Method, system, and non-transitory computer-readable recording medium for assisting biological signal analysis using clustering

Abstract

According to one aspect of the present invention, there is provided a method for assisting in biological signal analysis using clustering, the method including extracting features from a plurality of biological signal data related to biological signals, generating a plurality of feature vectors for each of the plurality of biological signal data based on the extracted features, and performing clustering on the plurality of biological signal data with reference to the plurality of feature vectors.

Description

Technical Field

[0001] The present invention relates to a method, a system, and a non-transitory computer-readable recording medium for assisting in biological signal analysis using clustering.

Background Art

[0002] Recently, wearable monitoring devices that can constantly monitor biological signals in daily life without visiting a hospital have been introduced. The biological signals measured through the wearable monitoring devices can be analyzed by automated or intelligent biological signal analysis models, and various latest technologies are being utilized to improve the accuracy of the analysis.

[0003] Despite the fact that the accuracy of analysis has been improved due to analysis models to which latest technologies such as artificial neural networks are applied, in order to make a final determination or diagnosis for biological signals, medical staff (such as doctors and reading experts) still have to go through the process of checking the analysis results of the biological signal analysis model.

[0004] However, since the biological signal data measured over several days or weeks or the analysis result data for the biological signal data are huge, reaching hundreds of thousands of pieces, there is a problem that a lot of time and effort are consumed for medical staff to check all the data one by one.

[0005] In particular, when the biological signal is an electrocardiogram signal, since the important features of the signal (such as the QRS complex form) appear differently depending on the characteristics of the measurement device and each individual being measured, even if an artificial intelligence-based analysis model is used, it is difficult to accurately analyze the electrocardiogram signal 100% only with that analysis model, and the process of medical staff directly checking the electrocardiogram signal is essentially required.

[0006] As an example, assuming that an electrocardiogram signal with a normal heart rate of 60 BPM is measured for 14 days, a total of more than 1.2 million beats (data units including QRS complexes) of data will be generated. Even if only about 1% of this data corresponds to abnormal data, 12,000 abnormal beat data will be generated. However, it takes a great deal of effort and time for medical staff to check and verify such a huge amount of data one by one.

[0007] Therefore, the inventor(s) of the present invention devised a technique to support medical staff in efficiently checking and verifying a plurality of biological signal data (or analysis results thereof) having similar characteristics (such as shape, rhythm, pattern, etc.) by accurately and efficiently clustering the biological signal data.

Summary of the Invention

Problems to be Solved by the Invention

[0008] An object of the present invention is to solve all of the above-mentioned problems.

[0009] Another object of the present invention is to enable medical staff to efficiently check and verify biological signal data and analysis result data thereof based on clusters without experiencing the inefficiency of checking all biological signal data, by extracting features from a plurality of biological signal data related to biological signals, generating a plurality of feature vectors for each of the plurality of biological signal data based on the extracted features, and performing clustering on the plurality of biological signal data with reference to the plurality of feature vectors.

[0010] Also, another object of the present invention is to improve the accuracy of clustering by utilizing various information obtained from biological signal data, and to adjust the number of clusters according to the situation, thereby enhancing all of the efficiency and reliability of medical staff's checking and verification of biological signal data analysis results.

Means for Solving the Problems

[0011] A representative configuration of the present invention for achieving the above object is as follows.

[0012] According to one aspect of the present invention, there is provided a method for assisting in biological signal analysis using clustering, the method comprising: extracting features from a plurality of biological signal data related to biological signals; generating a plurality of feature vectors for each of the plurality of biological signal data based on the extracted features; and performing clustering on the plurality of biological signal data with reference to the plurality of feature vectors.

[0013] According to another aspect of the present invention, there is provided a system for assisting in biological signal analysis using clustering, the system comprising: a feature extraction unit that extracts features from a plurality of biological signal data related to biological signals and generates a plurality of feature vectors for each of the plurality of biological signal data based on the extracted features; and a clustering management unit that performs clustering on the plurality of biological signal data with reference to the plurality of feature vectors.

[0014] In addition, there is further provided another method for implementing the present invention, a system, and a non-transitory computer-readable recording medium for recording a computer program for executing the method.

