Method for classifying arrhythmias from 12-lead electrocardiogram signals and apparatus for doing so.

Abstract

To provide an irregular pulse classification method of a 12-lead electrocardiogram signal and a device for performing the method.SOLUTION: A device for performing irregular pulse classification of a 12-lead electrocardiogram signal according to one embodiment of the present invention is a computing device provided with one or more processors and a memory for storing one or more programs executed by the one or more processors. The device includes a data input module which receives input of the 12-lead electrocardiogram signal and a classification module which outputs a classification result about the 12-lead electrocardiogram signal by using the technique of a machine learning base based on the input 12-lead electrocardiogram signal.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] Embodiments of the present invention relate to a technique for classifying arrhythmias from electrocardiogram signals.

Background Art

[0002] According to the World Health Organization (WHO), millions of people worldwide die annually due to heart diseases. Electrocardiogram (ECG) examination, known as a method for detecting the presence of heart diseases, is a non-invasive and most well-known examination method that attaches electrodes to the skin to sense and record the electrical activity state that appears in the heart during the cardiac cycle. Generally, 12 electrocardiogram signals are acquired and analyzed using 10 electrodes attached to the arms, legs, and chest to determine the presence of heart diseases such as arrhythmias.

[0003] Conventionally, doctors have determined heart diseases through the records of electrocardiogram signals recorded for each patient. Recently, attempts to use deep learning for the analysis and diagnosis of electrocardiogram signals have increased.

[0004] Conventional deep learning techniques for the analysis and diagnosis of electrocardiogram signals extracted time-frequency features from single-lead electrocardiogram signals to classify arrhythmias. However, some arrhythmias are observed only in specific electrocardiogram channels, so the classification accuracy may decrease depending on the type of arrhythmia to be classified.

[0005] Therefore, for an accurate diagnosis of arrhythmias, it is necessary to comprehensively and carefully check the electrocardiogram signals obtained from 12 channels. That is, since the electrocardiogram signals measured through 12 leads differ depending on the type of arrhythmia, it is necessary to comprehensively analyze the electrocardiogram signals obtained from 12 channels in order to accurately determine the patient's arrhythmia.

Prior Art Documents

Patent Documents

[0006] Korean Registered Patent Publication No. 10-2163217 (September 29, 2020) [Overview of the project] [Problems that the invention aims to solve]

[0007] Embodiments of the present invention provide a method for classifying arrhythmias from standard 12-lead electrocardiogram signals using machine learning techniques. [Means for solving the problem]

[0008] According to an exemplary embodiment of the present invention, a computing device is provided comprising one or more processors and a memory for storing one or more programs executed by the one or more processors, the computing device comprising a data input module for receiving input of a 12-lead electrocardiogram signal and a classification module for outputting a classification result for the 12-lead electrocardiogram signal using machine learning-based techniques based on the input 12-lead electrocardiogram signal.

[0009] The classification module may include an artificial neural network model that receives the input of the 12-lead electrocardiogram signal and is trained to classify arrhythmia information based on the 12-lead electrocardiogram signal.

[0010] The artificial neural network model may include: an initial feature block that outputs an initial feature map through a convolution operation based on the input 12-lead electrocardiogram signal; an attention block that outputs a concentrated feature map through element-wise weight calculations based on the initial feature map output from the initial feature block; a residual block that outputs a deep feature map through a shortcut operation based on the initial feature map output from the initial feature block; a combination block that combines the concentrated feature map output from the attention block and the deep feature map output from the residual block to output a final feature map; and a classification block that classifies the type of arrhythmia based on the final feature map output from the combination block.

[0011] The aforementioned attention block can perform maximum pooling and average pooling in parallel on the initial feature map output from the initial feature block through a maximum pooling layer and an average pooling layer, combine the pooling results output through the maximum pooling layer and the average pooling layer through a first join layer to output a weight feature map, and combine the initial feature map and the weight feature map through a second join layer to output a condensed feature map.

