Heart disease prediction model generation method, server, and computer program

Abstract

The embodiment of the invention discloses a heart disease prediction model generation method. The method may include a step of acquiring a plurality of electrocardiogram data and a step of generating a model for predicting a heart disease based on the plurality of electrocardiogram data.

Description

Technical Field

[0001] Multiple embodiments of the present invention relate to methods for providing a neural network model for predicting a patient's heart disease that may occur in the future, and more specifically, to a technique for using a deep learning model to evaluate and predict the risk level of arrhythmias including atrial fibrillation and other heart diseases based on electrocardiogram data. Background Art

[0002] Electrocardiogram (ECG) data is a widely used and important tool for diagnosing heart conditions by recording the electrical activity of the heart. Existing multi-lead electrocardiographs play a central role in diagnosing cardiac arrhythmias, i.e., heart diseases such as atrial fibrillation, by detecting abnormalities in cardiac rhythm and electrical conduction. Arrhythmias, including atrial fibrillation, are common diseases characterized by irregular or abnormal heart rates. Since the above-mentioned arrhythmias are usually asymptomatic or have mild symptoms, they are difficult to detect without proper examination. If such arrhythmias are not treated in time, they may cause serious health complications such as stroke, heart failure, and sudden cardiac death. In particular, atrial fibrillation is associated with an increased risk of stroke and heart failure, making it one of the important types of arrhythmias. Therefore, early detection and intervention of atrial fibrillation and arrhythmias are very important for effective disease management.

[0003] On the other hand, existing electrocardiographs only focus on diagnosis and have limitations in predicting or stratifying the risk of future heart disease. This is because existing electrocardiogram analysis methods mainly rely on statistical analysis. Statistical methodology cannot fully reflect nonlinear and complex patterns, making it difficult to accurately evaluate each patient's heart status and the risk of multiple arrhythmias.

[0004] Furthermore, statistical models are less capable of processing large amounts of data than probabilistic models based on deep learning. ECG data can be affected by overlapping signals such as muscle movement, breathing, and point noise from other devices. Such external factors can reduce the accuracy of statistical methodology, and as a result, this has become a major obstacle in accurately stratifying and quantifying the risk of heart disease.

[0005] Prior art literature

[0006] Patent Literature

[0007] Korean Patent Publication No. 10-2022-0104583 Summary of the invention

[0008] The technical problem to be solved by the present invention is proposed in view of the above-mentioned background technology. The purpose of the present invention is to provide the following technology, which can predict and stratify the risk level of future heart disease based on electrocardiogram data.

[0009] The technical problems to be solved by the present invention are not limited to the problems mentioned above, and ordinary technicians in the technical field to which the present invention belongs can clearly understand other technical problems not mentioned from the following description.

[0010] An embodiment of the present invention for solving the above problem discloses a method for generating a heart disease prediction model. The method may include the following steps: acquiring a plurality of electrocardiogram data; and generating a model for predicting heart disease based on the plurality of electrocardiogram data.

[0011] In an alternative embodiment, the step of generating a model for predicting the above-mentioned heart disease may include the following steps: constructing a learning data set based on the above-mentioned multiple electrocardiogram data; and learning at least one network function based on the above-mentioned learning data set to generate a heart disease risk level prediction model, the above-mentioned learning data set includes a first learning data set and a second learning data set classified according to different learning purposes, and the above-mentioned heart disease risk level prediction model is stratified based on the patient's electrocardiogram data and provides predictive information related to the risk level of future heart disease.

[0012] In an alternative embodiment, the first learning data set may include data for learning, wherein the learning corresponds to a process of transforming characteristics of electrocardiogram data into a characteristic space, and the second learning data set includes calibration data related to prediction of cardiac risk level.

[0013] In an alternative embodiment, the step of constructing the above-mentioned learning data set may include the step of performing preprocessing on the above-mentioned multiple ECG data, and the step of performing the above-mentioned preprocessing includes the following steps: performing noise preprocessing on the above-mentioned multiple ECG data; dividing the multiple ECG data after the above-mentioned noise preprocessing into predefined window sizes to generate multiple region of interest (ROI) signals; and extracting heart rate variability (HRV) characteristic information based on the leads of the above-mentioned multiple ECG data, and the above-mentioned heart rate variability characteristic information includes indicator information related to the variability between relevant heart rate intervals.

[0014] In an alternative embodiment, the step of generating the above-mentioned heart disease risk prediction model may include: an embedding model generating step, generating an embedding model by using the masked self-supervised learning based on the above-mentioned first learning data set; an embedding execution step, using the above-mentioned embedding model to extract multiple latent vectors (latent vectors) respectively corresponding to multiple region of interest signals, and the above-mentioned multiple region of interest signals correspond to the above-mentioned second learning data set; a clustering execution step, using the clustering model to perform clustering on the above-mentioned multiple latent vectors; and a vector database construction step, constructing a vector database (DB, DataBase) based on the above-mentioned multiple latent vectors and information related to each cluster corresponding to each latent vector.

[0015] In an alternative embodiment, the embedding model generating step may include the following steps: inputting data contained in the first learning data set into the self-reconstruction model to guide self-supervised learning, so that the self-reconstruction model generates an output similar to the input data; and extracting an encoder from the self-reconstruction model that has completed the learning and generating the embedding model, wherein the self-reconstruction model serves as a neural network model that masks a part of the input data and restores the masked part, including a dimensionality reduction network function and a dimensionality restoration network function.

[0016] In an alternative embodiment, the clustering model may perform clustering based on the similarity distance between multiple potential vectors respectively corresponding to the multiple region of interest signals, and the clustering execution step includes the step of extracting a representative vector for each cluster corresponding to the clustering result.

[0017] In an alternative embodiment, the method may include the following steps: acquiring electrocardiogram data of a predicted object; performing preprocessing on the electrocardiogram data of the predicted object; using the embedding model to generate a plurality of latent vectors corresponding to the preprocessed electrocardiogram data; respectively comparing the plurality of latent vectors with representative vectors corresponding to the plurality of clusters to classify the plurality of latent vectors into one of the plurality of clusters; and generating hierarchical information related to the risk level of heart disease based on the classification results of the plurality of latent vectors into the plurality of clusters.

[0018] In an alternative embodiment, the step of generating the above-mentioned heart disease risk prediction model may include the following steps: obtaining user meta-information corresponding to the above-mentioned multiple electrocardiogram data respectively; obtaining the above-mentioned multiple latent vectors and cluster information corresponding to the above-mentioned multiple latent vectors from the above-mentioned vector database; and training multiple tree models based on the above-mentioned heart rate variability characteristic information, the above-mentioned user meta-information, the above-mentioned multiple latent vectors and the cluster information corresponding to the above-mentioned latent vectors, and performing ensemble learning to merge the outputs of each tree model to generate the above-mentioned heart disease risk prediction model.

[0019] In an alternative embodiment, the step of generating a model for predicting the above-mentioned heart disease may also include the following steps: classifying the above-mentioned multiple electrocardiogram data obtained into a training data set and a verification data set; respectively obtaining a plurality of region-of-interest electrocardiogram data divided into predefined window sizes from the multiple electrocardiogram data classified into the above-mentioned training data set and the verification data set; training the above-mentioned deep learning model correspondingly to inputting a plurality of first region-of-interest electrocardiogram data obtained from the multiple electrocardiogram data included in the above-mentioned training data set into a deep learning model for predicting a patient's heart disease so as to predict the heart disease category of each of the above-mentioned multiple first region-of-interest electrocardiogram data; and using a plurality of second region-of-interest electrocardiogram data obtained from the multiple electrocardiogram data included in the above-mentioned verification data set for the trained deep learning model to determine a threshold for distinguishing the above-mentioned heart disease categories.

[0020] In an alternative embodiment, the step of determining the threshold value for distinguishing the above-mentioned heart disease categories may include the following steps: collecting probability scores associated with multiple region-of-interest electrocardiogram data separated from the same electrocardiogram data for the above-mentioned multiple second region-of-interest electrocardiogram data; averaging the probability scores associated with the above-mentioned multiple second region-of-interest electrocardiogram data collected to derive a final probability score; and fine-tuning the threshold value for distinguishing the above-mentioned heart disease categories based on the derived final probability score.

[0021] In an alternative embodiment, the determined threshold value may be stored together with the weighted value of the deep learning model.

[0022] According to another embodiment of the present invention, the present invention discloses a server for executing a method for generating a heart disease prediction model. The server may include: a memory storing at least one instruction; and a processor executing the at least one instruction stored in the memory, wherein the processor executes the at least one instruction to execute the method for generating a heart disease prediction model.

[0023] According to another embodiment of the present invention, the present invention discloses a computer program stored in a computer-readable recording medium. The computer program can be combined with a computer as hardware to execute a method for generating a heart disease prediction model.

[0024] Other details of the present invention are included in the detailed description and the accompanying drawings.

[0025] According to multiple embodiments of the present invention, region-of-interest electrocardiogram data can be extracted from the patient's electrocardiogram data, and the extracted region-of-interest electrocardiogram data can be applied to a deep learning model to improve the accuracy of predicting the patient's heart disease.

[0026] Furthermore, the present invention can provide the following effects: using multiple machine learning models to predict the risk of heart disease in the future, so that patients can understand their own health status and prevent heart disease in advance. Thus, heart disease can be detected early and treated quickly, thereby protecting the patient's life and improving the treatment results.

[0027] Furthermore, the present invention can provide the following effect: through an AI-based automated processing process, the physical time of medical staff required to analyze electrocardiogram data one by one is greatly reduced, thereby improving the effectiveness of screening and reducing the opportunity cost of the medical site.

