Electrocardio multi-task analysis method, device and equipment and storage medium

Abstract

The invention relates to the technical field of electrocardiosignal processing, in particular to an electrocardiosignal multi-task analysis method and device, equipment and a storage medium. The electrocardiosignals are input into a feature extraction network for feature extraction, and electrocardiosignal features are obtained; inputting the electrocardiosignal features into a deep learning network corresponding to cardiac beat positioning, datum point detection and cardiac beat classification for calculation to obtain scores and uncertainty corresponding to cardiac beat positioning, datum point detection and cardiac beat classification; and performing calculation according to the heart beat positioning score, the reference point detection score, the heart beat classification score, the heart beat positioning uncertainty, the reference point detection uncertainty and the heart beat classification uncertainty to obtain a heart beat positioning result, a reference point detection result and a heart beat classification result. According to the method, the technical problem that when an existing electrocardiogram multi-task analysis method is deployed in a platform with limited computing resources, the calculated amount and accuracy are difficult to balance can be solved.

Description

Technical Field

[0001] The present application relates to the technical field of electrocardiogram signal processing, and in particular to an electrocardiogram multi-task analysis method, device, equipment and storage medium. Background Art

[0002] As an important means of checking cardiovascular status, electrocardiogram (ECG) is increasingly used in wearable devices (such as smart watches). Compared with professional medical equipment such as electrocardiographs, monitors, and holters, wearable devices are limited by power consumption, size, cost, and complex usage scenarios. Their computing resources are quite limited, and the collected ECG signals are often interfered by noise. Heartbeat positioning, reference point detection, and heartbeat classification are three important tasks of ECG analysis. How to accurately, efficiently, and reliably complete these tasks on wearable devices is a challenge.

[0003] Traditional ECG analysis methods based on rules or machine learning have low computational complexity, but are limited by the model's expressiveness, and their accuracy is relatively low compared to deep learning methods. Deep learning methods have high accuracy, but most deep learning methods use multiple single-task networks, that is, use multiple corresponding independent feature extraction networks to analyze ECG multi-tasks, and the disadvantage is that the computational complexity is too large. Therefore, some methods use a single single-task network, that is, use a feature extraction network to perform ECG multi-task analysis, sacrificing some accuracy to save a lot of computing resources. However, due to the uneven quality of the input ECG signal, it may not be used for all analysis tasks. For example, the signal quality of a certain section of ECG is poor, and it can only be used for heartbeat positioning, not for benchmark detection and heartbeat classification. If the signal of poor quality ECG is forcibly used for analysis, the analysis results will be unreliable.

[0004] Therefore, it is difficult to balance the computational size and accuracy when the existing ECG multi-task analysis methods are deployed in platforms with limited computing resources. Summary of the invention

[0005] The main purpose of this application is to provide an ECG multi-task analysis method, device, equipment and storage medium, aiming to solve the technical problem that it is difficult to balance the amount of calculation and accuracy when the existing ECG multi-task analysis method is deployed in a platform with limited computing resources.

[0006] In order to achieve the above-mentioned invention object, the present application proposes a multi-task ECG analysis method, which comprises:

[0007] Obtain ECG signals;

[0008] Inputting the ECG signal into a feature extraction network to extract features and obtain ECG signal features;

[0009] Inputting the ECG signal features into the deep learning networks corresponding to heartbeat positioning, reference point detection and heartbeat classification respectively for calculation, and obtaining the heartbeat positioning score, reference point detection score, heartbeat classification score, heartbeat positioning uncertainty, reference point detection uncertainty and heartbeat classification uncertainty;

[0010] Calculations are performed based on the heartbeat positioning score, the reference point detection score, the heartbeat classification score, the heartbeat positioning uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty to obtain a heartbeat positioning result, a reference point detection result and a heartbeat classification result.

