Abnormal Discharge Detection Method, Abnormal Discharge Detection Device, Product and Equipment

Abstract

Embodiments of the present disclosure relate to an abnormal discharge detection method, an abnormal discharge detection device, a computer program product, and an electronic device, and relate to the field of artificial intelligence technology. Among them, the above method includes: determining a biomedical signal to be detected of a subject to be tested, inputting the biomedical signal to be detected into a first abnormal discharge detection sub-model to perform a first abnormal discharge detection on the biomedical signal to be detected; performing a formal transformation on the biomedical signal to be detected of the subject to be tested to obtain class image biomedical signal data corresponding to the biomedical signal to be detected; inputting the class image biomedical signal data into a second abnormal discharge detection sub-model to perform a second abnormal discharge detection on the class image biomedical signal data; and fusing the detection situations of the first abnormal discharge detection and the second abnormal discharge detection to obtain an abnormal discharge detection result of the subject to be tested. The present disclosure can improve the efficiency and accuracy of abnormal discharge detection and recognition.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology. Specifically, it relates to an abnormal discharge detection method, an abnormal discharge detection device, a computer program product, and an electronic device. Background Art

[0002] This section aims to provide background or context for the embodiments of the present disclosure stated in the claims. The descriptions herein are not admitted to be prior art merely because they are included in this section.

[0003] Electroencephalogram is an important basis for judging whether the brain function of a patient is abnormal, and it has important value for clinical diagnosis and treatment and the evaluation of the severity of the condition. The rapid development of artificial intelligence technology provides the possibility for the realization of automatic electroencephalogram analysis. In particular, neural network technology shows good application potential in detecting abnormal electroencephalogram discharges. Summary of the Invention

[0004] However, the detection accuracy and reliability of the electroencephalogram abnormal automatic detection methods in the related art are insufficient, and it is difficult to achieve truly independent electroencephalogram signal abnormal detection.

[0005] Therefore, there is a great need for an abnormal discharge detection method to improve the accuracy and reliability of electroencephalogram abnormal automatic detection.

[0006] In this context, embodiments of the present invention are expected to provide an abnormal discharge detection method, an abnormal discharge detection device, a computer-readable storage medium, a computer program product, and an electronic device.

[0007] According to a first aspect of an embodiment of the present disclosure, there is provided an abnormal discharge detection method, including: determining a biomedical signal to be detected of a subject to be tested, inputting the biomedical signal to be detected into a first abnormal discharge detection sub-model in a pre-trained abnormal discharge detection model, and performing a first abnormal discharge detection on the biomedical signal to be detected according to the first abnormal discharge detection sub-model; performing a form transformation on the biomedical signal to be detected of the subject to be tested to obtain image-like biomedical signal data corresponding to the biomedical signal to be detected;

[0008] inputting the image-like biomedical signal data into a second abnormal discharge detection sub-model in the pre-trained abnormal discharge detection model, and performing a second abnormal discharge detection on the image-like biomedical signal data according to the second abnormal discharge detection sub-model; fusing the detection situations of the first abnormal discharge detection and the second abnormal discharge detection to obtain an abnormal discharge detection result of the subject to be tested.

[0009] Optionally, the determining of the biomedical signal to be detected for the subject to be tested includes: acquiring the biomedical signal of the subject to be tested that is collected; segmenting the biomedical signal of the subject to be tested according to a preset duration, and determining at least one biomedical signal to be detected for the subject to be tested.

[0010] Optionally, the fusing of the detection situations of the first abnormal discharge detection and the second abnormal discharge detection to obtain the abnormal discharge detection result for the subject to be tested includes: fusing the first abnormal discharge detection result and the second abnormal discharge detection result to obtain the abnormal discharge detection result of the biomedical signal to be detected; in the case where the abnormal discharge detection result of the biomedical signal to be detected is abnormal discharge, marking the biomedical signal to be detected; and obtaining the abnormal discharge detection result for the subject to be tested according to the marked biomedical signal to be detected.

[0011] Optionally, the first abnormal discharge detection sub-model includes a first multi-head attention sub-model, a first deep convolutional neural network sub-model, and a first fully connected layer; the inputting of the biomedical signal to be detected into the first abnormal discharge detection sub-model of the pre-trained abnormal discharge detection model, and performing the first abnormal discharge detection on the biomedical signal to be detected according to the first abnormal discharge detection sub-model includes: inputting the biomedical signal to be detected into the first multi-head attention sub-model, and obtaining the first feature of the biomedical signal to be detected according to the output of the first multi-head attention sub-model; inputting the first feature into the first deep convolutional neural network sub-model, and obtaining the second feature according to the output of the first deep convolutional neural network sub-model; inputting the second feature into the first fully connected layer, and obtaining the first fully connected feature according to the output of the first fully connected layer; and processing the first fully connected feature through a preset activation function, and obtaining the first abnormal discharge detection result of the biomedical signal to be detected according to the processing result.

[0012] Optionally, the first abnormal discharge detection sub-model includes a second deep convolutional neural network sub-model and a second fully connected layer; the inputting of the image-like biomedical signal data into the second abnormal discharge detection sub-model of the pre-trained abnormal discharge detection model, and performing the second abnormal discharge detection on the image-like biomedical signal data according to the second abnormal discharge detection sub-model includes: inputting the image-like biomedical signal data into the second deep convolutional neural network sub-model, and obtaining the third feature according to the output of the second deep convolutional neural network sub-model; inputting the third feature into the second fully connected layer, and obtaining the second fully connected feature according to the output of the second fully connected layer; and processing the second fully connected feature through a preset activation function, and obtaining the second abnormal discharge detection result of the biomedical signal to be detected according to the processing result.

[0013] Optionally, the biomedical signal to be detected includes a first matrix corresponding to a first matrix dimension. A formal transformation is performed on the biomedical signal to be detected of the subject to obtain image-like biomedical signal data corresponding to the biomedical signal to be detected. The first matrix is deformed to obtain a second matrix corresponding to a second matrix dimension, and the difference between the number of rows and the number of columns of the second matrix dimension is less than a preset value. The image-like biomedical signal data is determined according to the second matrix.

[0014] Optionally, any one of the deep convolutional neural network submodels includes any one of a dilated deep convolutional network submodel and a multi-level convolutional neural network submodel.

[0015] According to a second aspect of the embodiments of the present disclosure, there is provided an abnormal discharge detection device, including: a first detection module configured to determine a biomedical signal to be detected of a subject and input the biomedical signal to be detected into a first abnormal discharge detection submodel in a pre-trained abnormal discharge detection model, and perform a first abnormal discharge detection on the biomedical signal to be detected according to the first abnormal discharge detection submodel; a signal preprocessing module configured to perform a formal transformation on the biomedical signal to be detected of the subject to obtain image-like biomedical signal data corresponding to the biomedical signal to be detected; a second detection module configured to input the image-like biomedical signal data into a second abnormal discharge detection submodel in a pre-trained abnormal discharge detection model, and perform a second abnormal discharge detection on the image-like biomedical signal data according to the second abnormal discharge detection submodel; and a fusion module configured to fuse the detection situations of the first abnormal discharge detection and the second abnormal discharge detection to obtain an abnormal discharge detection result of the subject.

[0016] According to a third aspect of the present disclosure, there is provided a computer program product containing instructions, which when running on a computer, causes the computer to execute the steps of the abnormal discharge detection method as described in the first aspect.

[0017] According to a fourth aspect of the embodiments of the present disclosure, there is provided a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the abnormal discharge detection method as described in the first aspect in the above embodiments.

[0018] According to a fifth aspect of the embodiments of the present disclosure, there is provided an electronic device, including: a processor; and a storage device for storing one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the abnormal discharge detection method as described in the first aspect in the above embodiments.