[0012] The residual block is configured to include N short residual blocks (where N is a natural number greater than or equal to 2) sequentially linked to reflect the features of the initial feature map output from the initial feature block, and the N short residual blocks can be input with a previous feature map (the feature map output from the N-1th short residual block), output a new feature map from the previous feature map through a convolutional layer, and output a combined version of the previous feature map and the new feature map through a third join layer.

[0013] The first and third join layers can use element-wise sum, while the second join layer can use element-wise multiplication.

[0014] The classification block can perform global max pooling and global average pooling in parallel on the final feature map through a global max pooling layer and a global average pooling layer, and can concatenate the pooling results output through the global max pooling layer and the global average pooling layer through a concatenation layer.

[0015] According to another exemplary embodiment of the present invention, there is provided a method performed by a computing device including one or more processors and a memory storing one or more programs executed by the one or more processors, the method comprising: receiving an input of a 12-lead electrocardiogram signal; and outputting a classification result for the 12-lead electrocardiogram signal using a machine learning-based technique based on the input 12-lead electrocardiogram signal. A method for classifying arrhythmias in a 12-lead electrocardiogram signal is provided.

Advantages of the Invention

[0016] According to an embodiment of the present invention, by classifying arrhythmias from a standard 12-lead electrocardiogram signal using machine learning techniques, there is an effect that can quickly and accurately classify from an electrocardiogram signal.

Brief Description of the Drawings

[0017] [Figure 1] It is a diagram showing the configuration of an arrhythmia classification device for a 12-lead electrocardiogram signal according to an embodiment of the present invention.

[0018] [Figure 2] It is a diagram schematically showing the structure of a classification module constituting an arrhythmia classification device for a 12-lead electrocardiogram signal according to an embodiment of the present invention.

[0019] [Figure 3] It is a diagram showing a 12-lead electrocardiogram signal according to an embodiment of the present invention.

[0020] [Figure 4] It is a diagram for explaining the caution block 320 of an arrhythmia classification device for a 12-lead electrocardiogram signal according to an embodiment of the present invention.

[0021] [Figure 5] It is a diagram for explaining the residual block 330 of an arrhythmia classification device for a 12-lead electrocardiogram signal according to an embodiment of the present invention.

[0022] [Figure 6] FIG. is a diagram showing a layer structure of an artificial neural network model in a classification module of an arrhythmia classification device for 12-lead electrocardiogram signals according to an embodiment of the present invention.

[0023] [Figure 7] FIG. is a flowchart for explaining an arrhythmia classification method for 12-lead electrocardiogram signals according to an embodiment of the present invention.

[0024] [Figure 8] FIG. is a block diagram for exemplarily explaining a computing environment including a computing device suitable for use in an exemplary embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0025] Hereinafter, specific embodiments of the present invention will be described with reference to the drawings. The following detailed description is provided to assist in a comprehensive understanding of the methods, apparatuses, and / or systems described herein. However, this is merely an example and the present invention is not limited thereto.

[0026] In describing embodiments of the present invention, if a specific description of known technology related to the present invention is deemed to unnecessarily obscure the gist of the invention, such detailed description will be omitted. Furthermore, terms used hereafter are defined in consideration of the function of the present invention, and these may vary depending on the intent or convention of the user or operator. Therefore, their definitions should be based on the overall content of this specification. Terms used in the detailed description are solely for describing embodiments of the present invention and should not be restrictive. Unless explicitly used otherwise, singular expressions imply the plural meaning. In this description, expressions such as “implication” or “presence” are intended to indicate a certain characteristic, number, step, action, element, part or combination thereof, and should not be interpreted in a way that excludes the existence or possibility of one or more other characteristics, numbers, steps, actions, elements, part or combination thereof other than those described.