[0028] Additionally, the present invention may use multiple tree series machine learning models to perform integration to integrate the latent vectors extracted from the embedding model of the heart rate variability characteristic information and metadata such as the patient's age, thereby further improving the prediction accuracy of the risk of heart disease.

[0029] The effects of the present invention are not limited to the effects mentioned above, and those skilled in the art can clearly understand other effects not mentioned from the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 The figure is a structural diagram showing a computing device for executing a method for generating a heart disease prediction model according to an embodiment of the present invention.

[0031] Figure 2 The flowchart is a method for learning a deep learning model for predicting heart disease according to an embodiment of the present invention.

[0032] Figure 3 A diagram illustrating a method for classifying a plurality of electrocardiogram data according to an embodiment of the present invention.

[0033] Figure 4 The figure shows a method of subdividing electrocardiogram data into individual heart rates according to an embodiment of the present invention.

[0034] Figure 5a to Figure 5c A diagram illustrating types of deep learning models according to an embodiment of the present invention.

[0035] Figure 6 A diagram illustrating the learning steps of a deep learning model according to an embodiment of the present invention.

[0036] Figure 7 The flowchart is a method for predicting heart disease using a trained deep learning model according to an embodiment of the present invention.

[0037] Figure 8 A diagram illustrating the steps of predicting heart disease using a deep learning model according to an embodiment of the present invention.

[0038] Fig. 9 An exemplary flow chart related to a method for generating a heart disease prediction model according to an embodiment of the present invention is shown.

[0039] Fig.10 The flowchart exemplarily shows the pre-processing process of a plurality of electrocardiogram data according to an embodiment of the present invention.

[0040] Fig.11 The flowchart exemplarily shows a process of embedding multiple potential vectors according to an embodiment of the present invention and constructing a vector database according to the execution results.

[0041] Fig.12 The following is a flowchart exemplarily showing a process of generating an embedding model according to an embodiment of the present invention.

[0042] Fig.13 The diagram is an example diagram for explaining a self-reconstruction model and a process of generating an embedding model from a trained self-reconstruction model according to an embodiment of the present invention.

[0043] Fig.14 The flowchart is an exemplary flowchart for explaining the process of generating a heart disease risk prediction model through integrated learning of multiple tree series models according to an embodiment of the present invention.

[0044] Description of Reference Numerals

[0045] 100: Computing Device

[0046] 110: Processor

[0047] 120: Memory

[0048] 130: Procedure

[0049] 140: Display DETAILED DESCRIPTION

[0050] The specific structural or functional descriptions of the embodiments are only for illustration, and the above descriptions can be implemented in various forms. Therefore, the actual implementation form is not limited to the specific embodiments disclosed, and the scope of this specification includes the changes, equivalent technical solutions or alternative technical solutions included in the technical ideas described in the embodiments.

[0051] The terms "first" or "second" can be used to describe multiple structural elements, but these terms are only used to describe two structural elements. For example, the first structural element can be named as the second structural element, and similarly, the second structural element can also be named as the first structural element.

[0052] When it is mentioned that a structural element is “connected” to other structural elements, it can be directly connected or coupled to the other structural elements, or other structural elements may exist in between.

[0053] Unless the context clearly indicates otherwise, a singular expression includes a plural expression. In this document, sentences such as "A or B", "at least one of A and B", "at least one of A or B", "A, B or C", "at least one of A, B and C" and "at least one of A, B or C" may respectively include one or all possible combinations of the multiple items listed together in the corresponding sentences in the above multiple sentences. In this specification, the terms "including" or "having" are used to specify the existence of the described features, numbers, steps, actions, structural elements, components or these combinations, and do not preclude the existence or additional possibilities of one or more other features, numbers, steps, actions, structural elements, components or these combinations.

[0054] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as those generally understood by ordinary technicians in the technical field to which the present invention belongs. The meanings of terms defined in commonly used dictionaries are the same as those in the context of the relevant technology and cannot be interpreted as ideal or excessive meanings unless clearly defined in this specification.

[0055] In the following, the embodiments are described in detail with reference to the drawings. In the description with reference to the drawings, the same components are given the same reference numerals regardless of the reference numerals, and repeated description thereof is omitted.

[0056] The present invention can obtain multiple electrocardiogram data corresponding to multiple patients, and generate a model for predicting heart disease based on the obtained multiple electrocardiogram data. In an embodiment, the model for predicting heart disease can be a neural network model that extracts important patterns and features from the electrocardiogram data and predicts the possibility of heart disease based on them. The computing device of the present invention processes the multiple electrocardiogram data into a form suitable for learning, and uses the processed data to train the model for predicting heart disease.

[0057] According to one embodiment, the present invention can generate and provide a deep learning model and a heart disease risk prediction model as a model for predicting heart disease. In an embodiment, the present invention can process multiple electrocardiogram data into multiple forms and generate and provide multiple prediction models by training neural networks in different ways. Below, with reference to multiple figures, a method for generating multiple models for predicting heart disease is described in detail.

[0058] Figure 1 The figure is a structural diagram showing a computing device for executing a method for generating a heart disease prediction model according to an embodiment of the present invention.

[0059] like Figure 1As shown, the computing device 100 may include at least one processor 110 and a memory 120 that may load or store a program 130 executed by the processor 110 . Figure 1 The structural elements included in the computing device 100 are only an example, and any person skilled in the art of the present invention can understand that Figure 1 In addition to the structural elements shown, other common structural elements may also be included.

[0060] The processor 110 controls the entire operation of each structural element of the computing device 100. The processor 110 may include a central processing unit (CPU), a microprocessor unit (MPU), a microcontroller unit (MCU), a graphics processing unit (GPU), a neural processing unit (NPU), a digital signal processor (DSP), or at least one of any form of processor known in the technical field of the present invention. In addition, the processor 110 may perform operations on at least one application or program for running the methods / operations of various embodiments of the present invention. The computing device 100 may include at least one processor.

[0061] The memory 120 stores at least one or a combination of two or more of various data, instructions, and information used by structural elements (eg, the processor 110) included in the computing device 100. The memory 120 may include a volatile memory and / or a nonvolatile memory.

[0062] The program 130 may include at least one action for implementing the methods / actions of various embodiments of the present invention, and is stored in the memory 120 in the form of software. The action corresponds to the instructions implemented in the program 130. For example, the program 130 may include multiple instructions for performing the following actions: an action of classifying a plurality of electrocardiogram data obtained from a plurality of patients into a training data set, a validation data set, and an experimental data set, an action of respectively obtaining a plurality of electrocardiogram data of regions of interest divided into predefined window sizes from the plurality of electrocardiogram data classified into the training data set, the validation data set, and the experimental data set, an action of training a deep learning model corresponding to inputting a plurality of first electrocardiogram data of regions of interest obtained from a plurality of electrocardiogram data included in the training data set into a deep learning model for predicting a patient's heart disease, so as to predict the heart disease category of each first electrocardiogram data of regions of interest, and an action of applying a plurality of second electrocardiogram data of regions of interest obtained from a plurality of electrocardiogram data included in the validation data set to the trained deep learning model to determine a threshold for distinguishing heart disease categories.

[0063] According to the embodiment, the region of interest electrocardiogram data divided into a predefined window size can represent an individual heart rate, but is not limited thereto. Hereinafter, according to the embodiment, for the convenience of description, the region of interest electrocardiogram data divided into a predefined window size can be described by taking an individual heart rate as an example, but the type of region of interest electrocardiogram data applicable to each embodiment is not limited thereto.

[0064] When the program 130 is loaded into the memory 120 , the processor 110 may execute multiple tasks for implementing the program 130 , thereby executing the methods / tasks of multiple embodiments of the present invention.

[0065] The running screen of the program 130 can be displayed through the display 140. Figure 1 In the case of a computer 100, the display 140 is a separate device connected to the computing device 100. In the case of a computing device 100 such as a terminal that a user can carry, such as a smart phone or a tablet computer, the display 140 may be a component of the computing device 100. The screen displayed on the display 140 may be before information is input to the program or the result of the program running.

[0066] Figure 2 The flowchart is a method for learning a deep learning model for predicting heart disease according to an embodiment of the present invention.

[0067] Figure 2 The learning method of the deep learning model shown is through Figure 1The processor of the computing device shown in the figure is executed. In step 210, the processor can classify the multiple electrocardiogram data obtained from multiple patients into a training data set, a verification data set and an experimental data set. First, the multiple electrocardiogram data obtained from multiple patients can be grouped into normal sinus rhythm (True-Normal Sinus Rhythm, T-NSR), atrial fibrillation-normal sinus rhythm (Afib-Normal Sinus Rhythm, AF-NSR) and clinically important arrhythmia-normal sinus rhythm (Clinically Important Arrhythmia-Normal Sinus Rhythm, CIA-NSR). In this case, the multiple electrocardiogram data obtained from multiple patients can be 10 seconds 12-lead electrocardiogram data, but this type of electrocardiogram data is only an example and is not limited to the above example.

[0068] On the other hand, clinically important arrhythmias (CIA) may include atrial arrhythmias, ventricular arrhythmias, atrial fibrillation, and incomplete bundle branch block (BBB). Atrial arrhythmias, as arrhythmias originating from the atria, refer to non-sustained arrhythmias that last for more than 30 seconds or less than 30 seconds, and may include atrial premature beats or sustained / non-sustained atrial rhythms, etc. Ventricular arrhythmias, as arrhythmias originating from the ventricles, refer to non-sustained arrhythmias that last for more than 30 seconds or less than 30 seconds, and may include ventricular premature beats or sustained / non-sustained ventricular arrhythmias, etc. Atrial fibrillation may refer to an arrhythmia in which the atria have no regular electrical signals and contractions, thereby causing irregular contractions of the ventricles. Finally, incomplete bundle branch block can refer to an arrhythmia in which the right or left bundle of cardiac signals passing through the ventricles is blocked, resulting in a characteristic electrocardiogram shape.