[0011] Furthermore, the step of inputting the ECG signal features into the deep learning networks corresponding to heartbeat positioning, reference point detection and heartbeat classification for calculation to obtain the heartbeat positioning score, reference point detection score, heartbeat classification score, heartbeat positioning uncertainty, reference point detection uncertainty and heartbeat classification uncertainty includes:

[0012] Inputting the electrocardiogram signal characteristics into a heartbeat localization network for calculation to obtain a heartbeat localization score and a heartbeat localization uncertainty;

[0013] Inputting the electrocardiogram signal characteristics into a reference point detection network for calculation to obtain a reference point detection score and a reference point detection uncertainty;

[0014] The electrocardiogram signal features are input into a heartbeat classification network for calculation to obtain a heartbeat classification score and a heartbeat classification uncertainty.

[0015] Further, the step of calculating according to the heartbeat positioning score, the reference point detection score, the heartbeat classification score, the heartbeat positioning uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty to obtain the heartbeat positioning result, the reference point detection result and the heartbeat classification result includes:

[0016] Calculating according to the heartbeat localization score and the heartbeat localization uncertainty to obtain a heartbeat localization result;

[0017] Calculating according to the reference point detection score, the heartbeat positioning uncertainty and the reference point detection uncertainty to obtain a reference point detection result;

[0018] A heartbeat classification result is obtained by calculating the heartbeat classification score, the heartbeat positioning uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty.

[0019] Furthermore, the step of calculating according to the heartbeat localization score and the heartbeat localization uncertainty to obtain the heartbeat localization result includes:

[0020] The heartbeat localization result is calculated by the following formula:

[0021]

[0022] Among them, p loc is the heartbeat positioning result, s loc is the heartbeat positioning score, u loc is the heartbeat positioning uncertainty, and the softmax function is a normalized exponential function.

[0023] Furthermore, the step of calculating the reference point detection result according to the reference point detection score, the heartbeat positioning uncertainty and the reference point detection uncertainty includes:

[0024] The benchmark detection result is calculated using the following formula:

[0025]

[0026] Among them, p del is the benchmark detection result, s del is the benchmark detection score, u loc is the heart beat location uncertainty, u del is the uncertainty of benchmark detection, and the softmax function is a normalized exponential function.

[0027] Furthermore, the step of calculating according to the heartbeat classification score and the heartbeat positioning uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty to obtain a heartbeat classification result includes:

[0028] The heart beat classification result is calculated by the following formula:

[0029]

[0030] Among them, p cls is the heart beat classification result, s cls is the heartbeat classification score, u cls is the heart beat classification uncertainty, u loc is the heart beat location uncertainty, u del is the uncertainty of benchmark detection, and the softmax function is a normalized exponential function.

[0031] The present application also provides an electrocardiogram multi-task analysis device, the device comprising:

[0032] An acquisition module, used for acquiring electrocardiogram signals;

[0033] A feature extraction module, used for inputting the ECG signal into a feature extraction network to extract features and obtain ECG signal features;

[0034] A multi-task network computing module, used for inputting the ECG signal features into the deep learning networks corresponding to heartbeat positioning, reference point detection and heartbeat classification for computing, and obtaining a heartbeat positioning score, a reference point detection score, a heartbeat classification score, a heartbeat positioning uncertainty, a reference point detection uncertainty and a heartbeat classification uncertainty;

[0035] The ECG analysis module is used to calculate according to the heartbeat positioning score, the reference point detection score, the heartbeat classification score, the heartbeat positioning uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty to obtain the heartbeat positioning result, the reference point detection result and the heartbeat classification result.

[0036] Furthermore, the ECG multi-task analysis device comprises:

[0037] The multi-task network calculation submodule is used to input the ECG signal features into the heartbeat localization network for calculation to obtain the heartbeat localization score and the heartbeat localization uncertainty; input the ECG signal features into the reference point detection network for calculation to obtain the reference point detection score and the reference point detection uncertainty; input the ECG signal features into the heartbeat classification network for calculation to obtain the heartbeat classification score and the heartbeat classification uncertainty.

[0038] The present application also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0039] The present application also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of any of the above methods when executed by a processor.

