Electrocardiogram reading method, program and device using artificial intelligence

Abstract

An electrocardiogram reading method, program and device using artificial intelligence are disclosed according to one embodiment of the present disclosure, the method being performed by a computing device. The method may comprise the steps of: acquiring electrocardiogram data from an electrocardiogram measurement device; performing first classification according to the readability of the acquired electrocardiogram data on the basis of an analysis of metadata related to the noise of the acquired electrocardiogram data; and, if the acquired electrocardiogram data is determined to be valid data requiring additional analysis according to the first classification, performing second classification according to the readability of the valid data on the basis of an analysis of an electrocardiogram signal included in the valid data.

Description

Method, program, and device for reading electrocardiograms using artificial intelligence

[0001] The present disclosure relates to artificial intelligence technology in the medical field, and more specifically, to a method for selecting readable electrocardiograms by considering different characteristics of each electrocardiogram measuring device, and analyzing diseases or health conditions using the selected electrocardiograms.

[0002]

[0003] Electrocardiograms (ECGs) are a crucial biosignal for diagnosing and monitoring cardiac conditions, playing a vital role in modern medicine. Recent advances in wearable devices have significantly improved the accessibility of ECG data, enabling personal health monitoring not only in clinical settings but also in everyday life. In particular, cloud-based ECG interpretation services are emerging as powerful tools that efficiently collect, process, and analyze ECG data, providing useful information to both users and healthcare professionals.

[0004] However, the diversity of hardware characteristics and noise processing methods across ECG measurement devices poses numerous challenges in providing these services to users. Due to the varying signal quality, noise removal methods, and data formats across devices, existing services struggle to properly select and analyze readable ECGs and provide consistent results regardless of device type. In particular, failure to consider the unique characteristics of each device during noise assessment and data selection can significantly reduce diagnostic accuracy and reliability. Consequently, many existing ECG analysis services are either optimized for specific devices or fail to consider the unique data collection and processing characteristics of each device. Consequently, they fail to ensure consistent service quality across device types.

[0005]

[0006] The present disclosure aims to provide an integrated system that enables consistent electrocardiogram data processing and analysis regardless of the characteristics and types of various electrocardiogram measuring devices.

[0007] However, the problems to be solved in this disclosure are not limited to the problems mentioned above, and other problems not mentioned can be clearly understood based on the description below.

[0008]

[0009] According to one embodiment of the present disclosure for achieving the aforementioned task, a method for predicting the prognosis of heart failure, performed by a computing device, is disclosed. The method may include the steps of: acquiring electrocardiogram data from an electrocardiogram measuring device; performing a first classification according to the readability of the acquired electrocardiogram data based on an analysis of metadata related to noise in the acquired electrocardiogram data; and performing a second classification according to the readability of the valid data based on an analysis of an electrocardiogram signal included in the valid data, if the acquired electrocardiogram data is determined to be valid data requiring further analysis based on the first classification.

[0010] Alternatively, metadata related to noise in the electrocardiogram data may vary depending on the type of electrocardiogram measuring device.

[0011] Alternatively, the analysis of metadata related to noise in the acquired electrocardiogram data may be to analyze the presence or absence of metadata related to the noise and the expression included in the metadata related to the noise.

[0012] Alternatively, the step of performing a first classification according to the readability of the acquired electrocardiogram data based on an analysis of metadata related to noise of the acquired electrocardiogram data may include a step of determining the acquired electrocardiogram data as invalid data with a low readability if metadata related to the noise exists and the metadata related to the noise includes an expression indicating unreadability.

[0013] Alternatively, the step of performing a first classification according to the readability of the acquired electrocardiogram data based on an analysis of metadata related to noise of the acquired electrocardiogram data may include a step of determining the acquired electrocardiogram data as data to be read or the valid data if metadata related to the noise exists and the metadata related to the noise does not include an expression indicating unreadability.

[0014] Alternatively, the step of performing a first classification according to the readability of the acquired electrocardiogram data based on an analysis of metadata related to noise of the acquired electrocardiogram data may include a step of determining the acquired electrocardiogram data as valid data if metadata related to the noise does not exist or it is impossible to confirm whether it exists.

[0015] Alternatively, if the acquired electrocardiogram data is determined as valid data requiring further analysis according to the first classification, the step of performing a second classification according to the readability of the valid data based on an analysis of the electrocardiogram signal included in the valid data may include a step of performing a second classification according to the readability of the valid data using a first neural network model that is pre-trained based on learning data in which a human reader performs labeling by judging readability based on an artificial intelligence reading result.

[0016] Alternatively, if the acquired electrocardiogram data is determined to be valid data requiring further analysis according to the first classification, the step of performing a second classification according to the readability of the valid data based on an analysis of the electrocardiogram signal included in the valid data may include a step of performing a second classification according to the readability of the valid data using a second neural network model pre-trained based on learning data that has been labeled by comparing the readability determined by a human reader with the readability determined by artificial intelligence.