[0069] The normal sinus rhythm group can be composed of patients without atrial fibrillation or arrhythmia pathology and with normal sinus rhythm ECG within 1 year. Figure 3The atrial fibrillation-normal sinus rhythm group can be composed of the electrocardiograms of patients with normal sinus rhythm and the electrocardiograms of atrial fibrillation or atrial flutter that occurred within 14 days from the corresponding normal sinus rhythm electrocardiogram. Similarly, the clinically important arrhythmia-normal sinus rhythm group can be composed of the electrocardiograms of patients with normal sinus rhythm and the electrocardiograms of arrhythmias that occurred within 14 days from the corresponding normal sinus rhythm electrocardiogram.

[0070] The processor can classify multiple electrocardiogram data belonging to a normal sinus rhythm group, an atrial fibrillation-normal sinus rhythm group, and a clinically important arrhythmia-normal sinus rhythm group into a training data set, a validation data set, and an experimental data set that can be used in a deep learning model for predicting heart disease.

[0071] As an example, Figure 3 The figure is a diagram showing a classification method of a plurality of electrocardiogram data according to an embodiment of the present invention. The processor can distinguish the heart diseases of the plurality of electrocardiogram data 310 according to the prediction of belonging to the normal sinus rhythm group, the atrial fibrillation-normal sinus rhythm group, and the clinically important arrhythmia-normal sinus rhythm group. As an example, the processor can distinguish the plurality of electrocardiogram data 310 into the plurality of electrocardiogram data 320 belonging to the normal sinus rhythm group and the atrial fibrillation-normal sinus rhythm group in order to predict atrial fibrillation, and can distinguish them into the plurality of electrocardiogram data 330 belonging to the normal sinus rhythm group and the clinically important arrhythmia-normal sinus rhythm group in order to predict arrhythmia.

[0072] Then, the processor may classify the plurality of electrocardiogram data 320 , 330 classified for each heart disease into a training data set, a validation data set, and an experimental data set based on an arbitrary date when the corresponding plurality of electrocardiogram data 320 , 330 is generated.

[0073] As an example, Figure 3 As shown, when multiple electrocardiogram data 320, 330 are acquired during the period from 2017-05-23 to 2022-05-23, for the multiple electrocardiogram data 320, 330, the processor can classify the multiple electrocardiogram data before the corresponding date into training data sets and verification data sets according to a specified ratio based on any date (for example, 2021-06-11), and classify the multiple electrocardiogram data after the corresponding date (including the corresponding date) into experimental data sets. Figure 3 In the example, multiple electrocardiogram data 320, 330 can be classified according to the ratio of training data set (60%), verification data set (20%) and experimental data set (20%). This classification ratio is only an example, and the present invention is not limited to the above example.

[0074] In step 220, the processor may obtain individual heart rates from the plurality of electrocardiogram data classified into the training data set, the validation data set, and the experimental data set. The processor may perform preprocessing on the 10-second 12-lead electrocardiogram data classified into the training data set, the validation data set, and the experimental data set in order to obtain accurate and reliable data.

[0075] More specifically, the processor can receive multiple electrocardiogram data classified into a training data set, a validation data set, and an experimental data set in the form of an extensible markup language (XML) file. The processor can parse the input data into a stereotyped data part such as the patient's name, age, and gender and an unstereotyped part consisting of a continuous signal. Afterwards, the processor can perform noise removal preprocessing on the unstereotyped data part consisting of the continuous signal and distinguish the key markers of the heart rate. The processor can obtain discontinuous individual heart rates from the unstereotyped data consisting of the continuous signal through the key markers of the heart rate distinguished as above.

[0076] As an example, refer to Figure 4 After decoding the 10-second 12-lead ECG data through Base64 encryption, the processor may pass the decoded data through an IIR Butterworth SOS filter with a moving average kernel and a power line noise filter for noise removal and purification. Afterwards, the processor may use a QRS peak detection algorithm to segment the de-noised 10-second 12-lead ECG data into individual heart rates.

[0077] Results, such as Figure 3 As shown, the processor can obtain multiple discontinuous individual heart rates from 10 seconds of 12-lead electrocardiogram data. The individual heart rates obtained as above can be used to train a deep learning model to more accurately predict heart disease.

[0078] In step 230, corresponding to inputting the first individual heart rate obtained from a plurality of electrocardiogram data included in the training data set into the deep learning model for predicting the patient's heart disease, the processor can train the deep learning model to predict the heart disease category of each first individual heart rate.

[0079] In this case, at least one deep learning model among ResNet-18, Conv1D including LSTM (Long Short-Term Memory), and Conv1D including Transformer can be used for training to predict the heart disease category of each first individual heart rate.

[0080] As an example, Figure 5aThe ResNet-18 shown can be a deep learning model that uses convolution operations such as various convolutional neural networks (CNNs) to extract necessary features from the input. In order to solve the vanishing gradient problem of the convolutional neural network architecture, such as Figure 5a As shown, ResNet-18 can perform residual learning through skip connections, which is a way of directly connecting input data to the output layer through multiple layers of the skip network.

[0081] The deep learning model of ResNet-18 requires a fixed-length input, so the processor may fix the length of the individual heart rate to the average length of all individual heart rates. For example, when the average length of all individual heart rates is 700, the processor may perform slicing on individual heart rates having a heart rate length longer than 700, and may perform zero padding on individual heart rates having a heart rate length shorter than 700 to fix their length to 700.

[0082] As another example, Figure 5b The Conv1D shown, including a Long Short-Term Memory (LSTM), can both capture local and long-range temporal patterns from sequential data. In this case, the Conv1D layer is good at detecting local temporal patterns, and the Long Short-Term Memory layer is good at modeling long-term dependencies.

[0083] As another example, Figure 5c The Conv1D with transformer shown can capture both local patterns and global dependencies of the input data. In this case, the transformer layer is good at modeling global dependencies, and the Conv1D layer is effective at detecting local patterns. Unlike ResNet-18, which has a fixed input length, the advantage of Conv1D with transformer is that it can accommodate a variety of input sizes.

[0084] on the other hand, Figure 6 FIG. 1 is a diagram showing the learning steps of a deep learning model according to an embodiment of the present invention. Figure 6 , the processor can use binary cross entropy with logarithmic loss and AdamW optimizer with initial learning rate of 0.0001 to optimize the parameters of the deep learning model. In this case, binary cross entropy is a loss function that reduces the difference between the predicted result and the actual answer during the learning process of the deep learning model, and AdamW optimizer is an algorithm that participates in the update of the actual deep learning model based on this loss function. The processor can apply a sigmoid function to the output of the deep learning model optimized by the above-mentioned binary cross entropy and AdamW optimizer, thereby obtaining a probability value in the range of 0 to 1 for each heart disease category of the first individual heart rate.

[0085] In step 240 , the processor may apply a second individual heart rate obtained from a plurality of electrocardiogram data included in the validation data set to the trained deep learning model to determine a threshold for distinguishing a category of heart disease.

[0086] More specifically, if Figure 6 As shown, the processor can collect the probability value of each individual heart rate separated from the same electrocardiogram data for each second individual probability value, average all the collected probability values ​​of the second individual heart rates and derive a final probability score, and then fine-tune the optimal threshold for distinguishing the categories of heart diseases based on the derived final probability score.

[0087] In this case, the optimal threshold can be obtained by applying a threshold between 0 and 1 in units of 0.01 and achieving the highest F1 score in the inflammation dataset related to cardiac disease categories, i.e., normal sinus rhythm and atrial fibrillation-normal sinus rhythm category or normal sinus rhythm and clinically important arrhythmias-normal sinus rhythm category.

[0088] In this case, the F1 score can be defined by the following formula 1.

[0089] Formula 1

[0090]

[0091] Among them, Precision is Recall

[0092] Among them, Precision is the proportion of predicted positives that are accurately classified, which can indicate the accuracy of the detected results, that is, how many actual objects are included in the detection results. Recall represents the proportion of actual positives that are accurately classified, which is an indicator to measure the degree to which the actual object is captured without being lost. In addition, TruePositive and TrueNegative represent samples classified as actual positives and actual negatives, respectively, and FalsePositive and FalseNegative refer to samples classified as false positives and false negatives, respectively.

[0093] The processor may store the optimal threshold for the heart disease category along with the weighted values ​​of the deep learning model.

[0094] Figure 7 The flowchart is a method for predicting heart disease using a trained deep learning model according to an embodiment of the present invention.

[0095] use Figure 7The deep learning model for predicting heart disease is shown in Figure 1 As shown in step 710, the processor receives electrocardiogram data of a patient whose heart disease category needs to be predicted. As shown in step 720, individual heart rates can be obtained from the received electrocardiogram data. In this case, the received electrocardiogram data and the obtained individual heart rates can be compared with Figure 3 The experimental dataset shown corresponds to .

[0096] In step 730 , the processor inputs the acquired individual heart rates into a deep learning model for predicting heart disease of the patient to derive a probability value of predicting a specific heart disease category for each individual heart rate.

[0097] In step 740, the processor may predict the patient's heart disease using the probability value derived for each individual heart rate. More specifically, the processor compares the average of the probability values ​​derived for all individual heart rates with a threshold to distinguish the patient's heart disease category.