[0019] An abnormal discharge detection method, an abnormal discharge detection device, a computer-readable storage medium, a computer program product, and an electronic device according to an embodiment of the present disclosure perform a first abnormal discharge detection on an original biomedical signal to be detected through a first abnormal discharge detection sub-model, and perform a second abnormal discharge detection on the image-like biomedical signal data corresponding to the original biomedical signal to be detected through a second abnormal discharge detection sub-model. Then, the detection situations of the first discharge detection and the second abnormal discharge detection are fused to obtain an abnormal discharge detection result. On the one hand, by performing abnormal discharge detection on the original biomedical signal to be detected, the present disclosure can retain the time information in the original biomedical signal to be detected. By performing abnormal discharge detection on the image-like biomedical signal data, the present disclosure can utilize the feature extraction advantage of the neural network in the image-like biomedical signal data to extract deeper biomedical signal features. The combination of the two not only retains the time information of the biomedical signal to be detected but also extracts richer biomedical signal features, which can improve the accuracy and reliability of the abnormal discharge detection of the biomedical signal to be detected. On the other hand, the abnormal discharge detection result obtained according to the method of the present disclosure can provide relatively accurate reference information for doctors, enabling doctors to quickly and accurately judge the diagnosis result of the patient to be diagnosed without carefully observing all the data, but only roughly browsing some key information and then combining the abnormal discharge detection result, which saves the workload of doctors and is convenient for wide application in clinical practice. On the further hand, based on the abnormal discharge detection model, the present disclosure can realize the automatic detection and recognition of abnormal discharges in biomedical signals and improve the efficiency of abnormal discharge detection. Description of the Drawings

[0020] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features, and advantages of the exemplary embodiments of the present invention will become readily understood. In the drawings, several embodiments of the present invention are shown in an exemplary rather than restrictive manner, wherein:

[0021] Figure 1 A schematic diagram showing an exemplary system architecture to which the embodiments of the present disclosure can be applied;

[0022] Figure 2 A schematic flowchart showing an abnormal discharge detection method in an exemplary embodiment of the present disclosure;

[0023] Figure 3 A schematic flowchart showing a method for obtaining a biomedical signal to be detected in an exemplary embodiment of the present disclosure;

[0024] Figure 4 A schematic diagram showing the model structure of a pre-trained abnormal discharge detection model in an exemplary embodiment of the present disclosure;

[0025] Figure 5 A schematic flowchart showing a method for performing a first abnormal discharge detection in an exemplary embodiment of the present disclosure;

[0026] Figure 6 A schematic flowchart showing a method for performing a second abnormal discharge detection in an exemplary embodiment of the present disclosure;

[0027] Figure 7 A schematic flowchart showing a method for obtaining an abnormal discharge detection result of a subject under test in an exemplary embodiment of the present disclosure;

[0028] Figure 8 A schematic diagram showing the composition of an abnormal discharge detection device in an exemplary embodiment of the present disclosure;

[0029] Figure 9 A schematic diagram showing the structure of an electronic device in an exemplary embodiment of the present disclosure.

[0030] In the drawings, the same or corresponding reference numerals denote the same or corresponding parts. Detailed implementation manners

[0031] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and then implement the present invention, and do not limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to be able to fully convey the scope of the present invention to those skilled in the art.

[0032] Those skilled in the art know that the embodiments of the present invention can be implemented as a system, a device, an equipment, a medium, a method, or a computer program product. Therefore, the present invention can be specifically implemented in the following forms: completely hardware, completely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0033] According to the embodiments of the present invention, an abnormal discharge detection method, an abnormal discharge detection device, a computer-readable storage medium, a computer program product, and an electronic device are provided.

[0034] In this article, any number of elements in the drawings is used for illustration rather than limitation, and any naming is only used for distinction and does not have any limiting meaning.

[0035] The principles and spirit of the present invention will be elaborated in detail below with reference to several representative embodiments of the present invention. Summary of the Invention

[0037] The inventors of the present disclosure have found that the abnormal discharge detection methods in the related art have the problem of low accuracy and reliability in detecting abnormal discharges in biomedical signals such as electroencephalograms.

[0038] In view of the above, the basic idea of the present disclosure is as follows: to provide an abnormal discharge detection method, an abnormal discharge detection device, a computer-readable storage medium, a computer program product, and an electronic device. The first abnormal discharge detection sub-model is used to perform the first abnormal discharge detection on the original biomedical signal to be detected, and the second abnormal discharge detection sub-model is used to perform the second abnormal discharge detection on the image-like biomedical signal data corresponding to the original biomedical signal to be detected. Then, the detection situations of the first discharge detection and the second abnormal discharge detection are fused to obtain the abnormal discharge detection result. On the one hand, by performing abnormal discharge detection on the original biomedical signal to be detected, the time information in the original biomedical signal to be detected can be retained. By performing abnormal discharge detection on the image-like biomedical signal data, the advantage of feature extraction of the neural network can be utilized in the image-like biomedical signal data, and deeper biomedical signal features can be extracted. The combination of the two not only retains the time information of the biomedical signal to be detected but also extracts richer biomedical signal features, which can improve the accuracy and reliability of the abnormal discharge detection of the biomedical signal to be detected. On the other hand, the abnormal discharge detection result obtained according to the method of the present disclosure can provide relatively accurate reference information for doctors, enabling doctors to quickly and accurately judge the diagnosis result of the patient to be diagnosed without carefully observing all the data, but only roughly browsing some key information and then combining the abnormal discharge detection result, saving the workload of doctors and facilitating wide application in clinical practice. On the other hand, based on the abnormal discharge detection model, the present disclosure can realize the automatic detection and recognition of abnormal discharges in biomedical signals and improve the efficiency of abnormal discharge detection.

[0039] After introducing the basic principle of the present invention, the various non-limiting embodiments of the present invention will be specifically introduced below.

[0040] Overview of Application Scenarios

[0041] It should be noted that the following application scenarios are only shown for the convenience of understanding the spirit and principle of the present invention, and the embodiments of the present invention are not limited in this regard. On the contrary, the embodiments of the present invention can be applied to any applicable scenario.

[0042] Embodiments of the present disclosure can be applied to the scenario of detecting abnormal electrical discharges in the human brain. For example, when a patient feels uncomfortable in the head, such as having a headache, the brain electrical signals of the patient can be collected. Then, the collected original brain electrical signals are converted into corresponding image-like brain electrical signals. The original brain electrical signals are input into the first abnormal discharge detection sub-model, and the image-like brain electrical signals are input into the second abnormal discharge detection sub-model to obtain the first abnormal discharge detection result and the second abnormal discharge detection result respectively. Then, the first abnormal discharge detection result and the second abnormal discharge detection result are fused, and the brain abnormal discharge detection result of the patient is obtained according to the fusion result. The abnormal discharge detection result of the patient can be fed back to the doctor, such as sending the abnormal discharge detection result and the corresponding patient identifier to the client logged in by the doctor, so as to assist the doctor in making a rapid and accurate clinical diagnosis of the patient based on the abnormal discharge detection result combined with other key index data, reducing the workload of the doctor and improving the work efficiency of the doctor.

[0043] Exemplary System Architecture

[0044] First, refer to Figure 1 to describe the system architecture of the exemplary application environment of the present disclosure.

[0045] As Figure 1 shown, the system architecture 100 may include a terminal device 110 and a server 120. Among them, the terminal device 110 may be a terminal device such as a smart phone, a tablet computer, a desktop computer, a laptop computer, a smart wearable device, etc. The server 120 generally refers to a background system that provides services related to the abnormal discharge detection method in this exemplary embodiment, and may be a single server or a cluster formed by multiple servers. A connection can be formed between the terminal device 110 and the server 120 through a wired or wireless communication link for data interaction.

[0046] In an exemplary embodiment, the above abnormal discharge detection method may be executed by the server 120. Correspondingly, the abnormal discharge detection device may be disposed in the server 120 to implement the corresponding module functions. For example, a biomedical signal of a subject to be measured is collected by an electroencephalograph, and the collected biomedical signal is sent to the server 120. The server 120 performs segmentation processing on the received biomedical signal to obtain a biomedical signal to be detected, duplicates the biomedical signal to be detected, and performs a form transformation on the duplicated biomedical signal to obtain biomedical signal data in the form of an image. Then, the biomedical signal to be detected is input into the first abnormal discharge detection sub-model, and the biomedical signal data in the form of an image is input into the second abnormal discharge detection sub-model. A first abnormal discharge detection result is obtained according to the first abnormal discharge detection sub-model, and a second abnormal discharge detection result is obtained according to the output of the second abnormal discharge detection sub-model. The first abnormal discharge detection result and the second abnormal discharge detection result are fused to obtain a final abnormal discharge detection result. The server 120 may send the final abnormal discharge detection result to the terminal device 110, and the final abnormal discharge detection result is displayed in the graphical user interface of the terminal device 110.

