KR102576829B1 - Method, system and non-transitory computer-readable recording medium for managing a learning model based on biometric signals - Google Patents

Abstract

According to one aspect of the present invention, there is provided a method for managing a learning model based on biometric signals, comprising: acquiring data on biosignals measured from a user's body, and when first data on the biosignals are obtained, A method comprising updating a learning model learned based on other data acquired before the first data using the first data, wherein the learning model is a learning model for detecting a cardiac abnormality event of the user. This is provided.

Description

METHOD, SYSTEM AND NON-TRANSITORY COMPUTER-READABLE RECORDING MEDIUM FOR MANAGING A LEARNING MODEL BASED ON BIOMETRIC SIGNALS}

The present invention relates to a method, system, and non-transitory computer-readable recording medium for managing a learning model based on biological signals.

In recent years, smart devices have emerged that allow users to easily and conveniently measure biosignals such as electrocardiograms at home without going to the hospital and even diagnose cardiac abnormalities such as arrhythmia based on this.

An example of the prior art in this regard is the technology disclosed in Korean Patent Publication No. 2007-96620, which includes a biosignal measuring unit that measures biosignals including an electrocardiogram, and an input from the biosignal measuring unit. An electrocardiogram abnormality sign detection unit that analyzes the electrocardiogram signal and detects electrocardiogram abnormality signs, a user activity status acquisition unit that acquires user activity status information when an abnormality detection signal is input from the electrocardiogram abnormality symptom detection unit, and acquires the user activity status An emergency situation determination unit that determines the presence or absence of an ECG abnormality based on the user activity status information input from the unit and the ECG signal input from the biosignal measuring unit, and an emergency situation that notifies the outside of the ECG abnormality input from the emergency situation determination unit. An electrocardiogram measurement device characterized by including a situation notification unit has been introduced.

However, according to the technologies introduced so far, including the above-described prior art, although the normal pattern of the electrocardiogram signal may vary depending on the user's age, physical structure (e.g., tilted position of the heart), etc., there is no uniform pattern. Because the presence or absence of cardiac abnormalities was determined based on (or demographic) criteria, there were many cases of false detection.

Accordingly, the present inventor(s) have developed a novel and advanced method that can accurately monitor cardiac abnormality events of the user using a learning model that can be dynamically managed using data on biosignals measured from the user's body. I am proposing a technology.

The purpose of the present invention is to solve all the problems of the prior art described above.

In addition, the present invention uses data about the user's biological signals to dynamically update the learned learning model based on other data acquired before the data, thereby continuously managing the learning model suitable for the user. The purpose.

In addition, another purpose of the present invention is to build and continuously update an individual learning model suited to the user's characteristics (e.g., age, body structure, aging, etc.) and accurately monitor cardiac abnormality events based on this. do.

A representative configuration of the present invention to achieve the above object is as follows.

According to one aspect of the present invention, there is provided a method for managing a learning model based on biometric signals, comprising: acquiring data on biosignals measured from a user's body, and when first data on the biosignals are obtained, A method comprising updating a learning model learned based on other data acquired before the first data using the first data, wherein the learning model is a learning model for detecting a cardiac abnormality event of the user. This is provided.

According to another aspect of the present invention, there is provided a system for managing a learning model based on biological signals, comprising: a biological signal acquisition unit that acquires data regarding biological signals measured from a user's body; and first data regarding the biological signals. When acquired, it includes a learning model management unit that updates a learning model learned based on other data acquired before the first data using the first data, wherein the learning model is configured to detect a cardiac abnormality event of the user. A system that is a learning model for

In addition, another method for implementing the present invention, another system, and a non-transitory computer-readable recording medium for recording a computer program for executing the method are further provided.

According to the present invention, it is possible to continuously manage a learning model suitable for the user by dynamically updating a learning model learned based on other data acquired before the data using data about the user's biological signals.

In addition, according to the present invention, it is possible to build and continuously update an individual learning model suited to the user's characteristics (e.g., age, body structure, aging, etc.) and accurately monitor cardiac abnormality events based on this.

Figure 1 is a diagram illustrating a schematic configuration of an entire system for managing a learning model based on biological signals according to an embodiment of the present invention.
Figure 2 is a diagram illustrating in detail the internal configuration of a learning model management system according to an embodiment of the present invention.
Figures 3 and 4 are diagrams illustrating a process in which a learning model is managed according to an embodiment of the present invention.