[0098] As an example, Figure 8 FIG. 1 is a diagram showing the steps of predicting heart disease using a deep learning model according to an embodiment of the present invention. Figure 8 After the processor loads the weighted values ​​and thresholds of the trained deep learning model, the probability value of each individual heart rate is derived. The processor can calculate the average value of the probability values ​​of all individual heart rates derived as above.

[0099] As an example, when the average normal sinus rhythm Logit value, which is the probability value for judging as normal sinus rhythm, is greater than the first threshold θ1 and the average clinically important arrhythmia-normal sinus rhythm Logit value, which is the probability value for judging as clinically important arrhythmia-normal sinus rhythm, is less than the second threshold θ2, the processor can judge the corresponding patient's heart disease category as a normal state, that is, normal sinus rhythm.

[0100] Alternatively, when the normal sinus rhythm Logit value is greater than a first threshold value θ1 and the clinically important arrhythmia-normal sinus rhythm Logit value is greater than a second threshold value θ2, the processor may compare the normal sinus rhythm Logit value and the clinically important arrhythmia-normal sinus rhythm Logit value to select the larger value as the final prediction. As an example, when the clinically important arrhythmia-normal sinus rhythm Logit value is greater than the normal sinus rhythm Logit value, the heart disease category of the corresponding patient may be determined as clinically important arrhythmia-normal sinus rhythm with the possibility of arrhythmia.

[0101] Furthermore, when the normal sinus rhythm Logit value is less than a first threshold θ1 and the clinically important arrhythmia-normal sinus rhythm Logit value is less than a second threshold θ2, the processor may compare the normal sinus rhythm Logit value and the clinically important arrhythmia-normal sinus rhythm Logit value to select the larger value as the final prediction.

[0102] Finally, when the normal sinus rhythm Logit value is less than the first threshold θ1 and the clinically important arrhythmia-normal sinus rhythm Logit value is greater than the second threshold θ2, the processor can judge the corresponding patient's heart disease category as clinically important arrhythmia-normal sinus rhythm with the possibility of arrhythmia.

[0103] The four methods for distinguishing this type of heart disease can also be applied equally to distinguishing between normal sinus rhythm and atrial fibrillation-normal sinus rhythm.

[0104] Below, refer to Figures 9 to 14 , specifically describing a method for using heart rate variability characteristic information, latent vectors, cluster information and metadata corresponding to the latent vectors, and a method for predicting heart disease by integrating multiple tree model outputs related to an embodiment of the present invention.

[0105] Fig. 9 An exemplary flow chart related to a method for generating a heart disease prediction model according to an embodiment of the present invention is shown.

[0106] According to an embodiment of the present invention, the method for generating a heart disease prediction model may include a step S1000 of acquiring a plurality of electrocardiogram data.

[0107] According to an embodiment, multiple electrocardiogram data may be obtained from multiple patients and may be collected in a variety of ways. In one embodiment, the electrocardiogram data of the present invention may refer to multi-lead electrocardiogram (ECG) data related to recording the electrical activity of the heart to evaluate the state of the heart. The multi-lead electrocardiogram data includes cardiac electrical signals obtained through multiple leads (i.e., leads), thereby providing comprehensive information related to cardiac rhythm, conductive state, and other cardiac functions.

[0108] In an embodiment, the multiple electrocardiogram data obtained from multiple patients may include normal electrocardiogram data, electrocardiogram data related to arrhythmia, and electrocardiogram data related to other heart states. In a specific embodiment, the electrocardiogram data obtained from multiple patients can be divided into groups such as True-Normal Sinus Rhythm (T-NSR), Atrial Fibrillation-Normal Sinus Rhythm (AF-NSR), and Clinically Important Arrhythmia-Normal Sinus Rhythm (CIA-NSR). In this case, the multiple electrocardiogram data obtained from multiple patients can be 10 seconds 12-lead electrocardiogram data, but this type of electrocardiogram data is only an example, and the present invention is not limited to this.

[0109] In one embodiment, the acquisition of multiple electrocardiogram data may be receiving or loading data stored in the memory 120. The acquisition of multiple electrocardiogram data may be receiving or loading multiple electrocardiogram data from other storage media, other computing devices, or separate processing modules within the same computing device based on a wired and wireless communication unit. As an example, a user (e.g., a patient or a medical staff member) may access the computing device 100 through a user terminal and receive a user interface for heart disease prediction from the computing device 100, and may transfer the electrocardiogram data to the computing device 100 by dragging and dropping the electrocardiogram data on the provided user interface.

[0110] According to various embodiments, the computing device 100 of the present invention may perform data enhancement on the acquired multiple electrocardiogram data. Specifically, the computing device 100 may perform data enhancement using lead pairs to enhance learning data for learning a neural network.

[0111] Specifically, each electrocardiogram data may be acquired by lead. For example, electrocardiogram data may be collected by various lead structures such as 1-lead, 3-lead, and 12-lead, which is used to more comprehensively evaluate the heart state by recording the electrical activities of multiple parts of the heart.

[0112] For example, for 12-lead signals, many signals can be acquired in a relatively short time, but the total amount of data is limited. On the contrary, although 1-lead (single lead) signals are recorded for a longer time, the diversity of data is limited. When the above two types of data are used together, a more accurate evaluation of the heart state can be performed, but in reality, the amount of such data pairing is not sufficient.

[0113] In order to solve this problem, the computing device 100 of the present invention can generate a reconstruction model, which can generate other lead data based on one lead data. The reconstruction model plays the role of receiving single lead data and generating multi-lead data based on it. In this process, the single lead data can be expanded to 3-leads or more leads (12-leads) to have an effect such as enhancing the learning data to have more lead data.

[0114] The learning of the reconstruction model includes the process of learning the correlation between single-lead and multi-lead data. To this end, the reconstruction model is trained using paired data in which both single-lead and multi-lead data exist. During the learning process, the model learns a method of reproducing multi-lead data from single-lead data, in which case the features of the single-lead data and the pattern of the multi-lead signal that can be inferred from it are learned.

[0115] After the model has completed learning, the reconstructed model can only generate multi-lead signals through single-lead data, thereby improving the deficiencies of actual multi-lead data. As a result, the diversity of data will increase and the generalization performance of the model will be improved. The multi-lead data generated in this way (i.e., enhanced multi-lead data) is used together with the existing multi-lead data for neural network learning, and the model can learn more diverse lead structures, which can improve the prediction performance of various cardiac states.

[0116] As a result, data augmentation of the reconstructed model substantially increases the amount of learning data to improve the model's prediction accuracy and strengthen the reliability of the heart disease prediction model. As a result, medical staff can more accurately diagnose the patient's health status and help establish more effective treatment and prevention plans.

[0117] According to an embodiment of the present invention, the method for generating a heart disease prediction model may include step S2000 of constructing a learning data set based on a plurality of electrocardiogram data.

[0118] In the embodiment, constructing a learning dataset from multiple electrocardiogram data is a process of collecting and composing multiple data in order to learn important features and patterns of electrocardiogram data, thereby generating a dataset including multiple examples and patterns required for heart disease prediction.

[0119] According to an embodiment, the learning data set of the present invention includes a variety of data related to the prediction of heart disease, which may include a training data set for training a neural network model, a validation data set for evaluating and optimizing model performance, and experimental data for evaluating the generalization performance of the model.

[0120] In one embodiment, the learning data set may include a first learning data set and a second learning data set classified according to different learning purposes. The first learning data set and the second learning data set may be data classified for training different neural network models. The first learning data set and the second learning data set may be classified by the computing device 100.

[0121] For example, the first learning dataset is used in the learning process of the embedding model, and the corresponding dataset focuses on extracting and representing the main features of the electrocardiogram data. The second learning dataset is used to learn a model for predicting the risk of heart disease, and the corresponding dataset can be used to calibrate and evaluate the accuracy of the prediction model.

[0122] Specifically, the first learning data set may include data for learning corresponding to the following process, that is, transforming the characteristics of the electrocardiogram data into a characteristic space. The first learning data set is data for electrocardiogram representation learning (ECGRepresentation Learning), which can be used to extract and embed various characteristics and patterns of electrocardiogram signals. In addition, the second learning data set may include data for calibration related to the prediction of cardiac risk level. That is, the second learning data set, as data for correcting the cardiac risk level, can be used to more accurately match the risk level predicted by the model.

[0123] According to an embodiment, the step of constructing the learning dataset may include the step of performing preprocessing on the plurality of electrocardiogram data.

[0124] Fig.10 The flowchart exemplarily shows the pre-processing process of a plurality of electrocardiogram data according to an embodiment of the present invention.

[0125] Reference Fig.10 , the preprocessing method for multiple electrocardiogram data may include: step S2100, performing noise preprocessing on multiple electrocardiogram data; step S2200, dividing the multiple electrocardiogram data after noise preprocessing into predefined window sizes to obtain multiple region of interest signals; and step S2300, extracting heart rate variability characteristic information (HRV, Heart Rate Variability) based on the leads of the multiple electrocardiogram data. In one embodiment, the heart rate variability characteristic information may include indicator information related to the variability between heart rate intervals.

[0126] In more detail, first, preprocessing is performed to remove noise from the multiple electrocardiogram data obtained in step S2100. Among them, the preprocessing for removing noise includes low-pass filtering and noise removal, which are used to remove the baseline noise of the electrocardiogram signal. Specifically, low-pass filtering improves signal purity by removing high-frequency components from the electrocardiogram data through frequency band limitation, and noise removal uses a timing analysis method to reduce unnecessary noise from the electrocardiogram data to make important signal components clearer. This preprocessing process minimizes the distortion of the electrocardiogram signal and contributes to improving the signal quality for more accurate evaluation of cardiac activity.