[0027] In the following description, similar terms such as “transfer,” “communicate,” “transmit,” and “receive” of signals or information include not only direct transmission of signals or information from one component to another, but also transmission via other components. In particular, “transferring” or “transmitting” a signal or information to one component indicates the final destination of that signal or information, not the direct destination. The same is true for “receiving” a signal or information. Furthermore, in this specification, two or more pieces of data or information being “related” means that by obtaining one piece of data (or information), it is possible to obtain at least a portion of other pieces of data (or information) based on it.

[0028] On the other hand, embodiments of the present invention may include a program for performing the methods described herein on a computer, and a computer-readable recording medium containing the program. The computer-readable recording medium may include program instructions, local data files, local data structures, etc., individually or in combination. The medium may be specially designed and configured for the present invention, or may be one that is commonly available in the field of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs and DVDs, and hardware devices such as ROMs, RAMs, and flash memory that are specially configured to store and execute program instructions. Examples of the program may include not only machine code, such as that produced by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like.

[0029] Figure 1 shows the configuration of an arrhythmia classification device for 12-lead electrocardiogram signals according to one embodiment of the present invention, and Figure 2 shows a schematic diagram of the structure of a classification module that constitutes an arrhythmia classification device for 12-lead electrocardiogram signals according to one embodiment of the present invention.

[0030] Referring to Figures 1 and 2, the arrhythmia classification device 100 may include a data input module 200 and a classification module 300. The arrhythmia classification device 100 can classify the type of arrhythmia from a 12-lead electrocardiogram signal using machine learning techniques. On the other hand, the neural network constituting the arrhythmia classification device 100 according to one embodiment of the present invention may use a one-dimensional convolutional neural network (1D Convolutional Neural Network) to process the 12-lead electrocardiogram signal.

[0031] In one embodiment, the data input module 200 and the classification module 300 may be implemented using one or more physically distinct devices, or by one or more processors or a combination of one or more processors and software, and may not be clearly distinguished in specific operation, unlike the illustrated example.

[0032] The data input module 200 can receive input of a 12-lead electrocardiogram (ECG) signal. For example, the data input module 200 can acquire an ECG signal measured by medical staff. The data input module 200 can also acquire an ECG signal from an external medical institution or research institution. Here, the ECG signal can be used as training data for the arrhythmia classification device 100. When the ECG signal is used as training data, arrhythmia information can be labeled. The arrhythmia information may be the type of arrhythmia diagnosed by the ECG signal.

[0033] The data input module 200 can provide the input 12-lead electrocardiogram signal to the classification module 300.

[0034] On the other hand, the 12-lead electrocardiogram signal can be acquired using 10 skin surface sensors, including four limb leads (right arm (RA), left arm (LA), right bundle branch (RL), and left bundle branch (LL)) and six chest leads (V1, V2, V3, V4, V5, and V6), as shown in Figure 3. The 12-lead electrocardiogram signal has the characteristic that the signal morphology in each lead differs depending on the type of arrhythmia, and abnormal signs are observed only in specific leads among the 12 leads. Accordingly, the arrhythmia classification device 100 of the 12-lead electrocardiogram signal according to one embodiment of the present invention can classify the type of arrhythmia for the input 12-lead electrocardiogram signal by learning the correlation between the 12-lead electrocardiogram signal and the type of arrhythmia using machine learning technology.

[0035] The classification module 300 can classify arrhythmia information related to an input 12-lead electrocardiogram (ECG) signal using machine learning-based techniques. For example, the classification module 300 may include an artificial neural network model that receives an ECG signal input from the data input module 200 and is trained to classify arrhythmia information based on the input ECG signal. The artificial neural network model can be trained to classify and output arrhythmia information for a given ECG signal.

[0036] In an exemplary embodiment, the classification module 300 may include an Initial Feature Block 310, an Attention Block 320, a Residual Block 330, a Sum Block 340, and a Classification Block 350.

[0037] The Initial Feature Block 310 can output an initial feature map through convolution operations based on the input 12-lead electrocardiogram signal.