[0047] In another exemplary embodiment, the above abnormal discharge detection method may also be executed by the terminal device 110. Correspondingly, the abnormal discharge detection device may be disposed in the terminal device 110 to implement the corresponding module functions. For example, a user uploads the biomedical signal of the subject to be measured collected to the terminal device 110. The terminal device 110 first segments the collected biomedical signal according to the pre-configured algorithm logic to obtain a biomedical signal to be detected, then duplicates the biomedical signal to be detected, performs a form transformation on the duplicated biomedical signal to obtain biomedical signal data in the form of an image. Then, the biomedical signal to be detected is input into the first abnormal discharge detection sub-model pre-configured in the terminal device 110, and the biomedical signal data in the form of an image is input into the second abnormal discharge detection sub-model pre-configured in the terminal device 110. A first discharge detection result is obtained according to the output of the first abnormal discharge detection sub-model, and a second abnormal discharge detection result is obtained according to the output of the second abnormal discharge detection sub-model. The terminal device 110 may fuse the first abnormal discharge detection result and the second abnormal discharge detection result according to the pre-configured algorithm, and obtain the abnormal discharge detection result of the subject to be measured according to the fusion result.

[0048] It should be understood, Figure 1The numbers of the terminal devices and servers therein are merely illustrative. According to the implementation requirements, there can be any number of terminal devices and servers. For example, the server 120 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms.

[0049] Those skilled in the art can easily understand that the above application scenarios are only for illustration, and the exemplary embodiments are not limited thereto.

[0050] Exemplary Method

[0051] Figure 2 The flowchart showing a method for detecting abnormal discharge in an exemplary embodiment of the present disclosure is referred to Figure 2 , and the method includes:

[0052] Step S210: Determine the biomedical signal to be detected of the object under test, input the biomedical signal to be detected into the first abnormal discharge detection sub-model in the pre-trained abnormal discharge detection model, and perform the first abnormal discharge detection on the biomedical signal to be detected according to the first abnormal discharge detection sub-model;

[0053] Step S220: Perform a form transformation on the biomedical signal to be detected of the object under test to obtain the image-like biomedical signal data corresponding to the biomedical signal to be detected;

[0054] Step S230: Input the image-like biomedical signal data into the second abnormal discharge detection sub-model in the pre-trained abnormal discharge detection model, and perform the second abnormal discharge detection on the image-like biomedical signal data according to the second abnormal discharge detection sub-model;

[0055] Step S240: Integrate the detection results of the first abnormal discharge detection and the second abnormal discharge detection to obtain the abnormal discharge detection result of the object under test.

[0056] Next, a detailed description will be given of the specific implementation of "Step S210: Determine the biomedical signal to be detected of the object under test, input the biomedical signal to be detected into the first abnormal discharge detection sub-model in the pre-trained abnormal discharge detection model, and perform the first abnormal discharge detection on the biomedical signal to be detected according to the first abnormal discharge detection sub-model".

[0057] In an exemplary embodiment, a biomedical signal is an active signal spontaneously generated by the physiological processes of a living body, which can accurately reflect the physiological state or physical signs of a person. Therefore, whether there is an abnormality in the living body can be determined through the biomedical signal. The living body in the present disclosure may include any living body, such as a human body, an animal body, etc., and no special limitation is made in this exemplary embodiment.

[0058] Exemplarily, the above-mentioned biomedical signal may include any one or more of the following: physiological signals such as electrocardiogram (ECG), electroencephalogram (EEG), electromyogram (EMG), electrooculogram (EOG), and electrogastrogram (EGG), or may also include non-electrophysiological signals such as body temperature, blood pressure, pulse, and respiration, which can be set according to actual situations, and no special limitation is made in the present disclosure.

[0059] Among them, the electrocardiogram signal may be an electrocardiogram collected by a multi-channel electrocardiograph. An electrocardiogram (ECG) refers to a graph of various forms of potential changes led out from the body surface through an electrocardiograph along with the bioelectric changes during each cardiac cycle when the pacemaker, atrium, and ventricle of the heart are successively excited. The electrocardiogram is an objective index for the occurrence, propagation, and recovery process of cardiac excitation.

[0060] The electroencephalogram signal may be an electroencephalogram collected by a multi-channel electroencephalograph. An electroencephalogram (EEG) is a graph obtained by amplifying and recording the spontaneous bioelectric potential of the cerebral cortex of the brain from the scalp through a precise instrument, and is the spontaneous and rhythmic electrical activity of a group of brain cells recorded through electrodes. This electrical activity takes the potential as the vertical axis and time as the horizontal axis, thereby recording the plane graph of the mutual relationship between potential and time. The frequency (period), amplitude, and phase of the brain wave constitute the basic characteristics of the electroencephalogram.

[0061] Electromyogram (EMG) is the superposition of motor unit action potentials in time and space among numerous muscle fibers, and can be obtained by pasting electromyography sensors on the skin.

[0062] Electrooculogram (EOG) is a bioelectric signal caused by the potential difference between the cornea and retina of the eye, and is extremely convenient to collect, and can be completed through a small number of electrodes.

[0063] Electrogastrogram (EGG) is an electric signal generated by the contraction of the stomach muscles, and can be collected using electrodes on the abdominal skin surface of the human body.

[0064] In an alternative embodiment, the subject to be tested may include patients seeking medical treatment with epilepsy or epilepsy-like symptoms. The subject to be tested may also include other people seeking medical treatment due to discomfort in the brain, such as those with brain pain, etc. This exemplary embodiment does not make special limitations on this.

[0065] Exemplarily, Figure 3 FIG. shows a schematic flowchart of a method for obtaining a biomedical signal to be detected in an exemplary embodiment of the present disclosure. Refer to Figure 3 This method may include steps S310 to S320. Among them:

[0066] In step S310, the biomedical signal of the subject to be tested collected is obtained.

[0067] For example, for each subject to be tested, a long-term biomedical signal may be collected, such as collecting a 10-minute biomedical signal. Since abnormal discharges are not continuous, the biomedical signal of the subject to be tested can be segmented to obtain the biomedical signal to be detected, and whether there is an abnormal discharge in the subject to be tested is determined according to the biomedical signal to be detected, improving the efficiency and accuracy of abnormal discharge detection.

[0068] In step S320, the biomedical signal of the subject to be tested is segmented according to a preset duration, and at least one biomedical signal to be detected of the subject to be tested is determined.

[0069] Taking the preset duration as 4 seconds as an example, the biomedical signal collected in step S210 can be segmented in units of 4 seconds, and each 4-second long segmented signal is used as a biomedical signal to be detected. In this way, 15 biomedical signals to be detected can be obtained in 1 minute, and 150 biomedical signals to be detected can be obtained in 10 minutes.

[0070] In another exemplary embodiment, the biomedical signal collected by the biomedical signal device can also be directly used as the biomedical signal to be detected without segmenting it. For example, a 1-minute biomedical signal can be directly collected as needed, and this 1-minute biomedical signal collected is used as the biomedical signal to be detected. This exemplary embodiment does not make special limitations on this.

[0071] For example, what the biomedical signal acquisition device directly collects is a digital sequence of biomedical signals. For example, a multi-channel electrocardiograph, a multi-channel electroencephalograph, and a multi-channel electromyograph can be used to collect the corresponding biomedical signals at a preset acquisition frequency, such as 500HZ, and a numerical sequence of biomedical signals is obtained.

[0072] Taking the electroencephalogram signal as an example, the electroencephalogram signal collected by the electroencephalograph can be a numerical sequence of electroencephalogram signals shown in the following formula (1).

[0073] $x_t=(x_{t1},x_{t2},x_{t3},x_{t4},\ldots,x_{tn})$ (1)

[0074] In formula (1), $x_t$ is the numerical sequence of EEG signals. $x_{t1},x_{t2},x_{t3},x_{t4},\ldots,x_{tn}$ represent the signal values of the EEG signal at different sampling times. For example, $x_{t1}$ represents the signal value at the 1st sampling time in the numerical sequence $x_t$ of the EEG signal, and $x_{tn}$ represents the signal value at the $n$th sampling time in the numerical sequence $x_t$ of the EEG signal.

[0075] It can be seen from formula (1) that the form of the original biomedical signal collected by the biomedical signal can be regarded as a one-dimensional or multi-dimensional long matrix. Taking the example that the number of channels of the multi-channel EEG signal is 29 and the total number of samples is 2000 in 4 seconds, the biomedical signal to be detected is a long matrix of 29 by 2000 dimensions. The abnormal discharge detection model is usually a neural network model, especially the convolutional neural network. It has natural processing advantages for image signals and can extract deeper feature expressions of image signals when processing image signals. Most image signals are matrices with more or less the same number of rows and columns, such as 256 by 256, 512 by 512, etc. In other words, the original biomedical signal is similar to a rectangular input, while the image signal is similar to a square input, and their forms are quite different, which makes it impossible for the neural network, especially the convolutional neural network, to fully exert its feature extraction advantages in abnormal discharge detection.