[0017] Alternatively, if the acquired electrocardiogram data is determined to be valid data requiring further analysis according to the first classification, the step of performing a second classification according to the readability of the valid data based on an analysis of the electrocardiogram signal included in the valid data may include a step of performing a second classification according to the readability of the valid data using a third neural network model pre-trained to detect reversal of electrocardiogram leads based on correlations between electrocardiogram leads.

[0018] Alternatively, if the acquired electrocardiogram data is determined as valid data requiring further analysis according to the first classification, the step of performing a second classification according to the readability of the valid data based on an analysis of the electrocardiogram signal included in the valid data may include a step of performing a second classification according to the readability of the valid data using a fourth neural network model that is pre-trained to estimate whether the electrocardiogram waveform has a voltage within a normal range by comparing the magnitude of the voltage in a characteristic section of the electrocardiogram waveform.

[0019] Alternatively, if the acquired electrocardiogram data is determined as valid data requiring further analysis according to the first classification, the step of performing a second classification according to the readability of the valid data based on an analysis of the electrocardiogram signal included in the valid data may include, when the second classification is performed using a plurality of neural network models, a step of performing the second classification according to the readability of the valid data by combining outputs of the plurality of neural network models.

[0020] Alternatively, when the second classification is performed using a plurality of neural network models, the step of performing the second classification according to the readability of the valid data by combining the outputs of the plurality of neural network models may include a step of reclassifying the valid data as invalid data having a low readability if any one of the plurality of neural network models estimates that the readability of the valid data is low.

[0021] Alternatively, the method may further include a step of performing a reading on the data to be read by using a fifth neural network model pre-trained to perform a prediction on a disease or health state based on the characteristics of the electrocardiogram waveform, if the valid data is determined as the data to be read according to the second classification.

[0022] According to one embodiment of the present disclosure for achieving the above-described task, a computer program stored in a computer-readable storage medium is disclosed. When the computer program is executed on one or more processors, the computer program performs operations for predicting the prognosis of heart failure. In this case, the operations may include: an operation of acquiring electrocardiogram data from an electrocardiogram measuring device; an operation of performing a first classification according to the readability of the acquired electrocardiogram data based on an analysis of metadata related to noise of the acquired electrocardiogram data; and an operation of performing a second classification according to the readability of the valid data based on an analysis of an electrocardiogram signal included in the valid data, if the acquired electrocardiogram data is determined to be valid data requiring further analysis based on the first classification.

[0023] According to one embodiment of the present disclosure for achieving the above-described task, a computing device for predicting the prognosis of heart failure is disclosed. The device may include a processor including at least one core; a memory including program codes executable by the processor; and a network unit for acquiring electrocardiogram data from an electrocardiogram measuring device. In this case, the processor performs a first classification according to the readability of the acquired electrocardiogram data based on an analysis of metadata related to noise in the acquired electrocardiogram data, and if the acquired electrocardiogram data is determined to be valid data requiring further analysis based on the first classification, the processor may perform a second classification according to the readability of the valid data based on an analysis of an electrocardiogram signal included in the valid data.

[0024]

[0025] According to the method of the present disclosure, the quality of electrocardiogram data can be improved by performing optimized noise removal and signal refinement processes that take into account the signal processing characteristics of different electrocardiogram measuring devices. This can contribute to more accurate reading results and increased user confidence.

[0026] Additionally, by processing data collected from various devices in a single system, users can receive the same level of electrocardiogram analysis service regardless of the device used by the subject of the electrocardiogram measurement.

[0027]

[0028] FIG. 1 is a block diagram of a computing device according to one embodiment of the present disclosure.

[0029] FIG. 2 is a block diagram illustrating a first classification process according to one embodiment of the present disclosure.

[0030] FIG. 3 is a block diagram illustrating a second classification process according to one embodiment of the present disclosure.

[0031] FIG. 4 is a flowchart illustrating a method for reading an electrocardiogram using artificial intelligence according to one embodiment of the present disclosure.

[0032]

[0033] Below, embodiments of the present disclosure are described in detail with reference to the attached drawings so that those skilled in the art can easily implement the present disclosure. The embodiments presented in this disclosure are provided to enable those skilled in the art to utilize or implement the contents of the present disclosure. Accordingly, various modifications to the embodiments of the present disclosure will be apparent to those skilled in the art. That is, the present disclosure may be implemented in various different forms and is not limited to the embodiments described below.

[0034] Throughout the specification of this disclosure, identical or similar drawing numbers refer to identical or similar components. Furthermore, for the purpose of clearly describing the disclosure, drawing numbers for parts in the drawings that are not relevant to the description of the disclosure may be omitted.

[0035] The term "or" as used herein is intended to mean an inclusive "or" rather than an exclusive "or." That is, unless otherwise specified herein or clear from context, "X employs A or B" should be understood to mean either of the natural inclusive permutations. For example, unless otherwise specified herein or clear from context, "X employs A or B" can be interpreted to mean either X employs A, X employs B, or X employs both A and B.

[0036] The term "and / or" as used herein should be understood to refer to and include all possible combinations of one or more of the related concepts listed.

[0037] The terms "comprises" and / or "comprising" as used herein should be understood to mean the presence of certain features and / or components. However, it should be understood that the terms "comprises" and / or "comprising" do not exclude the presence or addition of one or more other features, other components, and / or combinations thereof.

