputs PPG data; and
After performing preprocessing on the ECG data and the PPG data, apply a predetermined deep learning model learned in advance to estimate the user's blood pressure,
The predetermined deep learning model is a model learned by binding a CNN (Convolutional Neural Network) model, Bi-GRU (Bidirectional Gated Recurrent Unit) model, and Attention model,
The processor,
Input the preprocessed ECG and PPG data into the CNN model to output CNN weight parameters,
Apply the output CNN weight parameters as input data of the Bi-GRU model to output Bi-GRU weight parameters,
A wearable device for measuring blood pressure, characterized in that the blood pressure value is estimated after applying the Bi-GRU weight parameters as input data to an attention model. According to paragraph 1,
The processor is a wearable for measuring blood pressure that re-learns the estimated blood pressure value through recalibration when the average absolute error between the blood pressure value estimated from the predetermined deep learning model and the reference value is greater than a predetermined value. device. According to paragraph 1,
A wearable device for measuring blood pressure, wherein the processor performs preprocessing on the ECG data by applying a bandpass filter of 0.5 to 35 hertz (Hz). According to paragraph 3,
The processor performs preprocessing on the PPG data by applying a bandpass filter of 0.5 to 15 hertz (Hz). According to paragraph 2,
A wearable device for measuring blood pressure, wherein the predetermined value is 5. According to paragraph 1,
A wearable device for measuring blood pressure, wherein the processor divides the preprocessed ECG and PPG data into 5-second increments and inputs them into the CNN model. According to paragraph 1,
The wearable device is a wearable device for measuring blood pressure, including a smart watch. According to paragraph 2,
A wearable device for measuring blood pressure, where the reference value is a blood pressure value measured based on a cuff. In a method for a wearable device to measure blood pressure,
Sensing the user's electrocardiogram (ECG) signal and outputting ECG data;
Sensing the user's pulse wave (Photoplethysmogram, PPG) signal and outputting PPG data; and
Preprocessing the ECG data and the PPG data and then applying them to a predetermined deep learning model to estimate the user's blood pressure,
The predetermined deep learning model is a model learned by binding a CNN (Convolutional Neural Network) model, a Bi-GRU (Bidirectional Gated Recurrent Unit) model, and an attention model,
The step of estimating the blood pressure is,
Inputting the preprocessed ECG and PPG data into a CNN model and outputting CNN weight parameters;
Applying the output CNN weight parameters as input data of the Bi-GRU model to output Bi-GRU weight parameters; and
A method of measuring blood pressure, comprising the step of applying the Bi-GRU weight parameters as input data to an attention model and then estimating the blood pressure value. According to clause 9,
A method of measuring blood pressure, further comprising relearning the magnitude of the average absolute error between the blood pressure value estimated from the predetermined deep learning model and the reference value by recalibrating the estimated blood pressure value. According to clause 10,
A method of measuring blood pressure, characterized in that the predetermined value is 5. According to clause 9,
The step of estimating blood pressure includes dividing the preprocessed ECG and PPG data into 5-second increments and inputting them into the CNN model. According to clause 9,
The step of estimating the blood pressure includes performing preprocessing on the ECG data by applying a bandpass filter of 0.5 to 35 hertz (Hz). According to clause 9,
The step of estimating the blood pressure includes performing preprocessing on the PPG data by applying a bandpass filter of 0.5 to 15 hertz (Hz). A computer-readable recording medium recording a computer program for performing the method according to any one of claims 9 to 14 on a computer.