[0040] The present application provides a multi-task ECG analysis method, which extracts ECG signal features through a feature extraction network, and then inputs the features into a deep learning network corresponding to multiple ECG analysis tasks, respectively, to obtain scores and uncertainties corresponding to multiple ECG analysis tasks, which can reduce the amount of calculation while improving the analysis accuracy. According to the sensitivity of different tasks to noise, when calculating the analysis results of each task, the uncertainty of each task is adjusted, so that tasks with high noise sensitivity are less affected by noise samples, and tasks with low noise sensitivity are prevented from losing too many samples, further improving the accuracy of ECG analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flowchart of a multi-task ECG analysis method according to an embodiment of the present application;

[0042] Figure 2This is a schematic block diagram of the structure of an ECG multi-task analysis device according to an embodiment of the present application;

[0043] Figure 3 A schematic block diagram of the structure of a computer device according to an embodiment of the present application.

[0044] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0046] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "a", "an", "above", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of features, integers, steps, operations, elements, modules, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, modules, components, and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be an intermediate element. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any module and all combinations of one or more associated listed items.

[0047] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as generally understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with the meanings in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined as herein.

[0048] Reference Figure 1 The embodiment of the present invention provides a multi-task ECG analysis method, including steps S1-S4, specifically:

[0049] S1, obtaining ECG signals;

[0050] S2, inputting the ECG signal into a feature extraction network to extract features and obtain ECG signal features;

[0051] The ECG signal features such as amplitude, peak value, waveform index and power spectrum density are extracted through the feature extraction network. The extracted ECG signal features are used to input the deep learning networks corresponding to heartbeat positioning, reference point detection and heartbeat classification for calculation. The steps of inputting the same ECG signal features into multiple single-task networks reduce the steps of feature extraction of multiple single-task networks, thereby reducing the amount of calculation. Multiple single-task networks can independently and parallelly calculate the scores and uncertainties of the corresponding ECG tasks, improving the efficiency of ECG multi-task analysis.

[0052] S3, inputting the ECG signal features into the deep learning networks corresponding to heartbeat positioning, reference point detection and heartbeat classification for calculation, and obtaining the heartbeat positioning score, reference point detection score, heartbeat classification score, heartbeat positioning uncertainty, reference point detection uncertainty and heartbeat classification uncertainty;

[0053] S4. Calculate according to the heartbeat positioning score, the reference point detection score, the heartbeat classification score, the heartbeat positioning uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty to obtain a heartbeat positioning result, a reference point detection result and a heartbeat classification result.

[0054] The ECG signal features are respectively input into the deep learning networks corresponding to heartbeat positioning, fiducial point detection and heartbeat classification to obtain the scores and uncertainties corresponding to heartbeat positioning, fiducial point detection and heartbeat classification. By setting up multiple single-task networks to process the corresponding ECG analysis tasks respectively, and multiple single-task networks can be independently calculated in parallel, the accuracy and efficiency of ECG analysis can be improved. According to the sensitivity of each task of heartbeat positioning, fiducial point detection and heartbeat classification to noise, the uncertainty of each task is increased or decreased when calculating the analysis results, so that the output uncertainty size is more in line with the prior cognition, making the analysis results of each task more accurate, and further improving the accuracy of ECG analysis.

[0055] As described above, the features of the ECG signal are extracted through the feature extraction network, and then the features are respectively input into the deep learning networks corresponding to multiple ECG analysis tasks to obtain the scores and uncertainties corresponding to the multiple ECG analysis tasks, which can reduce the amount of calculation while improving the analysis accuracy. According to the sensitivity of different tasks to noise, when calculating the analysis results of each task, the uncertainty of each task is adjusted, so that tasks with high noise sensitivity are less disturbed by noise samples, and tasks with low noise sensitivity are prevented from losing too many samples, further improving the accuracy of ECG analysis.

