Training framework for multi-group electrocardiogram (MG-ECG) analysis

Abstract

A method of performing electrocardiogram (ECG) analysis by at least one processor, the method comprising: receiving ECG data from a plurality of leads; grouping the ECG data into a plurality of data groups; generating, according to each data group of the plurality of data groups, a feature vector using a respective machine learning model; and performing the ECG analysis using a plurality of feature vectors generated according to each data group of the plurality of data groups.

Description

[0001] Cross-reference to related applications

[0002] This application claims priority to U.S. Patent Application No. 16 / 556,491, filed August 30, 2019, with the United States Patent and Trademark Office, the disclosure of which is incorporated herein by reference in its entirety. Background Technology

[0003] Electrocardiogram (ECG) examinations are one of the most common medical procedures that help doctors diagnose many heart conditions, including atrial fibrillation, myocardial infarction, and acute coronary syndrome (ACS). Approximately 300 million ECGs are recorded annually. Traditional methods for ECG analysis tend to use digital signal processing algorithms, such as wavelet transform, to compute features from the ECG signal. However, such methods are not comprehensive, and therefore may be insufficient to distinguish between many types of arrhythmias. Recent methods employ deep neural networks, such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs), and can achieve good accuracy in multi-class classification tasks based on ECG signals. However, most existing work is only effective for single-lead ECG data, which does not provide comprehensive information about the heartbeat. Summary of the Invention

[0004] Current training frameworks for ECG data rely on single-lead signals and do not take into account the geometry of electrodes and leads. Some embodiments of this disclosure accept multi-lead signals (e.g., 12-lead) as a setup and apply multiple axis-specific feature extraction modules, followed by a finely tuned analysis model to achieve multiple objectives of ECG analysis, such as ECG monitoring and alarm, and computer-aided diagnostics.

[0005] According to some embodiments, a method for performing electrocardiogram (ECG) analysis via at least one processor is provided, comprising: receiving ECG data from multiple leads; grouping the ECG data into multiple data groups; generating feature vectors using a corresponding machine learning model for each of the multiple data groups; and performing ECG analysis using the multiple feature vectors generated for each of the multiple data groups.

[0006] According to some embodiments, an apparatus for performing electrocardiogram (ECG) analysis is provided, comprising at least one memory configured to store computer program code; at least one processor configured to access the computer program code and operate as instructed by the computer program code. The computer program code includes grouping code configured to cause the at least one processor to group ECG data received by the at least one processor and from a plurality of leads into a plurality of data groups; generating code configured to cause the at least one processor to generate, according to each data group of the plurality of data groups, a feature vector using a respective machine learning model stored in the at least one memory; and performing code configured to cause the at least one processor to perform the ECG analysis using the plurality of feature vectors generated according to each data group of the plurality of data groups.

[0007] According to some embodiments, a non-transitory computer-readable medium stores computer instructions that, when executed by at least one processor of an apparatus, cause the at least one processor to: receive electrocardiogram (ECG) data from a plurality of leads; group the ECG data into a plurality of data groups; generate, according to each data group of the plurality of data groups, a feature vector using a respective machine learning model stored in a memory; and perform an ECG analysis using the plurality of feature vectors generated according to each data group of the plurality of data groups. BRIEF DESCRIPTION OF DRAWINGS

[0008] Figure 1 is a diagram of an environment in which the methods, apparatuses, and systems described herein can be implemented according to embodiments.

[0009] Figure 2 is a diagram of example components of one or more devices of Figure 1

[0010] Figure 3 is a diagram of the main architecture of an MG-ECG analysis framework of embodiments.

[0011] Figure 4 is a flowchart of a method for performing ECG analysis according to embodiments.

[0012] Figure 5 is a diagram of an apparatus for performing ECG analysis according to embodiments. DETAILED DESCRIPTION

[0013] ​Some embodiments of the present disclosure are designed to implement multiple data analysis tasks by single-lead and multi-lead ECG data. In one embodiment, a multi-group electrocardiogram (MG-ECG) analysis framework uses a grouping module that groups data streams from multiple ECG leads into multiple groups based on different criteria. Two criteria can be, for example, (1) to form all leads into a single group; (2) to form each lead into a specific group. In one embodiment, a multi-axis feature extraction module employs multiple models for each pre-defined group from the grouping module, and data features from the multiple models are collected for a final analysis module. Thus, the MG-ECG analysis framework of embodiments can be widely applied to various types of analysis tasks. In analyzing, the MG-ECG analysis framework can also consider background knowledge, such as geometric properties, ontologies.