[0038] Unless otherwise specified in this disclosure or unless the context makes it clear that the singular form is intended to be referred to, the singular should generally be construed to include “one or more.”

[0039] The term "Nth (N is a natural number)" used in the present disclosure can be understood as an expression used to mutually distinguish components of the present disclosure based on a predetermined standard such as a functional perspective, a structural perspective, or convenience of explanation. For example, components performing different functional roles in the present disclosure can be distinguished as a first component or a second component. However, components that are substantially the same within the technical spirit of the present disclosure but must be distinguished for convenience of explanation may also be distinguished as a first component or a second component.

[0040] The term "acquisition" as used in this disclosure may be understood to mean not only receiving data through a wired or wireless communication network with an external device or system, but also generating data in an on-device form.

[0041] Meanwhile, the term "module" or "unit" used in the present disclosure can be understood as a term referring to an independent functional unit that processes computing resources, such as a computer-related entity, firmware, software or a part thereof, hardware or a part thereof, or a combination of software and hardware. At this time, the "module" or "unit" may be a unit composed of a single element, or a unit expressed as a combination or set of multiple elements. For example, as a narrow concept, a "module" or "unit" may refer to a hardware element of a computing device or a set thereof, an application program that performs a specific function of software, a processing process implemented through software execution, or a set of instructions for program execution, etc. In addition, as a broad concept, a "module" or "unit" may refer to the computing device itself that constitutes the system, or an application running on the computing device, etc. However, since the above-described concept is only an example, the concept of “module” or “part” may be defined in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.

[0042] The term "model" as used herein may be understood as a system implemented using mathematical concepts and language to solve a specific problem, a set of software units to solve a specific problem, or an abstract model of a processing process to solve a specific problem. For example, a neural network "model" may refer to the entire system implemented as a neural network that has problem-solving capabilities through learning. In this case, the neural network can have problem-solving capabilities by optimizing the parameters connecting nodes or neurons through learning. A neural network "model" may include a single neural network or a set of neural networks that are a combination of multiple neural networks.

[0043] The explanation of the above terms is intended to aid understanding of the present disclosure. Therefore, unless explicitly stated as limiting the contents of the present disclosure, it should be noted that the above terms are not intended to limit the technical ideas of the contents of the present disclosure.

[0044] FIG. 1 is a block diagram of a computing device according to one embodiment of the present disclosure.

[0045] A computing device (100) according to one embodiment of the present disclosure may be a hardware device or a part of a hardware device that performs comprehensive processing and calculation of data, or may be a software-based computing environment connected to a communication network. For example, the computing device (100) may be a server that performs intensive data processing functions and shares resources, or may be a client that shares resources through interaction with a server. In addition, the computing device (100) may be a cloud system in which multiple servers and clients interact to comprehensively process data. Since the above description is only one example related to the type of computing device (100), the type of computing device (100) may be configured in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.

[0046] Referring to FIG. 1, a computing device (100) according to one embodiment of the present disclosure may include a processor (110), a memory (120), and a network unit (130). However, FIG. 1 is merely an example, and the computing device (100) may include other components for implementing a computing environment. In addition, only some of the disclosed components may be included in the computing device (100).

[0047] The processor (110) according to one embodiment of the present disclosure may be understood as a configuration unit including hardware and / or software for performing computing operations. For example, the processor (110) may read a computer program to perform data processing for machine learning. The processor (110) may process computational processes such as processing input data for machine learning, feature extraction for machine learning, and error calculation based on backpropagation. The processor (110) for performing such data processing may include a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), a tensor processing unit (TPU), an application specific integrated circuit (ASIC), or a field programmable gate array (FPGA). The above-described type of the processor (110) is only one example, and thus, the type of the processor (110) may be configured in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.

[0048] The processor (110) can evaluate the readability of electrocardiogram data by analyzing metadata included in electrocardiogram data collected from an electrocardiogram measuring device. At this time, readability can be understood as a probability indicating whether the electrocardiogram signal is a high-quality signal that allows diagnosis of a disease or judgment of a health condition through at least one of a reader or artificial intelligence. The information collected or the method of identifying and removing noise on its own may vary depending on the type of electrocardiogram measuring device. At this time, the type of electrocardiogram measuring device may vary depending on the electrocardiogram measuring method (e.g., single-lead measurement, 12-lead measurement, etc.) and may also vary depending on the manufacturer of the electrocardiogram measuring device. Therefore, the processor (110) can analyze metadata included in the electrocardiogram data to determine the readability of the electrocardiogram data by reflecting the data characteristics that vary depending on the electrocardiogram measuring device. The processor (110) can, through the analysis of metadata included in the electrocardiogram data, determine how the device that measured the electrocardiogram data identified signal quality or processed noise, etc. Furthermore, the processor (110) can determine, based on the verified information, whether the electrocardiogram data is valid data requiring further analysis or invalid data with low readability. Here, low readability can be understood as meaning that the electrocardiogram signal itself is unreadable, or even if readability is performed, the results are expected to have low accuracy or reliability.