[0038] In an exemplary embodiment, the Initial Feature Block 310 may include a Convolution Layer, an Activation Layer (PReLU), and an Average Pooling Layer. That is, the Initial Feature Block 310 can output an initial feature map from a 12-lead electrocardiogram signal through the Convolution Layer, Activation Layer, and Average Pooling Layer. Here, the Activation Layer can use the PReLU (Parametric ReLU) function as its activation function. An activation function is necessary to readjust the signal intensity of a neuron, and the PReLU function outputs values ​​less than 0 multiplied by a parameter (a) which is adjusted through learning, and outputs values ​​greater than 0 as the input value.

[0039] The attention block 320 can output a focus feature map through element-wise weight calculations based on the initial feature map output from the initial feature block 310.

[0040] In an exemplary embodiment, as shown in Figure 4, the attention block 320 may include a convolutional layer, a Max Pooling Layer, an Average Pooling Layer, a Dilated Convolution Layer, a Activation Layer, and a Combined Layer.

[0041] Specifically, attention block 320 can perform max pooling and average pooling in parallel and simultaneously through the max pooling layer and average pooling layer. Attention block 320 can also perform an extended convolution operation on each pooling result calculated through the max pooling layer and average pooling layer, and combine the results through the first join layer (element-wise sum). Here, the output value as the result of the operation may be a weight feature map (i.e., element-wise weight values). Attention block 320 can also perform an operation (element-wise multiplication) to combine the initial feature map and the weight feature map through the second join layer. Attention block 320 can also output a condensed feature map through an activation function (ReLU) as the result of the operation. In this case, before performing the operation through the second join layer, attention block 320 can prevent overfitting by ensuring that the element-wise weight values ​​of the weight feature map are in the range of 0 to 1 through an activation layer (Sigmoid Function) and dropout.

[0042] In other words, the attention block 320 considers the spacing between elements using max pooling, average pooling, and extended convolution operations, assigns weight values ​​to each element according to its importance, outputs a weighted feature map, and allows for the emphasis of important elements among the elements through the combination of the weighted feature map and the initial feature map. Here, elements can be pixels in the feature map. On the other hand, in this invention, the dilation rate of the extended convolution layer is set to 2, but is not limited to this.

[0043] The residual block 330 can output a deep feature map through a shortcut operation based on the initial feature map output from the initial feature block 310.

[0044] In an exemplary embodiment, the residual block 330 may include a plurality of short residual blocks. As shown in Figure 5, the short residual block may include a plurality of convolutional layers, a plurality of batch normalization layers, an activation layer (ReLU), and a join layer.

[0045] In other words, each short residual block outputs a new feature map from the previous feature map (the feature map output from the N-1th short attention block, in which case the first short attention block's initial feature map is the previous feature map) through a convolutional layer, and the previous and new feature maps can be combined (element-wise sum) through a join layer. Therefore, each short residual block can solve the problem of feature information disappearing while extracting deep features by combining the new feature map with the previous feature map, thereby ensuring a smooth flow of information among all short residual blocks of the residual block 330. On the other hand, overfitting can be prevented by applying dropout to the feature map output from the short residual block activation layer.

[0046] The sum block 340 can combine the lumped feature map output from the attention block 320 and the deep feature map output from the residual block 330 to output a final feature map. In this case, the sum block 340 can use element-wise sum to combine the lumped feature map and the final feature map.

[0047] The Classification Block 350 can classify the type of arrhythmia based on the final feature map output from the Combination Block 340. In this case, the Classification Block 350 estimates the probability for each class (arrhythmia information) based on the final feature map and can classify the type of arrhythmia into the class with the highest probability.

[0048] In an exemplary embodiment, the classification block 350 can perform global max pooling and global average pooling in parallel and simultaneously through a global max pooling layer and a global average pooling layer. The classification block 350 can also perform an operation to concatenate the pooling results calculated through the global max pooling layer and the global average layer through a concatenation layer. This allows the output data of the mutually separate layers to be combined into a single output. By using the global max pooling layer and the global average pooling layer, the final feature map can be output as a pre-set number of one-dimensional matrix feature maps. Furthermore, the classification block 350 can estimate probabilities for each class through a fully connected layer using the output pre-set number of one-dimensional matrix feature maps, and classify the class with the highest probability into a type of arrhythmia. Here, a Softmax function or the like can be used as the fully connected layer. The pre-set number here may be the number of arrhythmia types.