[0014] Figure 1 is a diagram of an environment 100 in which methods, apparatuses and systems described herein can be implemented in accordance with embodiments. As shown, the environment 100 can include a user device 110, a platform 120 and a network 130. Devices of the environment 100 can be interconnected by wired connections, wireless connections, or a combination of wired and wireless connections. Figure 1

[0015] The user device 110 includes one or more devices capable of receiving, generating, storing, processing, and / or providing information associated with the platform 120. For example, the user device 110 can include a computing device (e.g., a desktop computer, a laptop computer, a tablet computer, a handheld computer, a smart speaker, a server, etc.), a mobile phone (e.g., a smart phone, a wireless phone, etc.), a wearable device (e.g., a pair of smart glasses or a smart watch), or a similar device. In some implementations, the user device 110 can receive information from and / or send information to the platform 120.

[0016] The platform 120 includes one or more devices, as described elsewhere herein. In some implementations, the platform 120 can include a cloud server or a group of cloud servers. In some implementations, the platform 120 can be designed as a modular platform, such that software components can be swapped in or out depending on specific needs. Thus, the platform 120 can be easily and / or quickly reconfigured for different uses.

[0017] In some implementations, as shown, the platform 120 can be hosted in a cloud computing environment 122. It should be noted that while the implementations described herein describe the platform 120 as being hosted in the cloud computing environment 122, in some implementations, the platform 120 is not cloud-based (i.e., can be implemented outside of a cloud computing environment) or can be partially cloud-based.

[0018] ​The cloud computing environment 122 includes the environment of the hosting platform 120. The cloud computing environment 122 can provide computing, software, data access, storage, and other services without requiring end users (e.g., user equipment 110) to know the physical location and configuration of the systems and / or devices of the hosting platform 120. As shown in the figure, the cloud computing environment 122 may include a group of computing resources 124 (this group of computing resources is collectively referred to as "computing resources 124", and a single computing resource is referred to as "computing resource 124").

[0019] Computing resource 124 includes one or more personal computers, workstations, server devices, or other types of computing and / or communication devices. In some implementations, computing resource 124 may control platform 120. Cloud resources may include computing instances running in computing resource 124, storage devices provided in computing resource 124, data transmission devices provided by computing resource 124, etc. In some implementations, computing resource 124 may communicate with other computing resources 124 via wired connections, wireless connections, or a combination of wired and wireless connections.

[0020] Further as Figure 1 As shown, computing resources 124 include a set of cloud resources, such as one or more applications (“APP”) 124-1, one or more virtual machines (“VM”) 124-2, virtualized storage (“VS”) 124-3, one or more hypervisors (“HYP”) 124-4, etc.

[0021] Application 124-1 includes one or more software applications that can be provided to or accessed by user equipment 110 and / or platform 120. Application 124-1 eliminates the need to install and run software applications on user equipment 110. For example, application 124-1 may include software associated with platform 120 and / or any other software that can be provided through cloud computing environment 122. In some implementations, an application 124-1 may send information to / receive information from one or more other applications 124-1 via virtual machine 124-2.

[0022] Virtual machine 124-2 includes a software implementation of a machine (e.g., a computer) that runs programs like a physical machine. Depending on the extent to which virtual machine 124-2 corresponds to any actual machine and its purpose, virtual machine 124-2 can be a system virtual machine or a process virtual machine. A system virtual machine can provide a complete system platform supporting the operation of a full operating system (“OS”). A process virtual machine can run a single program and can support a single process. In some implementations, virtual machine 124-2 can run on behalf of a user (e.g., user device 110) and can manage the infrastructure of cloud computing environment 122, such as data management, synchronization, or long-duration data transfer.