[0127] Furthermore, according to an embodiment, the computing device 100 may perform data enhancement for improving the diversity and robustness of model learning. The computing device 100 may perform data enhancement by simulating a variety of noise patterns, thereby enabling the model to learn and adapt to a variety of noise conditions that may occur in actual environments.

[0128] In a specific embodiment, during the preprocessing of the electrocardiogram data, the data set is enhanced by artificially adding various noises that may occur in the actual electrocardiogram measurement environment, such as Gaussian noise, jitter noise, random time warping, scaling, amplitude modulation, and phase shifting, so that the model can learn and adapt to the noise that may occur in various environments. The above-mentioned noise addition and removal process has the advantages of improving data quality and improving the generalization performance of the model.

[0129] Afterwards, in step S2200, the noise-removed electrocardiogram data is segmented into predefined window sizes and organized into a plurality of region of interest signals. According to an embodiment, the region of interest signal is defined as a range including at least 3 to 5 QRS complexes (heart rate) as a changeable window size unit including a meaningful heart rate from the electrocardiogram data, respectively. In one embodiment, the QRS complex represents the depolarization of the ventricle in the heart's electrical activity, which, as a main feature of the electrocardiogram, helps to identify the beginning and end of the heart rate. For example, each region of interest signal can be segmented into units from 30 seconds to 1 minute, which is useful for analyzing important patterns in the electrocardiogram signal and identifying the regularity and abnormality of the heart rate.

[0130] The computing device 100 extracts the region of interest signal and organizes it into data that can more accurately analyze the main characteristics of the heart rate, which plays an important role in identifying specific patterns and abnormal symptoms of the electrocardiogram signal and evaluating the risk of heart disease.

[0131] Furthermore, in step S2300, a heart rate variability characteristic information extraction process is performed. The heart rate variability characteristic information is an index for measuring the periodic changes of the heart rate over time, which is an important signal characteristic reflecting the activity of the heart. For example, the computing device 100 divides the electrocardiogram data into units of 30 seconds to 1 minute and obtains a plurality of region of interest signals, and extracts the heart rate variability characteristics corresponding to each region of interest signal. This is to analyze the data within a range longer than the bit unit analysis time to more accurately evaluate the overall activity of the heart.

[0132] The computing device 100 derives the heart rate variability characteristics from each region of interest signal by frequency analysis or time region analysis. In a specific embodiment, the computing device 100 may apply a variety of analysis methods to extract the heart rate variability characteristic information corresponding to each region of interest signal. As a specific example, the computing device may extract the heart rate variability characteristic information by evaluating the high-frequency and low-frequency components based on spectrum analysis and statistical analysis of normal to normal intervals.

[0133] Spectral analysis evaluates heart rate variability in the frequency domain, with high-frequency components primarily reflecting the activity of the parasympathetic nervous system and low-frequency components representing mixed activity of the sympathetic and parasympathetic nervous systems. This is useful for understanding the state of equilibrium of the autonomic nervous system. Statistical analysis of normal-to-normal intervals quantifies the activity of the heart's autonomic nervous system by evaluating the variability of heart rate intervals in time regions. This can be an important indicator for evaluating heart rate irregularity, stress response, heart health status, etc.

[0134] In one embodiment, the heart rate variability characteristic information extracted in lead units may include successive root mean square difference (RMSSD), standard deviation of normal to normal intervals (SDNN), standard deviation of average normal to normal intervals (SDANN), etc. Specifically, the successive root mean square difference represents the square root of the mean square deviation of continuous heart rate intervals, which mainly reflects the influence of the parasympathetic nervous system. The standard deviation of normal to normal intervals represents the standard deviation of all normal heart rate intervals, which represents the overall variability of the autonomic nervous system. The standard deviation of average normal to normal intervals measures the standard deviation of the scoring heart rate intervals within a specified period to evaluate long-term heart rate variability. This indicator plays an important role in evaluating the activity of the autonomic nervous system of the heart and comprehensively analyzing the health status of the heart. That is, the heart rate variability characteristic information can respectively reflect various aspects of heart rate variability, which is used to evaluate the irregular shape of the heart rate, the activity of the autonomic nervous system, etc., and is the basis for predicting the risk level of heart disease.

[0135] According to an embodiment of the present invention, the method for generating a heart disease prediction model may include step S3000 of learning at least one network function based on a learning data set to generate a heart disease risk level prediction model.

[0136] Fig.11 The flowchart exemplarily shows a process of embedding multiple potential vectors according to an embodiment of the present invention and constructing a vector database according to the execution results.

[0137] Reference Fig.11 The steps of generating a heart disease risk prediction model may include: step S3100, generating an embedding model by using a masked self-supervised learning method using a first learning data set; embedding execution step S3200, using the embedding model to extract multiple latent vectors corresponding to multiple region of interest signals respectively, and the above latent vectors correspond to the above second learning data set; step S3300, using a clustering model to perform clustering on the multiple latent vectors; and step S3400, constructing a vector database (DB, DataBase) based on the above multiple latent vectors and information related to each cluster corresponding to each latent vector.

[0138] In more detail, in step S3100, the computing device 100 may generate an embedding model based on the first learning data set.

[0139] The embedding model may be a model trained to represent complex patterns of electrocardiogram data in a low-dimensional space. When a plurality of region of interest signals corresponding to the electrocardiogram data are input, the embedding model may generate a latent vector corresponding to each region of interest signal.

[0140] The latent vector may be a low-dimensional vector that compresses and represents important features of the electrocardiogram data. In order to make the electrocardiogram signal, which is high-dimensional data, easy to understand and analyze, it is transformed into a data representation having the main information form of the signal. The latent vector can evaluate the heart health status or provide useful information for detecting symptoms of a disease based on the pattern of the electrocardiogram data.

[0141] The detailed instructions for generating the embedding model will be referred to Fig.12 and Fig.13 To explain.

[0142] Fig.12 The following is a flowchart exemplarily showing a process of generating an embedding model according to an embodiment of the present invention. Fig.13 The diagram is an example diagram for explaining a self-reconstruction model and a process of generating an embedding model from a trained self-reconstruction model according to an embodiment of the present invention.

[0143] Reference Fig.12 , the step of generating an embedding model may include: step S3110, inputting data contained in a first learning data set into the self-reconstruction model to guide self-supervised learning, so that the self-reconstruction model generates an output similar to the above-mentioned input data; and step S3120, extracting an encoder from the self-reconstruction model that has completed the above-mentioned learning and generating the above-mentioned embedding model.

[0144] A self-reconstructing model may be a model that is designed to inherently learn important features of data during the process of learning through input and output.

[0145] In one embodiment, the self-reconstruction model is a neural network model that masks a portion of the input data and restores the masked portion, and may include an encoder (or a dimensionality reduction network function) and a decoder (or a dimensionality restoration network function). In one embodiment, the self-reconstruction model of the present invention may be a masked autoencoder (MAE) that learns important patterns in electrocardiogram data and accurately restores masked signals, but is not limited thereto.

[0146] The self-reconstruction model can effectively learn important features of the data by actively blocking a portion of the electrocardiogram signal (specifically, the signal of each region of interest) and reconstructing the overall signal based on the remaining portion.

[0147] Self-reconstruction models can learn by minimizing the restoration error between input and output, thereby effectively extracting and representing the important characteristics of the signal. In particular, when the data is complex or incomplete, self-reconstruction models have an advantage in understanding and utilizing the inherent structure of the data.

[0148] In more detail, Fig.13 As shown, after the self-reconstruction model 800 reduces the dimension of the input data, it learns important features by restoring it again. The self-reconstruction model may include an encoder 810 (or a dimension reduction network function) and a decoder 820 (or a dimension restoration network function). The self-reconstruction model 800 transforms high-dimensional data into a low-dimensional latent space through the encoder 810, compressing the core information of the data and reducing noise in the process. Afterwards, the decoder 820 is used to try to restore the compressed low-dimensional representation to the original high-dimensional space, thereby learning and reproducing the important patterns of the data.

[0149] According to an embodiment, the computing device 100 can guide the self-reconstruction model 800 to learn by a shielded learning method. In this method, a part of the input data is actively shielded so that the model can learn by restoring the shielded part. For example, when it is not shielded, the self-reconstruction model can focus on learning noise instead of QRS complexes, so it can focus on learning important features by shielding. As a result, the model will develop the ability to extract important pattern features from the remaining part of the input data and accurately restore the shielded information. The advantage is that it can work robustly even in noisy or incomplete parts of the data.

[0150] Furthermore, by restoring the masked parts, the context of the data can be understood and important features can be learned, which is extremely useful in analyzing and interpreting complex medical data such as electrocardiogram data. Since the mask is applied, the model can focus on important signals such as the QRS complex, which enables highly reliable data analysis for accurate diagnosis and evaluation of heart disease.

[0151] That is, in the present invention, in order to improve reliability in the process of analyzing and parsing electrocardiogram data, the self-reconstruction model is trained by a shielded learning method to focus on important cardiac signals. This allows the model to understand the context of the data, learn the main features, and thus accurately grasp and restore core signals such as the QRS complex.

[0152] The self-reconstruction model aims to reproduce the entire signal. After shielding the signal in a specific meaningful range, it will focus on restoring the shielded part. As a specific example, after shielding the meaningful part in the entire range of the electrocardiogram signal, about 3 QRS wave groups are included in the shielded area. The self-reconstruction model focuses on learning this area, and the loss function is only calculated for this area. In this way, the model learns the characteristics of the data with important signals as the center, thereby more accurately understanding and reproducing the entire context of the data.

[0153] As a result, it is possible to operate robustly even when there is noise or incompleteness in the ECG data, and to perform highly reliable data analysis required for early diagnosis and evaluation of heart disease. This access method can minimize the loss of important information in the analysis of ECG signals, support accurate evaluation of heart conditions, and thus significantly improve the analysis of medical data.