The detailed description of the present invention described below refers to the accompanying drawings, which show by way of example specific embodiments in which the present invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention. It should be understood that the various embodiments of the present invention are different from one another but are not necessarily mutually exclusive. For example, specific shapes, structures and characteristics described herein may be implemented with changes from one embodiment to another without departing from the spirit and scope of the invention. Additionally, it should be understood that the location or arrangement of individual components within each embodiment may be changed without departing from the spirit and scope of the present invention. Accordingly, the detailed description described below is not to be taken in a limiting sense, and the scope of the present invention should be taken to encompass the scope claimed by the claims and all equivalents thereof. Like reference numbers in the drawings indicate identical or similar elements throughout various aspects.

Hereinafter, several preferred embodiments of the present invention will be described in detail with reference to the attached drawings in order to enable those skilled in the art to easily practice the present invention.

Configuration of the entire system

Figure 1 is a diagram illustrating a schematic configuration of an entire system for managing a learning model based on biological signals according to an embodiment of the present invention.

As shown in FIG. 1, the entire system according to an embodiment of the present invention may be configured to include a communication network 100, a learning model management system 200, and a device 300.

First, the communication network 100 according to an embodiment of the present invention can be configured regardless of communication mode, such as wired communication or wireless communication, and can be used as a local area network (LAN) or a metropolitan area network (MAN). ), and a wide area network (WAN). Preferably, the communication network 100 referred to in this specification may be the known Internet or World Wide Web (WWW). However, the communication network 100 is not necessarily limited thereto and may include at least a portion of a known wired or wireless data communication network, a known telephone network, or a known wired or wireless television communication network.

For example, the communication network 100 is a wireless data communication network, including radio frequency (RF) communication, WiFi communication, cellular (LTE, etc.) communication, and Bluetooth communication (more specifically, Bluetooth low energy (BLE) communication. It may implement conventional communication methods such as Bluetooth Low Energy), infrared communication, ultrasonic communication, etc. at least in part.

Next, the learning model management system 200 according to an embodiment of the present invention may be a digital device equipped with a memory means and a microprocessor to have computing capabilities. This learning model management system 200 may be a server system.

The learning model management system 200 according to an embodiment of the present invention can communicate with a device 300, which will be described later, through the communication network 100, and obtain data about biological signals measured from the user's body. , when first data about the biosignal is acquired, a function of updating a learning model learned based on other data acquired before the first data using the first data can be performed. According to one embodiment of the present invention, this learning model may include a learning model for detecting a user's cardiac abnormality event.

In addition, cardiac abnormalities according to an embodiment of the present invention include premature atrial complex, premature ventricular complex, atrial fibrillation, atrial flutter, and polygenic atrial tachycardia. (multifocal atrial tachycardia), paroxysmal supraventricular tachycardia, Wolff-Parkinson-White syndrome, ventricular tachycardia, ventricular fibrillation, AV block ) may include a variety of abnormal cardiac events associated with arrhythmias. Meanwhile, it should be noted that the cardiac abnormality event according to an embodiment of the present invention is not necessarily limited to the embodiments related to arrhythmia listed above, and may also be changed to various cardiac abnormality events related to heart health, such as cardiac arrest.

The configuration and functions of the learning model management system 200 according to the present invention will be discussed in more detail below. Meanwhile, although the learning model management system 200 has been described as above, this description is illustrative, and at least some of the functions or components required for the learning model management system 200 may be used in the device 300, which will be described later, as needed. Alternatively, it is obvious to those skilled in the art that it may be implemented within an external system (not shown) or included within the device 300 or an external system.

Next, the device 300 according to an embodiment of the present invention is a digital device including a function that allows communication after accessing the learning model management system 200 through the communication network 100, such as a smartphone or tablet. Any portable digital device, such as a PC, equipped with a memory means, equipped with a microprocessor, and equipped with computing power can be adopted as the device 300 according to the present invention. In addition, according to an embodiment of the present invention, the device 300 includes a biometric signal measurement sensor (e.g., electrocardiogram sensor, electrocardiogram sensor, heart rate sensor, brain wave sensor, A pulse sensor) may be further included. Meanwhile, according to an embodiment of the present invention, the bio-signal measurement sensor discussed above may be included in a wearable device 400 to be described later.

Meanwhile, the device 300 according to an embodiment of the present invention may include an application to support the function according to the present invention for managing a learning model based on biometric signals. Such an application may be downloaded from the learning model management system 200 or an external application distribution server (not shown).

Configuration of the learning model management system

Below, we will look at the internal structure of the learning model management system 200 and the function of each component, which performs important functions for implementing the present invention.

FIG. 2 is a diagram illustrating in detail the internal configuration of the learning model management system 200 according to an embodiment of the present invention.