[0023] Virtualized storage 124-3 includes one or more storage systems and / or one or more devices that utilize virtualization technology within the storage system or device of computing resource 124. In some implementations, the type of virtualization in the storage system environment may include block virtualization and file virtualization. Block virtualization may refer to the abstraction (or separation) of logical storage from physical storage, enabling access to the storage system without considering physical storage or heterogeneous architecture. Separation allows storage system administrators flexibility in how they manage end-user storage. File virtualization eliminates the dependency between data accessed at the file level and the location where the file is physically stored. This allows for optimization of storage usage, server consolidation, and / or interference-free file migration performance.

[0024] Hypervisor 124-4 provides hardware virtualization technology, which allows multiple operating systems (e.g., "guest operating systems") to run simultaneously on a host computer, such as computing resource 124. Hypervisor 124-4 can present a virtual operating platform to the guest operating systems and manage their operation. Multiple instances of each operating system can share virtualized hardware resources.

[0025] Network 130 includes one or more wired and / or wireless networks. For example, network 130 may include cellular networks (e.g., fifth-generation (5G) networks, long-term evolution (LTE) networks, third-generation (3G) networks, code division multiple access (CDMA) networks, etc.), public land mobile networks (PLMN), local area networks (LAN), wide area networks (WAN), metropolitan area networks (MAN), telephone networks (e.g., public switched telephone network (PSTN)), private networks, ad hoc networks, intranets, the Internet, fiber-optic networks, etc., and / or combinations of these or other types of networks.

[0026] Figure 1 The number and arrangement of devices and networks shown are provided as examples. In practice, additional devices and / or networks, fewer devices and / or networks, different devices and / or networks, or networks may exist. Figure 1 The devices and / or networks shown are arranged differently. Furthermore, Figure 1 The two or more devices shown can be implemented within a single device, or Figure 1 The single device shown can be implemented as multiple distributed devices. Alternatively, a group of devices in environment 100 (e.g., one or more devices) can perform one or more functions described as being performed by another group of devices in environment 100.

[0027] Figure 2 yes Figure 1FIG. 1 illustrates an example system including one or more devices. The devices of system 100 can communicate over one or more wired and / or wireless networks. Each device can comprise a user device 110 and / or a platform 120. As shown, system 100 can include a plurality of user devices 110 and a plurality of platforms 120. For example, system 100 can include one or more of the user devices 110 and one or more of the platforms 120. In some implementations, one or more of the user devices 110 can be configured to perform one or more functionalities of the platforms 120. In some implementations, one or more of the platforms 120 can be configured to perform one or more functionalities of the user devices 110. Figure 2 As shown, device 200 can include bus 210, processor 220, memory 230, storage component 240, input component 250, output component 260, and communication interface 270.

[0028] Bus 210 includes a component that permits communication among the components of device 200. Processor 220 is implemented in hardware, firmware, or a combination of hardware and software. Processor 220 is a central processing unit (CPU), a graphics processing unit (GPU), an accelerated processing unit (APU), a microprocessor, a microcontroller, a digital signal processor (DSP), a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), or another type of processing component. In some implementations, processor 220 includes one or more processors capable of being programmed to perform a function. Memory 230 includes a random access memory (RAM), a read only memory (ROM), and / or another type of dynamic or static storage device (e.g., a flash memory, a magnetic storage device, and / or an optical storage device) that stores information and / or instructions for use by processor 220.

[0029] Storage component 240 stores information and / or software related to the operation and use of device 200. For example, storage component 240 can include a hard disk (e.g., a magnetic disk, an optical disk, a magneto-optic disk, and / or a solid state disk), a compact disc (CD), a digital versatile disc (DVD), a floppy disk, a cartridge, a magnetic tape, and / or another type of non-transitory computer-readable medium, along with a corresponding drive.

[0030] Input component 250 includes a component that permits device 200 to receive information, such as through user input (e.g., a touch screen display, a keyboard, a keypad, a mouse, a button, a switch, and / or a microphone). Additionally, or alternatively, input component 250 can include a sensor for sensing information (e.g., a global positioning system (GPS) component, an accelerometer, a gyroscope, and / or an actuator). Output component 260 includes a component that provides output information from device 200 (e.g., a display, a speaker, and / or one or more light-emitting diodes (LEDs)).