[0154] According to one embodiment, the self-reconstruction model 800 may include at least one hidden layer, and an odd number of hidden layers may be configured between the input and output layers. The number of nodes in each layer is reduced from the number of nodes in the input layer to the middle layer of the bottleneck layer (encoding), and is reduced from the bottleneck layer and symmetrically extended to the output layer (symmetrical with the input layer). In this case, Fig.13 The layers of the encoder 810 and the layers of the decoder 820 are shown to be symmetrical, but the present invention is not limited to this, and the nodes of the layers of the encoder 810 and the layers of the decoder 820 can be symmetrical, and can also be asymmetrical. The self-reconstruction model 800 can perform nonlinear dimensionality reduction. The number of input layers and output layers can correspond to the number of items of input data remaining after preprocessing the input data. In the structure of the self-reconstruction model 800, the number of nodes of the hidden layer included in the encoder 810 can be gradually reduced in the direction away from the input layer. When the number of nodes in the bottleneck layer (the layer with the least nodes between the encoder and the decoder) is too small, a sufficient amount of information cannot be transmitted, and therefore, a specific number or more (for example, more than half of the input layer, etc.) can be maintained.

[0155] In an embodiment, the computing device 100 trains a self-reconstruction model through a plurality of region of interest signals corresponding to a first learning data set, whereby the model learns important features and patterns of the electrocardiogram data. The trained self-reconstruction models are respectively applied to a plurality of electrocardiogram data, thereby extracting and analyzing the main features of the electrocardiogram signal in the process of masking a portion of the input data and restoring it.

[0156] In summary, the computing device 100 guides the self-reconstruction model to learn through the shielded learning method in order to ensure high reliability during the process of analyzing and parsing the electrocardiogram data. In this process, the self-reconstruction model focuses on the main signals such as the QRS complex to learn the inherent distribution of the heart rate. During the period when the self-reconstruction model reproduces the original information of the electrocardiogram data, it will effectively learn the important features (or characteristics) of the electrocardiogram data itself, such as the peak value of the QRS complex.

[0157] Afterwards, in step S3120, the encoder is extracted from the self-reconstruction model that has completed learning and an embedding model is generated. In the trained self-reconstruction model, the encoder part is trained (i.e., pre-learned by the first learning data set) to compress important features of high-dimensional data and remove noise, so that a low-dimensional latent vector can be output corresponding to the input data (e.g., multiple region of interest signals included in the second learning data set).

[0158] According to an embodiment, for the generated embedding model, since it is expressed by the characteristics of the compressed data, the similarity between the input data can be effectively reflected in the latent space. For example, corresponding to similar input data, the embedding model outputs a latent vector on a similar latent space. For example, a second latent vector of a second region of interest signal similar to the first region of interest signal is located on a latent space similar to the first latent vector corresponding to the first region of interest signal. That is, it is configured at a close position on the latent space.

[0159] In contrast, when the input data is different (i.e., the difference is large), the embedding model places the data at positions far apart from each other in the latent space. Thus, the similarities and differences between the data can be clearly distinguished based on the characteristic patterns of the electrocardiogram data, which can be used in the diagnosis and evaluation of heart diseases.

[0160] That is, the present invention uses the first learning data set to guide the autonomous learning of the self-reconstruction model, thereby completing the learned self-reconstruction model extraction encoder to form an embedding model.

[0161] After using the first learning data set to generate an embedding model, the computing device 100 can use the embedding model to perform embedding to extract multiple latent vectors corresponding to each region of interest signal, where the multiple region of interest signals correspond to the second learning data set.

[0162] More specifically, each region of interest signal included in the second learning data set is input to the embedding model through the computing device 100, and the embedding model compresses the main features of the corresponding signal to output a low-dimensional latent vector. The latent vector simply expresses the complex pattern of the region of interest signal and reflects the similarities and differences between the region of interest signals in the latent space. For example, signals of similar patterns in the region of interest signals of the electrocardiogram data included in the second learning data set are output as multiple latent vectors close to each other in the latent space.

[0163] Thereafter, in step S3300, the computing device 100 may use a clustering model to perform clustering on a plurality of latent vectors.

[0164] According to an embodiment, the computing device 100 may use a clustering model to perform clustering on the plurality of latent vectors.

[0165] As an example, the clustering model may be a model trained to perform clustering using a plurality of latent vectors extracted by the embedding model, which may be a K-Nearest Neighbors (KNN) model.

[0166] In one embodiment, the clustering model may perform clustering based on the similarity distance between a plurality of latent vectors corresponding to each ROI signal.

[0167] More specifically, in the clustering process, the distance or similarity between the potential vectors is calculated to determine the proximity or similarity of these vectors. In general, the distance is calculated using a measurement method such as Euclidean distance or cosine similarity.

[0168] For example, the distance between the latent vector A extracted from the signal of a specific region of interest and the latent vector B extracted from the signal of other regions of interest is calculated to evaluate the similarity of the two vectors. When the distance between A and B is close, the two vectors can represent electrocardiogram signals including the same or similar patterns. The clustering model groups multiple latent vectors into multiple clusters based on this distance information.

[0169] In the case of the KNN model, clustering is performed by finding the K nearest neighbor vectors from a particular vector and including the corresponding vector in the cluster to which the above multiple vectors belong. For example, when K=3, the three nearest neighbor vectors of the potential vector A are found and A is assigned to the cluster including the three neighbor vectors in the cluster to which the above neighbor vectors belong. This process is repeated for all potential vectors, and eventually, vectors that share similar features are grouped into the same cluster.

[0170] The process is useful for classifying multiple patterns in electrocardiogram data, analyzing how each pattern is associated with a specific type or risk level of heart disease. For example, when a particular cluster contains information related to irregularities in heart rate, the corresponding cluster can be used as an indicator of a disease such as arrhythmia.

[0171] As a specific example, when clustering is performed on each latent vector using a clustering model, four clusters may be formed. In the above clustering process, multiple vectors are clustered based on the similarity between the latent vectors.

[0172] When it is assumed that there are four clusters, namely, a first cluster, a second cluster, a third cluster, and a fourth cluster, these clusters may represent different electrocardiogram patterns. For example, the first cluster may include signals with a normal heart rate pattern, and the second cluster may include signals with specific abnormalities such as arrhythmia. The third cluster and the fourth cluster may respectively include signals reflecting other types of heart diseases or conditions. In the above description, it is described that four clusters (or groups) can be formed, but this is only an example, and the actual number of clusters may vary depending on the characteristics of the data and the settings of the clustering model.

[0173] The above clustering can automatically identify specific patterns or anomalies in ECG data and analyze the characteristics of each cluster to provide information useful for early diagnosis and management of heart disease. The clustering results provide important data for medical experts to comprehensively evaluate the patient's condition and establish a customized treatment plan.

[0174] Furthermore, in step S3400 , the computing device 100 may construct a vector database based on information related to the plurality of potential vectors and the clusters corresponding to the potential vectors.

[0175] The vector database of the present invention may be a NoSQL database server for storing and managing information related to embedded vectors and their cluster information and distances between vectors.

[0176] The vector database of the present invention is optimized for the analysis and management of multiple potential vectors extracted from multiple electrocardiogram data. Thus, characteristic patterns of electrocardiogram signals can be systematically stored and searched for use when needed.

[0177] As an example, the vector database of the present invention can support risk assessment and status monitoring based on electrocardiogram data. When receiving newly collected electrocardiogram data from a specific patient, the potential vector extracted from the corresponding data can be compared with the existing vectors stored in the vector database to evaluate the similarity. Through this similarity evaluation, it is possible to understand which cluster the patient's electrocardiogram data belongs to and which pattern of the previously observed pattern the corresponding cluster is similar to.

[0178] Furthermore, in an embodiment, the computing device 100 may extract representative vectors related to the multiple clusters corresponding to the clustering results, that is, principal vectors. Such principal vectors are vectors representing each cluster, which may represent the center of multiple potential vectors within the cluster.

[0179] More specifically, when the data stored in the vector database is huge, for example, when 300,000 pieces of ECG data need to be searched, it may take a long time. Usually, when new ECG data of a specific patient is input, it is segmented into multiple region of interest signals during the preprocessing process, and 80 to 120 embedding vectors are generated through the embedding process of each region of interest signal. For example, when individually comparing new ECG data with all vectors stored in the database, the amount of calculation is very large, which is inefficient. This will take a long time and is not suitable for situations where real-time analysis is required.

[0180] Therefore, in order to shorten the search time and improve efficiency, multiple latent vectors generated based on the patient's ECG data can be compared with the principal vector of each cluster to quickly identify similar clusters. In this process, the inner product between each embedded vector and the principal vector is calculated to evaluate the similarity, and based on this information, it is determined to which cluster the corresponding ECG data belongs.

[0181] Compared with comparing the entire data one by one, this method compares the main vectors representing each cluster, thereby quickly identifying related clusters. As a result, it plays an important role in the rapid analysis and effective management of electrocardiogram data, rapid evaluation of the risk level of heart disease, and real-time monitoring of patient status.

[0182] Furthermore, in the embodiment, the heart disease prediction information related to the electrocardiogram data of the prediction object can be generated through the following process.

[0183] In an embodiment, a method for generating heart disease prediction information related to electrocardiogram data of a prediction object may include: acquiring the electrocardiogram data of the prediction object; performing preprocessing on the electrocardiogram data of the prediction object; using an embedding model to generate multiple latent vectors corresponding to the preprocessed electrocardiogram data; respectively comparing each latent vector with a representative vector corresponding to each cluster to classify each latent vector into one of each of the above-mentioned clusters; and generating hierarchical information related to the risk level of heart disease based on the classification results of each latent vector into multiple clusters.

