A method and apparatus for processing electrocardiosignal

Abstract

Embodiments of the present application relate to a kind of electrocardiosignal processing method and device, the method includes: receiving first electrocardiosignal;Filtering processing generates second electrocardiosignal;Baseline drift elimination processing is respectively carried out to first, second electrocardiosignal, and third, fourth electrocardiosignal is generated;R point identification processing is carried out to fourth electrocardiosignal, and first R point sequence is generated;Electrocardiosignal segment intercepting processing is carried out to third, fourth electrocardiosignal, and first, second electrocardiosignal segment sequence is generated;Heart beat classification processing is carried out to first electrocardiosignal segment sequence, and first segment type sequence is generated;Interference classification processing is carried out to second electrocardiosignal segment sequence, and second segment type sequence is generated;Segment type fusion processing is carried out to first, second segment type sequence, and third segment type sequence is generated;First R point sequence and third segment type sequence are as electrocardiosignal processing result output.The processing efficiency of electrocardiosignal processing can be improved in the present application and the stability of processing quality is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and apparatus for processing electrocardiogram signals. Background Technology

[0002] Conventional electrocardiogram (ECG) signal analysis mainly relies on manual analysis. Due to limitations in human experience and workload, the quality of this conventional processing method cannot be guaranteed, and the processing efficiency is also low. Summary of the Invention

[0003] The purpose of this invention is to address the shortcomings of existing technologies by providing a method, apparatus, electronic device, and computer-readable storage medium for electrocardiogram (ECG) signal processing. The method involves first filtering the input raw ECG signal to generate a corresponding filtered ECG signal; then performing baseline drift elimination processing on both the raw and filtered ECG signals; next, identifying R-point sequences based on the baseline-drift-eliminated filtered ECG signals; further, extracting ECG signal segments from the baseline-drift-eliminated raw and filtered ECG signals according to different ECG signal segments based on the R-point sequences to generate corresponding raw and filtered ECG segment sequences; then, identifying corresponding heartbeat and interference types based on different artificial intelligence models on the raw and filtered ECG segment sequences; finally, performing type fusion processing on the identified heartbeat and interference types and outputting the final fusion result as the ECG signal processing result. This invention improves processing efficiency and ensures the stability of processing quality.

[0004] To achieve the above objectives, a first aspect of the present invention provides an electrocardiogram (ECG) signal processing method, the method comprising:

[0005] Receive the first electrocardiogram signal;

[0006] The first electrocardiogram (ECG) signal is filtered to generate a corresponding second ECG signal; and the first and second ECG signals are respectively processed to eliminate baseline drift to generate corresponding third and fourth ECG signals.

[0007] The fourth electrocardiogram signal is subjected to R-point identification processing to generate a corresponding first R-point sequence; the first R-point sequence includes multiple first R-points.

[0008] Based on the first R-point sequence, the third and fourth ECG signals are processed to extract ECG signal segments, generating corresponding first and second ECG segment sequences;

[0009] The first ECG segment sequence is processed by heartbeat classification to generate a corresponding first segment type sequence; and the second ECG segment sequence is processed by interference classification to generate a corresponding second segment type sequence.

[0010] The first and second fragment type sequences are fused to generate a corresponding third fragment type sequence.

[0011] The first R-point sequence and the third segment type sequence are output as the ECG signal processing results.

[0012] Preferably, the step of performing R-point identification processing on the fourth electrocardiogram signal to generate a corresponding first R-point sequence specifically includes:

[0013] The fourth electrocardiogram signal is subjected to QRS wave recognition processing to generate multiple corresponding first QRS wave signals; and each first QRS wave signal is subjected to R point recognition processing to generate corresponding first R points; and all the obtained first R points are sorted in chronological order to generate a corresponding first R point sequence.

[0014] Preferably, the step of performing ECG signal segmentation processing on the third and fourth ECG signals based on the first R-point sequence to generate corresponding first and second ECG segment sequences specifically includes:

[0015] On the fourth electrocardiogram (ECG) signal, an ECG signal segment of a preset first length is extracted with each of the first R points as the center point as the corresponding first ECG segment; and the obtained first ECG segments are sorted in chronological order to generate the corresponding first ECG segment sequence.

[0016] On the third electrocardiogram (ECG) signal, an ECG signal segment of a preset second length is extracted with each of the first R points as the center point as the corresponding second ECG segment; and all the obtained second ECG segments are sorted in chronological order to generate a corresponding second ECG segment sequence; the second length is less than the first length.

[0017] Preferably, the step of performing heartbeat classification processing on the first ECG segment sequence to generate a corresponding first segment type sequence specifically includes:

[0018] The first ECG segment of the first ECG segment sequence is traversed; during traversal, the first ECG segment currently traversed is fed into a preset artificial intelligence model for one-dimensional ECG feature extraction and heartbeat type classification based on the extracted ECG features to generate the corresponding first segment type; at the end of the traversal, all the obtained first segment types are sorted in chronological order to generate the corresponding first segment type sequence.

[0019] Furthermore, the aforementioned artificial intelligence model includes a first input module, a first fully connected network, a first classifier, and a first output module; the first input module is connected to the first fully connected network; the first fully connected network is connected to the first classifier; the first classifier is connected to the first output module; the first fully connected network includes one or more first fully connected layers.