[0031] The communication interface 270 includes a transceiver-like component (e.g., a transceiver and / or a separate receiver and transmitter) that enables the device 200 to communicate with other devices, such as via a wired connection, a wireless connection, or a combination of both. The communication interface 270 can allow the device 200 to receive information from another device and / or provide information to another device. For example, the communication interface 270 can include an Ethernet interface, an optical interface, a coaxial interface, an infrared interface, a radio frequency (RF) interface, a universal serial bus (USB) interface, a Wi-Fi interface, a cellular network interface, and / or the like.

[0032] The device 200 can perform one or more processes described herein. The device 200 can perform these processes in response to processor 220 executing software instructions stored by a non-transitory computer-readable medium, such as memory 230 and / or storage component 240. In this context, a computer-readable medium is defined as a non-transitory memory device. A memory device includes memory space within a single physical storage device or memory space spread across multiple physical storage devices.

[0033] The software instructions can be read into the memory 230 and / or storage component 240 from another computer-readable medium or from another device via the communication interface 270. When executed, the software instructions stored in the memory 230 and / or storage component 240 can cause the processor 220 to perform one or more processes described herein. Additionally, or alternatively, hardwired circuitry can be used in place of or in combination with software instructions to perform one or more processes described herein. Thus, implementations described herein are not limited to any specific combination of hardware circuitry and software.

[0034] Figure 2 The number and arrangement of components shown in FIG. 8 are provided as an example. In practice, device 800 can include additional components, fewer components, different components, or differently arranged components than those shown in FIG. 8. Additionally, or alternatively, a set of components (e.g., one or more components) of device 800 can perform one or more functions described as being performed by another set of components of device 800. Figure 2 The number and arrangement of components shown in FIG. 8 are provided as an example. In practice, device 800 can include additional components, fewer components, different components, or differently arranged components than those shown in FIG. 8. Additionally, or alternatively, a set of components (e.g., one or more components) of device 800 can perform one or more functions described as being performed by another set of components of device 800.

[0035] Figure 3 A main architecture of the MG-ECG analysis framework 300 of an embodiment is shown.

[0036] The framework 300 can include three main modules: (1) a grouping module 320, (2) one or more feature extraction modules 340, (3) an ECG analysis module 360. The grouping module 320 can group the multi-lead ECG data into multiple groups. In one embodiment, a pool of grouping criteria 325 can provide different strategies that can help the grouping module 320 determine group assignments. Each feature extraction module can extract a group of ECG data with the same set of parameters. The ECG analysis module 360 can accomplish specific tasks such as aggregation, classification, prediction, etc., and then achieve the ultimate goal of the framework 300.

[0037] In one embodiment, the grouping module 320 can receive a plurality of ECG data 310, including, for example, ECG data 310-1, ECG data 310-2, ECG data 310-3,... ECG data 310-n, where each ECG data is from a corresponding lead. The grouping module 320 can receive the plurality of ECG data 310 from, for example, an ECG device that includes a plurality of electrodes and a plurality of leads for an ECG examination. The grouping module 320 can group the plurality of ECG data 310 into a plurality of groups 330. In one embodiment, the grouping module 320 can group the plurality of ECG data 310 into a plurality of groups 330-1, 330-2,... 330-n. In each group 330, the data shares certain types of features or has common characteristics. The grouping module 320 can use the pool of grouping criteria 325 or a set of built-in rules designed to decide how the ECG data 310 is grouped. For example, assuming the geometric characteristics of each lead of the ECG data 310 are used as criteria for grouping, by assuming that similar directions of ECG signals provide similar features, such that the ECG data 300 from different leads can be grouped according to the way the electrodes are placed. The grouping module 320, particularly the pool of grouping criteria 325, can require domain knowledge and expert opinions to create rules and criteria for grouping. A number of examples of criteria include, but are not limited to, electrode placement that defines the axis of ECG leads, contiguity of leads, and random grouping strategies. A pre-processing procedure for the ECG data of each group 330 can be performed in the grouping module 320 to generate normalized and clean data observations.