[0049] Figure 6 shows the layer structure of an artificial neural network model in the classification module of an arrhythmia classification device for 12-lead electrocardiogram signals according to one embodiment of the present invention.

[0050] Figure 6 shows the layer structure of the optimized artificial neural network model. Considering the procedure described above, we designed a CNN model capable of classifying 45 types of arrhythmias. The main structure of the CNN model is an attention block 320 to highlight important features and a residual block 330 to improve classification accuracy.

[0051] Figure 7 is a flowchart illustrating a method for classifying arrhythmias from a 12-lead electrocardiogram signal according to one embodiment of the present invention. The method shown in Figure 7 can be performed, for example, by the aforementioned arrhythmia classification device for 12-lead electrocardiogram signals. In the illustrated flowchart, the method is described in several steps, but at least some of the steps can be performed in a different order, combined with other steps, omitted, divided into more detailed steps, or with the addition of one or more steps not shown.

[0052] The 12-lead electrocardiogram signal arrhythmia classification device 100 is an S710 that can receive a 12-lead electrocardiogram signal input. For example, the 12-lead electrocardiogram signal arrhythmia classification device 100 can acquire a 12-lead electrocardiogram signal measured by medical staff. In addition, the 12-lead electrocardiogram signal arrhythmia classification device 100 can acquire a 12-lead electrocardiogram signal from an external medical institution or research institution.

[0053] Next, the arrhythmia classification device 100 for 12-lead electrocardiogram signals can output classification results for the 12-lead electrocardiogram signal using machine learning-based techniques S720. For example, the arrhythmia classification device 100 for 12-lead electrocardiogram signals may include an artificial neural network model that is trained to classify arrhythmia information based on an input 12-lead electrocardiogram signal. The artificial neural network model can be trained to classify and output arrhythmia information for a 12-lead electrocardiogram signal when a 12-lead electrocardiogram signal is input.

[0054] Finally, the arrhythmia classification device 100 of the 12-lead electrocardiogram signal can generate arrhythmia information based on the classification result and provide it to the user (e.g., medical staff) S730.

[0055] Figure 8 is a block diagram illustrating a computing environment, including computing equipment suitable for use in an exemplary embodiment. In the illustrated embodiment, each component may have different functions and capabilities in addition to those described below, and may include additional components in addition to those described below.

[0056] The illustrated computing environment 10 includes a computing device 12. In one embodiment, the computing device 12 may be an arrhythmia classification device 100 for 12-lead electrocardiogram signals.

[0057] The computing device 12 includes at least one processor 14, a computer-readable storage medium 16, and a communication bus 18. The processor 14 can be operated by the computing device 12 in accordance with the exemplary embodiments described above. For example, the processor 14 can execute one or more programs stored in the computer-readable storage medium 16. The one or more programs may include one or more computer-executable instructions, which, when executed by the processor 14, can be configured by the computing device 12 to operate in accordance with the exemplary embodiments.

[0058] The computer-readable storage medium 16 is configured to store computer-executable instructions or program code, program data, and / or other suitable forms of information. The program 20 stored in the computer-readable storage medium 16 includes a set of instructions that can be executed by the processor 14. In one embodiment, the computer-readable storage medium 16 may be memory (volatile memory such as random-access memory, non-volatile memory, or a suitable combination thereof), one or more magnetic disk storage devices, optical disk storage devices, flash memory devices, other forms of storage media that can be accessed by the computing device 12 and can store desired information, or a suitable combination thereof.

[0059] The communication bus 18 includes a processor 14 and a computer-readable storage medium 16, and interconnects various other components of the computing device 12.