[0020] Furthermore, the step of feeding the currently traversed first ECG segment into a preset type of artificial intelligence model for one-dimensional ECG feature extraction and performing heartbeat type classification processing based on the extracted ECG features to generate the corresponding first segment type specifically includes:

[0021] The first ECG segment is fed into the aforementioned artificial intelligence model; the first input module performs a one-dimensional model vector transformation on the first ECG segment to generate a corresponding first input vector, which is then output to the first fully connected network; the first fully connected network performs a fully connected operation based on the first input vector to generate a corresponding first ECG feature vector, which is then output to the first classifier; the first classifier performs heartbeat type classification processing based on the first ECG feature vector to generate a corresponding first classification vector, which is then output to the first output module; and the first output module extracts the heartbeat classification type corresponding to the first vector data with the highest classification probability value from the first classification vector as the corresponding first segment type, which is then output; the first classification vector includes multiple first vector data; each first vector data corresponds to a heartbeat classification type, and the first vector data represents the classification probability of the corresponding heartbeat classification type.

[0022] Preferably, the step of performing interference classification processing on the second ECG segment sequence to generate a corresponding second segment type sequence specifically includes:

[0023] The second ECG segment of the second ECG segment sequence is traversed; during traversal, the currently traversed second ECG segment is fed into a preset two-type artificial intelligence model for one-dimensional interference feature extraction, and interference type classification is performed based on the extracted interference features to generate the corresponding second segment type; at the end of the traversal, all the obtained second segment types are sorted in chronological order to generate the corresponding second segment type sequence.

[0024] Furthermore, the second type of artificial intelligence model comprises a second input module, a first convolutional neural network, a first data conversion module, a second classifier, and a second output module; the second input module is connected to the first convolutional neural network; the first convolutional neural network is connected to the first data conversion module; the first data conversion module is connected to the second classifier; and the second classifier is connected to the second output module.

[0025] Furthermore, the step of feeding the currently traversed second ECG segment into a preset two-type artificial intelligence model for one-dimensional interference feature extraction and performing interference type classification based on the extracted interference features to generate the corresponding second segment type specifically includes:

[0026] The current second ECG segment is fed into the two-class artificial intelligence model; the second input module performs high-dimensional tensor transformation on the first ECG segment to generate a corresponding first input tensor, which is output to the first convolutional neural network; the first convolutional neural network performs interference feature extraction on the first input tensor to generate a corresponding first feature tensor, which is output to the first data conversion module; the first data conversion module performs one-dimensional vector transformation on the first feature tensor to generate a corresponding first feature vector, which is output to the second classifier; the second classifier performs interference type classification based on the first feature vector to generate a corresponding second classification vector, which is output to the second output module; and the second output module extracts the interference classification type corresponding to the second vector data with the highest classification probability value from the second classification vector as the corresponding second segment type, which is output; the second classification vector includes multiple second vector data; each second vector data corresponds to one interference classification type, and the second vector data is the classification probability of the corresponding interference classification type; the interference classification type includes non-interference type, weak interference type, and strong interference type.

[0027] Preferably, the step of performing fragment type fusion processing on the first and second fragment type sequences to generate the corresponding third fragment type sequence specifically includes:

[0028] The first and second fragment types with the same sequence index in the first and second fragment type sequences are taken as the corresponding first fragment type pairs; and the fragment type fusion processing of each first fragment type pair is performed according to the preset fusion mode to generate the corresponding third fragment type; and all the obtained third fragment types are sorted in chronological order to generate the corresponding third fragment type sequence.

[0029] Furthermore, the step of performing fragment type fusion processing on each of the first fragment type pairs according to a preset fusion mode to generate a corresponding third fragment type specifically includes:

[0030] The fusion pattern is identified;

[0031] When the fusion mode is the first mode, it is determined whether the second segment type of the current first segment type pair is a non-interference type; if yes, the corresponding third segment type is set to the first segment type of the current first segment type pair; if no, the corresponding third segment type is set to the interference segment type.

[0032] When the fusion mode is the second mode, it is determined whether the second segment type of the current first segment type pair is a strong interference type; if yes, the corresponding third segment type is set as the interference segment type; if no, the corresponding third segment type is set as the first segment type of the current first segment type pair.

[0033] A second aspect of the present invention provides an apparatus for implementing the method described in the first aspect above, the apparatus comprising: a receiving module, a preprocessing module, an R-point identification module, an electrocardiogram signal segment extraction and processing module, an artificial intelligence model processing module, a segment type fusion processing module, and an output module;

[0034] The receiving module is used to receive the first electrocardiogram signal;

[0035] The preprocessing module is used to filter the first electrocardiogram (ECG) signal to generate a corresponding second ECG signal; and to perform baseline drift elimination processing on the first and second ECG signals to generate corresponding third and fourth ECG signals, respectively.

[0036] The R-point identification module is used to perform R-point identification processing on the fourth electrocardiogram signal to generate a corresponding first R-point sequence; the first R-point sequence includes multiple first R-points;

[0037] The ECG signal segment extraction and processing module is used to perform ECG signal segment extraction and processing on the third and fourth ECG signals according to the first R-point sequence to generate corresponding first and second ECG segment sequences.

[0038] The artificial intelligence model processing module is used to perform heartbeat classification processing on the first ECG segment sequence to generate a corresponding first segment type sequence; and to perform interference classification processing on the second ECG segment sequence to generate a corresponding second segment type sequence;

[0039] The fragment type fusion processing module is used to perform fragment type fusion processing on the first and second fragment type sequences to generate a corresponding third fragment type sequence;

[0040] The output module is used to output the first R-point sequence and the third segment type sequence as the electrocardiogram signal processing result.

[0041] A third aspect of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;

[0042] The processor is used to couple with the memory, read and execute instructions in the memory to implement the steps of the method described in the first aspect above;

[0043] The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

[0044] A fourth aspect of the present invention provides a computer-readable storage medium storing computer instructions that, when executed by a computer, cause the computer to perform the instructions described in the first aspect.