[0038] In one embodiment, the framework 300 can include one or more feature extraction modules 340. For example, the framework 300 can include feature extraction modules 340-1, 340-2,... 340-n. The grouped ECG data generally share similar feature sets and common characteristics. A group-specific feature extraction module 340 can be designed for a respective one of the data groups 330. For example, a feature extraction module 340-1 can be designed for the data group 330-1. In one embodiment, the feature extraction modules 340 can have similar structures such that the extracted features are comparable. The feature extraction modules 340 can accept the pre-processed ECG data 330 as input and generate feature vectors 350 as output. For example, the feature vectors 350 can include feature vectors 350-1, 350-2,... 350-n. Each feature vector 350 can be, for example, a set of features. In one embodiment, each feature extraction module can output a respective one (or more) of the feature vectors 350. For example, the feature extraction module 340-1 can output the feature vector 350-1. In terms of models, the feature extraction modules 340 can use any machine learning method, including, for example, support vector machines (SVM), random forests (RF), or deep learning models, such as CNNs and RNNs. Each feature extraction module 340 can use a respective one or multiple models. The parameters of each feature extraction module 340 can be trained individually to obtain a group-specific extraction method.

[0039] The ECG analysis module 360 can accept the extracted features and produce final results 370, such as classification results, outlier alerts, predictive diagnoses including, for example, signs of pathology. The results 370 can be used for or include, for example, monitoring and alarming, computer-aided diagnosis, and detection of signs of pathology. The ECG analysis module 360 can be provided with a pool of task-specific modules 365. The pool of task-specific modules 365 can be a collection of different models for various ECG-related tasks. For example, such models can include several statistical process control algorithms for ECG monitoring and alarming, several predictive models and classifier models for computer-aided diagnosis, and some statistical tools for common pathology state calculations. Depending on the goal of using the framework 300, the ECG analysis module 360 can deploy appropriate tools from the pool of task-specific modules 365 to complete the end-to-end framework and achieve the final goal.

[0040] The at least one processor can be configured to group the modules 320, the feature extraction modules 340, and the ECG analysis module 360 such that the at least one processor performs the functions of the respective modules. For example, one or more of the at least one processor can together perform the functions of one or more of the modules, or a respective one or more of the at least one processor can perform the functions of each of the modules.

[0041] Figure 4A method executed by at least one processor according to an embodiment of the present disclosure is shown.

[0042] In this embodiment, at least one processor may receive multiple ECG data sets 310 (410). The at least one processor may then group the multiple ECG data sets 310 into multiple groups 330 (420) by executing the functions of the grouping module 320. For example, when grouping the multiple ECG data sets 310, the at least one processor may use a pool of supporting grouping criteria 325 stored in memory. The at least one processor may then generate a feature vector 350 (430) from each group 330 by executing the functions of each feature extraction module 340. The at least one processor may then analyze the feature vectors 350 and produce a result 370 (440) by executing the functions of the ECG analysis module 360. For example, when analyzing the feature vectors 350, the at least one processor may use a task-specific module pool 365 stored in one or more memories.

[0043] Figure 5 This is a diagram of an apparatus 500 for performing ECG analysis according to an embodiment. Figure 5 As shown, the apparatus 500 includes data grouping code 510, feature generation code 520, and execution code 540. The apparatus 500 may include at least one processor to execute one or more of these codes.

[0044] Data grouping code 510 can be configured to cause at least one processor to group multiple ECG data 310 into multiple groups 330 by causing at least one processor to perform the functions of grouping module 320. Feature generation code 520 may include multiple feature generation codes. Each of the multiple feature generation codes can be configured to cause at least one processor to generate a feature vector 350 from a corresponding group among the multiple groups 330 by causing at least one processor to perform the functions of corresponding feature extraction module 340. Execution code 530 can be configured to cause at least one processor to analyze the feature vector 350 and produce a result 370 by causing at least one processor to perform the functions of ECG analysis module 360.

[0045] Embodiments of this disclosure, for example Figure 1 The illustrated embodiments may be end-to-end frameworks. Embodiments of this disclosure are improvements to existing methods for ECG analysis models because, for example, embodiments of this disclosure can consider the similarities and differences between different leads by defining an extraction module 340 with the same structure, and include a grouping module 320 that divides multiple leads into multiple groups 330.

[0046] Furthermore, embodiments of the present disclosure can accept 12-lead ECG data, which can provide comprehensive information to achieve better performance of the model. In addition, the grouping module of the embodiments can reduce the size of the entire model, so that the model can be applied to smaller devices such as laptops and mobile devices.