[0056] In one embodiment, the step of inputting the ECG signal features into the deep learning networks corresponding to heartbeat positioning, reference point detection and heartbeat classification for calculation to obtain the heartbeat positioning score, reference point detection score, heartbeat classification score, heartbeat positioning uncertainty, reference point detection uncertainty and heartbeat classification uncertainty includes:

[0057] S301, inputting the ECG signal feature into a heartbeat localization network for calculation to obtain a heartbeat localization score and a heartbeat localization uncertainty;

[0058] S302, inputting the ECG signal feature into a reference point detection network for calculation to obtain a reference point detection score and a reference point detection uncertainty;

[0059] S303: Input the electrocardiogram signal features into a heartbeat classification network for calculation to obtain a heartbeat classification score and a heartbeat classification uncertainty.

[0060] When setting the output of the beat localization network, channel 0 can be set to represent non-beats and channel 1 can be set to represent beats. When setting the output of the reference point detection network, channel 0 can be set to represent non-characteristic waves, channel 1 can be set to represent P waves, channel 2 can be set to represent QRS waves, and channel 3 can be set to represent T waves. When setting the output of the beat classification network, channel 0 can be set to represent non-beats, channel 1 can be set to represent sinus beats, channel 2 can be set to represent supraventricular beats, channel 3 can be set to represent ventricular beats, channel 4 can be set to represent fusion beats, and channel 5 can be set to represent other types of beats, such as pacemaker beats. The network structures of the beat localization network, the reference point detection network, and the beat classification network can be changed according to actual conditions. The uncertainty evaluation module of the corresponding task is added to the beat localization network, the reference point detection network, and the beat classification network, and the uncertainty of its output is accumulated according to the noise sensitivity of the task. When applied to the training of deep learning networks corresponding to multiple tasks, it can reduce the interference of noise samples on tasks with high signal quality requirements, while allowing tasks with low signal quality requirements to make full use of training data; in actual ECG multi-task analysis, the deep learning network corresponding to multiple tasks can output the uncertainty of each task to evaluate the reliability of the analysis results.

[0061] In one embodiment, the step of calculating according to the heartbeat positioning score, the reference point detection score, the heartbeat classification score, the heartbeat positioning uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty to obtain the heartbeat positioning result, the reference point detection result and the heartbeat classification result includes:

[0062] S401, calculating according to the heartbeat positioning score and the heartbeat positioning uncertainty to obtain a heartbeat positioning result;

[0063] S402, calculating according to the reference point detection score, the heartbeat positioning uncertainty and the reference point detection uncertainty to obtain a reference point detection result;

[0064] S403 , calculating according to the heartbeat classification score, the heartbeat positioning uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty to obtain a heartbeat classification result.

[0065] Since the three tasks of heartbeat localization, fiducial point detection, and heartbeat classification have different sensitivity to noise, heartbeat localization is the least sensitive to noise, followed by fiducial point detection, and heartbeat classification is the highest. Therefore, if the uncertainty of heartbeat localization of a certain input signal is very high, then the uncertainty of fiducial point detection and heartbeat classification of this signal should be higher. According to the sensitivity of different tasks to noise, the uncertainty of each task is accumulated, that is, the uncertainty corresponding to the calculation of heartbeat classification result is the highest, the uncertainty corresponding to the calculation of fiducial point detection result is second, and the uncertainty corresponding to the calculation of heartbeat localization result is the lowest, so that the output uncertainty size is more consistent with prior knowledge. When applied to the training of deep learning networks corresponding to multiple tasks, the gradient of the return can be reduced by giving higher uncertainty to noise samples, thereby reducing the interference of noise samples on training; at the same time, the uncertainty of tasks with high noise sensitivity is made higher by accumulation, and the task will be less affected by noise samples. In actual ECG multi-task analysis, the uncertainty of the output of the deep learning network corresponding to multiple tasks can be used to evaluate the reliability of the analysis results, such as screening out reliable analysis results by setting uncertainty thresholds.

[0066] In one embodiment, the step of calculating the heartbeat localization score and the heartbeat localization uncertainty to obtain the heartbeat localization result includes:

[0067] S404, calculating the heartbeat positioning result by the following formula:

[0068]

[0069] Among them, p loc is the heartbeat positioning result, s loc is the heartbeat positioning score, u loc is the heartbeat positioning uncertainty, and the softmax function is a normalized exponential function.