[0184] In an embodiment, the prediction subject may be an individual who receives an electrocardiogram for the purpose of evaluating and managing the risk of heart disease. This may be a person who receives a general health diagnosis, a person with a family history of heart disease, or a patient who has been previously diagnosed with heart disease or presents with related symptoms. The prediction subject may continuously evaluate the heart state through regular electrocardiogram monitoring, and may utilize the computing device 100 of the present invention for the early detection and prevention of heart disease.

[0185] In more detail, after acquiring the electrocardiogram data of the prediction object, the data is preprocessed to remove noise and purify the signal. Wherein, as shown in the above description, the preprocessing may include a process of removing unnecessary signal components and emphasizing important signal features in order to improve the quality of the electrocardiogram data, and may include a process of dividing each electrocardiogram signal into appropriate units (predefined window sizes) for processing. Typically, the predefined window size can be defined as an interval containing important information in the electrocardiogram data, for example, a length from 30 seconds to about 1 minute. In this process, the electrocardiogram data can be divided according to specified time intervals, and each interval can be analyzed independently.

[0186] The preprocessed electrocardiogram data is input to an embedding model (e.g., an encoder extracted from an autonomous regeneration model pre-learned by a first learning data set), thereby generating a latent vector (or embedding vector) in each data. The generated plurality of latent vectors are compared with existing vectors stored in a vector database, and each vector is classified into a specific cluster according to the similarity. Each cluster has a specific pattern reflecting the state of the heart, and therefore, the risk of heart disease is evaluated according to which cluster the latent vector belongs to.

[0187] Furthermore, the computing device 100 may generate hierarchical information related to the risk level of heart disease based on the classification results of the plurality of latent vectors being classified into the plurality of clusters. More specifically, for the electrocardiogram data of the prediction object, a plurality of region of interest signals are segmented and acquired according to a specific time unit during the preprocessing process, and as each region of interest signal is input into the embedding model, a plurality of latent vectors related to each region of interest signal are generated. In this case, prediction information related to the risk level of heart disease may be generated based on which cluster each latent vector is classified into and the number of classifications.

[0188] As a specific example, when the ECG data of the prediction object is composed of 10 30-second region of interest signals, the corresponding 10 region of interest signals can be transformed into 10 latent vectors through the embedding model, and the corresponding 10 vectors can be classified into one of the four predefined clusters. Among them, each predefined cluster represents a specific heart state, the first cluster is the cluster representing the highest risk level, and conversely, the fourth cluster is the cluster representing the lowest risk level.

[0189] The computing device 100 can predict the overall risk level based on how the 10 input signals are classified into clusters.

[0190] For example, when 8 signals belong to the first cluster and 2 signals belong to the second cluster, it can be evaluated that the patient has a high risk of heart disease. On the contrary, when 5 signals belong to the second cluster, 3 signals belong to the third cluster, and 2 signals belong to the fourth cluster, it can be evaluated as a medium risk. When 2 signals belong to the third cluster and 8 signals belong to the fourth cluster, it can be evaluated as a low risk. When all 10 signals belong to the fourth cluster, it can be evaluated as an extremely low risk of heart disease.

[0191] That is, the computing device 100 of the present invention can use the embedding model to analyze the region of interest signal of the electrocardiogram data to obtain multiple latent vectors, and each latent vector is evaluated based on the learning data set to which cluster of multiple clusters pre-defined by the clustering model it belongs, thereby being able to predict the patient's risk of heart disease in a hierarchical form. In other words, the computing device 100 compares the multiple latent vectors generated by the embedding model with the pre-defined clusters to grasp the cluster to which each vector belongs, and evaluates the patient's heart condition in multiple dimensions based on this information.

[0192] This stratified risk level prediction method improves the accuracy of electrocardiogram data analysis and helps to comprehensively understand the patient's heart condition. Therefore, the calculation device of the present invention systematically and clearly stratifies and provides the patient's heart disease risk level, thereby providing important information for medical experts to take appropriate preventive measures or establish a treatment plan. As a result, early detection and appropriate response to heart disease can be achieved, thereby contributing to the patient's health management.

[0193] Fig.14 The flowchart is an exemplary flowchart for explaining the process of generating a heart disease risk prediction model through integrated learning of multiple tree series models according to an embodiment of the present invention.

[0194] Reference Fig.14 The steps of generating a heart disease risk prediction model may include: step S3500, obtaining user meta-information corresponding to a plurality of electrocardiogram data respectively; step S3600, obtaining a plurality of latent vectors and cluster information corresponding to the plurality of latent vectors from a vector database; and step S3700, training a plurality of tree models based on heart rate variability characteristic information, user meta-information, a plurality of latent vectors and cluster information corresponding to the latent vectors, and performing ensemble learning to merge the outputs of the respective tree models to generate a heart disease risk prediction model.

[0195] More specifically, in step S3500, the step of obtaining user meta information may include the process of collecting personal health information such as the patient's age, gender, medical history, and living habits. This information normally reflects the health status and characteristics of each patient and has an important impact on the analysis of electrocardiogram data. In particular, this meta information is an important variable for more accurately evaluating the risk of heart disease. Thus, patient-customized prediction and evaluation can be achieved by taking into account the patient's personal health factors.

[0196] In step S3600, multiple latent vectors and cluster information corresponding to each vector are obtained from the vector database. The latent vector is a summary of the characteristics of the electrocardiogram data, and the cluster information represents the pattern of a specific heart state or disease to which the corresponding vector belongs. Thus, it is possible to confirm which cluster the patient's electrocardiogram data belongs to and evaluate the risk level of the heart disease represented by the cluster.

[0197] In step S3700, multiple tree models are trained based on the heart rate variability characteristic information, user meta information, latent vectors and cluster information. Each tree model uses a variety of tree-based machine learning algorithms such as decision trees, random forests, gradient boosting trees, etc. to predict the risk level of heart disease.

[0198] Among them, cluster information can be obtained from a vector database, and can include information about the cluster to which each vector belongs and information related to the distance between vectors. For example, cluster information provides information about which cluster each potential vector belongs to and distance information representing the similarity or difference between multiple vectors in a cluster. This information plays an important role in evaluating which cluster a specific ECG pattern belongs to and the similarity of the pattern to other ECG data. Thus, the characteristics of the ECG data can be analyzed and whether the corresponding data is related to a specific disease or condition can be accurately grasped.

[0199] More specifically, the process of training multiple tree series models based on heart rate variability characteristic information, user meta information, latent vectors and cluster information is as follows.

[0200] First, the heart rate variability characteristic information is the heart rate variability index extracted from the electrocardiogram data, which reflects the activity of the cardiac autonomic nervous system. The user meta-information is information indicating the patient's personal health status such as age, gender, medical history, and lifestyle. The latent vector is a value that represents the characteristics of the electrocardiogram data in a low-dimensional space, and the cluster information indicates which heart state or disease the vector is related to.

[0201] This variety of information is used as input variables for training multiple tree-based models. For example, the random forest model uses this input variable to form multiple decision trees, combining the results of each tree to provide a final prediction. Gradient boosting trees correct the errors of the previous model at each step through continuous model learning, and aim for more accurate predictions.

[0202] In this case, cluster information is particularly important. The distance information between the cluster to which each latent vector belongs and other clusters is used to better understand the characteristics of the ECG data and to assess the risk of heart disease. This information plays an important role in decision making in multiple tree models, which learn data patterns based on this information.

[0203] In the final ensemble process, the predictions of the individual tree models are combined to derive a comprehensive prediction. During this process, the weights of each model are adjusted to improve the reliability and accuracy of the final prediction. The ensemble process improves the shortcomings of the individual models and combines their strengths to provide more robust and consistent predictions.

[0204] As a result, this comprehensive learning and integration process can reflect the multidimensional characteristics of ECG data to accurately and precisely predict the risk level of heart disease.

[0205] After that, ensemble learning is performed to merge the outputs of each tree model, which combines the prediction results of the individual models to improve the accuracy and reliability of the final prediction. Ensemble learning can more accurately predict the risk of heart disease, thereby comprehensively evaluating the health status of patients and establishing early diagnosis and preventive treatment plans.

[0206] That is, in the present invention, the advantages of multiple models are combined through ensemble learning of multiple tree series models to improve the prediction performance higher than that of a single model, and the robustness to noise or changes in data can be improved. Thus, the model can provide consistent predictions for multiple patient data and reduce the uncertainty of the predictions.

[0207] In summary, the computing device of the present invention uses three neural network models to more accurately predict the risk level of heart disease. First, the embedding model transforms the electrocardiogram data into low-dimensional latent vectors to summarize important signal features. Second, the clustering model clusters multiple latent vectors to determine the cluster to which each vector belongs. Finally, the heart disease risk level prediction model combines heart rate variability characteristic information, user meta-information, latent vectors, and cluster information to predict the risk level of heart disease. Through this combined access method, the computing device can predict the risk level of heart disease more accurately and consistently.

[0208] In particular, in the final prediction process of the risk level of heart disease, multiple tree series models using four variables, such as heart rate variability characteristic information, user meta information, latent vector, and cluster information, are used to generate their own outputs. The above outputs are combined through ensemble learning to generate the final output, that is, the result of predicting the risk level of heart disease. Through this structure, the accuracy of predicting the risk level of heart disease can be improved, and a multi-dimensional and comprehensive evaluation can be achieved by combining multiple variables. This plays an important role in more accurately reflecting the individual health status of patients, allowing medical staff to clearly understand the risk level of patients and establish the optimal treatment plan.