[0060] The computing device 12 may also include one or more input / output interfaces 22 and one or more network communication interfaces 26 that provide interfaces for one or more input / output devices 24. The input / output interfaces 22 and network communication interfaces 26 are connected to a communication bus 18. The input / output devices 24 may be connected to other components of the computing device 12 via the input / output interfaces 22. Exemplary input / output devices 24 may include input devices such as pointing devices (such as a mouse or trackpad), keyboards, touch input devices (such as a touchpad or touchscreen), voice or sound input devices, various types of sensor devices and / or imaging devices, and / or output devices such as display devices, printers, speakers and / or network cards. Exemplary input / output devices 24 may be included inside the computing device 12 as one component of the computing device 12, or they may be connected to the computing device 12 as separate devices distinct from the computing device 12.

[0061] Although typical embodiments of the present invention have been described in detail above, those with ordinary skill in the art to which the present invention pertains will understand that various modifications are possible to the embodiments described above without departing from the scope of the present invention. Therefore, the scope of the rights of the present invention should not be limited to the embodiments described, but should be defined not only by the claims described below, but also by equivalent claims or the like. [Note 1] One or more processors, A computing device comprising memory for storing one or more programs executed by the one or more processors, A data input module that receives 12-lead electrocardiogram signals, A computing device comprising: a classification module that outputs a classification result for the input 12-lead electrocardiogram signal using machine learning-based technology based on the input 12-lead electrocardiogram signal. [Note 2] The aforementioned classification module is The computing device according to Appendix 1, which includes an artificial neural network model that receives the input of the 12-lead electrocardiogram signal and is trained to classify arrhythmia information based on the 12-lead electrocardiogram signal. [Note 3] The aforementioned artificial neural network model is An initial feature block that outputs an initial feature map through a convolution operation based on the input 12-lead electrocardiogram signal, A attention block that outputs a concentrated feature map through element-wise weight calculations based on the initial feature map output from the initial feature block, A residual block that outputs a deep feature map through shortcut calculations based on the initial feature map output from the initial feature block, A merging block that combines the concentrated feature map output from the attention block and the deep feature map output from the residual block to output a final feature map, A computing device according to Appendix 2, comprising a classification block that classifies the type of arrhythmia based on the final feature map output from the aforementioned combination block. [Note 4] The aforementioned warning block is, The computing device according to Appendix 3, which performs maximum pooling and average pooling in parallel on the initial feature map output from the initial feature block through a maximum pooling layer and an average pooling layer, combines the pooling results output through the maximum pooling layer and the average pooling layer through a first join layer to output a weight feature map, and combines the initial feature map and the weight feature map through a second join layer to output a lumped feature map. [Note 5] The aforementioned residual block is, It is configured to include N short residual blocks (where N is a natural number greater than or equal to 2) that are sequentially linked to reflect the features of the initial feature map output from the initial feature block, The aforementioned N short residual blocks are, The computing device described in Appendix 4, which takes a previous feature map (a feature map output from the N-1th short residual block) as input, outputs a new feature map from the previous feature map through a convolutional layer, and outputs the previous feature map and the new feature map combined through a third joining layer. [Note 6] The first and third join layers use element-wise sums. The second joining layer is a computing device as described in Appendix 5, which uses element-wise multiplication. [Note 7] The aforementioned classification block is, The computing device described in Appendix 3, which performs global maximum pooling and global average pooling in parallel on the final feature map through a global maximum pooling layer and a global average pooling layer, and concatenates the pooling results output through the global maximum pooling layer and the global average pooling layer through a concatenation layer. [Note 8] One or more processors, A method performed on a computing device having memory for storing one or more programs executed by one or more processors, The steps include receiving the input of a 12-lead electrocardiogram signal, A method for classifying arrhythmias from a 12-lead electrocardiogram signal, comprising the step of outputting a classification result for the 12-lead electrocardiogram signal using machine learning-based techniques based on the input 12-lead electrocardiogram signal. [Note 9] The step of outputting the classification result is: The method for classifying arrhythmias from a 12-lead electrocardiogram signal according to Appendix 8, which receives the input of the 12-lead electrocardiogram signal through an artificial neural network model and classifies arrhythmia information based on the 12-lead electrocardiogram signal. [Note 10] The aforementioned classification step is, The initial feature block outputs an initial feature map through a convolution operation based on the input 12-lead electrocardiogram signal, The attention block includes the step of outputting a concentrated feature map through element-wise weight calculations based on the output initial feature map, The residual block provides a step of outputting a deep feature map through a shortcut operation based on the output initial feature map, The process involves a merging block to combine the output condensed feature map and the output deep feature map to output a final feature map, A method for classifying arrhythmias from 12-lead electrocardiogram signals according to Appendix 9, further comprising the step of classifying the type of arrhythmia based on the output final feature map using a classification block. [Note 11] The step of outputting the aforementioned concentrated feature map is: Steps include performing maximum pooling and average pooling in parallel on the initial feature map through a maximum pooling layer and an average pooling layer, The steps include: combining the pooling results output through the maximum pooling layer and the average pooling layer through a first join layer to output a weight feature map; A method for classifying arrhythmias from a 12-lead electrocardiogram signal according to Appendix 10, further comprising the step of combining the initial feature map and the weight feature map through a second coupling layer to output the concentrated feature map. [Note 12] The aforementioned residual block is, It is configured to include N short residual blocks (where N is a natural number greater than or equal to 2) that are sequentially linked to reflect the features of the initial feature map, The aforementioned N short residual blocks are, A method for classifying arrhythmias in a 12-lead electrocardiogram signal as described in Appendix 11, wherein a previous feature map (a feature map output from the N-1 short residual block) is input, a new feature map is output from the previous feature map through a convolutional layer, and the previous feature map and the new feature map are combined and output through a third combining layer. [Note 13] The first and third join layers use element-wise sums. The second coupling layer is an arrhythmia classification method for 12-lead electrocardiogram signals described in Appendix 12, which uses element-wise multiplication. [Note 14] The step of classifying the type of arrhythmia is, The steps include performing global maximum pooling and global average pooling in parallel on the aforementioned final feature map through a global maximum pooling layer and a global average pooling layer, A method for classifying arrhythmias from 12-lead electrocardiogram signals according to Appendix 10, further comprising the step of concatenating each pooling result output through the global maximum pooling layer and the global average pooling layer through a concatenation layer. [Explanation of symbols]