[0070] Since heartbeat localization is the least sensitive to noise, the corresponding uncertainty when calculating the heartbeat localization result is set to the lowest, and the exponential coefficient is only the heartbeat localization uncertainty. The softmax function is a normalized exponential function, and the output value can be converted into a probability distribution in the range of [0, 1] and 1 through the softmax function.

[0071] In one embodiment, the step of calculating the reference point detection result according to the reference point detection score, the heartbeat positioning uncertainty and the reference point detection uncertainty includes:

[0072] S405, calculate the benchmark detection result by the following formula:

[0073]

[0074] Among them, p ael is the benchmark detection result, s del is the benchmark detection score, u loc is the heart beat location uncertainty, u del is the uncertainty of benchmark detection, and the softmax function is a normalized exponential function.

[0075] Since the sensitivity of fiducial point detection to noise is higher than that of heartbeat localization, when the corresponding uncertainty is set when calculating the fiducial point detection result, the exponential coefficient is set to the sum of the heartbeat localization uncertainty and the fiducial point detection uncertainty. The softmax function is a normalized exponential function, which can convert the output value into a probability distribution in the range of [0, 1] and 1.

[0076] In one embodiment, the step of calculating the heartbeat classification score and the heartbeat positioning uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty to obtain the heartbeat classification result includes:

[0077] S406, calculating the heartbeat classification result by the following formula:

[0078]

[0079] Among them, p cls is the heart beat classification result, s cls is the heartbeat classification score, u cls is the heart beat classification uncertainty, u loc is the heart beat location uncertainty, u del is the uncertainty of benchmark detection, and the softmax function is a normalized exponential function.

[0080] Since heartbeat classification is most sensitive to noise, the corresponding uncertainty when calculating the heartbeat classification result is set to the highest, and the exponential coefficient is the sum of the heartbeat positioning uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty. The softmax function is a normalized exponential function, and the output value can be converted into a probability distribution ranging from [0, 1] and 1 through the softmax function.

[0081] Reference Figure 2 , is a structural block diagram of an ECG multi-task analysis device in one embodiment of the present application, the device includes:

[0082] An acquisition module 100 is used to acquire an electrocardiogram signal;

[0083] The feature extraction module 200 is used to input the ECG signal into a feature extraction network to extract features and obtain ECG signal features;

[0084] The multi-task network computing module 300 is used to input the ECG signal features into the deep learning networks corresponding to heartbeat positioning, reference point detection and heartbeat classification for calculation, and obtain the heartbeat positioning score, reference point detection score, heartbeat classification score, heartbeat positioning uncertainty, reference point detection uncertainty and heartbeat classification uncertainty;

[0085] The ECG analysis module 400 is used to calculate according to the heartbeat positioning score, the reference point detection score, the heartbeat classification score, the heartbeat positioning uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty to obtain the heartbeat positioning result, the reference point detection result and the heartbeat classification result.

[0086] In one embodiment, the above-mentioned ECG multi-task analysis device further includes:

[0087] The multi-task network calculation submodule is used to input the ECG signal features into the heartbeat localization network for calculation to obtain the heartbeat localization score and the heartbeat localization uncertainty; input the ECG signal features into the reference point detection network for calculation to obtain the reference point detection score and the reference point detection uncertainty; input the ECG signal features into the heartbeat classification network for calculation to obtain the heartbeat classification score and the heartbeat classification uncertainty.

[0088] In one embodiment, the above-mentioned ECG multi-task analysis device further includes:

[0089] The ECG analysis submodule is used to calculate according to the heartbeat localization score and the heartbeat localization uncertainty to obtain a heartbeat localization result; to calculate according to the reference point detection score, the heartbeat localization uncertainty and the reference point detection uncertainty to obtain a reference point detection result; and to calculate according to the heartbeat classification score, the heartbeat localization uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty to obtain a heartbeat classification result.