[0049] If the electrocardiogram data is determined to be valid data requiring further analysis, the processor (110) can further evaluate the readability of the electrocardiogram data by analyzing the electrocardiogram signal included in the electrocardiogram data. After completing the first filtering of the electrocardiogram signal quality through metadata analysis that takes into account the characteristics of each type of electrocardiogram measuring device, the processor (110) can perform a second filtering of the signal quality by additionally performing analysis of the electrocardiogram signal itself. For example, the processor (110) can estimate the readability of the valid data by evaluating the noise level of the signal included in the valid data as a continuous variable through a pre-trained neural network model to analyze the quality of the electrocardiogram signal. The processor (110) can estimate the readability of the valid data by analyzing whether the valid data includes a reversed lead signal through a pre-trained neural network model to detect whether the electrocardiogram lead is reversed. The processor (110) can estimate the readability of the valid data by determining whether the voltage level of the signal included in the valid data is within the normal range through a pre-trained neural network model that analyzes the voltage level of the electrocardiogram signal. At this time, the normal range of the voltage level may be a value predetermined according to clinical criteria or a value set according to a user command. The processor (110) can perform at least one of the plurality of estimation methods for the readability described above. When performing two or more estimation methods, the processor (110) can derive a final result by combining the estimation results for the readability of each method. At this time, the combination can be implemented by an average operation for the estimation results, a weighted sum operation (e.g., weighting and summing the outputs according to the importance set based on the frequency of use of each of the plurality of models, etc.). Through these two filtering processes, the processor (110) can accurately select readable electrocardiogram data and improve the quality of the reading service.

[0050] The memory (120) according to one embodiment of the present disclosure may be understood as a configuration unit including hardware and / or software for storing and managing data processed in the computing device (100). That is, the memory (120) may store any type of data generated or determined by the processor (110) and any type of data received by the network unit (130). For example, the memory (120) may include at least one type of storage medium among a flash memory type, a hard disk type, a multimedia card micro type, a card type memory, a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, and an optical disk. In addition, the memory (120) may also include a database system that controls and manages data in a predetermined system. The type of memory (120) described above is only one example, and thus the type of memory (120) can be configured in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.

[0051] The memory (120) can structure and organize and manage data, combinations of data, and program codes executable by the processor (110) required for the processor (110) to perform operations. For example, the memory (120) can store medical data received through the network unit (130) described below. The memory (120) can store program codes that operate a machine learning model to receive medical data as input and perform learning, program codes that operate a machine learning model to receive medical data as input and perform inference according to the purpose of use of the computing device (100), and processed data generated as the program codes are executed.

[0052] The network unit (130) according to one embodiment of the present disclosure may be understood as a component that transmits and receives data through any type of known wired or wireless communication system. For example, the network unit (130) may perform data transmission and reception using a wired or wireless communication system such as a local area network (LAN), wideband code division multiple access (WCDMA), long term evolution (LTE), wireless broadband internet (WiBro), fifth generation mobile communication (5G), ultra wide-band, ZigBee, radio frequency (RF) communication, wireless LAN, wireless fidelity, near field communication (NFC), or Bluetooth. Since the above-described communication systems are only examples, the wired and wireless communication system for data transmission and reception of the network unit (130) may be applied in various ways other than the above-described examples.

[0053] The network unit (130) can receive data necessary for the processor (110) to perform calculations through wired or wireless communication with any system or any client, etc. In addition, the network unit (130) can transmit data generated through calculations of the processor (110) through wired or wireless communication with any system or any client, etc. For example, the network unit (130) can receive medical data through communication with a cloud server that performs tasks such as databases in a hospital environment, standardization of medical data, clients such as smart watches, or medical computing devices, etc. The network unit (130) can transmit output data of a machine learning model, intermediate data derived from the calculation process of the processor (110), processed data, etc. through communication with the aforementioned database, server, client, or computing device, etc.

[0054] FIG. 2 is a block diagram illustrating a first classification process according to an embodiment of the present disclosure. FIG. 3 is a block diagram illustrating a second classification process according to an embodiment of the present disclosure.

[0055] A computing device (100) according to one embodiment of the present disclosure can obtain electrocardiogram data from an electrocardiogram measuring device. Regardless of the type of the electrocardiogram measuring device, the computing device (100) can obtain electrocardiogram data collected and processed by the electrocardiogram measuring device through communication with the electrocardiogram measuring device. For example, referring to FIG. 2, the computing device (100) can obtain electrocardiogram data A (11), electrocardiogram data B (15), and electrocardiogram data C (19) by performing communication with electrocardiogram measuring device A (200), electrocardiogram measuring device B (250), and electrocardiogram measuring device C (290), respectively. At this time, the electrocardiogram measuring device A (200), the electrocardiogram measuring device B (250), and the electrocardiogram measuring device C (290) can be distinguished depending on the manufacturer. And, electrocardiogram data A (11), electrocardiogram data B (15), and electrocardiogram data C (19) may be data in XML format including electrocardiogram signals and metadata.