[0045] This invention provides a method, apparatus, electronic device, and computer-readable storage medium for electrocardiogram (ECG) signal processing. The invention first filters the input raw ECG signal to generate a corresponding filtered ECG signal. Then, baseline drift elimination processing is performed on both the raw and filtered ECG signals. Next, R-point identification is performed on the baseline-drift-eliminated filtered ECG signal to obtain a corresponding R-point sequence. Then, based on the R-point sequence, ECG signal segments are extracted from the baseline-drift-eliminated raw and filtered ECG signals according to different ECG signal segments to generate corresponding raw and filtered ECG segment sequences. Next, different artificial intelligence models are used to identify the corresponding heartbeat and interference types in the raw and filtered ECG segment sequences. Finally, type fusion processing is performed on the identified heartbeat and interference types, and the final fusion result is output as the ECG signal processing result. This invention improves the processing efficiency of ECG signal processing and ensures the stability of processing quality. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of an electrocardiogram signal processing method provided in Embodiment 1 of the present invention;

[0047] Figure 2 This is a module structure diagram of an electrocardiogram signal processing device provided in Embodiment 2 of the present invention;

[0048] Figure 3 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0050] Embodiment 1 of the present invention provides a method for processing electrocardiogram signals. Figure 1 This is a schematic diagram of an electrocardiogram signal processing method provided in Embodiment 1 of the present invention, as shown below. Figure 1 As shown, this method mainly includes the following steps:

[0051] Step 1: Receive the first electrocardiogram signal.

[0052] Here, the first electrocardiogram signal is the electrocardiogram (ECG) waveform signal.

[0053] Step 2: Filter the first ECG signal to generate the corresponding second ECG signal; and perform baseline drift elimination processing on the first and second ECG signals respectively to generate the corresponding third and fourth ECG signals;

[0054] Specifically, this includes: Step 21, filtering the first electrocardiogram signal to generate a corresponding second electrocardiogram signal;

[0055] Here, filtering the first ECG signal is performed to eliminate power frequency noise, electromyographic noise, etc. in the original ECG signal, i.e., the first ECG signal. When performing filtering, this embodiment of the invention first eliminates white noise based on a bandpass filter, and then filters power frequency noise and electromyographic noise based on wavelet transform. The filtered ECG signal obtained after filtering is the second ECG signal.

[0056] Step 22: Baseline drift elimination processing is performed on the first and second ECG signals to generate the corresponding third and fourth ECG signals.

[0057] Here, the embodiments of the present invention support multiple baseline drift elimination methods; one of which is a baseline drift elimination processing method based on median filtering; another of which is a baseline drift elimination processing method based on wavelet transform; and yet another of which is a baseline drift elimination processing method based on a simple integer coefficient filter. It should be noted that, in this embodiment of the invention, a pre-set ECG duration threshold is used to identify long-duration ECGs. If the duration of the first ECG signal exceeds the preset ECG duration threshold, a baseline drift elimination processing method based on wavelet transform is preferably used to perform baseline drift elimination processing on the first and second ECG signals. If the duration of the first ECG signal does not exceed the ECG duration threshold, a baseline drift elimination processing method based on median filtering or a simple integer coefficient filter is selected to perform baseline drift elimination processing on the first and second ECG signals. It should be further noted that, because the processing method based on simple integer coefficient filters is less likely to cause waveform distortion than the processing method based on median filtering, when the duration of the first ECG signal does not exceed the ECG duration threshold, a baseline drift elimination processing method based on simple integer coefficient filters is preferred for the second ECG signal, while a baseline drift elimination processing method based on median filtering or a simple integer coefficient filter is selected for the first ECG signal. The signal obtained after baseline drift elimination processing of the original ECG signal (i.e., the first ECG signal) is the third ECG signal. The signal obtained after baseline drift elimination processing of the filtered ECG signal (i.e., the second ECG signal) is the fourth ECG signal. The fourth ECG signal will be used in subsequent steps for heart rate classification. The third ECG signal, compared to the fourth ECG signal, still contains original noise interference, which will be used in subsequent steps for interference classification.

[0058] Step 3: Perform R-point identification processing on the fourth ECG signal to generate the corresponding first R-point sequence;

[0059] The sequence of first R points includes multiple first R points;

[0060] Specifically, this includes: performing QRS wave recognition processing on the fourth electrocardiogram signal to generate multiple corresponding first QRS wave signals; performing R-point recognition processing on each first QRS wave signal to generate corresponding first R points; and sorting all the obtained first R points in chronological order to generate a corresponding first R point sequence.

[0061] Here, based on the characteristics of ECG signals, we know that a segment of an electrocardiogram signal consists of multiple heartbeat signals, each corresponding to one heartbeat. Typical waveforms of heartbeat signals include P waves, QRS waves, T waves, U waves, etc.; among them, the QRS wave is actually a QRS complex composed of Q, R, and S waves. The R wave in the QRS complex has the highest peak value, and the corresponding peak point is the R point. Based on the above description, in the current step of detecting the R point of the fourth electrocardiogram signal, the QRS wave is first detected to obtain all QRS complex signal segments in the electrocardiogram signal, which is the first QRS wave signal. Then, the R point is identified for each first QRS wave signal to obtain the corresponding first R point. Finally, the first R point sequence is generated by sorting the first R points.

[0062] Step 4: Based on the first R-point sequence, perform ECG signal segmentation processing on the third and fourth ECG signals to generate the corresponding first and second ECG segment sequences;

[0063] Specifically, it includes: Step 41, on the fourth electrocardiogram signal, taking each first R point as the center point, extracting an electrocardiogram signal segment of a preset first length as the corresponding first electrocardiogram segment; and generating a corresponding first electrocardiogram segment sequence by sorting all the obtained first electrocardiogram segments in chronological order.