Advantages of the Invention

[0015] According to the present invention, an effect is achieved in that medical staff can efficiently inspect biological signal data and analysis result data thereof based on clusters without experiencing the inefficiency of fully inspecting all biological signal data.

[0016] Also, according to the present invention, by utilizing various information obtained from biological signal data to improve the accuracy of clustering and adjusting the number of clusters according to the situation, an effect is achieved in that both the efficiency and reliability of the inspection by medical staff of the analysis results of biological signal data can be enhanced.

Brief Description of the Drawings

[0017]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Modes for Carrying Out the Invention

[0018] The detailed description of the present invention to be described below refers to the accompanying drawings that illustrate specific embodiments in which the present invention can be implemented as examples. These embodiments are described in detail so that those skilled in the art can fully implement the present invention. It should be understood that various embodiments of the present invention are different from each other but do not necessarily have to be mutually exclusive. For example, the specific shapes, structures, and characteristics described herein can be embodied in other embodiments without departing from the spirit and scope of the present invention in relation to one embodiment. Also, it should be understood that the position or arrangement of individual components within each disclosed embodiment can be changed without departing from the spirit and scope of the present invention. Therefore, the detailed description to be described below is not used in a limiting sense, and the scope of the present invention is limited only by the appended claims together with all scopes equivalent to what the claims claim, provided that it is appropriately described. Similar reference numerals in the drawings refer to the same or similar functions across various aspects.

[0019] Hereinafter, in order to enable those having ordinary knowledge in the technical field to which the present invention pertains to easily implement the present invention, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

[0020] Configuration of the entire system Hereinafter, a preferred embodiment of the biological signal analysis system according to the present invention will be described in detail as follows.

[0021] FIG. 1 is a drawing schematically showing the configuration of the overall system according to the present invention.

[0022] As shown in FIG. 1, the overall system according to one embodiment of the present invention can be composed of a communication network 100, a biological signal analysis system 200, and a device 300.

[0023] First, the communication network 100 according to an embodiment of the present invention can be configured regardless of the mode of communication such as wired communication or wireless communication, and can be composed of various communication networks such as a short-distance communication network (LAN, Local Area Network), a metropolitan area communication network (MAN, Metropolitan Area Network), and a wide area communication network (WAN, Wide Area Network). Preferably, the communication network 100 referred to in this specification can include well-known short-distance wireless communication networks such as Wi-Fi, Wi-Fi Direct, LTE Direct, and Bluetooth. However, the communication network 100 is not necessarily limited thereto, and may include at least a part of a well-known wired / wireless data communication network, a well-known telephone network, or a well-known wired / wireless television communication network.

[0024] For example, the communication network 100 can be a wireless data communication network that implements at least a part of conventional communication methods such as Wi-Fi communication, Wi-Fi Direct communication, Long Term Evolution (LTE) communication, Bluetooth communication (including Bluetooth Low Energy (BLE)), infrared communication, and ultrasonic communication. As another example, the communication network 100 can be an optical communication network that implements at least a part of conventional communication methods such as LiFi (Light Fidelity).

[0025] Next, the biological signal analysis system 200 according to an embodiment of the present invention can perform a function of extracting features from a plurality of biological signal data related to biological signals, generating a plurality of feature vectors for each of the plurality of biological signal data based on the extracted features, and performing clustering on the plurality of biological signal data with reference to the plurality of feature vectors.

[0026] On the one hand, the biosignal analysis system 200 has been described as above. However, such a description is exemplary, and it is obvious to those skilled in the art that at least some of the functions and components required for the biosignal analysis system 200 may be implemented or included within the device 300 as needed.

[0027] Finally, the device 300 according to an embodiment of the present invention is a digital device including a function capable of communicating after being connected to the biosignal analysis system 200, and any digital device equipped with a memory means and a microprocessor and having computing power can be adopted as the device 300 according to the present invention. The device 300 may be a wearable device such as smart glasses, a smartwatch, a smart patch, a smart band, a smart ring, a smart necklace, etc., or may be a somewhat traditional device such as a smartphone, a smart pad, a desktop computer, a notebook computer, a workstation, a PDA, a web pad, a mobile phone, etc.