[0047] In embodiments of the present disclosure, the grouping criteria of the grouping module 320 can be loaded as built-in rules instead of a standard pool. This method is more effective for specific tasks. In embodiments of the present disclosure, the grouping results of the grouping module 320 can be mutually exclusive groups or overlapping groups.

[0048] In embodiments of the present disclosure, the feature extraction module 340 can apply different types of models for different groups. This method is a better option for the overlapping grouping strategy. In embodiments of the present disclosure, similar models of the feature extraction module 340 can share a subset of parameters to consider the similarity between groups.

[0049] Embodiments of the present disclosure can provide a framework designed as an end-to-end process, so that the optimization and modification of the entire framework are carried out simultaneously. Alternatively, the process can be a step-by-step training process, in which the feature extraction module 340 can be trained separately, for example, using an encoder and decoder structure.

[0050] Embodiments of the present disclosure are not limited to application to ECG analysis. That is, embodiments of the present disclosure can be extended to other applications with multiple input sources.

[0051] The foregoing disclosure provides illustration and description, but is not intended to be exhaustive or to limit the embodiments to the precise form disclosed. Modifications and variations can be possible in light of the above disclosure or can be acquired from practice of the embodiments.

[0052] As used herein, the term "component" is intended to be broadly interpreted to encompass hardware, firmware, or a combination of hardware and software.

[0053] It will be apparent that systems and / or methods, described herein, can be implemented in different forms of hardware, firmware, or a combination of hardware and software. The actual specialized control hardware or software code used to implement these systems and / or methods is not limiting of the embodiments. Thus, the operation and behavior of the systems and / or methods were described herein without reference to specific software code — it being understood that software and hardware can be designed to implement the systems and / or methods based on the description herein.

[0054] Even if a combination is recited in the claims and / or described in the specification, such a combination is not intended to limit the disclosure to a combination of features unless explicitly described as such. Rather, the disclosure can be practiced with a wide variety of combinations of features, even if not explicitly described in the specification. Although each dependent claim listed below can only directly depend on one claim, the disclosure of possible implementations includes each dependent claim in combination with any other claim in the set of claims.

[0055] No element, act, or instruction used herein should be construed as critical or essential unless explicitly described as such. Also, as used herein, the articles “a” and “an” are intended to include one or more items, and can be used interchangeably with “one or more.” Furthermore, as used herein, the term “set” is intended to include one or more items (e.g., related items, unrelated items, a combination of related and unrelated items, etc.), and can be used interchangeably with “one or more.” Where only one item is intended, the term “one” or similar language is used. Also, as used herein, the terms “has,” “have,” “having,” or the like are intended to be open-ended terms. Further, the phrase “based on” is intended to mean “based, at least in part, on” unless explicitly stated otherwise.

Claims

1. A method for performing electrocardiogram (ECG) analysis via at least one processor, the method comprising: Receive multiple ECG data points, wherein each ECG data point comes from multiple leads; The shared features or common characteristics of the multiple ECG data are determined, and the multiple ECG data are grouped according to the shared features or common characteristics, and the multiple ECG data are grouped into multiple data groups. The grouping strategy for grouping the multiple ECG data includes a continuous lead grouping strategy and a random grouping strategy. For each of the plurality of data groups, a feature vector is generated using a corresponding machine learning model, wherein the corresponding machine learning model for each of the plurality of data groups is trained independently of each other to obtain a machine learning model specific to each data group; and ECG analysis is performed using multiple feature vectors generated based on each of the multiple data groups.

2. The method according to claim 1, characterized in that, Data within one of the multiple data groups shares the same features or common characteristics.

3. The method according to claim 1, characterized in that, The grouping includes: grouping the ECG data according to the shared features or common characteristics, such that data from the first lead of the plurality of leads and data from the second lead of the plurality of leads are grouped in the same data group within the plurality of data groups.

4. The method according to claim 1, characterized in that, The grouping includes determining the plurality of data groups by using a pool of grouping criteria or rules stored in memory.

5. The method according to claim 4, characterized in that, The grouping includes determining the plurality of data groups by using the grouping standard pool, the grouping standard pool including the geometric characteristics of each of the plurality of leads, such that the ECG data from the plurality of leads are grouped according to the placement of the electrodes during the ECG examination.