[0076] That is to say, when the form of the biomedical signal input into the abnormal discharge detection model is similar to the form of the image signal, the advantages of the neural network, especially the convolutional neural network, in feature extraction can be fully applied to the field of abnormal discharge detection to extract deeper biomedical signal features and improve the accuracy of biomedical signal detection.

[0077] However, for biomedical signals, they are related to time. As mentioned above, they are time series. After converting them into image-form signals, the time information will be lost, and the time information is also an important feature for abnormal discharge detection.

[0078] Based on this, in the present disclosure, the original biomedical signal to be detected is processed by the first abnormal discharge detection sub-model to perform the first abnormal discharge detection. In this way, the time information of the biomedical signal to be detected is retained in the first abnormal discharge detection. At the same time, the class-image biomedical signal data is processed by the second abnormal discharge detection sub-model, so as to extract deeper biomedical signal features by taking advantage of the natural advantage of the neural network model in processing image data, and perform the second abnormal discharge detection. Then, the first abnormal discharge detection situation and the second abnormal discharge detection situation are fused to obtain the abnormal discharge detection result of the subject under test. Obviously, the time information of the biomedical signal to be detected is retained in the fusion result, and richer and deeper biomedical signal features are obtained through the class-image biomedical signal data. Therefore, the accuracy and reliability of the abnormal discharge detection of biomedical signals are improved.

[0079] In an exemplary embodiment, the pre-trained abnormal discharge detection model includes a first abnormal discharge detection sub-model and a second abnormal discharge detection sub-model. Among them, the first abnormal discharge detection sub-model is used to process the original biomedical signal, and the second abnormal discharge detection sub-model is used to process the class-image biomedical signal corresponding to the original biomedical signal.

[0080] For example, biomedical signals can be collected from the bodies of different subjects, and then doctors label whether there is abnormal discharge in the collected biomedical signals, so as to obtain a labeled training data set. The initial abnormal discharge detection model is trained with supervised learning according to the labeled training data set. For example, according to the binary cross-entropy loss between the prediction result of the initial abnormal discharge detection model and the label, iterative training is carried out in the direction of reducing the binary cross-entropy loss, and the model parameters are updated. When the binary cross-entropy loss is less than a certain preset value or the number of iterations reaches a certain threshold, the training is stopped, so as to obtain the pre-trained abnormal discharge detection model.

[0081] Among them, the initial abnormal discharge detection model includes an initial first abnormal discharge detection sub-model and an initial second abnormal discharge detection sub-model. The input of the initial first abnormal discharge detection sub-model is the original biomedical signal in the training data set, and the output is the first abnormal discharge detection result. The input of the initial second abnormal discharge detection sub-model is the class-image biomedical data corresponding to the original biomedical signal in the training data set, and the output is the second abnormal discharge detection result. The output of the initial abnormal discharge detection model is the fusion result of the first abnormal discharge detection result and the second abnormal discharge detection result. In this way, the finally obtained pre-trained abnormal discharge detection model includes a first abnormal discharge detection sub-model and a second abnormal discharge detection sub-model.

[0082] In an exemplary implementation, the pre-trained abnormal discharge detection model is determined based on a convolutional neural network. For example, the first abnormal discharge detection sub-model includes a first convolutional neural network, and the second abnormal discharge detection sub-model includes a second convolutional neural network.

[0083] For example, the model structure of the pre-trained abnormal discharge detection model in the present disclosure can be as Figure 4 shown. Referring to Figure 4 , the pre-trained abnormal discharge detection model includes a first abnormal discharge detection sub-model 41 and a second abnormal discharge detection sub-model 42. Among them, the first abnormal discharge detection sub-model 41 may include a first multi-head attention sub-model 411, a first deep convolutional neural network sub-model 412, a first fully connected layer 413, and a first activation layer 414. The second abnormal discharge detection sub-model 42 may include a second deep convolutional neural network sub-model 421, a second fully connected layer 422, and a second activation layer 423. Among them, both the first activation layer and the second activation layer can be preset activation functions.

[0084] Exemplarily, Figure 5 shows a schematic flowchart of a method for performing the first abnormal discharge detection in an exemplary embodiment of the present disclosure. Referring to Figure 5 , the method may include steps S510 to S540. Among them:

[0085] In step S510, the biomedical signal to be detected is input into the first multi-head attention sub-model, and the first feature of the biomedical signal to be detected is obtained according to the output of the first multi-head attention sub-model.

[0086] In an exemplary implementation, the QKV of the first multi-head attention sub-model is the same, that is, all are the original signals of the biomedical signal to be detected. Q, K, and V respectively correspond to the query, key, and value matrices in the multi-head attention model.

[0087] The specific structure of the first multi-head attention sub-model in the present disclosure can all refer to the structure of the multi-head attention model in the existing natural language processing field, and will not be elaborated here.

[0088] After the original signal of the biomedical signal to be detected passes through the first multi-head attention sub-model, the time information features of the biomedical signal to be detected can be extracted through the first multi-head attention sub-model. However, the first feature output by the first multi-head attention sub-model is a shallow biomedical signal understanding feature, and this shallow biomedical signal feature cannot represent the deeper signal meaning of the biomedical signal. Therefore, in this disclosure, inspired by the field of computer vision, the shallow features are input into a deep convolutional neural network to extract deeper features of the biomedical signal, thereby improving the accuracy of abnormal discharge detection.

[0089] In step S520, the first feature is input into the first deep convolutional neural network sub-model, and a second feature is obtained according to the output of the first deep convolutional neural network sub-model.

[0090] In an exemplary implementation, the first deep convolutional neural network sub-model may include any one of a first depthwise dilated convolutional network sub-model and a first multi-level convolutional neural network sub-model.

[0091] In the case where the first deep convolutional neural network sub-model is the first depthwise dilated convolutional network sub-model, since the original biomedical signal, such as the above-mentioned 29-by-2000 electroencephalogram signal, is not very sensitive to convolution, the convolution dilation rate of each layer in the first depthwise dilated convolutional network sub-model can be the same, such as all being 1, that is, the first depthwise dilated convolutional network sub-model can be an ordinary multi-layer convolutional neural network sub-model, so that while ensuring the processing effect, the processing efficiency of the model can be improved. Of course, the dilation rate of each layer of the first depthwise dilated convolutional network sub-model can also be different. For example, the model structure of the first depthwise dilated convolutional network sub-model is the same as that of the second depthwise dilated convolutional network sub-model described below. This exemplary implementation does not make special limitations on this.

[0092] The number of layers and the dilation rate of each layer of the first depthwise dilated convolutional network sub-model can be determined according to requirements or experimental results. For example, if the experimental results show that when the number of layers is 3 and the dilation rate of each layer is 1, the training result is the best, then the number of layers of the first depthwise dilated convolutional network sub-model can be 3, and the dilation rate of each layer is 1.

[0093] In the case where the first deep convolutional neural network sub-model is the first multi-level convolutional neural network sub-model, the convolutional kernels of each layer of the first multi-level convolutional neural network sub-model can be the same, but the number of convolutional kernels of each layer is different. The convolutional kernels of each layer of the first multi-level convolutional neural network sub-model can also be different. This exemplary implementation does not make special limitations on this.

[0094] In step S530, the second feature is input into the first fully connected layer, and a first fully connected feature is obtained according to the output of the first fully connected layer.

[0095] For example, the second feature can be input into the first fully connected layer, and the first fully connected layer can perform dimensionality reduction processing on the second feature to obtain the feature after dimensionality reduction.

[0096] In step S540, the first fully connected feature is processed through a preset activation function, and a first abnormal discharge detection result of the to-be-detected biomedical signal is obtained according to the processing result.

[0097] In an exemplary embodiment, the preset activation function may include a binary classification activation function, such as the sigmoid activation function. Of course, the preset activation function may also include a multi-classification activation function, such as the softmax activation function.

[0098] Taking the binary classification activation function as an example, the feature after dimensionality reduction can be mapped through the binary classification activation function to map the feature after dimensionality reduction to a probability value between 0 and 1, obtaining a first abnormal probability value, and taking the first abnormal probability value as the first abnormal discharge detection result.