[0056] The computing device (100) can perform a first classification according to the readability of the electrocardiogram data based on an analysis of metadata related to noise in the electrocardiogram data acquired from the electrocardiogram measuring device. The computing device (100) can analyze metadata related to noise among the metadata included in the electrocardiogram data and classify the electrocardiogram data into valid data requiring further analysis, invalid data with low readability, or data to be read, depending on the readability of the electrocardiogram data. In this case, the metadata related to noise can be understood as metadata including information about noise in the electrocardiogram signal generated by the electrocardiogram measuring device.

[0057] Specifically, the computing device (100) can analyze the presence or absence of metadata related to noise (S110). Each electrocardiogram measuring device may or may not be able to identify noise on its own. Furthermore, even if metadata related to noise exists in each electrocardiogram measuring device, there may be cases where the information contained in the metadata cannot be confirmed for various reasons. Therefore, the computing device (100) can determine whether noise-related metadata generated by the electrocardiogram measuring device exists and whether it can be confirmed. If noise-related metadata does not exist in the electrocardiogram data or cannot be confirmed, quality evaluation must be performed through analysis of the electrocardiogram signal, and therefore, the computing device (100) can first classify the electrocardiogram data as valid data (20) requiring further analysis. If noise-related metadata exists in the electrocardiogram data in a identifiable state, the computing device (100) can perform analysis on the expression contained in the noise-related metadata.

[0058] The computing device (100) can analyze noise-related expressions included in noise-related metadata. Specifically, the computing device (100) can analyze whether the noise-related metadata includes an expression indicating unreadability (S120). At this time, the expression indicating unreadability can be understood as an expression including the concept that the result of the reading of a disease or health condition is likely to have low accuracy or reliability. When the electrocardiogram measuring device itself generates noise information through noise analysis, an expression indicating whether the signal quality is at a readable level can be included in the metadata. The computing device (100) can identify the expression and determine whether the electrocardiogram measuring device itself has determined that the electrocardiogram data is unreadable, thereby assessing the readability of the electrocardiogram data. If the noise-related metadata includes an expression indicating unreadability, the computing device (100) can classify the electrocardiogram data as invalid data (30) with a low readability. That is, if the electrocardiogram measuring device determines that the data is unreadable, the computing device (100) may also classify the data as invalid data (30) and decide not to use it for reading. If the noise-related metadata does not include an expression indicating unreadability, the computing device (100) may classify the electrocardiogram data as valid data (20) requiring additional analysis or data to be read (50). Since the electrocardiogram measuring device determines that the electrocardiogram data is readable, it may be classified as data to be read (50) without additional analysis. However, for more accurate reading, additional analysis may be required even if the electrocardiogram measuring device determines that the data is readable.Accordingly, if there is metadata related to noise in the electrocardiogram data and the metadata related to the noise does not include an expression indicating that the data is unreadable, the computing device (100) can selectively classify the electrocardiogram data into valid data (20) or data to be read (50) according to a user command. For example, if metadata included in the electrocardiogram data in XML format includes metadata related to expressions such as “poor recording,” “noise,” “missing,” “inconclusive,” etc., the computing device (100) can identify the above-described expressions and the metadata associated therewith and classify the data as invalid data (30). If metadata included in the electrocardiogram data in XML format does not include metadata related to the above-described expressions, the computing device (100) can classify the data as data to be read (50) by default. At this time, if it is preset to be classified as valid data (20) by a user command, the computing device (100) can classify the data as valid data (20).

[0059] The computing device (100) can primarily select readable data without having to evaluate the quality of signals contained in all electrocardiogram data collected from the electrocardiogram measuring device through this first classification. In other words, the computing device (100) can efficiently optimize the computing resources required for selecting data for reading through the first classification.

[0060] If the electrocardiogram data acquired from the electrocardiogram measuring device is determined as valid data (20) requiring further analysis according to the first classification, the computing device (100) can perform a second classification according to the readability of the valid data (20) based on the analysis of the electrocardiogram signal included in the valid data (20). The analysis of the electrocardiogram signal included in the valid data (20) can be understood as a task of analyzing the quality of the signal based on the correlation of the features existing in the electrocardiogram signal or the signals for each lead using a pre-learned neural network model. For example, referring to FIG. 3, the computing device (100) can input the valid data (20) into the pre-learned neural network model (300) and generate output data regarding the readability of the valid data (20). And, the computing device (100) can reclassify valid data (20) into invalid data (30) with low readability based on the output data of the neural network model (300), or classify it into read target data (50) to be used for reading.

[0061] Specifically, the computing device (100) inputs valid data (20) into a first neural network model (310) that is pre-trained based on learning data that a human reader has labeled and judged to have readability based on the artificial intelligence reading result, thereby generating first output data (41) regarding the readability of the valid data (20). When there are electrocardiograms that are judged to be readable and electrocardiograms that are judged to be unreadable when judged by the human reader with reference to the artificial intelligence reading result, labeling of the electrocardiograms may be performed according to each judgment, thereby providing learning data for the first neural network model (310). The first neural network model (310) may perform learning based on the learning data provided in this manner. When the training of the first neural network model (310) is completed, the computing device (100) can input valid data (20) into the trained first neural network model (310) and generate a probability value representing the readability of the valid data (20) as first output data (41).