6. The method according to claim 1, characterized in that, The corresponding machine learning model for each of the plurality of data sets has the same subset of parameters.

7. The method according to any one of claims 1 to 6, characterized in that, The ECG analysis includes: Selecting a model from multiple models stored in memory; and The model generates an output using the plurality of feature vectors generated based on each of the plurality of data groups.

8. The method according to any one of claims 2 to 5, characterized in that, The corresponding machine learning model for each of the plurality of data sets has the same subset of parameters.

9. An apparatus for performing electrocardiogram (ECG) analysis, the apparatus comprising: At least one memory configured to store computer program code; At least one processor is configured to access and operate according to the instructions of the computer program code, the computer program code comprising: Grouping codes are configured to cause the at least one processor to receive multiple ECG data, wherein each ECG data comes from multiple leads; The code generates code configured to cause the at least one processor to determine shared features or common characteristics of the plurality of ECG data, and to group the plurality of ECG data into multiple data groups based on the shared features or common characteristics. For each of the multiple data groups, a feature vector is generated using a corresponding machine learning model stored in the at least one memory. The grouping strategy for the plurality of ECG data includes a continuous lead grouping strategy and a random grouping strategy. The corresponding machine learning model for each of the multiple data groups is trained independently to obtain a machine learning model specific to each data group. The execution code is configured such that the at least one processor performs ECG analysis using multiple feature vectors generated based on each of the multiple data groups.

10. The device according to claim 9, characterized in that, The grouping code is configured to cause the at least one processor to group the ECG data based on shared features or common characteristics, such that data within one of the multiple data groups share the same features or common characteristics.

11. The device according to claim 10, characterized in that, The grouping code is configured such that the at least one processor groups the ECG data according to the shared feature or the common characteristic, such that data from the first lead of the plurality of leads and data from the second lead of the plurality of leads are grouped within the same data group of the plurality of data groups.

12. The device according to claim 9, characterized in that, The grouping code is configured to cause the at least one processor to determine the plurality of data groups by using a pool of grouping criteria or rules stored in the at least one memory.

13. The device according to claim 12, characterized in that, The grouping code is configured such that the at least one processor determines the plurality of data groups by using the grouping standard pool, the grouping standard pool including the geometric characteristics of each of the plurality of leads, such that the ECG data from the plurality of leads are grouped according to the placement of the electrodes during the ECG examination.

14. The device according to claim 9, characterized in that, The corresponding machine learning models for each of the plurality of data sets are trained independently of each other.

15. The device according to claim 9, characterized in that, The corresponding machine learning model for each of the plurality of data sets has the same subset of parameters.

16. The device according to any one of claims 9 to 15, characterized in that, The executable code includes: Model selection code, configured to cause the at least one processor to select a model from a plurality of models stored in the at least one memory; and Output generation code, configured to cause the at least one processor to generate output using the model and the plurality of feature vectors generated based on each of the plurality of data groups.

17. The device according to any one of claims 10 to 13, characterized in that, The corresponding machine learning models for each of the plurality of data sets are trained independently of each other.

18. The device according to any one of claims 10 to 13, characterized in that, The corresponding machine learning model for each of the plurality of data sets has the same subset of parameters.

19. A non-transitory computer-readable medium storing computer instructions that, when executed by at least one processor of a device, cause the at least one processor to perform the method according to any one of claims 1 to 8.

20. An apparatus for performing electrocardiogram (ECG) analysis, the apparatus comprising: The receiving module is configured to receive multiple ECG data, wherein each ECG data comes from multiple leads; The grouping module is configured to determine the shared features or common characteristics of the multiple ECG data, and to group the multiple ECG data into multiple data groups according to the shared features or common characteristics. The grouping strategy for grouping the multiple ECG data includes a lead continuity grouping strategy and a random grouping strategy. A generation module is configured to generate feature vectors for each of the plurality of data groups using a corresponding machine learning model, wherein the corresponding machine learning models for each of the plurality of data groups are trained independently of each other to obtain a machine learning model specific to each data group; and The analysis module is configured to perform ECG analysis using multiple feature vectors generated based on each of the multiple data groups.

Images