[0064] Here, the first length is a preset time length, such as 2 seconds; each first ECG segment corresponds to a first R point, the center point of the first ECG segment is the corresponding R point, and the segment length of the first ECG segment is the first length * 2; through the first ECG segment truncation method centered on the R point in this embodiment of the invention, the complete signal of the heartbeat signal at the current R point (including P wave, QRS wave, T wave, U wave, etc.) can be obtained to the maximum extent, and one or two heartbeat signals before and after it can also be obtained;

[0065] Step 42: On the third ECG signal, take each first R point as the center point and extract an ECG signal segment of a preset second length as the corresponding second ECG segment; and generate the corresponding second ECG segment sequence by sorting all the obtained second ECG segments in chronological order.

[0066] The second length is less than the first length.

[0067] Here, the second length is a preset time length, which is less than the first length, such as 1 second; each second ECG segment corresponds to a first R point, the center point of the second ECG segment is the corresponding R point, and the segment length of the second ECG segment is the second length * 2; the method of extracting the second ECG segment is similar to the method of extracting the first ECG segment in step 41, and will not be described further here.

[0068] Step 5: Perform heartbeat classification processing on the first ECG segment sequence to generate the corresponding first segment type sequence; and perform interference classification processing on the second ECG segment sequence to generate the corresponding second segment type sequence;

[0069] Specifically, this includes: Step 51, performing heartbeat classification processing on the first ECG segment sequence to generate a corresponding first segment type sequence;

[0070] Specifically, this includes: traversing the first ECG segment of the first ECG segment sequence; during traversal, the currently traversed first ECG segment is fed into a preset type of artificial intelligence model for one-dimensional ECG feature extraction and heartbeat type classification based on the extracted ECG features to generate the corresponding first segment type; at the end of the traversal, all the obtained first segment types are sorted in chronological order to generate the corresponding first segment type sequence.

[0071] One type of artificial intelligence model includes a first input module, a first fully connected network, a first classifier, and a first output module; the first input module is connected to the first fully connected network; the first fully connected network is connected to the first classifier; the first classifier is connected to the first output module; the first fully connected network includes one or more first fully connected layers;

[0072] Here, one type of artificial intelligence model is an AI model used for heartbeat type classification; the first input module is the model's data preprocessing module used to perform one-dimensional vector transformation on the input data; the first fully connected network is implemented using a multilayer perceptron (MLP) neural network, consisting of one or more fully connected layers (FC), which extract heartbeat morphology features from the input vector through fully connected computation; the first classifier is used to calculate classification probabilities based on the heartbeat morphology features output by the first fully connected network and output a classification vector with multi-class probability information. The first classifier is implemented by default using a classifier based on the softmax function, but other classifiers that can achieve the same function can also be used, which are not listed here; the first output module selects the classification type corresponding to the highest classification probability from the multiple classification probabilities output by the first classifier as the output;

[0073] The current step, based on the model structure of the aforementioned type of artificial intelligence model, inputs the first ECG segment currently being traversed into a preset type of artificial intelligence model for one-dimensional ECG feature extraction. Based on the extracted ECG features, it performs heartbeat type classification processing to generate the corresponding first segment type. Specifically:

[0074] The current first ECG segment is fed into an artificial intelligence model; the first input module performs a one-dimensional model vector transformation on the first ECG segment to generate a corresponding first input vector, which is then output to the first fully connected network; the first fully connected network performs a fully connected operation based on the first input vector to generate a corresponding first ECG feature vector, which is then output to the first classifier; the first classifier performs heartbeat type classification processing based on the first ECG feature vector to generate a corresponding first classification vector, which is then output to the first output module; and the first output module extracts the heartbeat classification type corresponding to the first vector data with the highest classification probability value from the first classification vector as the corresponding first segment type and outputs it.

[0075] The first classification vector includes multiple first vector data; each first vector data corresponds to a heartbeat classification type, and the first vector data is the classification probability of the corresponding heartbeat classification type; here, the heartbeat classification types include at least sinus heartbeat type, atrial fibrillation heartbeat type, ventricular heartbeat type, atrial heartbeat type, and junctional heartbeat type;

[0076] It should be noted that, before using the aforementioned type of artificial intelligence model, in order to ensure that the model's analysis and processing quality remains within a stable range, this embodiment of the invention also requires training another type of artificial intelligence model, specifically:

[0077] Step A1: Construct a loss function for a class of artificial intelligence models based on the cross-entropy loss function, denoted as the first loss function; set a corresponding first loss value convergence interval for the first loss function; set a corresponding first training segment number for the current training; and initialize the corresponding first counter to 0.

[0078] Step A2: Select the ECG segment for training from the preset training set as the corresponding first training ECG segment, and perform heartbeat type labeling on the heartbeat signal located in the center of the first training ECG segment to generate the corresponding first type label;

[0079] The first type of label includes at least the sinus beat label, atrial fibrillation beat label, ventricular beat label, atrial beat label, and junctional beat label;

[0080] Step A3: Input the first training ECG segment into an artificial intelligence model for training to generate the corresponding first training type; and calculate the loss value based on the first type label, the first training type and the first loss function to generate the corresponding first loss value;

[0081] Step A4: If the first loss value does not meet the convergence interval of the first loss value, then the model parameters of the first fully connected network and the first classifier in the first type of artificial intelligence model are modulated, and after the modulation is completed, return to step A3 for training; if the first loss value meets the convergence interval of the first loss value, then the first counter is incremented by 1. If the first counter after incrementing by 1 is less than or equal to the number of the first training segments, then return to step A2 to continue training; if the first counter after incrementing by 1 is greater than the number of the first training segments, then the training of the first type of artificial intelligence model ends.