[0062] The computing device (100) inputs valid data (20) into a second neural network model (330) that has been pre-trained based on training data that has been labeled by comparing the readability determined by a human reader with the readability determined by artificial intelligence, thereby generating second output data (43) regarding the readability of the valid data (20). If the first neural network model (310) was trained based on labeled training data that determined readability from a human perspective, the second neural network model (330) is different in that it was trained based on labeled training data that determined readability from an artificial intelligence perspective. For example, if data that a human reader determined to be unreadable due to noise is readable by artificial intelligence, the data can be labeled as readable data and included in the training data of the second neural network model (330). In other words, the second neural network model (330) can be trained through labeled data with a different perspective on readability than the first neural network model (310). Therefore, the first neural network model (310) and the second neural network model (330) can be used complementarily. When the training of the second neural network model (330) is completed, the computing device (100) can input valid data (20) into the trained second neural network model (330) and generate a probability value representing the readability of the valid data (20) as second output data (43).

[0063] The computing device (100) can input valid data (20) into a third neural network model (350) that is pre-trained to detect inversion of an electrocardiogram lead based on correlations between electrocardiogram leads, and generate third output data (45) regarding the readability of the valid data (20). Inversion of an electrocardiogram lead can occur due to a sensor being placed in an incorrect position during electrocardiogram measurement, and can cause a problem in that an unreadable signal is measured, or even if a readable signal is possible, the possibility of a misreading increases. In other words, since inversion of an electrocardiogram lead causes errors in electrocardiogram reading, it is desirable to classify a signal in which a lead inversion has occurred as data with a low readability. For this processing, the third neural network model (350) can be trained to derive correlations that exist between electrocardiogram leads and output the probability that an inversion will occur in a specific lead based on the derived correlations. At this time, the learning can be performed in a supervised learning manner using learning data that is labeled with whether or not a lead is inverted. When the learning of the third neural network model (350) is completed, the computing device (100) inputs valid data (20) into the learned third neural network model (330) and generates a probability value of the presence of a signal in which a lead inversion has occurred in the valid data (20) as third output data (45).

[0064] The computing device (100) inputs valid data (20) into a fourth neural network model (370) that has been pre-trained to estimate whether the electrocardiogram waveform has a voltage within a normal range by comparing the voltage magnitude in the characteristic section of the electrocardiogram waveform, thereby generating fourth output data (47) regarding the readability of the valid data (20). If the voltage magnitude of the electrocardiogram signal is small, it is difficult to accurately analyze the electrocardiogram. Therefore, in order to enable measures such as amplifying the electrocardiogram signal with a small voltage magnitude or inducing filming in another way, it is necessary to classify signals with a voltage magnitude smaller than the normal range as data with a low readability. For this processing, the fourth neural network model (370) can be trained to detect a characteristic section such as QRS in the electrocardiogram waveform, calculate the difference between the point where the voltage is highest and the point where the voltage is lowest within the detected characteristic section, and determine whether the calculated difference falls within the normal range. At this time, the normal range may be a voltage level that is clinically judged to be capable of normal reading, or may be a value set by the user's request. In addition, the learning of the fourth neural network model (370) may be performed through not only supervised learning but also unsupervised learning, self-supervised learning, etc. When the learning of the fourth neural network model (370) is completed, the computing device (100) may input the valid data (20) into the learned fourth neural network model (370) to generate the fourth output data (47) including information on whether the signal included in the valid data (20) is a low-voltage signal that is outside the normal range.

[0065] When multiple output data (41, 43, 45, 47) are generated through multiple neural network models (310, 330, 350, 370) as shown in FIG. 3, the computing device (100) can perform a second classification on valid data by combining the result values ​​included in the multiple output data (41, 43, 45, 47). When any one of the multiple output data (41, 43, 45, 47) includes a result value corresponding to a judgment that the readability of the valid data is low, the computing device (100) can reclassify the valid data (20) into invalid data (30) with a low readability. For example, if a probability value indicating the readability of valid data (20) included in the first output data (41) is less than a preset reference value, while a result value included in other output data (43, 45, 47) corresponds to a judgment that the readability is high, the computing device (100) can reclassify the valid data (20) as invalid data (30) with a low readability. Conversely, if all of the plurality of output data (41, 43, 45, 47) include result values ​​corresponding to a judgment that the readability of valid data is high, the computing device (100) can classify the valid data (20) as data to be read (50).

[0066] Meanwhile, as shown in FIG. 3, the second classification can be performed using all of the plurality of neural network models (310, 330, 350, 370), but the second classification can also be performed using some of the plurality of neural network models (310, 330, 350, 370). That is, depending on the user's needs, the readability of valid data (20) can be determined using only one of the plurality of neural network models (310, 330, 350, 370), or the readability can be determined using all or some of them. This selective operation has the advantage of enabling data selection according to the user's needs. In addition, since this selective operation does not necessarily require the use of all models to select the data to be read (50), it can contribute to the efficient use of computing resources.

[0067] Through this second classification, the computing device (100) can accurately select high-quality data to be used for reading, thereby enhancing the quality and reliability of the reading service. Furthermore, the computing device (100) can process data collected from various devices into a single, integrated system, regardless of device type, thereby providing users with an efficient reading service environment.