[0090] In one embodiment, the above-mentioned ECG multi-task analysis device further includes:

[0091] The first ECG analysis submodule is used to calculate the heartbeat positioning result using the following formula:

[0092]

[0093] Among them, p loc is the heartbeat positioning result, s loc is the heartbeat positioning score, u loc is the heartbeat positioning uncertainty, and the softmax function is a normalized exponential function.

[0094] In one embodiment, the above-mentioned ECG multi-task analysis device further includes:

[0095] The second ECG analysis submodule is used to calculate the benchmark detection results using the following formula:

[0096]

[0097] Among them, p del is the benchmark detection result, s del is the benchmark detection score, u loc is the heart beat location uncertainty, u del is the uncertainty of benchmark detection, and the softmax function is a normalized exponential function.

[0098] In one embodiment, the above-mentioned ECG multi-task analysis device further includes:

[0099] The third ECG analysis submodule is used to calculate the heart beat classification result using the following formula:

[0100]

[0101] Among them, p cls is the heart beat classification result, s cls is the heartbeat classification score, u cls is the heart beat classification uncertainty, u loc is the heart beat location uncertainty, u del is the uncertainty of benchmark detection, and the softmax function is a normalized exponential function.

[0102] Reference Figure 3 In an embodiment of the present application, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store usage data during the ECG multi-task analysis method, etc. The network interface of the computer device is used to communicate with an external terminal via a network connection. Furthermore, the above-mentioned computer device can also be provided with an input device and a display screen, etc. When the above-mentioned computer program is executed by the processor, it uses an ECG multi-task analysis method, including the following steps: acquiring an ECG signal; inputting the ECG signal into a feature extraction network for feature extraction to obtain ECG signal features; inputting the ECG signal features into deep learning networks corresponding to heartbeat positioning, reference point detection and heartbeat classification for calculation to obtain a heartbeat positioning score, a reference point detection score, a heartbeat classification score, a heartbeat positioning uncertainty, a reference point detection uncertainty and a heartbeat classification uncertainty; performing calculations based on the heartbeat positioning score, the reference point detection score, the heartbeat classification score, the heartbeat positioning uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty to obtain a heartbeat positioning result, a reference point detection result and a heartbeat classification result.

[0103] Those skilled in the art will understand that Figure 3 The structure shown in is merely a block diagram of a portion of the structure related to the present application solution and does not constitute a limitation on the computer device to which the present application solution is applied.

[0104] An embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, a multi-task ECG analysis method is implemented, including the following steps: obtaining an ECG signal; inputting the ECG signal into a feature extraction network for feature extraction to obtain ECG signal features; inputting the ECG signal features into deep learning networks corresponding to heartbeat positioning, reference point detection, and heartbeat classification for calculation to obtain heartbeat positioning scores, reference point detection scores, heartbeat classification scores, heartbeat positioning uncertainty, reference point detection uncertainty, and heartbeat classification uncertainty; calculating according to the heartbeat positioning score, the reference point detection score, the heartbeat classification score, the heartbeat positioning uncertainty, the reference point detection uncertainty, and the heartbeat classification uncertainty to obtain heartbeat positioning results, reference point detection results, and heartbeat classification results. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0105] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided in this application and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0106] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, device, article or method including the element.

[0107] The above description is only a preferred embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A multi-task ECG analysis method, characterized in that: The method comprises: Obtain ECG signals; Inputting the ECG signal into a feature extraction network to extract features and obtain ECG signal features; Inputting the ECG signal features into the deep learning networks corresponding to heartbeat positioning, reference point detection and heartbeat classification respectively for calculation, and obtaining the heartbeat positioning score, reference point detection score, heartbeat classification score, heartbeat positioning uncertainty, reference point detection uncertainty and heartbeat classification uncertainty; Calculations are performed based on the heartbeat positioning score, the reference point detection score, the heartbeat classification score, the heartbeat positioning uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty to obtain a heartbeat positioning result, a reference point detection result and a heartbeat classification result.