[0099] Of course, the first multi-head attention submodel may not be used. For example, the original to-be-detected biomedical signal is directly input into the first deep convolutional neural network submodel, and the first abnormal discharge detection is performed according to the first deep convolutional neural network submodel. For example, the original to-be-detected biomedical signal is first input into the first deep convolutional neural network submodel to obtain output 1, then output 1 is input into the first fully connected layer to obtain output 2, and then output 2 is input into the preset activation function, and a first abnormal probability value is obtained according to the output of the preset activation function.

[0100] The deep convolutional neural network submodel in the present disclosure may also be other convolutional neural network structures, and this exemplary embodiment does not make special limitations thereon.

[0101] Through the above steps S510 to S540, the first abnormal discharge detection submodel can perform abnormal discharge detection on the original to-be-detected biomedical signal to obtain a first abnormal discharge detection result.

[0102] Next, a detailed description will be given of the specific implementation manner of "step S220, performing a form transformation on the to-be-detected biomedical signal of the subject to obtain class image biomedical signal data corresponding to the to-be-detected biomedical signal".

[0103] In an exemplary embodiment, the biomedical signal to be detected includes a first matrix, and the first matrix corresponds to a first matrix dimension. In other words, the biomedical signal to be detected can be represented by the first matrix, and the first matrix can be a matrix with N rows and M columns, such as the matrix of 29 rows multiplied by 2000 columns mentioned above. N represents the number of channels of the biomedical signal, and 2000 represents the number of samples for each channel of the biomedical signal to be detected.

[0104] Based on this, exemplarily, an embodiment of step S220 may include: performing matrix transformation on the first matrix to obtain a second matrix, the second matrix corresponding to a second matrix dimension, and the difference between the number of rows and the number of columns of the second matrix dimension being less than a preset value; determining the image-like biomedical signal data according to the second matrix.

[0105] For example, the first matrix can be converted into a second matrix with K rows and L columns through a matrix transformation function, such as the reshape function, where the absolute value of the difference between K and L is less than the preset value. The preset value can be custom-determined according to requirements, such as 0, 50, 100, etc. The preset value cannot be set too large, otherwise the difference in form between the converted matrix and the image data matrix will still be large. The purpose of setting the preset value is to make the converted matrix more suitable for convolution operations. For example, converting the above-mentioned first matrix of 29 multiplied by 2000 into a second matrix of 290 multiplied by 200. In this way, it can be ensured that the form of the second matrix is similar to the form of the image signal, and the second matrix is used as the image-like biomedical signal data of the biomedical signal to be detected.

[0106] It should be noted that in the process of converting from the first matrix to the second matrix, only the row and column dimensions of the matrix change, and the elements in the matrix do not change at all.

[0107] Next, a detailed description will be given of the specific implementation manner of "step S230, inputting the image-like biomedical signal data into the second abnormal discharge detection sub-model of the pre-trained abnormal discharge detection model, and performing a second abnormal discharge detection on the image-like biomedical signal data according to the second abnormal discharge detection sub-model".

[0108] In an exemplary embodiment, as shown in the previous Figure 4 The second abnormal discharge detection sub-model includes a second deep convolutional neural network sub-model and a second fully connected layer.

[0109] Next, a description will be given of the specific implementation manner of step S230 in combination with Figure 6 Exemplarily, Figure 6 shows a schematic flowchart of a method for performing a second abnormal discharge detection in an exemplary embodiment of the present disclosure. Refer to Figure 6, the method may include steps S610 to S630. Among them:

[0110] In step S610, the class image biomedical signal data is input into the second deep convolutional neural network sub-model, and a third feature is obtained according to the output of the second deep convolutional neural network sub-model.

[0111] In an exemplary embodiment, the second deep convolutional neural network sub-model may include a second deep dilated convolutional network sub-model and a second deep multi-level convolutional neural network sub-model.

[0112] When the second deep convolutional neural network sub-model is the second deep dilated convolutional network sub-model, the second deep dilated convolutional network sub-model may be a multi-layer dilated convolutional neural network with different dilation rates for each layer, such as a 3-layer dilated convolutional network with dilation rates of 1, 2, and 3 respectively. In this way, through the second deep dilated convolutional network sub-model, deep feature extraction can be performed on the class image biomedical signal data to obtain richer biomedical signal features.

[0113] When the second deep convolutional neural network sub-model is the second multi-level convolutional neural network sub-model, the convolutional kernels of each layer of the second multi-level convolutional neural network sub-model may be different. For example, the first layer includes 3 convolutional kernels 1, the second layer includes 2 convolutional kernels 2, and the third layer includes 1 convolutional kernel 3. Among them, convolutional kernel 1, convolutional kernel 2, and convolutional kernel 3 are all different, and this exemplary embodiment does not make special limitations on this.

[0114] In step S620, the third feature is input into the second fully connected layer, and a second fully connected feature is obtained according to the output of the second fully connected layer.

[0115] For example, the third feature can be input into the second fully connected layer, and the second fully connected layer performs dimensionality reduction processing on the third feature to obtain the dimensionality-reduced feature, that is, the second fully connected feature.

[0116] In step S630, the second fully connected feature is processed through a preset activation function, and a second abnormal discharge detection result of the biomedical signal to be detected is obtained according to the processing result.

[0117] Exemplarily, the preset activation function in step S630 and the preset activation function in step S540 may be the same. Continuing with the binary classification activation function as an example, the dimensionality-reduced feature obtained in step S620 above can be mapped through the binary classification activation function, and the dimensionality-reduced feature is mapped to a probability value between 0 and 1 to obtain a second abnormal probability value, and the second abnormal probability value is used as the second abnormal discharge detection result.

[0118] For example, the class image biomedical signal data is first input into the second deep convolutional neural network sub-model to obtain output 11, then output 11 is input into the second fully connected layer to obtain output 21, and then output 21 is input into a preset activation function, and a second abnormal probability value is obtained according to the output of the preset activation function.

[0119] Through the above steps S610 to S630, the second abnormal discharge detection sub-model can perform abnormal discharge detection on the class image biomedical signal data corresponding to the original biomedical signal to be detected, and obtain a second abnormal discharge detection result.

[0120] Next, a detailed description will be given of the specific implementation manner of "step S240, fusing the detection situations of the first abnormal discharge detection and the second abnormal discharge detection to obtain the abnormal discharge detection result of the subject to be tested".

[0121] Exemplarily, in combination with Figure 7 An exemplary implementation manner of step S240 will be described. Figure 7 The flowchart shows a method for obtaining an abnormal discharge detection result of a subject to be tested in an exemplary embodiment of the present disclosure. Refer to Figure 7 and the method may include steps S710 to S730. Among them:

[0122] In step S710, the detection situations of the first abnormal discharge detection and the second abnormal discharge detection are fused to obtain the abnormal discharge detection result of the biomedical signal to be detected.

[0123] For example, the above first abnormal discharge detection result and the second abnormal discharge detection result may be fused. As Figure 4 shown, the above first abnormal probability value and the second abnormal probability value may be fused to obtain a fusion probability value. When the fusion probability value is greater than or equal to a probability threshold, it is determined that there is an abnormal discharge in the biomedical signal to be detected; otherwise, it is determined that there is no abnormal discharge in the biomedical signal to be detected, so as to obtain the abnormal discharge detection result of the biomedical signal to be detected.

[0124] Exemplarily, for the first abnormal discharge detection and the second abnormal discharge detection, it may further include: fusing the output features of the above first fully connected layer and the output features of the above second fully connected layer to obtain fused fully connected features, and then inputting the fused fully connected features into a preset activation function, and obtaining the abnormal discharge detection result of the biomedical signal to be detected according to the probability value output by the preset activation function. If the probability value is greater than or equal to the probability threshold, it is determined that there is an abnormal discharge; otherwise, it is determined that there is no abnormal discharge.

[0125] Of course, the above second feature and third feature can also be fused to obtain a fused feature, and then a fully connected layer is used to perform dimensionality reduction processing on the fused feature to obtain the fused feature after dimensionality reduction. The fused feature after dimensionality reduction is input into a preset activation function, and an abnormal discharge detection result of the biomedical signal to be detected is obtained according to the probability value output by the preset activation function. If the probability value is greater than or equal to the probability threshold, it is determined that there is abnormal discharge; otherwise, it is determined that there is no abnormal discharge.