[0062] 10: Computing Environment 12: Computing devices 14: Processor 16: Computer-readable storage media 18: Communications bus 20: Program 22: Input / Output Interface 24: Input / Output Devices 26: Network communication interface 100:12 lead electrocardiogram signal arrhythmia classification device 200: Data Input Module 300: Classification Module 310: Initial Feature Block 320: Warning Block 330: Residual Block 340: Joining Block 350: Classification Block

Claims

1. One or more processors, A computing device comprising memory for storing one or more programs executed by the one or more processors, A data input module that receives input of a 12-lead electrocardiogram signal, The system includes a classification module that outputs a classification result for the input 12-lead electrocardiogram signal using machine learning-based techniques based on the input 12-lead electrocardiogram signal, The aforementioned classification module is It includes an artificial neural network model that receives the input of the 12-lead electrocardiogram signal and learns to classify arrhythmia information based on the 12-lead electrocardiogram signal, The aforementioned artificial neural network model is An initial feature block that outputs an initial feature map through a convolution operation based on the input 12-lead electrocardiogram signal, A attention block that outputs a concentrated feature map through element-wise weight calculations based on the initial feature map output from the initial feature block, A residual block that outputs a deep feature map through shortcut calculations based on the initial feature map output from the initial feature block, A merging block that combines the concentrated feature map output from the attention block and the deep feature map output from the residual block to output a final feature map, A computing device comprising: a classification block that classifies the type of arrhythmia based on the final feature map output from the aforementioned combination block.