As shown in Figure 2, the learning model management system 200 according to an embodiment of the present invention includes a biometric signal acquisition unit 210, a learning model management unit 220, a monitoring unit 230, a communication unit 240, and It may be configured to include a control unit 250. According to one embodiment of the present invention, at least some of the biosignal acquisition unit 210, learning model management unit 220, monitoring unit 230, communication unit 240, and control unit 250 communicate with an external system. It may be a program module that does. These program modules may be included in the learning model management system 200 in the form of an operating system, application module, or other program module, and may be physically stored in various known storage devices. Additionally, these program modules may be stored in a remote memory device capable of communicating with the learning model management system 200. Meanwhile, such program modules include, but are not limited to, routines, subroutines, programs, objects, components, data structures, etc. that perform specific tasks or execute specific abstract data types according to the present invention.

First, the biosignal acquisition unit 210 according to an embodiment of the present invention may perform a function of acquiring data regarding biosignals measured from the user's body. Data related to biological signals according to an embodiment of the present invention may include data related to electrocardiogram, heart rate, brain waves, pulse, etc. Meanwhile, the data regarding biological signals according to the present invention is not necessarily limited to those listed above, and any biological signal that may be associated with cardiac abnormalities may be expanded within the scope of achieving the purpose of the present invention. reveal.

For example, the biosignal acquisition unit 210 according to an embodiment of the present invention acquires data about the user's electrocardiogram signal through a device (e.g., wearable device 400) attached to the user's body. can do. Devices attached to the body (eg, smart watches, smart patches, etc.) may be wired or wireless devices and may include a sensor for measuring electrocardiogram signals.

For another example, the biosignal acquisition unit 210 according to an embodiment of the present invention may acquire data related to biosignals from the user's body at predetermined unit time intervals. The bio-signals acquired at such predetermined unit time intervals may include at least one of the user's bio-signals continuously measured during the unit interval time and the user's bio-signals temporarily measured at each unit interval time. Meanwhile, it should be noted that the unit time interval according to an embodiment of the present invention is not fixed to a specific interval and may change depending on time zone, situation, or user settings.

For another example, the biosignal acquisition unit 210 according to an embodiment of the present invention converts biosignals from the user's body into a predetermined unit size (e.g., data on an electrocardiogram signal including a predetermined number of cycles). You can obtain data about. More specifically, the bio-signal acquisition unit 210 may acquire data on bio-signals measured from the user's body whenever the data accumulates to a predetermined level.

Next, when the first data related to the biosignal is acquired, the learning model management unit 220 according to an embodiment of the present invention applies a learning model learned based on other data acquired before the first data to the first data. You can perform the update function using . For example, at least one of probability, vector, matrix, logit, and coordinate associated with cardiac abnormality may be output from the learning model according to an embodiment of the present invention. This output can be classified or clustered into specific cardiac abnormality events (e.g., normal, abnormal, etc.) according to predetermined criteria (such clustering may include distance (e.g., K-means), density (DB-SCAN), etc. Clustering can be achieved by, etc.). Additionally, these predetermined standards may be preset or may be dynamically updated during the learning process.

For example, a learning model according to an embodiment of the present invention may be composed of an input layer, a hidden layer, and an output layer based on an artificial neural network, and a learning model management unit ( 220) may perform pre-training on the learning model based on other data acquired before the first data about the biosignal is acquired and update the weights of the pre-trained learning model using the first data. . According to an embodiment of the present invention, such learning models may include Autoencoder, Generative Adversarial Nets (GAN), U-NET, etc. Meanwhile, the learning model according to the present invention is not necessarily limited to the learning models listed above, and supervised learning (in this case, labels for data are added in the learning process) is used within the scope of achieving the purpose of the present invention. may be provided), and can be changed to various learning models included in unsupervised learning or reinforcement learning.

More specifically, the learning model management unit 220 performs transfer learning using the first data on a learning model learned based on other data acquired before the first data about the biosignal is acquired, thereby performing the above learning. You can update the model's weights. That is, transfer learning may be performed in which the weights of a learning model pre-trained based on previous data are updated using newly acquired data.

Additionally, the learning model management unit 220 may allow learning of the learning model to start when a learning start trigger occurs. For example, this learning start trigger is based on certain conditions (e.g., the number of data is 10 or more, the time of first acquisition of data) based on the point at which the minimum (or appropriate level) of data required to train the learning model is accumulated. This may occur when (10 minutes after) is satisfied.