[0082] It should also be noted that one type of artificial intelligence model in the embodiments of the present invention can also be implemented by classification models based on neural network structures such as Support Vector Machine (SVM) and decision tree;

[0083] It should also be noted that, in this embodiment of the invention, after each extraction of one-dimensional electrocardiogram features based on a class of artificial intelligence models and classification of heartbeat types based on the extracted electrocardiogram features, the first electrocardiogram segment with a low classification probability corresponding to the first segment type in the current first electrocardiogram segment sequence is stored as training data in the training set; and the class of artificial intelligence models are periodically trained according to new batch training data, thereby ensuring that the analysis and processing accuracy of the class of artificial intelligence models can be continuously improved;

[0084] Step 52: Perform interference classification processing on the second ECG segment sequence to generate the corresponding second segment type sequence;

[0085] Specifically, this includes: traversing the second ECG segment of the second ECG segment sequence; during traversal, sending the currently traversed second ECG segment into a preset two-type artificial intelligence model to extract one-dimensional interference features and performing interference type classification based on the extracted interference features to generate the corresponding second segment type; at the end of the traversal, sorting all the obtained second segment types in chronological order to generate the corresponding second segment type sequence.

[0086] The second type of artificial intelligence model consists of a second input module, a first convolutional neural network, a first data conversion module, a second classifier, and a second output module; the second input module is connected to the first convolutional neural network; the first convolutional neural network is connected to the first data conversion module; the first data conversion module is connected to the second classifier; and the second classifier is connected to the second output module.

[0087] Here, the second type of artificial intelligence model is an AI model used for interference type classification; the second input module is the model's data preprocessing module, used to perform high-dimensional tensor transformation on the input one-dimensional data for subsequent use in Convolutional Neural Networks (CNNs); the first CNN consists of multiple convolutional layers, and uses convolution operations to extract interference features from the input tensor; the first data transformation module performs one-dimensional vector transformation on the high-dimensional tensor output by the first CNN for subsequent use in the classifier; the second classifier calculates the classification probability based on the one-dimensional vector output by the first data transformation module and outputs a classification vector with multi-class probability information. The second classifier is implemented by default using a classifier based on the softmax function, but other classifiers that can achieve the same function can also be used, which are not listed here; the second output module selects the classification type corresponding to the highest classification probability from the multiple classification probabilities output by the second classifier as the output.

[0088] The current step, based on the model structure of the aforementioned two types of artificial intelligence models, inputs the currently traversed second ECG segment into a preset two-type artificial intelligence model for one-dimensional interference feature extraction. Based on the extracted interference features, it performs interference type classification processing to generate the corresponding second segment type. Specifically:

[0089] The current second ECG segment is fed into a second-class artificial intelligence model; the second input module performs high-dimensional tensor transformation on the first ECG segment to generate a corresponding first input tensor, which is output to the first convolutional neural network; the first convolutional neural network performs interference feature extraction on the first input tensor to generate a corresponding first feature tensor, which is output to the first data conversion module; the first data conversion module performs one-dimensional vector transformation on the first feature tensor to generate a corresponding first feature vector, which is output to the second classifier; the second classifier performs interference type classification based on the first feature vector to generate a corresponding second classification vector, which is output to the second output module; and the second output module extracts the interference classification type corresponding to the second vector data with the highest classification probability value from the second classification vector as the corresponding second segment type and outputs it.

[0090] The second classification vector includes multiple second vector data; each second vector data corresponds to an interference classification type, and the second vector data is the classification probability of the corresponding interference classification type; the interference classification types include non-interference type, weak interference type, and strong interference type; here, if the interference classification type is non-interference type, it means that the heartbeat signal at the center of the current second ECG segment is a normal heartbeat signal; if the interference classification type is weak interference type, it means that the heartbeat signal at the center of the current second ECG segment is a heartbeat signal with interference but the signal-to-noise ratio is still within an acceptable range; if the interference classification type is strong interference type, it means that the heartbeat signal at the center of the current second ECG segment is actually an interference signal with an extremely low signal-to-noise ratio.

[0091] It should be noted that, before using the above two types of artificial intelligence models, in order to ensure that the analysis and processing quality of the models remains within a stable quality range, this embodiment of the invention also requires training the two types of artificial intelligence models, specifically as follows:

[0092] Step B1: Construct a loss function for the two types of artificial intelligence models based on the cross-entropy loss function, denoted as the second loss function; set a corresponding second loss value convergence interval for the second loss function; set a corresponding second training segment number for the current training; and initialize the corresponding second counter to 0.

[0093] Step B2: Select the ECG segment for training from the preset training set as the corresponding second training ECG segment, and mark the heartbeat signal located in the center of the second training ECG segment with interference type to generate the corresponding second type label;

[0094] The second type of label includes non-interference type, weak interference type, and strong interference type;

[0095] Step B3: Input the second training ECG segment into the second type of artificial intelligence model for training to generate the corresponding second training type; and calculate the corresponding second loss value based on the second type label, the second training type and the second loss function.

[0096] Step B4: If the second loss value does not meet the convergence interval of the second loss value, then the model parameters of the first convolutional neural network and the second classifier in the two-class artificial intelligence model are modulated, and after the modulation is completed, return to step B3 for training; if the second loss value meets the convergence interval of the second loss value, then the second counter is incremented by 1. If the second counter after incrementing by 1 is less than or equal to the number of the second training segments, then return to step B2 to continue training; if the second counter after incrementing by 1 is greater than the number of the second training segments, then the training of the two-class artificial intelligence model ends.

[0097] It should also be noted that the two types of artificial intelligence models in the embodiments of the present invention can also be implemented by classification models based on neural network structures such as fully connected networks, support vector machines, and decision trees;

[0098] It should also be noted that, in this embodiment of the invention, after each extraction of one-dimensional interference features based on the two-class artificial intelligence model and classification of interference types based on the extracted interference features, the second ECG segment with a low classification probability corresponding to the second segment type in the current second ECG segment sequence is stored as training data in the training set; and the two-class artificial intelligence model is periodically trained according to new batch training data, thereby ensuring that the analysis and processing accuracy of the two-class artificial intelligence model can be continuously improved.