[0028] In particular, the device 300 according to an embodiment of the present invention may include sensing means (e.g., contact electrodes, infrared sensors, etc.) for acquiring a predetermined biosignal from the human body, and may include display means for providing various information related to the measurement of the biosignal to the user.

[0029] Also, according to an embodiment of the present invention, the device 300 may further include an application program for performing the functions according to the present invention. Such an application can exist in the form of a program module within the corresponding device 300. The nature of such a program module may generally be similar to the feature extraction unit 210, the clustering management unit 220, the communication unit 230, and the control unit 240 of the biosignal analysis system 200 as described later. Here, at least a part of the application may be replaced by a hardware device or a firmware device capable of performing substantially the same or equivalent functions as needed.

[0030] Configuration of the biological signal analysis system Hereinafter, the internal configuration of the biosignal analysis system 200 that performs important functions for the implementation of the present invention and the functions of each component will be described in detail.

[0031] FIG. 2 is a drawing exemplarily showing the internal configuration of a biosignal analysis system according to an embodiment of the present invention.

[0032] Referring to FIG. 2, the biosignal analysis system 200 according to an embodiment of the present invention may include a feature extraction unit 210, a clustering management unit 220, a communication unit 230, and a control unit 240. According to an embodiment of the present invention, at least a part of the feature extraction unit 210, the clustering management unit 220, the communication unit 230, and the control unit 240 of the biosignal analysis system 200 may be program modules that communicate with an external system (not shown). Such program modules may be included in the biosignal analysis system 200 in the form of an operating system, an application program module, and other program modules, and may be physically stored on various known storage devices. Also, such program modules may be stored in a remote storage device that can communicate with the biosignal analysis system 200. On the other hand, such program modules include, but are not limited to, routines, subroutines, programs, objects, components, data structures, etc. that perform specific operations described later according to the present invention or execute specific abstract data types.

[0033] On the other hand, according to an embodiment of the present invention, it is clarified that the biosignal analysis system 200 can operate independently without communication with the outside, and in such a case, the communication unit 230 may not be included in the biosignal analysis system 200.

[0034] First, the feature extraction unit 210 according to an embodiment of the present invention can acquire a plurality of biosignal data measured from the body of the subject. Here, the biosignal data may be electrocardiogram signal data measured from the body of the subject, and such electrocardiogram signal data may be composed of beats including QRS complexes.

[0035] For example, the electrocardiogram signal data to be clustered can be composed of five beat sequences, and an interval from the R peak of the first beat to the R peak of the last beat can be defined as one electrocardiogram signal data (see FIG. 3).

[0036] And, the feature extraction unit 210 according to an embodiment of the present invention can extract features from a plurality of biological signal data related to biological signals, and can generate a plurality of feature vectors for each of the plurality of biological signal data based on the extracted features.

[0037] For example, the feature extraction unit 210 according to an embodiment of the present invention can extract features from biological signal data by using a feature extraction model learned through supervised learning.

[0038] According to an embodiment of the present invention, in the QRS complex included in the electrocardiogram signal, the change in the signal value is significantly represented. Therefore, when clustering based on the raw data of the biological signal data, problems may occur in that clustering is not correctly performed due to beat position differences, beat height differences, baseline fluctuations, noise, etc. According to the present invention, features are extracted from the raw data to generate feature vectors, and clustering is performed based on the feature vectors. Therefore, it becomes possible to reduce the problems that may occur when clustering based on raw data.

[0039] Furthermore, the feature extraction unit 210 according to an embodiment of the present invention can normalize a plurality of raw data and extract features from the normalized data. By doing so, even when the length of the raw data changes due to various factors such as heart rate, it becomes possible to make the length of the data to be clustered the same (uniform).

[0040] In addition, the feature extraction unit 210 according to an embodiment of the present invention can extract features from a plurality of raw data using a feature extraction model, and the plurality of raw data can be normalized by the method described above and input into the feature extraction model. For example, the feature extraction model can be a model obtained by optimizing the SE (Squeeze & Excitation) ResNet model according to the object of the present invention.

[0041] On the other hand, the feature extraction unit 210 according to an embodiment of the present invention can extract a scaled R-R interval in which features related to the inter-beat rhythm within a plurality of raw data related to an electrocardiogram signal are reflected, and the scaled R-R interval extracted in this way can be used as a criterion for clustering described later.