2. The ECG multi-task analysis method according to claim 1, characterized in that: The step of inputting the electrocardiogram signal features into the deep learning networks corresponding to heartbeat positioning, reference point detection and heartbeat classification for calculation to obtain the heartbeat positioning score, reference point detection score, heartbeat classification score, heartbeat positioning uncertainty, reference point detection uncertainty and heartbeat classification uncertainty comprises: Inputting the electrocardiogram signal characteristics into a heartbeat localization network for calculation to obtain a heartbeat localization score and a heartbeat localization uncertainty; Inputting the electrocardiogram signal characteristics into a reference point detection network for calculation to obtain a reference point detection score and a reference point detection uncertainty; The electrocardiogram signal features are input into a heartbeat classification network for calculation to obtain a heartbeat classification score and a heartbeat classification uncertainty.

3. The ECG multi-task analysis method according to claim 1, characterized in that: The step of calculating according to the heartbeat positioning score, the reference point detection score, the heartbeat classification score, the heartbeat positioning uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty to obtain the heartbeat positioning result, the reference point detection result and the heartbeat classification result comprises: Calculating according to the heartbeat localization score and the heartbeat localization uncertainty to obtain a heartbeat localization result; Calculating according to the reference point detection score, the heartbeat positioning uncertainty and the reference point detection uncertainty to obtain a reference point detection result; A heartbeat classification result is obtained by calculating the heartbeat classification score, the heartbeat positioning uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty.

4. The ECG multi-task analysis method according to claim 3, characterized in that: The step of calculating according to the heartbeat localization score and the heartbeat localization uncertainty to obtain the heartbeat localization result comprises: The heartbeat localization result is calculated by the following formula: Among them, p loc is the heartbeat positioning result, s loc is the heartbeat positioning score, u loc is the heartbeat positioning uncertainty, and the softmax function is a normalized exponential function.

5. The ECG multi-task analysis method according to claim 3, characterized in that: The step of calculating the reference point detection result according to the reference point detection score, the heartbeat positioning uncertainty and the reference point detection uncertainty comprises: The benchmark detection result is calculated using the following formula: Among them, p del is the benchmark detection result, s del is the benchmark detection score, u loc is the heart beat location uncertainty, u del is the uncertainty of benchmark detection, and the softmax function is a normalized exponential function.

6. The ECG multi-task analysis method according to claim 3, characterized in that: The step of calculating the heartbeat classification score, the heartbeat positioning uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty to obtain the heartbeat classification result comprises: The heart beat classification result is calculated by the following formula: Among them, p cls is the heart beat classification result, s cls is the heartbeat classification score, u cls is the heart beat classification uncertainty, u loc is the heart beat location uncertainty, u del is the uncertainty of benchmark detection, and the softmax function is a normalized exponential function.

7. An electrocardiogram multi-task analysis device, characterized in that: The device comprises: An acquisition module, used for acquiring electrocardiogram signals; A feature extraction module, used for inputting the ECG signal into a feature extraction network for feature extraction to obtain ECG signal features; A multi-task network computing module, used for inputting the ECG signal features into the deep learning networks corresponding to heartbeat positioning, reference point detection and heartbeat classification for computing, and obtaining a heartbeat positioning score, a reference point detection score, a heartbeat classification score, a heartbeat positioning uncertainty, a reference point detection uncertainty and a heartbeat classification uncertainty; The ECG analysis module is used to calculate according to the heartbeat positioning score, the reference point detection score, the heartbeat classification score, the heartbeat positioning uncertainty, the reference point detection uncertainty and the heartbeat classification uncertainty to obtain the heartbeat positioning result, the reference point detection result and the heartbeat classification result.

8. The ECG multi-task analysis device according to claim 7, characterized in that: include: The multi-task network calculation submodule is used to input the ECG signal features into the heartbeat localization network for calculation to obtain the heartbeat localization score and the heartbeat localization uncertainty; input the ECG signal features into the reference point detection network for calculation to obtain the reference point detection score and the reference point detection uncertainty; input the ECG signal features into the heartbeat classification network for calculation to obtain the heartbeat classification score and the heartbeat classification uncertainty.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images