[0209] Thus, the heart disease risk prediction model generated by completing the final learning can be stratified based on the patient's electrocardiogram data and provide prediction information related to the risk of future heart disease. The stratified information plays an important role in multi-dimensionally evaluating the patient's heart health status and establishing an individual treatment plan. This allows medical staff to more clearly understand the patient's risk level and take more precise treatment and prevention measures, thereby facilitating the early detection and management of heart disease.

[0210] Specifically, the system comprehensively analyzes the patient's meta-information of the electrocardiogram data to classify the possibility of heart disease into multiple risk levels. For example, the risk level is stratified into low, medium, high, etc., so that medical staff can clearly understand which patient to focus on. As a result, the possibility of high-risk group patients receiving early diagnosis and treatment increases, and low-risk group patients can reduce unnecessary examinations and carry out effective health management. As a result, it can play an important role in achieving effective allocation of medical resources, preventing complications caused by heart disease, and improving the quality of life of patients.

[0211] The embodiments described above can be implemented by hardware structural elements, software structural elements and / or a combination of hardware structural elements and software structural elements. For example, the devices, methods and structural elements described in multiple embodiments can be implemented using a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor or any other device that runs instructions and can respond to common computers or special purpose computers. The processor device can run an operating system (OS) and software applications executed on the above operating system. In addition, the processing device can access, store, operate, process and generate data in response to the operation of the software. For ease of understanding, there is a case where a processing device is described as being used, and ordinary technicians in the corresponding technical field can know that the processing device includes multiple processing elements and / or multiple types of processing elements. For example, the processing device may include multiple processors or a processor and a controller. In addition, it may also have other processing structures (processing configuration) such as parallel processors.

[0212] Software may include a computer program, code, instruction, or a combination of at least one of these, to configure a processing device in a manner to perform work as required or to issue instructions to a processing device independently or collectively. Software and / or data may be stored in any type of machinery, component, physical device, virtual device, computer storage medium or device for interpretation by a processing device or for providing instructions or data to a processing device. Software may be distributed on computer systems connected by a network and stored or run by a distributed method. Software and data may be stored in a computer readable recording medium.

[0213] The method of the embodiment is implemented and recorded in a computer-readable medium in the form of program instructions that can be run by a variety of computer units. Computer-readable media can store program instructions, data files, data structures, etc. alone or in combination, and the program instructions recorded in the medium can be specially designed or constructed for the embodiment or commonly used by ordinary technicians in the field of computer software. As an example of a computer-readable recording medium, it includes magnetic media (magnetic media) such as hard disks, floppy disks, and tapes, optical recording media (optical media) such as CD-ROMs and DVDs, magneto-optical media (magneto-optical media) such as optical floppy disks (floptical disks), and read-only memories (ROMs), random access memories (RAMs), flash memories, etc., which are specially constructed hardware devices in a manner of storing and running program instructions. As an example of program instructions, it includes mechanical codes formed by compilers and high-level language codes that are run by computers using interpreters, etc.

[0214] The hardware devices described above may be operated as one or more software modules in order to execute the tasks of the embodiments, and vice versa.

[0215] As described above, although the embodiments are described by limited drawings, a person skilled in the art can adopt various technical modifications and variations based on the embodiments. For example, the described techniques can be performed in a different order than the described methods and / or the described systems, structures, devices, circuits and other structural elements can be combined or combined in a different form than the described methods, and even if they are replaced or replaced by other structural elements or equivalent technical solutions, appropriate results can be achieved.

[0216] Therefore, these examples, other embodiments and contents equivalent to the scope of protection of the invention claims belong to the scope of protection of the invention claims described later.

Claims

1. A method for generating a heart disease prediction model, executed by at least one processor of a computing device, characterized in that: The steps include: acquiring a plurality of electrocardiogram data; and A model for predicting heart disease is generated based on the above-mentioned multiple electrocardiogram data.

2. The method for generating a heart disease prediction model according to claim 1, characterized in that: The steps for generating a model for predicting the above-mentioned heart disease include the following steps: Building a learning dataset based on the above multiple electrocardiogram data; and Based on the above learning data set, at least one network function is learned to generate a heart disease risk prediction model. The above learning data set includes a first learning data set and a second learning data set classified according to different learning purposes. The above-mentioned heart disease risk level prediction model stratifies based on the electrocardiogram data of the patient and provides prediction information related to the risk level of future heart disease.

3. The method for generating a heart disease prediction model according to claim 2, characterized in that: The first learning data set includes data for learning, and the learning corresponds to a process of transforming the characteristics of the electrocardiogram data into a characteristic space. The second learning data set includes calibration data related to the prediction of cardiac risk level.

4. The method for generating a heart disease prediction model according to claim 2, characterized in that: The step of constructing the above-mentioned learning data set includes the step of performing preprocessing on the above-mentioned multiple electrocardiogram data. The steps of performing the above preprocessing include the following steps: performing noise preprocessing on the multiple electrocardiogram data; Segmenting the plurality of electrocardiogram data subjected to the noise preprocessing into predefined window sizes to generate a plurality of region of interest signals; and Extracting heart rate variability characteristic information based on the leads of the above-mentioned multiple electrocardiogram data, The heart rate variability characteristic information includes index information related to the variability between heart rate intervals.

5. The method for generating a heart disease prediction model according to claim 4, characterized in that: The steps for generating the above-mentioned heart disease risk prediction model include: an embedding model generating step, generating an embedding model by masked self-supervised learning using the first learning dataset; An embedding execution step, using the embedding model to extract a plurality of latent vectors respectively corresponding to a plurality of region of interest signals, wherein the plurality of region of interest signals correspond to the second learning data set; A clustering execution step, performing clustering on the plurality of latent vectors using a clustering model; and The vector database construction step constructs a vector database based on the multiple potential vectors and information related to each cluster corresponding to each potential vector.

6. The method for generating a heart disease prediction model according to claim 5, characterized in that: The above embedding model generation step includes the following steps: Inputting data included in the first learning data set into the self-reconstruction model to guide self-supervised learning, so that the self-reconstruction model generates output similar to the input data; and Extract the encoder from the self-reconstructed model that completes the above learning and generate the above embedding model, The self-reconstruction model is a neural network model that masks a part of the input data and restores the masked part, including a dimensionality reduction network function and a dimensionality restoration network function.

7. The method for generating a heart disease prediction model according to claim 5, characterized in that: The clustering model performs clustering based on the similarity distance between multiple potential vectors corresponding to the multiple region of interest signals respectively. The clustering execution step includes the step of extracting a representative vector for each cluster corresponding to the clustering result.

8. The method for generating a heart disease prediction model according to claim 7, characterized in that: The above method comprises the following steps: Obtaining electrocardiogram data of the predicted object; Preprocessing the electrocardiogram data of the prediction object; Using the embedding model to generate a plurality of latent vectors corresponding to the electrocardiogram data preprocessed as described above; Comparing the plurality of latent vectors with representative vectors corresponding to the plurality of clusters respectively to classify the plurality of latent vectors into one of the plurality of clusters; as well as Hierarchical information on the risk level of heart disease is generated based on the classification results of the plurality of latent vectors into the plurality of clusters.

9. The method for generating a heart disease prediction model according to claim 5, characterized in that: The steps of generating the above-mentioned heart disease risk prediction model include the following steps: Acquire user meta information corresponding to the plurality of electrocardiogram data respectively; Acquire the plurality of latent vectors and cluster information corresponding to the plurality of latent vectors from the vector database; and Based on the above-mentioned heart rate variability characteristic information, the above-mentioned user meta-information, the above-mentioned multiple latent vectors and the cluster information corresponding to the above-mentioned latent vectors, multiple tree models are trained, and integrated learning is performed to merge the outputs of each tree model to generate the above-mentioned heart disease risk prediction model.

10. The method for generating a heart disease prediction model according to claim 1, characterized in that: The step of generating a model for predicting the above-mentioned heart disease also includes the following steps: Classifying the acquired multiple electrocardiogram data into a training data set and a verification data set; Acquire a plurality of regions of interest electrocardiogram data segmented into predefined window sizes from the plurality of electrocardiogram data classified into the training data set and the validation data set; The deep learning model is trained corresponding to the input of a plurality of first region-of-interest electrocardiogram data obtained from the plurality of electrocardiogram data included in the training data set into the deep learning model for predicting a patient's heart disease, so as to predict the heart disease category of each of the plurality of first region-of-interest electrocardiogram data; and A plurality of second region-of-interest electrocardiogram data obtained from a plurality of electrocardiogram data included in the validation data set is used to train the deep learning model to determine a threshold for distinguishing the categories of the heart diseases.

11. The method for generating a heart disease prediction model according to claim 10, characterized in that: The steps of determining the threshold values ​​for distinguishing the above cardiac disease categories include the following steps: collecting probability scores related to a plurality of region-of-interest electrocardiogram data separated from the same electrocardiogram data for the plurality of second region-of-interest electrocardiogram data; averaging the probability scores associated with the plurality of collected electrocardiogram data of the second region of interest to derive a final probability score; as well as The thresholds for distinguishing the above cardiac disease categories are fine-tuned based on the above derived final probability scores.

12. The method for generating a heart disease prediction model according to claim 10, characterized in that: The determined threshold is stored together with the weighted value of the deep learning model.

13. A server, characterized in that: include: a memory storing at least one instruction; and A processor, executing the at least one instruction stored in the memory, The processor executes the at least one instruction to perform the method according to claim 1.

14. A computer program, characterized in that The computer-readable recording medium is stored and combined with a computer as hardware to be able to execute the method according to claim 1.

Images