For example, the learning model management unit 220 accumulates and acquires data about the user's biosignals until the learning start trigger occurs, and when the learning start trigger occurs, the data about the accumulated biosignals is acquired. You can use it to learn a learning model (specifically, pre-training). In other words, learning can begin after data above a certain level required for learning is accumulated. Next, the learning model management unit 220 may update the learned learning model using the first data obtained after the accumulated biosignal data. Next, the learning model management unit 220 may update the updated learning model again using the second data obtained immediately after the first data. Meanwhile, data related to the above biosignals (eg, first data, second data) may be acquired at predetermined unit time intervals.

Additionally, the learning model management unit 220 may determine the extent to which the first data is reflected in the update of the learning model by referring to the number of other data used to learn the learning model before the first data. For example, the number of such data may be specified based on the number of data continuously acquired according to a predetermined unit time interval or a predetermined unit size as discussed above.

Specifically, the learning model management unit 220 learns by varying the importance (e.g., weight reflected in the learning model) of the first data according to the number of other data used to learn the learning model before the first data. The model can be updated.

For example, when the learning model management unit 220 updates the learning model using the first data, the first data above is greater than the number of other data used to learn the learning model before the first data. In the case of re-updating the above updated learning model using the second data acquired immediately after the acquisition point, the number of other data used to learn the above updated learning model before the second data is greater, When updating the weights of the learning model, the importance of the first data may be set higher than the importance of the second data.

Next, the monitoring unit 230 according to an embodiment of the present invention analyzes data on biosignals measured from the user's body using the learning model updated by the learning model management unit 220, thereby generating a cardiac abnormality event. It can perform a monitoring function.

For example, when the learning model is updated based on first data related to biological signals, the monitoring unit 230 refers to the result output when the first data is input to the updated learning model and detects a cardiac abnormality event. You can decide whether it will happen or not.

Additionally, if it is determined that a cardiac abnormality event has occurred, the monitoring unit 230 may store the cardiac abnormality event and provide notification information regarding the cardiac abnormality event. For example, such notification information may be provided based on at least one of visual (eg, text, image), auditory (eg, sound) and tactile (eg, vibration) methods.

Next, the communication unit 240 according to an embodiment of the present invention performs a function that enables data transmission and reception from/to the biosignal acquisition unit 210, the learning model management unit 220, and the monitoring unit 230. You can.

Lastly, the control unit 250 according to an embodiment of the present invention has a function of controlling the flow of data between the biosignal acquisition unit 210, the learning model management unit 220, the monitoring unit 230, and the communication unit 240. It can be done. That is, the control unit 250 according to the present invention controls the data flow to/from the outside of the learning model management system 200 or the data flow between each component of the learning model management system 200, thereby controlling the biometric signal acquisition unit 210. ), the learning model management unit 220, the monitoring unit 230, and the communication unit 240 can each be controlled to perform their own functions.

Figures 3 and 4 are diagrams illustrating a process in which a learning model is managed according to an embodiment of the present invention.

Referring to FIG. 3, according to an embodiment of the present invention, a smartphone 300 (i.e., device 300 according to the present invention) is connected to a wearable device 400, and the wearable device 400 ) can assume a situation in which biosignals are obtained from the user's body and provided to the smartphone 300. Here, the wearable device 400 according to an embodiment of the present invention is a digital device that includes a function that allows communication after connecting to the smartphone 300 or the learning model management system 200 through the communication network 100. As such, it may include various devices such as smart watches and wireless patches that are directly or indirectly attached to the user's body and can acquire biological signals measured from the user. Meanwhile, according to one embodiment of the present invention, all or part of the functions of the learning model management system 200 are included in the smartphone 300 or wearable device 400 according to the present invention (e.g., in the form of an application) may be included).

First, according to an embodiment of the present invention, the wearable device 400 attached to the user's body can obtain data about biosignals measured from the user's body.

Then, according to an embodiment of the present invention, data regarding the above measured biosignals are sent to the learning model management system 200 through the smartphone 300 at predetermined unit time intervals (or predetermined unit sizes). can be provided.

Next, referring to FIG. 4, according to an embodiment of the present invention, when the data related to the biosignal provided above is accumulated to a predetermined level, a learning start trigger may be generated, and the biosignal accumulated to the predetermined level may be A learning model 410 for monitoring cardiac abnormality events may be learned using the related data 401.

In addition, when the above learning is completed according to an embodiment of the present invention, the learned learning model 410 is used to determine whether a cardiac abnormality event exists in the data 401 regarding biosignals accumulated at a predetermined level. can be monitored (451).

Then, according to an embodiment of the present invention, if the first data 402 related to the biosignal is additionally acquired after the above learning start trigger occurs, other data 401 acquired before the first data is added to the first data 401. The learning model 410 learned based on the data may be updated (i.e., the updated learning model 420) using the first data 402.