[0042] FIG. 4 is a drawing exemplarily showing the structure of a model used for extracting features according to an embodiment of the present invention. However, it is clarified that the feature extraction unit 210 according to the present invention is not necessarily implemented by the model illustrated in FIG. 4 and can be changed as much as possible within the scope that can achieve the object of the present invention.

[0043] Next, the clustering management unit 220 according to an embodiment of the present invention can obtain analysis result data for a plurality of biological signal data from a biological signal analysis model. According to an embodiment of the present invention, the biological signal analysis model can be a model that outputs an analysis result as to whether the biological signal data to be analyzed corresponds to arrhythmia or an analysis result as to which type of arrhythmia it corresponds to.

[0044] For example, according to an embodiment of the present invention, the biological signal analysis model can calculate a score as to whether the biological signal data corresponds to a normal state (or does not correspond) from the perspective of arrhythmia by analyzing the biological signal data of the subject using a machine learning algorithm such as an artificial neural network.

[0045] As another example, according to one embodiment of the present invention, a biosignal analysis model can calculate a score regarding whether (or not) biosignal data corresponds to a specific type of arrhythmia by analyzing the biosignal data of a subject using a machine learning algorithm such as an artificial neural network.

[0046] Here, according to one embodiment of the present invention, the score calculated by the biosignal analysis model can be a concept including a value regarding at least one of probability, vector, matrix, and coordinate regarding whether (or not) it corresponds to a normal state or whether (or not) it corresponds to a specific type of arrhythmia.

[0047] On the other hand, biosignals that can be analyzed by the biosignal analysis model may include signals related to electrocardiogram (ECG), electromyogram (EMG), electroencephalogram (EEG), photoplethysmogram (PPG), heart beat, body temperature, blood glucose, pupil change, blood pressure, blood dissolved oxygen content, etc.

[0048] Next, a clustering management unit 220 according to one embodiment of the present invention can perform clustering on a plurality of first-type biosignal data analyzed to correspond to a first type by a biosignal analysis model among a plurality of biosignal data.

[0049] Here, according to one embodiment of the present invention, the first type is the broadest concept referring to a type that can be discriminated by a biosignal analysis model, and includes all of a normal state from the perspective of arrhythmia, an abnormal state from the perspective of arrhythmia, and a state corresponding to a specific type of arrhythmia (for example, Atrial Premature Contraction (APC), Ventricular Premature Complexes (VPC), Atrial fibrillation (A.Fib), Paroxysmal Supra Ventricular Tachycardia (PSVT), etc.).

[0050] Specifically, the clustering management unit 220 according to an embodiment of the present invention can cluster a plurality of first-type biological signal data analyzed as corresponding to the first type into at least one cluster. According to an embodiment of the present invention, by performing clustering in this way, the biological signal data that will belong to the same cluster can include common features (patterns, feature points, waveforms, etc.) with each other.

[0051] For example, as algorithms that can be used for biological signal data clustering according to an embodiment of the present invention, complete linkage clustering, k-means, mean shift, Gaussian Mixture Model (GMM), DBSCAN (Density-Based Spatial Clustering of Applications with Noise), Self-Organizing Map (SOM), etc. can be assumed. However, it should be clearly stated that the clustering algorithm according to the present invention is not necessarily limited to the above-listed ones and can be changed as much as possible within the range that can achieve the object of the present invention.

[0052] Also, the clustering management unit 220 according to an embodiment of the present invention can perform clustering on the extracted plurality of feature vectors.

[0053] Then, the clustering management unit 220 according to an embodiment of the present invention can perform clustering on a plurality of biological signal data corresponding to the plurality of feature vectors with reference to the plurality of feature vectors.

[0054] For example, the biological signal data (multiple pieces of raw data) according to an embodiment of the present invention may include data classified as APC (Atrial Premature Complex) or data classified as VPC (Ventricular Premature Complex). The clustering management unit 220 according to an embodiment of the present invention can perform clustering on the data classified as APC and clustering on the data classified as VPC separately.