[0126] In step S720, when the abnormal discharge detection result of the biomedical signal to be detected is abnormal discharge, the biomedical signal to be detected is marked.

[0127] For example, when there is abnormal discharge in the biomedical signal to be detected, an abnormal mark, such as a first mark, can be added to the biomedical signal to be detected. Since the biomedical signal to be detected is obtained in segments, in this way, the time period during which the subject has abnormal discharge can be located through the abnormal mark, which is convenient for doctors to perform subsequent diagnosis and confirmation.

[0128] Exemplarily, when the abnormal discharge detection result of the biomedical signal to be detected is normal discharge, it can be not marked at all, or a normal mark, such as a second mark, can be added to it.

[0129] In step S730, an abnormal discharge detection result of the subject is obtained according to the marked biomedical signal to be detected.

[0130] Exemplarily, the time periods corresponding to the marked biomedical signal to be detected can be summarized to obtain the time periods during which the subject has abnormal discharge, so that each time period with abnormal discharge is used as the abnormal discharge detection result and fed back to the client where the doctor is located. The doctor can quickly and accurately locate the time periods that may have abnormal discharge according to the abnormal discharge detection result, and study the biomedical signal images of the time periods that may have abnormal discharge to confirm whether there is really abnormal discharge. In this way, the doctor does not need to study the entire biomedical signal image as a whole, but only needs to confirm the abnormality of the biomedical signal images of the time periods with abnormalities summarized in the detection result, which improves the doctor's work efficiency.

[0131] On the one hand, by performing abnormal discharge detection on the original biomedical signal to be detected, the present disclosure can retain the time information in the original biomedical signal to be detected. By performing abnormal discharge detection on the image-like biomedical signal data, the present disclosure can utilize the feature extraction advantage of the neural network in the image-like biomedical signal data and extract deeper biomedical signal features. The combination of the two not only retains the time information of the biomedical signal to be detected but also extracts richer biomedical signal features, which can improve the accuracy and reliability of the abnormal discharge detection of the biomedical signal to be detected. On the other hand, the abnormal discharge detection result obtained according to the method of the present disclosure can provide relatively accurate reference information for doctors, enabling doctors to quickly and accurately determine the diagnosis result of the patient to be diagnosed without carefully observing all the data, but only roughly browsing some key information and then combining the abnormal discharge detection result, which saves the workload of doctors and is convenient for wide application in clinical practice. On the other hand, based on the abnormal discharge detection model, the present disclosure can realize the automatic detection and recognition of abnormal discharge of biomedical signals and improve the efficiency of abnormal discharge detection.

[0132] In addition, it should be noted that the above-mentioned drawings are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present invention, rather than for restrictive purposes. It is easy to understand that the processes shown in the above-mentioned drawings do not indicate or limit the time sequence of these processes. Additionally, it is also easy to understand that these processes can be executed synchronously or asynchronously in, for example, multiple modules.

[0133] Exemplary Device

[0134] The exemplary embodiment of the present disclosure further provides an abnormal discharge detection device. Refer to Figure 8 As shown, the abnormal discharge detection device 800 may include the following program modules: a first detection module 810, configured to determine the biomedical signal to be detected of the subject to be tested, input the biomedical signal to be detected into the first abnormal discharge detection sub-model in the pre-trained abnormal discharge detection model, and perform the first abnormal discharge detection on the biomedical signal to be detected according to the first abnormal discharge detection sub-model; a signal preprocessing module 820, configured to perform a form transformation on the biomedical signal to be detected of the subject to be tested to obtain the image-like biomedical signal data corresponding to the biomedical signal to be detected; a second detection module 830, configured to input the image-like biomedical signal data into the second abnormal discharge detection sub-model in the pre-trained abnormal discharge detection model, and perform the second abnormal discharge detection on the image-like biomedical signal data according to the second abnormal discharge detection sub-model; a fusion module 840, configured to fuse the detection situations of the first abnormal discharge detection and the second abnormal discharge detection to obtain the abnormal discharge detection result of the subject to be tested.

[0135] In an exemplary embodiment, determining a biomedical signal to be detected of a subject to be tested includes: acquiring the biomedical signal of the subject to be tested that is collected; segmenting the biomedical signal of the subject to be tested according to a preset duration, and determining at least one biomedical signal to be detected of the subject to be tested.

[0136] In an exemplary embodiment, fusing the detection situations of the first abnormal discharge detection and the second abnormal discharge detection to obtain the abnormal discharge detection result of the subject to be tested includes: fusing the detection situations of the first abnormal discharge detection and the second abnormal discharge detection to obtain the abnormal discharge detection result of the biomedical signal to be detected; marking the biomedical signal to be detected when the abnormal discharge detection result of the biomedical signal to be detected is abnormal discharge; and obtaining the abnormal discharge detection result of the subject to be tested according to the marked biomedical signal to be detected.

[0137] In an exemplary embodiment, the first abnormal discharge detection sub-model includes a first multi-head attention sub-model, a first deep convolutional neural network sub-model, and a first fully connected layer; inputting the biomedical signal to be detected into the first abnormal discharge detection sub-model of a pre-trained abnormal discharge detection model, and performing the first abnormal discharge detection on the biomedical signal to be detected according to the first abnormal discharge detection sub-model includes: inputting the biomedical signal to be detected into the first multi-head attention sub-model, and obtaining a first feature of the biomedical signal to be detected according to the output of the first multi-head attention sub-model; inputting the first feature into the first deep convolutional neural network sub-model, and obtaining a second feature according to the output of the first deep convolutional neural network sub-model; inputting the second feature into the first fully connected layer, and obtaining a first fully connected feature according to the output of the first fully connected layer; processing the first fully connected feature through a preset activation function, and obtaining the first abnormal discharge detection result of the biomedical signal to be detected according to the processing result.

[0138] In an exemplary embodiment, the second abnormal discharge detection sub-model includes a second deep convolutional neural network sub-model and a second fully connected layer; inputting the class image biomedical signal data into the second abnormal discharge detection sub-model of the pre-trained abnormal discharge detection model, and performing second abnormal discharge detection on the class image biomedical signal data according to the second abnormal discharge detection sub-model includes: inputting the class image biomedical signal data into the second deep convolutional neural network sub-model, and obtaining a third feature according to the output of the second deep convolutional neural network sub-model; inputting the third feature into the second fully connected layer, and obtaining a second fully connected feature according to the output of the second fully connected layer; processing the second fully connected feature through a preset activation function, and obtaining a second abnormal discharge detection result of the biomedical signal to be detected according to the processing result.

[0139] In an exemplary embodiment, the biomedical signal to be detected includes a first matrix, and the first matrix corresponds to a first matrix dimension. Performing a form transformation on the biomedical signal to be detected of the subject to obtain class image biomedical signal data corresponding to the biomedical signal to be detected; performing matrix transformation on the first matrix to obtain a second matrix, and the second matrix corresponds to a second matrix dimension, and the difference between the number of rows and the number of columns of the second matrix dimension is less than a preset value; determining the class image biomedical signal data according to the second matrix.

[0140] In an exemplary embodiment, any one of the deep convolutional neural network sub-models includes any one of a dilated deep convolutional network sub-model and a multi-level convolutional neural network sub-model.

[0141] The specific details of each part in the above device have been described in detail in the corresponding method part of the above embodiments. The details not disclosed can be seen in the embodiments of the above method part, and thus will not be repeated.

[0142] Exemplary Storage Medium

[0143] The storage medium of the exemplary embodiment of the present invention will be described below.

[0144] In this exemplary embodiment, the above method can be implemented by a program product. For example, a portable compact disc read-only memory (CD-ROM) can be used and includes program code, and can be run on a device, such as a personal computer. However, the program product of the present invention is not limited thereto. In this document, a readable storage medium can be any tangible medium that contains or stores a program, and the program can be used by or in combination with an instruction execution system, device, or device.

[0145] The program product may employ any combination of one or more readable media. The readable media may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0146] The computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which the readable program code is carried. Such a propagated data signal may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. The readable signal medium may also be any readable medium other than the readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device.

[0147] The program code contained on the readable medium may be transmitted by any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0148] The program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages such as Java, C++, etc., and also including conventional procedural programming languages such as the "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user computing device and partially on a remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., by using an Internet service provider to connect through the Internet).

[0149] Exemplary Computer Program Product

[0150] The exemplary embodiments of the present disclosure also provide a computer program product. The computer program product includes a computer program, which when executed by a processor implements the above-mentioned abnormal discharge detection method.