[0068] If the valid data (20) is determined as the data to be read (50) according to the second classification, the computing device (100) can perform reading on the data to be read (50) by using the fifth neural network model that has been pre-trained to perform prediction on a disease or health state based on the characteristics of the electrocardiogram waveform. For example, the computing device (100) can input the data to be read (50) into the fifth neural network model that has been pre-trained to analyze the possibility of the presence of structural heart disease based on the morphological characteristics of the electrocardiogram waveform, and generate a probability value regarding the possibility of the presence of structural heart disease as a reading result. At this time, the fifth neural network model may be a model that has been trained based on training data composed of electrocardiogram signals in which the presence or absence of structural heart disease is labeled. Meanwhile, the reading task of the fifth neural network model is not limited to the structural heart disease as an example described above, and can be expanded to various diagnostic areas that can be read using an electrocardiogram.

[0069] FIG. 4 is a flowchart illustrating a method for reading an electrocardiogram using artificial intelligence according to one embodiment of the present disclosure.

[0070] Referring to FIG. 4, a computing device (100) according to one embodiment of the present disclosure can obtain electrocardiogram data from an electrocardiogram measuring device (S210). For example, the computing device (100) can receive electrocardiogram data generated by the electrocardiogram measuring device through communication with the electrocardiogram measuring device. The electrocardiogram data generated by the electrocardiogram measuring device can include not only the electrocardiogram signal but also various metadata generated by the electrocardiogram measuring device itself.

[0071] The computing device (100) may perform a first classification according to the readability of the electrocardiogram data based on the analysis of the metadata related to noise in the electrocardiogram data acquired through step S210. At this time, the analysis of the metadata related to noise in the electrocardiogram data may be understood as analyzing the presence or absence of metadata related to noise and the expressions included in the metadata related to noise. For example, if metadata related to noise exists and includes an expression indicating unreadability, the computing device (100) may determine the electrocardiogram data as invalid data with a low readability. If metadata related to noise exists and does not include an expression indicating unreadability, the computing device (100) may determine the electrocardiogram data as data to be read or valid data requiring further analysis. If metadata related to noise does not exist or its presence cannot be confirmed, the computing device (100) may determine the electrocardiogram data as valid data requiring further analysis. Since the metadata related to the noise of the electrocardiogram data may vary depending on the type of electrocardiogram measuring device, the computing device (100) can quickly classify the data by determining the readability without directly evaluating the quality of the electrocardiogram signal through analysis of the metadata such as the example described above.

[0072] If the electrocardiogram data acquired through step S210 is determined as valid data requiring further analysis according to the first classification, the computing device (100) may perform a second classification according to the readability of the valid data based on the analysis of the electrocardiogram signal included in the valid data (S230). The computing device (100) may perform the second classification according to the readability of the valid data using a first neural network model that has been pre-trained based on training data that has been labeled by a human reader judging the readability based on the artificial intelligence reading result. The computing device (100) may perform the second classification according to the readability of the valid data using a second neural network model that has been pre-trained based on training data that has been labeled by comparing the readability judged by a human reader with the readability judged by artificial intelligence. The computing device (100) may perform the second classification according to the readability of the valid data using a third neural network model that has been pre-trained to detect the inversion of the electrocardiogram leads based on the correlation between the electrocardiogram leads. The computing device (100) can perform a second classification according to the readability of valid data by using a pre-trained fourth neural network model to estimate whether the electrocardiogram waveform has a voltage within a normal range by comparing the magnitude of the voltage in the characteristic section of the electrocardiogram waveform. The computing device (100) can perform the second classification by combining outputs generated using all or some of the four models described above, or can perform the second classification based on outputs generated using only one of the four models described above. For example, if any one of the plurality of neural network models estimates that the readability of the valid data is low, the computing device (100) can reclassify the valid data as invalid data with a low readability.If all of the multiple neural network models estimate that the valid data requires further analysis, the computing device (100) can classify the valid data as data to be read.

[0073] The various embodiments of the present disclosure described above can be combined with additional embodiments and modified within the scope understood by those skilled in the art in light of the detailed description above. It should be understood that the embodiments of the present disclosure are illustrative in all respects and not restrictive. For example, each component described as a single component may be implemented in a distributed manner, and likewise, components described as distributed may be implemented in a combined manner. Accordingly, all changes or modifications derived from the meaning, scope, and equivalent concepts of the claims of the present disclosure should be construed as being included within the scope of the present disclosure.

Claims

1. A method for reading an electrocardiogram using artificial intelligence, performed by a computing device including at least one processor, A step of acquiring electrocardiogram data from an electrocardiogram measuring device; A step of performing a first classification according to the readability of the acquired electrocardiogram data based on an analysis of metadata related to noise of the acquired electrocardiogram data; and If the acquired electrocardiogram data is determined as valid data requiring further analysis according to the first classification, a step of performing a second classification according to the readability of the valid data based on an analysis of the electrocardiogram signal included in the valid data; Including, method.

2. In paragraph 1, Metadata related to the noise in the above ECG data is: Depending on the type of electrocardiogram measuring device, method.