[0055] Also, the clustering management unit 220 according to an embodiment of the present invention can perform clustering on the biological signal data by using hierarchical clustering technology. According to the hierarchical clustering technology utilized in the present invention, there is an advantage that it is not necessary to determine the number of clusters in advance, and the structural relationship between clusters can be easily grasped. And only clusters at a level (number) suitable for the analysis or interpretation situation can be generated through hierarchical clustering, so that unnecessary clustering can be prevented from being executed, and the interpretation (verification) efficiency can be further improved.

[0056] Specifically, the clustering management unit 220 according to an embodiment of the present invention can perform clustering by referring to a distance map between a plurality of feature vectors generated based on features extracted from a plurality of biological signal data. For example, the biological signal data corresponding to a certain feature vector can be clustered such that the farther the distance between certain feature vectors, the more the biological signal data corresponding to the relevant feature vectors belong to different clusters, and the closer the distance between certain feature vectors, the more the biological signal data corresponding to the relevant feature vectors belong to the same cluster.

[0057] More specifically, the clustering management unit 220 according to an embodiment of the present invention can perform clustering by further referring to a distance map between a plurality of raw data. Incidentally, the distance map between a plurality of raw data can be treated as an index representing the difference between the forms or positions of QRS complexes included in the raw data. For example, when the distance difference of a certain raw data is equal to or greater than a predetermined value (that is, when the form or position of the QRS complex is different), the clustering management unit 220 according to an embodiment of the present invention can assign a predetermined penalty to the distance map between the feature vectors corresponding to the corresponding raw data, so that the morphological similarity between the biological signal data appearing in the raw data can be reflected in the clustering.

[0058] On the other hand, the clustering management unit 220 according to an embodiment of the present invention can perform complete linkage clustering by referring to the distance map calculated as described above.

[0059] Then, the clustering management unit 220 according to an embodiment of the present invention can determine the hierarchy (number) of the clusters output as a result of clustering based on a cut-off distance threshold determined by the ratio of the total number of raw data to the number of clusters in the hierarchical cluster constructed as a result of clustering. For example, as shown in FIG. 5, the increase ratio of the inspection efficiency by clustering (total number of raw data / number of clusters) must be 20 times or more. When the total number of raw data is 3,256, the hierarchy that generates all 132 clusters can be determined as the output hierarchy, and thereby the increase ratio of the inspection efficiency can be 24.7.

[0060] In other words, when the increase ratio of the inspection efficiency by clustering is 20, instead of the medical staff (inspector) inspecting 20 electrocardiogram signal data one by one, only one cluster including the 20 electrocardiogram signal data needs to be inspected. Therefore, the inspection efficiency can be increased by about 20 times by the clustering according to the present invention.

[0061] On the other hand, FIG. 6 is a drawing exemplarily showing the degree to which the verification efficiency of the checker is increased by performing clustering according to an embodiment of the present invention. Referring to FIG. 6, it can be confirmed that the higher the number of biological signal data to be clustered (the horizontal axis in FIG. 6), the higher the increase ratio of the verification efficiency by clustering (the vertical axis in FIG. 6).

[0062] On the other hand, FIGS. 7 to 10 are drawings exemplarily showing the results of performing clustering on electrocardiogram signal data according to an embodiment of the present invention.

[0063] Referring to FIGS. 7 and 8, the results of clustering 18 electrocardiogram signal data classified as APC as one cluster (see FIG. 7) and the results of clustering 24 electrocardiogram signal data classified as APC as one cluster (see FIG. 8) can be confirmed.

[0064] Referring to FIGS. 9 and 10, the results of clustering 16 electrocardiogram signal data classified as VPC as one cluster (see FIG. 9) and the results of clustering 9 electrocardiogram signal data classified as VPC as one cluster (see FIG. 10) can be confirmed.

[0065] Next, the communication unit 230 according to an embodiment of the present invention can perform a function of enabling data transmission and reception from / to the feature extraction unit 210 and the clustering management unit 220.

[0066] Finally, the control unit 240 according to an embodiment of the present invention can perform a function of controlling the data flow among the feature extraction unit 210, the clustering management unit 220, and the communication unit 230. That is, the control unit 240 according to the present invention can control the data flow from / to the outside of the biological signal analysis system 200 or the data flow among the components of the biological signal analysis system 200, so as to control the feature extraction unit 210, the clustering management unit 220, and the communication unit 230 to perform their respective specific functions.