[0151] In one embodiment, a computer program product may be a tangible product containing a computer program, such as a computer-readable storage medium storing the computer program. The readable storage medium may be a storage medium based on signals such as electricity, magnetism, light, electromagnetic, infrared, etc., including but not limited to: random access memory (RAM), read-only memory (ROM), magnetic tape, floppy disk, flash memory, mechanical hard disk (HDD), solid state drive (SSD), and so on. Exemplarily, the computer program product may be implemented as a non-volatile storage medium storing the computer program, such as read-only memory, Nand Flash, etc.

[0152] In one embodiment, a computer program product may be an intangible product containing a computer program. Exemplarily, the computer program product may be implemented as a virtual digital product, such as an executable file storing the computer program, digital files such as installation packages.

[0153] The code of the computer program can be written in one or more programming languages. Programming languages such as C, Java, C++, Python, etc. The program code can be executed entirely on the user's computing device, or partially on the user's computing device, or executed as an independent software package, or partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving a remote computing device, the remote computing device can be connected to the user's computing device through any type of network, such as a local area network (LAN), wide area network (WAN), etc., or can be connected to an external computing device (for example, through an Internet connection provided by an operator).

[0154] Computer programs can be carried or transmitted by means of electrical, magnetic, optical, electromagnetic, infrared, etc. signals. An electronic device can convert the signal carrying the computer program into a digital signal and then run the computer program. When the computer program runs on the electronic device, its code is used to cause the electronic device to execute (more specifically, to cause the processor of the electronic device to execute) the method steps of various exemplary embodiments of the present disclosure. For example, the above-mentioned abnormal discharge detection method can be executed, which includes the following steps: determining the biomedical signal to be detected of the subject to be tested, inputting the biomedical signal to be detected into the first abnormal discharge detection sub-model in the pre-trained abnormal discharge detection model, and performing the first abnormal discharge detection on the biomedical signal to be detected according to the first abnormal discharge detection sub-model; performing a formal transformation on the biomedical signal to be detected of the subject to be tested to obtain the image-like biomedical signal data corresponding to the biomedical signal to be detected; inputting the image-like biomedical signal data into the second abnormal discharge detection sub-model in the pre-trained abnormal discharge detection model, and performing the second abnormal discharge detection on the image-like biomedical signal data according to the second abnormal discharge detection sub-model; fusing the detection results of the first abnormal discharge detection and the second abnormal discharge detection to obtain the abnormal discharge detection result of the subject to be tested.

[0155] By executing the above method steps through a computer program, on the one hand, by performing abnormal discharge detection on the original biomedical signal to be detected, the time information in the original biomedical signal to be detected can be retained. By performing abnormal discharge detection on the image-like biomedical signal data, the advantage of feature extraction of the neural network can be exerted in the image-like biomedical signal data, and deeper biomedical signal features can be extracted. The combination of the two not only retains the time information of the biomedical signal to be detected but also extracts richer biomedical signal features, which can improve the accuracy and reliability of the abnormal discharge detection of the biomedical signal to be detected. On the other hand, the abnormal discharge detection result obtained according to the method of the present disclosure can provide relatively accurate reference information for doctors, enabling doctors to quickly and accurately judge the diagnosis result of the patient to be seen without carefully observing all the data, but only roughly browsing some key information and then combining the abnormal discharge detection result, saving the workload of doctors and facilitating wide application in clinical practice. On the other hand, based on the abnormal discharge detection model, automatic detection and recognition of abnormal discharges in biomedical signals can be realized, improving the efficiency of abnormal discharge detection.

[0156] Exemplary Electronic Device

[0157] Reference Figure 9An electronic device according to an exemplary embodiment of the present disclosure will be described. The electronic device is the above-mentioned terminal device 110 or server 120. The electronic device may include a processor and a memory. The memory stores executable instructions of the processor, such as a computer program. The processor executes the method steps of various exemplary embodiments of the present disclosure by executing the executable instructions. In addition, the electronic device may further include a display for displaying a graphical user interface.

[0158] Reference will be made below Figure 9 , and the electronic device will be described by way of example in the form of a general-purpose computing device. It should be understood that Figure 9 the electronic device 900 shown is merely an example and should not impose limitations on the functions and scope of use of the embodiments of the present disclosure.

[0159] As Figure 9 shown, the electronic device 900 may include: a processor 910, a memory 920, a bus 930, an I / O (input / output) interface 940, a network adapter 950, and a display 960.

[0160] The memory 920 may include volatile memory, such as RAM 921 and a cache unit 922, and may also include non-volatile memory, such as ROM 923. The memory 920 may further include one or more program modules 924, and such program modules 924 include but are not limited to: an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include the implementation of a network environment. For example, the program module 924 may include each module in the above-mentioned device.

[0161] The processor 910 may include one or more processing units. For example, the processor 910 may include an AP (Application Processor), a modem processor, a GPU (Graphics Processing Unit), an ISP (Image Signal Processor), a controller, an encoder, a decoder, a DSP (Digital Signal Processor), a baseband processor, and / or an NPU (Neural-Network Processing Unit), etc.

[0162] The processor 910 can be used to execute executable instructions stored in the memory 920. For example, it can execute the above-mentioned abnormal discharge detection method, which includes the following steps: determining a biomedical signal to be detected of a subject to be tested, inputting the biomedical signal to be detected into a first abnormal discharge detection sub-model in a pre-trained abnormal discharge detection model, and performing a first abnormal discharge detection on the biomedical signal to be detected according to the first abnormal discharge detection sub-model; performing a formal transformation on the biomedical signal to be detected of the subject to be tested to obtain class image biomedical signal data corresponding to the biomedical signal to be detected; inputting the class image biomedical signal data into a second abnormal discharge detection sub-model in the pre-trained abnormal discharge detection model, and performing a second abnormal discharge detection on the class image biomedical signal data according to the second abnormal discharge detection sub-model; fusing the detection situations of the first abnormal discharge detection and the second abnormal discharge detection to obtain an abnormal discharge detection result of the subject to be tested.

[0163] Implementing the above method through a computer program, on the one hand, by performing abnormal discharge detection on the original biomedical signal to be detected, the time information in the original biomedical signal to be detected can be retained. By performing abnormal discharge detection on the class image biomedical signal data, the advantage of feature extraction of the neural network can be utilized in the class image biomedical signal data to extract deeper biomedical signal features. The combination of the two not only retains the time information of the biomedical signal to be detected but also extracts richer biomedical signal features, which can improve the accuracy and reliability of the abnormal discharge detection of the biomedical signal to be detected. On the other hand, the abnormal discharge detection result obtained according to the method of the present disclosure can provide relatively accurate reference information for doctors, enabling doctors to quickly and accurately judge the diagnosis result of the patient to be seen without carefully observing all the data, but only roughly browsing some key information and then combining it with the abnormal discharge detection result, saving the workload of doctors and facilitating wide application in clinical practice. On the other hand, based on the abnormal discharge detection model, automatic detection and recognition of abnormal discharges in biomedical signals can be realized, improving the efficiency of abnormal discharge detection.

[0164] The bus 930 is used to realize the connection between different components of the electronic device 900 and can include a data bus, an address bus, and a control bus.

[0165] The electronic device 900 can communicate with one or more external devices 1000 (such as a keyboard, a mouse, an external controller, etc.) through the I / O interface 940.

[0166] The electronic device 900 can communicate with one or more networks through the network adapter 950. For example, the network adapter 950 can provide mobile communication solutions such as 3G / 4G / 5G, or wireless communication solutions such as wireless local area network, Bluetooth, and near field communication. The network adapter 950 can communicate with other modules of the electronic device 900 through the bus 930.

[0167] The electronic device 900 can display a graphical user interface through the display 960, such as a graphical user interface for displaying the abnormal discharge detection result, etc.

[0168] Although Figure 9 not shown in the figure, other hardware and / or software modules can also be set in the electronic device 900, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0169] In addition, the above-mentioned drawings are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present disclosure, rather than for limiting purposes. It is easy to understand that the processes shown in the above-mentioned drawings do not indicate or limit the time sequence of these processes. Additionally, it is also easy to understand that these processes can be executed synchronously or asynchronously in multiple modules, for example.