3. In paragraph 1, Analysis of metadata related to noise in the acquired electrocardiogram data is as follows: Analyzing the presence or absence of metadata related to the above noise and the expression included in the metadata related to the above noise. method.

4. In paragraph 3, Based on the analysis of the metadata related to the noise of the acquired electrocardiogram data, the step of performing the first classification according to the readability of the acquired electrocardiogram data is as follows. A step of determining the acquired electrocardiogram data as invalid data with a low possibility of being readable, when metadata related to the noise exists and the metadata related to the noise includes an expression indicating unreadability; Including, method.

5. In paragraph 3, Based on the analysis of the metadata related to the noise of the acquired electrocardiogram data, the step of performing the first classification according to the readability of the acquired electrocardiogram data is as follows. A step of determining the acquired electrocardiogram data as data to be read or the valid data, if metadata related to the noise exists and the metadata related to the noise does not include an expression indicating unreadability; Including, method.

6. In paragraph 3, Based on the analysis of the metadata related to the noise of the acquired electrocardiogram data, the step of performing the first classification according to the readability of the acquired electrocardiogram data is as follows. A step of determining the acquired electrocardiogram data as the valid data when the metadata related to the noise does not exist or it is impossible to confirm whether the metadata exists; Including, method.

7. In paragraph 1, If the acquired electrocardiogram data is determined as valid data requiring further analysis according to the first classification, the step of performing a second classification according to the readability of the valid data based on the analysis of the electrocardiogram signal included in the valid data is as follows. A step of performing a second classification according to the readability of the valid data using a first neural network model pre-trained on learning data in which a human reader performs labeling to determine readability based on the artificial intelligence reading results; Including, method.

8. In paragraph 1, If the acquired electrocardiogram data is determined as valid data requiring further analysis according to the first classification, the step of performing a second classification according to the readability of the valid data based on the analysis of the electrocardiogram signal included in the valid data is as follows. A step of performing a second classification according to the readability of the valid data using a pre-trained second neural network model based on training data that has been labeled by comparing the readability judged by a human reader with the readability judged by artificial intelligence; Including, method.

9. In paragraph 1, If the acquired electrocardiogram data is determined as valid data requiring further analysis according to the first classification, the step of performing a second classification according to the readability of the valid data based on the analysis of the electrocardiogram signal included in the valid data is as follows. A step of performing a second classification according to the readability of the valid data using a third neural network model pre-trained to detect reversal of the electrocardiogram leads based on correlations between the electrocardiogram leads; Including, method.

10. In paragraph 1, If the acquired electrocardiogram data is determined as valid data requiring further analysis according to the first classification, the step of performing a second classification according to the readability of the valid data based on the analysis of the electrocardiogram signal included in the valid data is as follows. A step of performing a second classification according to the readability of the valid data by using a pre-trained fourth neural network model to estimate whether the electrocardiogram waveform has a voltage within a normal range by comparing the voltage magnitude in a characteristic section of the electrocardiogram waveform; Including, method.

11. In paragraph 1, If the acquired electrocardiogram data is determined as valid data requiring further analysis according to the first classification, the step of performing a second classification according to the readability of the valid data based on the analysis of the electrocardiogram signal included in the valid data is as follows. When the second classification is performed using a plurality of neural network models, a step of performing the second classification according to the readability of the valid data by combining the outputs of the plurality of neural network models; Including, method.

12. In paragraph 11, In the case where the above second classification is performed using a plurality of neural network models, the step of performing the second classification according to the readability of the valid data by combining the outputs of the plurality of neural network models is: A step of reclassifying the valid data into invalid data with low readability if any one of the plurality of neural network models estimates that the valid data has low readability; Including, method.

13. In paragraph 1, When the valid data is determined as the target data for reading according to the second classification, a step of performing reading on the target data for reading by using a fifth neural network model that has been pre-learned to perform prediction on a disease or health state based on the characteristics of the electrocardiogram waveform; Including more, method.

14. A computer program stored in a computer-readable storage medium, wherein the computer program, when executed on one or more processors, performs operations for reading an electrocardiogram using artificial intelligence. The above actions are, An act of acquiring electrocardiogram data from an electrocardiogram measuring device; An operation of performing a first classification according to the readability of the acquired electrocardiogram data based on an analysis of metadata related to noise of the acquired electrocardiogram data; and If the acquired electrocardiogram data is determined as valid data requiring further analysis according to the first classification, an operation of performing a second classification according to the readability of the valid data based on an analysis of the electrocardiogram signal included in the valid data; Including Computer program.

15. A computing device for reading electrocardiograms using artificial intelligence, A processor comprising at least one core; a memory containing program codes executable by the processor; and A network unit that acquires electrocardiogram data from an electrocardiogram measuring device; Including, The above processor, Based on the analysis of the metadata related to the noise of the acquired electrocardiogram data, a first classification is performed according to the readability of the acquired electrocardiogram data, If the acquired electrocardiogram data is determined as valid data requiring further analysis according to the first classification above, a second classification is performed based on the analysis of the electrocardiogram signal included in the valid data according to the readability of the valid data. device.

Images