[0067] In the above, embodiments for analyzing electrocardiogram signal data have been mainly described. However, it should be clarified that the biological signals that can be the analysis targets according to the present invention are not necessarily limited to electrocardiogram signals only, and can be used for other types of biological signals as long as the object of the present invention can be achieved.

[0068] The embodiments according to the present invention described above can be embodied in the form of program instruction words that can be executed through various computer components and recorded on a non-transitory computer-readable recording medium. The non-transitory computer-readable recording medium can include program instruction words, data files, data structures, etc. alone or in combination. The program instruction words recorded on the non-transitory computer-readable recording medium may be those specially designed and configured for the present invention, or those known and usable to those skilled in the computer software field. Examples of the non-transitory computer-readable recording medium include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specially configured to store and execute program instruction words such as ROMs, RAMs, and flash memories. Examples of program instruction words include not only machine language codes generated by compilers but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware device can be configured to operate as one or more software modules for performing the processing according to the present invention, and vice versa.

[0069] As described above, the present invention has been described by way of specific examples and drawings such as specific components, etc., but this is only provided to assist in a more general understanding of the present invention, and the present invention is not limited to the above examples. Those with ordinary knowledge in the technical field to which the present invention pertains can attempt various modifications and variations from such descriptions.

[0070] Therefore, the idea of the present invention should not be defined as being limited to the above-described examples, and it can be said that not only the scope of the following claims but also all those equivalently or equivalently modified to this scope of claims belong to the category of the idea of the present invention.

Description of Reference Numerals

[0071] 100: Communication network 200: Biosignal analysis system 210: Feature extraction unit 220: Clustering management unit 230: Communication unit 240: Control unit 300: Device

Claims

1. A method for assisting in biological signal analysis using clustering, comprising: extracting features from a plurality of biological signal data related to biological signals, and generating a plurality of feature vectors for each of the plurality of biological signal data based on the extracted features; and performing clustering on the plurality of biological signal data with reference to the plurality of feature vectors.

2. The method according to claim 1, wherein in the feature vector generation step, the plurality of biological signal data are normalized, and features are extracted from the normalized plurality of biological signal data.

3. The method according to claim 1, wherein in the clustering performance step, clustering for a plurality of biological signal data related to biological signals corresponding to a first classification and clustering for a plurality of biological signal data related to biological signals corresponding to a first classification are performed separately.

4. The method according to claim 1, wherein in the clustering performance step, clustering for the plurality of biological signal data is performed with reference to a distance map between the feature vectors.

5. The method according to claim 4, wherein in the clustering performance step, clustering for the plurality of biological signal data is further performed with reference to a distance map between the raw data of the plurality of biological signal data.

6. The method according to claim 1, wherein in the clustering performance step, hierarchical clustering for the plurality of biological signal data is performed, and the number of clusters output as a result of the clustering is determined based on the biological signal data verification efficiency.

7. A non-transitory computer-readable recording medium recording a computer program for executing the method according to claim 1.

8. A system for assisting in biological signal analysis using clustering, comprising: a feature extraction unit that extracts features from a plurality of biological signal data related to biological signals, and generates a plurality of feature vectors for each of the plurality of biological signal data based on the extracted features; and a clustering management unit that performs clustering on the plurality of biological signal data with reference to the plurality of feature vectors.

9. The system according to claim 8, wherein the feature extraction unit normalizes the plurality of biological signal data, and extracts features from the normalized plurality of biological signal data.

10. The clustering management unit performs clustering on a plurality of biological signal data related to biological signals corresponding to the first classification and clustering on a plurality of biological signal data related to biological signals corresponding to the first classification separately. The system according to claim 8.

11. The clustering management unit performs clustering on the plurality of biological signal data by referring to the distance map between the feature vectors. The system according to claim 8.

12. The clustering management unit further performs clustering on the plurality of biological signal data by referring to the distance map between the raw data of the plurality of biological signal data. The system according to claim 11.

13. The clustering management unit performs hierarchical clustering on the plurality of biological signal data and determines the number of clusters output as a result of the clustering based on the biological signal data inspection efficiency. The system according to claim 8.

Images