[0099] Step 6: Perform fragment type fusion processing on the first and second fragment type sequences to generate the corresponding third fragment type sequence;

[0100] Specifically, this includes: taking the first and second segment types with the same sequence index in the first and second segment type sequences as corresponding first segment type pairs; performing segment type fusion processing on each first segment type pair according to a preset fusion mode to generate corresponding third segment types; and sorting all the obtained third segment types in chronological order to generate corresponding third segment type sequences.

[0101] Specifically, according to a preset fusion mode, each first fragment type pair undergoes fragment type fusion processing to generate a corresponding third fragment type, as follows:

[0102] Step C1: Identify the fusion pattern;

[0103] The integration mode includes the first mode and the second mode;

[0104] Step C2: When the fusion mode is the first mode, identify whether the second segment type of the current first segment type pair is a non-interference type; if yes, set the corresponding third segment type to the first segment type of the current first segment type pair; if no, set the corresponding third segment type to the interference segment type.

[0105] Here, in the first mode, the embodiments of the present invention preferably retain only the heartbeat classification information of the heartbeat signal corresponding to the non-interference type, i.e., the normal heartbeat signal, and regard the heartbeat signals corresponding to the weak and strong interference types as noise interference signals;

[0106] Step C3: When the fusion mode is the second mode, identify whether the second segment type of the current first segment type pair is a strong interference type; if yes, set the corresponding third segment type as the interference segment type; if no, set the corresponding third segment type as the first segment type of the current first segment type pair.

[0107] Here, in the second mode, the embodiments of the present invention preferably treat only the heartbeat signals corresponding to the strong interference type as noise interference signals, and retain the heartbeat classification information of the heartbeat signals corresponding to the non-interference type and the weak interference type.

[0108] Step 7: Output the first R-point sequence and the third segment type sequence as the ECG signal processing results.

[0109] Here, each first R point in the first R point sequence indicates the position of each heartbeat signal on the first ECG signal, and the third segment type in the third segment type sequence corresponds one-to-one with the first R point; by combining the first R point sequence and the third segment type sequence, the position of the signal marker point (R point) and the signal type of each heartbeat signal on the first ECG signal can be determined. The signal types include sinus heartbeat type, atrial fibrillation heartbeat type, ventricular heartbeat type, atrial heartbeat type, junctional heartbeat type and interference segment type.

[0110] Figure 2 This is a module structure diagram of an electrocardiogram signal processing device provided in Embodiment 2 of the present invention. This device can be a terminal device or server implementing the method of the embodiments of the present invention, or it can be a device connected to the aforementioned terminal device or server to implement the method of the embodiments of the present invention. For example, the device can be a device or chip system of the aforementioned terminal device or server. Figure 2 As shown, the electrocardiogram (ECG) signal processing device includes: a receiving module 201, a preprocessing module 202, an R-point identification module 203, an ECG signal segment extraction and processing module 204, an artificial intelligence model processing module 205, a segment type fusion processing module 206, and an output module 207.

[0111] The receiving module 201 is used to receive the first electrocardiogram signal.

[0112] The preprocessing module 202 is used to filter the first ECG signal to generate the corresponding second ECG signal; and to perform baseline drift elimination processing on the first and second ECG signals to generate the corresponding third and fourth ECG signals.

[0113] The R-point identification module 203 is used to perform R-point identification processing on the fourth electrocardiogram signal to generate a corresponding first R-point sequence; the first R-point sequence includes multiple first R-points.

[0114] The ECG signal segment extraction and processing module 204 is used to perform ECG signal segment extraction and processing on the third and fourth ECG signals according to the first R-point sequence to generate corresponding first and second ECG segment sequences.

[0115] The artificial intelligence model processing module 205 is used to perform heartbeat classification processing on the first electrocardiogram segment sequence to generate the corresponding first segment type sequence; and to perform interference classification processing on the second electrocardiogram segment sequence to generate the corresponding second segment type sequence.

[0116] The fragment type fusion processing module 206 is used to perform fragment type fusion processing on the first and second fragment type sequences to generate the corresponding third fragment type sequence.

[0117] The output module 207 is used to output the first R-point sequence and the third segment type sequence as the result of electrocardiogram signal processing.

[0118] The electrocardiogram signal processing device provided in this embodiment of the invention can execute the method steps in the above method embodiment. Its implementation principle and technical effect are similar, and will not be repeated here.

[0119] It should be noted that the division of the various modules in the above device is merely a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, these modules can be implemented entirely in software via processing element calls; they can be fully implemented in hardware; or some modules can be implemented by processing element calls to software, while others are implemented in hardware. For example, the receiving module can be a separate processing element, or it can be integrated into a chip in the above device. Alternatively, it can be stored as program code in the memory of the above device, and called and executed by a processing element of the device. The implementation of other modules is similar. Moreover, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed through integrated logic circuits in the hardware of the processor element or through software instructions.

[0120] For example, these modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more Digital Signal Processors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs). As another example, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together as a System-on-a-Chip (SOC).