For example, by performing transfer learning using the first data 402 on the learning model 410 learned based on other data 401 acquired before the first data 402 is acquired, The weight of the learning model 410 may be updated.

In addition, when the update of the above learning model is completed according to an embodiment of the present invention, whether a cardiac abnormality event exists in the above first data 402 will be monitored using the updated learning model 420. Can (452).

Then, if the second data 403 is additionally acquired after the above first data 402 is acquired according to an embodiment of the present invention, it is based on other data acquired before the second data 403. The learned learning model 430 can be updated using the second data 403 above.

In addition, when the update of the above learning model is completed according to an embodiment of the present invention, whether a cardiac abnormality event exists in the above second data 403 can be monitored using the updated learning model 430. Can (453).

Meanwhile, according to an embodiment of the present invention, if it is determined that a cardiac abnormality event has occurred during the above monitoring process, the cardiac abnormality event may be stored and notification information regarding the cardiac abnormality event may be provided (460).

As discussed above, according to an embodiment of the present invention, whenever new data regarding the user's biosignal is acquired, the learning model learned based on the data acquired just before that is dynamically updated, thereby causing the user's heart abnormality event. It is possible to continuously maintain a sophisticated learning model for monitoring. Furthermore, the extent to which new data is reflected in the update of the existing learning model (431, 432, 433) is adaptive depending on the number of data used to train the existing learning model (or the number of data acquired before the new data). By allowing it to be adjusted, the learning model can be updated stably.

The embodiments according to the present invention described above can be implemented in the form of program instructions that can be executed through various computer components and recorded on a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc., singly or in combination. Program instructions recorded on the computer-readable recording medium may be specially designed and configured for the present invention, or may be known and usable by those skilled in the computer software field. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs and DVDs, and magneto-optical media such as floptical disks. medium), and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, etc. Examples of program instructions include not only machine language code such as that created by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. A hardware device can be converted into one or more software modules to perform processing according to the invention and vice versa.

In the above, the present invention has been described in terms of specific details, such as specific components, and limited embodiments and drawings, but this is only provided to facilitate a more general understanding of the present invention, and the present invention is not limited to the above embodiments. Anyone with ordinary knowledge in the technical field to which the invention pertains can make various modifications and changes from this description.

Therefore, the spirit of the present invention should not be limited to the above-described embodiments, and the scope of the patent claims described below as well as all scopes equivalent to or equivalently changed from the scope of the claims are within the scope of the spirit of the present invention. It will be said to belong to

100: communication network
200: Learning model management system
210: Biological signal acquisition unit
220: Learning model management unit
230: Monitoring unit
240: Department of Communications
250: control unit
300: device
400: Wearable device

Claims (11)

A method implemented in a learning model management system for managing a learning model based on biological signals, wherein the learning model management system includes a biological signal acquisition unit and a learning model management unit,
a step of the bio-signal acquisition unit acquiring data about bio-signals measured from the user's body, and
When the learning model management unit obtains first data about the biosignal, updating a previously learned learning model based on other data acquired before the first data using the first data; ,
The learning model is a learning model for detecting cardiac abnormality events of the user,
The first data and the other data are data related to biological signals measured from the user's body, and the first data is data that follows the other data in time,
As the number of other data increases, the extent to which the first data is reflected in the update of the learning model decreases.
method. According to paragraph 1,
Data regarding the biological signals are acquired at predetermined unit time intervals.
method. According to paragraph 1,
In the update step, pre-training the learning model based on data about biosignals accumulated until the learning start trigger occurs.
method. According to paragraph 1,
In the update step, transfer learning is performed using the first data to update the weights of the learning model.
method. delete A non-transitory computer-readable recording medium recording a computer program for executing the method according to claim 1. A system for managing a learning model based on biological signals,
A biometric signal acquisition unit that acquires data about biosignals measured from the user's body, and
When first data about the biosignal is acquired, a learning model manager that updates a previously learned learning model based on other data acquired before the first data using the first data,
The learning model is a learning model for detecting cardiac abnormality events of the user,
The first data and the other data are data related to biological signals measured from the user's body, and the first data is data that follows the other data in time,
As the number of other data increases, the extent to which the first data is reflected in the update of the learning model decreases.
system. In clause 7,
Data regarding the biological signals are acquired at predetermined unit time intervals.
system. In clause 7,
The learning model management unit pre-trains the learning model based on data about biological signals accumulated until the learning start trigger occurs.
system. In clause 7,
The learning model management unit updates the weights of the learning model by performing transfer learning using the first data.
system. delete

Images