2. The caution block is The computing device according to claim 1, wherein maximum pooling and average pooling are performed in parallel on the initial feature map output from the initial feature block through a maximum pooling layer and an average pooling layer, the pooling results output through the maximum pooling layer and the average pooling layer are combined through a first combine layer to output a weight feature map, and the initial feature map and the weight feature map are combined through a second combine layer to output a concentrated feature map.

3. The residual block is It is configured to include N short residual blocks (where N is a natural number of 2 or more) that are sequentially linked to reflect the features of the initial feature map output from the initial feature block, The aforementioned N short residual blocks are, The computing device according to claim 2, wherein a previous feature map (a feature map output from the N-1th short residual block) is input, a new feature map is output from the previous feature map through a convolutional layer, and the previous feature map and the new feature map are combined and output through a third joining layer.

4. The first and third join layers use element-wise sums, The computing device according to claim 3, wherein the second combining layer uses element-wise multiplication.

5. The classification block is The computing device according to claim 1, wherein global maximum pooling and global average pooling are performed in parallel on the final feature map through a global maximum pooling layer and a global average pooling layer, and the pooling results output through the global maximum pooling layer and the global average pooling layer are concatenated through a concatenation layer.

6. One or more processors, A method performed on a computing device having memory for storing one or more programs executed by one or more processors, The steps include receiving the input of a 12-lead electrocardiogram signal, The process includes the step of outputting a classification result for the input 12-lead electrocardiogram signal using machine learning-based techniques based on the input 12-lead electrocardiogram signal, The step of outputting the classification result is: The system receives the input of the 12-lead electrocardiogram signal through an artificial neural network model, and classifies arrhythmia information based on the 12-lead electrocardiogram signal. The aforementioned classification step is, The initial feature block outputs an initial feature map through a convolution operation based on the input 12-lead electrocardiogram signal, The attention block includes the step of outputting a concentrated feature map through element-wise weight calculations based on the output initial feature map, The residual block provides a step of outputting a deep feature map through a shortcut operation based on the output initial feature map, The process involves a merging block to combine the output condensed feature map and the output deep feature map to output a final feature map, A method for classifying arrhythmias from a 12-lead electrocardiogram signal, further comprising the step of classifying the type of arrhythmia based on the output final feature map using a classification block.

7. The step of outputting the concentrated feature map is: Steps include performing maximum pooling and average pooling in parallel on the initial feature map through a maximum pooling layer and an average pooling layer, The steps include: combining the pooling results output through the maximum pooling layer and the average pooling layer through a first join layer to output a weight feature map; A method for classifying arrhythmias from a 12-lead electrocardiogram signal according to claim 6, further comprising the step of combining the initial feature map and the weight feature map through a second combining layer to output the concentrated feature map.

8. The residual block is It is configured to include N short residual blocks (where N is a natural number greater than or equal to 2) that are sequentially linked to reflect the features of the initial feature map, The aforementioned N short residual blocks are, A method for classifying arrhythmias in a 12-lead electrocardiogram signal according to claim 7, wherein a previous feature map (a feature map output from the N-1 short residual block) is input, a new feature map is output from the previous feature map through a convolutional layer, and the previous feature map and the new feature map are combined and output through a third combining layer.

9. The first and third join layers use element-wise sums, The method for classifying arrhythmias in a 12-lead electrocardiogram signal according to claim 8, wherein the second coupling layer uses element-wise multiplication.

10. The step of classifying the type of arrhythmia is: The steps include performing global maximum pooling and global average pooling in parallel on the aforementioned final feature map through a global maximum pooling layer and a global average pooling layer, The method for classifying arrhythmias from a 12-lead electrocardiogram signal according to claim 6, further comprising the step of concatenating each pooling result output through the global maximum pooling layer and the global average pooling layer through a concatenation layer.

Images