[0121] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. This computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer described above can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions described above can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, Bluetooth, microwave, etc.) means. The computer-readable storage medium described above can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The aforementioned available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0122] Figure 3 This is a schematic diagram of an electronic device provided in Embodiment 3 of the present invention. This electronic device can be the aforementioned terminal device or server, or it can be a terminal device or server connected to the aforementioned terminal device or server that implements the method of the embodiments of the present invention. Figure 3As shown, the electronic device may include: a processor 301 (e.g., CPU), a memory 302, and a transceiver 303; the transceiver 303 is coupled to the processor 301, and the processor 301 controls the transmission and reception operations of the transceiver 303. The memory 302 may store various instructions for performing various processing functions and implementing the methods and processes provided in the above embodiments of the present invention. Preferably, the electronic device involved in the embodiments of the present invention further includes: a power supply 304, a system bus 305, and a communication port 306. The system bus 305 is used to realize communication connections between components. The communication port 306 is used for communication between the electronic device and other peripherals.

[0123] exist Figure 3 The system bus mentioned can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This system bus can be divided into address bus, data bus, control bus, etc. For ease of representation, only one thick line is used in the diagram, but this does not indicate that there is only one bus or one type of bus. The communication interface is used to enable communication between the database access device and other devices (e.g., clients, read-write libraries, and read-only libraries). Memory may include Random Access Memory (RAM) and may also include non-volatile memory, such as at least one disk drive.

[0124] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0125] It should be noted that the embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when run on a computer, cause the computer to execute the methods and processes provided in the above embodiments.

[0126] This invention also provides a chip for executing instructions, which is used to execute the methods and processes provided in the above embodiments.

[0127] This invention provides a method, apparatus, electronic device, and computer-readable storage medium for electrocardiogram (ECG) signal processing. The method first filters the input raw ECG signal to generate a corresponding filtered ECG signal. Then, baseline drift elimination processing is performed on both the raw and filtered ECG signals. Next, R-point identification is performed on the baseline-drift-eliminated filtered ECG signal to obtain a corresponding R-point sequence. Then, based on the R-point sequence, ECG signal segments are extracted from the baseline-drift-eliminated raw and filtered ECG signals according to different ECG signal segments to generate corresponding raw and filtered ECG segment sequences. Next, different artificial intelligence models are used to identify the corresponding heartbeat and interference types in the raw and filtered ECG segment sequences. Finally, type fusion processing is performed on the identified heartbeat and interference types, and the final fusion result is output as the ECG signal processing result. This invention improves the processing efficiency of ECG signal processing and ensures the stability of processing quality.

[0128] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0129] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0130] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for processing electrocardiogram (ECG) signals, characterized in that, The method includes: Receive the first electrocardiogram signal; The first electrocardiogram (ECG) signal is filtered to generate a corresponding second ECG signal; and the first and second ECG signals are respectively processed to eliminate baseline drift to generate corresponding third and fourth ECG signals. The fourth electrocardiogram signal is subjected to R-point identification processing to generate a corresponding first R-point sequence; the first R-point sequence includes multiple first R-points. Based on the first R-point sequence, the third and fourth ECG signals are processed to extract ECG signal segments, generating corresponding first and second ECG segment sequences; The first ECG segment sequence is processed by heartbeat classification to generate a corresponding first segment type sequence; and the second ECG segment sequence is processed by interference classification to generate a corresponding second segment type sequence. The first and second fragment type sequences are fused to generate a corresponding third fragment type sequence. The first R-point sequence and the third segment type sequence are output as the electrocardiogram signal processing results; Specifically, the baseline drift elimination process involves: if the duration of the first ECG signal exceeds a preset ECG duration threshold, baseline drift elimination is performed on the first and second ECG signals based on wavelet transform; if the duration of the first ECG signal does not exceed the ECG duration threshold, baseline drift elimination is performed on the second ECG signal based on a simple integer coefficient filter, and baseline drift elimination is performed on the first ECG signal based on median filtering or a simple integer coefficient filter. The step of performing ECG signal segmentation processing on the third and fourth ECG signals based on the first R-point sequence to generate corresponding first and second ECG segment sequences specifically includes: On the fourth electrocardiogram (ECG) signal, an ECG signal segment of a preset first length is extracted with each of the first R points as the center point as the corresponding first ECG segment; and all the obtained first ECG segments are sorted in chronological order to generate a corresponding first ECG segment sequence; the first ECG segment includes the complete heartbeat signal of the current R point and one or two heartbeat signals before and after it. On the third electrocardiogram (ECG) signal, an ECG signal segment of a preset second length is extracted with each of the first R points as the center point as the corresponding second ECG segment; and all the obtained second ECG segments are sorted in chronological order to generate a corresponding second ECG segment sequence; the second length is less than the first length; The step of performing fragment type fusion processing on the first and second fragment type sequences to generate the corresponding third fragment type sequence specifically includes: In the first and second fragment type sequences, the first and second fragment types with the same sequence index are regarded as the corresponding first fragment type pairs; The system performs segment type fusion processing on each first segment type pair according to a preset fusion mode to generate a corresponding third segment type. Specifically, the fusion mode is identified; when the fusion mode is a first mode, it is identified whether the second segment type of the current first segment type pair is a non-interference type; if yes, the corresponding third segment type is set as the first segment type of the current first segment type pair; if no, the corresponding third segment type is set as an interference segment type; when the fusion mode is a second mode, it is identified whether the second segment type of the current first segment type pair is a strong interference type; if yes, the corresponding third segment type is set as an interference segment type; if no, the corresponding third segment type is set as the first segment type of the current first segment type pair. The obtained third fragment types are then sorted in chronological order to generate the corresponding third fragment type sequence.

2. The electrocardiogram signal processing method according to claim 1, characterized in that, The step of performing R-point identification processing on the fourth electrocardiogram signal to generate the corresponding first R-point sequence specifically includes: The fourth electrocardiogram signal is subjected to QRS wave recognition processing to generate multiple corresponding first QRS wave signals; and each first QRS wave signal is subjected to R point recognition processing to generate corresponding first R points; and all the obtained first R points are sorted in chronological order to generate a corresponding first R point sequence.