[0170] It should be understood that the present disclosure is not limited to the specific method steps or structures that have been described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. Those skilled in the art will easily think of other embodiments based on the specific implementation manners provided by the present disclosure. Therefore, the specific implementation manners provided by the present disclosure are only exemplary, and the scope and spirit of the present disclosure are pointed out by the claims, and should cover any variations, uses, or adaptations of the present disclosure, and these variations, uses, or adaptations follow the general principles of the present disclosure and include well-known common general knowledge or conventional technical means in the technical field not disclosed by the present disclosure.

Claims

1. An abnormal discharge detection method, characterized in that: include; Determine a biomedical signal to be detected of the subject, input the biomedical signal to be detected represented in the form of a first matrix into a first abnormal discharge detection sub-model in a pre-trained abnormal discharge detection model, and perform a first abnormal discharge detection on the biomedical signal to be detected according to the first abnormal discharge detection sub-model; Performing matrix deformation on the biomedical signal to be detected of the subject represented in the form of a first matrix to obtain image-like biomedical signal data corresponding to the biomedical signal to be detected, wherein the image-like biomedical signal data is represented by a second matrix, wherein the first matrix corresponds to a first matrix dimension, which is determined according to the number of sampling channels and the sampling frequency of the biomedical signal to be detected, and the second matrix corresponds to a second matrix dimension, wherein the difference between the number of rows and the number of columns of the second matrix dimension is less than a preset value; Inputting the image-like biomedical signal data into a second abnormal discharge detection submodel in a pre-trained abnormal discharge detection model, and performing a second abnormal discharge detection on the image-like biomedical signal data according to the second abnormal discharge detection submodel; The detection conditions of the first abnormal discharge detection and the second abnormal discharge detection are integrated to obtain an abnormal discharge detection result of the detected object; Among them, the first abnormal discharge detection submodel includes a first multi-head attention submodel, a first deep convolutional neural network submodel and a first fully connected layer; the second abnormal discharge detection submodel includes a second deep convolutional neural network submodel and a second fully connected layer; the detection result of the first abnormal discharge detection includes a first abnormal probability value, and the detection result of the second abnormal discharge detection includes a second abnormal probability value, and the fusion of the detection conditions of the first abnormal discharge detection and the second abnormal discharge detection to obtain the abnormal discharge detection result of the tested object includes: fusing the first abnormal probability value and the second abnormal probability value to obtain a fused probability value, and obtaining the abnormal discharge detection result of the tested object according to the fused probability value; or fusing the output features of the first fully connected layer and the output features of the second fully connected layer to obtain a fused connection feature, inputting the fused connection feature into a preset activation function, and obtaining the tested object according to the probability value output by the preset activation function. The abnormal discharge detection result of the image; or the biomedical signal to be detected is input into the first multi-head attention sub-model, and the first feature of the biomedical signal to be detected is obtained according to the output of the first multi-head attention sub-model, and the first feature is input into the first deep convolutional neural network sub-model, and the second feature is obtained according to the output of the first deep convolutional neural network sub-model, and the image-like biomedical signal data is input into the second deep convolutional neural network sub-model, and the third feature is obtained according to the output of the second deep convolutional neural network sub-model, and the second feature output by the first deep convolutional neural network sub-model and the third feature output by the second deep convolutional neural network sub-model are fused to obtain a fused feature, and the fused feature is reduced in dimension through a fully connected layer to obtain the reduced fused feature, and the reduced fused feature is input into a preset activation function, and the abnormal discharge detection result of the biomedical signal to be detected is obtained according to the probability value output by the preset activation function.

2. The method according to claim 1, characterized in that Determining the biomedical signal to be detected of the subject includes: Acquiring the collected biomedical signals of the subject; The biomedical signals of the subject are acquired in segments according to a preset time length, and at least one biomedical signal to be detected of the subject is determined.

3. The method according to claim 1 or 2, characterized in that: The step of fusing the detection conditions of the first abnormal discharge detection and the second abnormal discharge detection to obtain the abnormal discharge detection result of the detected object includes: fusing the detection results of the first abnormal discharge detection and the second abnormal discharge detection to obtain an abnormal discharge detection result of the biomedical signal to be detected; When the abnormal discharge detection result of the biomedical signal to be detected is abnormal discharge, marking the biomedical signal to be detected; The abnormal discharge detection result of the detected object is obtained according to the marked biomedical signal to be detected.

4. The method according to claim 1, characterized in that: The performing first abnormal discharge detection on the biomedical signal to be detected according to the first abnormal discharge detection sub-model includes: Inputting the second feature into the first fully connected layer, and obtaining a first fully connected feature according to an output of the first fully connected layer; The first fully connected feature is processed by a preset activation function, and a first abnormal discharge detection result of the biomedical signal to be detected is obtained according to the processing result.

5. The method according to claim 1, characterized in that The performing second abnormal discharge detection on the image-like biomedical signal data according to the second abnormal discharge detection sub-model includes: Inputting the third feature into the second fully connected layer, and obtaining a second fully connected feature according to an output of the second fully connected layer; The second fully connected feature is processed by a preset activation function, and a second abnormal discharge detection result of the biomedical signal to be detected is obtained according to the processing result.

6. The method according to claim 4 or 5, characterized in that: Any of the deep convolutional neural network sub-models includes any one of a deep dilated convolutional network sub-model and a multi-layer convolutional neural network sub-model.

7. An abnormal discharge detection device, characterized in that: include; A first detection module is configured to determine a biomedical signal to be detected of the subject, input the biomedical signal to be detected represented in the form of a first matrix into a first abnormal discharge detection sub-model in a pre-trained abnormal discharge detection model, and perform a first abnormal discharge detection on the biomedical signal to be detected according to the first abnormal discharge detection sub-model; A signal preprocessing module, configured to perform matrix deformation on the biomedical signal to be detected of the subject represented in the form of a first matrix, to obtain image-like biomedical signal data corresponding to the biomedical signal to be detected, wherein the image-like biomedical signal data is represented by a second matrix, wherein the first matrix corresponds to a first matrix dimension, which is determined according to the number of sampling channels and the sampling frequency of the biomedical signal to be detected, and the second matrix corresponds to a second matrix dimension, wherein the difference between the number of rows and the number of columns of the second matrix dimension is less than a preset value; a second detection module, configured to input the image-like biomedical signal data into a second abnormal discharge detection sub-model in a pre-trained abnormal discharge detection model, and perform a second abnormal discharge detection on the image-like biomedical signal data according to the second abnormal discharge detection sub-model; a fusion module configured to fuse the detection conditions of the first abnormal discharge detection and the second abnormal discharge detection to obtain an abnormal discharge detection result of the detected object; Among them, the first abnormal discharge detection submodel includes a first multi-head attention submodel, a first deep convolutional neural network submodel and a first fully connected layer; the second abnormal discharge detection submodel includes a second deep convolutional neural network submodel and a second fully connected layer; the detection result of the first abnormal discharge detection includes a first abnormal probability value, and the detection result of the second abnormal discharge detection includes a second abnormal probability value, and the fusion of the detection conditions of the first abnormal discharge detection and the second abnormal discharge detection to obtain the abnormal discharge detection result of the tested object includes: fusing the first abnormal probability value and the second abnormal probability value to obtain a fused probability value, and obtaining the abnormal discharge detection result of the tested object according to the fused probability value; or fusing the output features of the first fully connected layer and the output features of the second fully connected layer to obtain a fused connection feature, inputting the fused connection feature into a preset activation function, and obtaining the tested object according to the probability value output by the preset activation function. The abnormal discharge detection result of the image; or the biomedical signal to be detected is input into the first multi-head attention sub-model, and the first feature of the biomedical signal to be detected is obtained according to the output of the first multi-head attention sub-model, and the first feature is input into the first deep convolutional neural network sub-model, and the second feature is obtained according to the output of the first deep convolutional neural network sub-model, and the image-like biomedical signal data is input into the second deep convolutional neural network sub-model, and the third feature is obtained according to the output of the second deep convolutional neural network sub-model, and the second feature output by the first deep convolutional neural network sub-model and the third feature output by the second deep convolutional neural network sub-model are fused to obtain a fused feature, and the fused feature is reduced in dimension through a fully connected layer to obtain the reduced fused feature, and the reduced fused feature is input into a preset activation function, and the abnormal discharge detection result of the biomedical signal to be detected is obtained according to the probability value output by the preset activation function.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

9. An electronic device, characterized in that: include: one or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, enables the one or more processors to implement the method according to any one of claims 1 to 6.

Images