3. The electrocardiogram signal processing method according to claim 1, characterized in that, The step of performing heartbeat classification processing on the first electrocardiogram segment sequence to generate a corresponding first segment type sequence specifically includes: The first ECG segment of the first ECG segment sequence is traversed; during traversal, the first ECG segment currently traversed is fed into a preset artificial intelligence model for one-dimensional ECG feature extraction and heartbeat type classification based on the extracted ECG features to generate the corresponding first segment type; at the end of the traversal, all the obtained first segment types are sorted in chronological order to generate the corresponding first segment type sequence.

4. The electrocardiogram signal processing method according to claim 3, characterized in that, The artificial intelligence model includes a first input module, a first fully connected network, a first classifier, and a first output module; the first input module is connected to the first fully connected network; the first fully connected network is connected to the first classifier; the first classifier is connected to the first output module; the first fully connected network includes one or more first fully connected layers.

5. The electrocardiogram signal processing method according to claim 4, characterized in that, The step of feeding the currently traversed first ECG segment into a preset type of artificial intelligence model for one-dimensional ECG feature extraction and performing heartbeat type classification based on the extracted ECG features to generate the corresponding first segment type specifically includes: The first ECG segment is fed into the aforementioned artificial intelligence model; the first input module performs a one-dimensional model vector transformation on the first ECG segment to generate a corresponding first input vector, which is then output to the first fully connected network; the first fully connected network performs a fully connected operation based on the first input vector to generate a corresponding first ECG feature vector, which is then output to the first classifier; the first classifier performs heartbeat type classification processing based on the first ECG feature vector to generate a corresponding first classification vector, which is then output to the first output module; and the first output module extracts the heartbeat classification type corresponding to the first vector data with the highest classification probability value from the first classification vector as the corresponding first segment type, which is then output; the first classification vector includes multiple first vector data; each first vector data corresponds to a heartbeat classification type, and the first vector data represents the classification probability of the corresponding heartbeat classification type.

6. The electrocardiogram signal processing method according to claim 1, characterized in that, The step of performing interference classification processing on the second ECG segment sequence to generate a corresponding second segment type sequence specifically includes: The second ECG segment of the second ECG segment sequence is traversed; during traversal, the currently traversed second ECG segment is fed into a preset two-type artificial intelligence model for one-dimensional interference feature extraction, and interference type classification is performed based on the extracted interference features to generate the corresponding second segment type; at the end of the traversal, all the obtained second segment types are sorted in chronological order to generate the corresponding second segment type sequence.

7. The electrocardiogram signal processing method according to claim 6, characterized in that, The two types of artificial intelligence models consist of a second input module, a first convolutional neural network, a first data conversion module, a second classifier, and a second output module; the second input module is connected to the first convolutional neural network; the first convolutional neural network is connected to the first data conversion module; the first data conversion module is connected to the second classifier; and the second classifier is connected to the second output module.

8. The electrocardiogram signal processing method according to claim 7, characterized in that, The step of feeding the currently traversed second ECG segment into a preset two-type artificial intelligence model for one-dimensional interference feature extraction and performing interference type classification based on the extracted interference features to generate the corresponding second segment type specifically includes: The current second ECG segment is fed into the two-class artificial intelligence model; the second input module performs high-dimensional tensor transformation on the second ECG segment to generate a corresponding first input tensor, which is output to the first convolutional neural network; the first convolutional neural network performs interference feature extraction on the first input tensor to generate a corresponding first feature tensor, which is output to the first data conversion module; the first data conversion module performs one-dimensional vector transformation on the first feature tensor to generate a corresponding first feature vector, which is output to the second classifier; the second classifier performs interference type classification based on the first feature vector to generate a corresponding second classification vector, which is output to the second output module; and the second output module extracts the interference classification type corresponding to the second vector data with the highest classification probability value from the second classification vector as the corresponding second segment type, which is output; the second classification vector includes multiple second vector data; each second vector data corresponds to one interference classification type, and the second vector data is the classification probability of the corresponding interference classification type; the interference classification type includes non-interference type, weak interference type, and strong interference type.

9. An apparatus for implementing the electrocardiogram signal processing method according to any one of claims 1-8, characterized in that, The device includes: a receiving module, a preprocessing module, an R-point identification module, an electrocardiogram signal segment extraction and processing module, an artificial intelligence model processing module, a segment type fusion processing module, and an output module; The receiving module is used to receive the first electrocardiogram signal; The preprocessing module is used to filter the first electrocardiogram (ECG) signal to generate a corresponding second ECG signal; and to perform baseline drift elimination processing on the first and second ECG signals to generate corresponding third and fourth ECG signals, respectively. The R-point identification module is used to perform R-point identification processing on the fourth electrocardiogram signal to generate a corresponding first R-point sequence; the first R-point sequence includes multiple first R-points; The ECG signal segment extraction and processing module is used to perform ECG signal segment extraction and processing on the third and fourth ECG signals according to the first R-point sequence to generate corresponding first and second ECG segment sequences. The artificial intelligence model processing module is used to perform heartbeat classification processing on the first ECG segment sequence to generate a corresponding first segment type sequence; and to perform interference classification processing on the second ECG segment sequence to generate a corresponding second segment type sequence; The fragment type fusion processing module is used to perform fragment type fusion processing on the first and second fragment type sequences to generate a corresponding third fragment type sequence; The output module is used to output the first R-point sequence and the third segment type sequence as the electrocardiogram signal processing result.

10. An electronic device, characterized in that, include: Memory, processor, and transceiver; The processor is configured to be coupled to the memory, read and execute instructions in the memory to implement the method according to any one of claims 1-8; The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a computer, cause the computer to perform the method described in any one of claims 1-8.

Images