Search

CN-122004890-A - Analysis method and related device for lead electrocardiograph data

CN122004890ACN 122004890 ACN122004890 ACN 122004890ACN-122004890-A

Abstract

The application discloses an analysis method and a related device of lead electrocardiograph data, which relate to the technical field of data processing, and are characterized in that the lead electrocardiograph data to be analyzed is obtained, then a soft mask vector set corresponding to the data is calculated, weighting calculation is carried out based on the soft mask vector set, the operation is used for noise resistance, fine features in the electrocardiograph data are better reserved, so that the electrocardiograph data is analyzed more accurately, microscopic morphological features and macroscopic rhythm features of the lead electrocardiograph data are extracted respectively, feature decoupling is realized from macroscopic and microscopic angles, a cross attention mechanism is adopted to divide the microscopic morphological features into main view features and auxiliary view features, the overall spatial features are more comprehensively represented by the fused panoramic spatial features, so that the analysis result of the electrocardiograph data is more accurate, the panoramic spatial features and macroscopic rhythm features are fused, and the analysis result of the lead electrocardiograph data is obtained, and the analysis result is fused with the two features, so that the analysis accuracy of the lead electrocardiograph data is further improved.

Inventors

  • CHENG BINGHUI

Assignees

  • 蓬阳丰业(北京)医疗科技有限公司

Dates

Publication Date
20260512
Application Date
20260209

Claims (10)

  1. 1. A method of analyzing lead electrocardiographic data, comprising: Acquiring preset lead electrocardiographic data to be analyzed, wherein the preset lead electrocardiographic data to be analyzed comprises standard lead electrocardiographic data and extended lead electrocardiographic data; determining a soft mask vector set corresponding to the to-be-analyzed preset lead electrocardiograph data, wherein the soft mask vector set comprises soft mask vectors corresponding to each lead in the to-be-analyzed preset lead electrocardiograph data one by one; weighting calculation is carried out on the preset lead electrocardiograph data to be analyzed according to the soft mask vectors, and weighted preset lead electrocardiograph data are obtained; inputting the preset lead electrocardiograph data into a first branch for feature extraction to obtain microscopic morphological features; Inputting RR interval sequences corresponding to the preset lead electrocardiograph data into a second branch for feature extraction to obtain macro rhythm features; Invoking a preset topological structure of a preset cross attention mechanism, respectively processing main view features corresponding to the standard lead electrocardiograph data and auxiliary view features corresponding to the extended lead electrocardiograph data in the micro morphological features, and fusing the processed results to obtain panoramic space features; and fusing the panoramic space characteristics and the macro rhythm characteristics to obtain an analysis result of the electrocardio data of the preset lead number to be analyzed.
  2. 2. The method for analyzing the lead electrocardiographic data according to claim 1, wherein determining the soft mask vector set corresponding to the preset lead electrocardiographic data to be analyzed includes: Performing short-time Fourier transform on the electrocardiograph data of the preset leads to be analyzed to obtain a two-dimensional time-frequency representation set; calculating the power spectrum density of each lead according to each two-dimensional time-frequency representation in the two-dimensional time-frequency representation set to obtain a power spectrum density set; calculating the instantaneous signal-to-noise ratio of each lead according to each power spectrum density in the power spectrum density set to obtain an instantaneous signal-to-noise ratio set; and calculating a soft mask vector of each lead according to each instantaneous signal-to-noise ratio in the instantaneous signal-to-noise ratio set to obtain the soft mask vector set.
  3. 3. The method for analyzing cardiac data according to claim 1, wherein the weighting calculation is performed on each of the leads according to the soft mask vectors to obtain weighted preset lead number cardiac data, comprising: and carrying out Hadamard product on the electrocardiograph data corresponding to each lead in the electrocardiograph data of the preset lead number to be analyzed and the corresponding soft mask vector to obtain the electrocardiograph data of the preset lead number.
  4. 4. The method for analyzing the lead electrocardiographic data according to claim 1, wherein the step of inputting the preset lead electrocardiographic data into the first branch to perform feature extraction to obtain micro morphological features includes: And inputting the electrocardio data of the preset lead number into a deep convolution neural network in the first branch to extract the characteristics, so as to obtain the micro morphological characteristics.
  5. 5. The method for analyzing lead electrocardiographic data according to claim 1, wherein inputting the RR interval sequence corresponding to the preset lead electrocardiographic data into the second branch for feature extraction, obtaining macro-rhythm features, includes: r wave detection is carried out on the electrocardio data of the preset lead number to obtain the RR interval sequence; And calling a pre-trained time sequence feature extraction network to process the RR interval sequence to obtain the macro rhythm feature.
  6. 6. The method for analyzing the lead electrocardiographic data according to claim 1, wherein the invoking the preset topology structure of the preset cross-attention mechanism respectively processes the main view feature corresponding to the standard lead electrocardiographic data and the auxiliary view feature corresponding to the extended lead electrocardiographic data in the micro morphological feature, and fuses the processed results to obtain panoramic space features, including: Dividing morphological features corresponding to the standard lead electrocardiographic data in the microscopic morphological features into main view features, wherein the main view features are cardiac integral depolarization features; Dividing morphological features corresponding to the extended lead electrocardiographic data in the microscopic morphological features into auxiliary view features, wherein the auxiliary view features comprise local potential features of the left atrium back wall and the right atrium; Analyzing and fusing the features of the main view by adopting the preset cross attention mechanism; When the main view feature presents uncertainty or suspected abnormality, calculating the attention weight of the main view feature corresponding to the auxiliary view feature by adopting the preset cross attention mechanism, carrying out weighted fusion on the auxiliary view feature according to the attention weight, and analyzing whether an atrial fibrillation wave exists in the auxiliary view feature to obtain a query result; and generating the panoramic space feature by the query result and the fused main view feature.
  7. 7. The method for analyzing the lead electrocardiographic data according to claim 1, wherein the step of fusing the panoramic spatial feature and the macro-rhythm feature to obtain an analysis result of the lead electrocardiographic data comprises the steps of: and performing feature stitching on the panoramic space features and the macro rhythm features under the same electrocardiograph data to obtain an analysis result of the lead electrocardiograph data.
  8. 8. An apparatus for analyzing lead electrocardiographic data, comprising: The acquisition unit is used for acquiring the preset lead electrocardiograph data to be analyzed, wherein the preset lead electrocardiograph data to be analyzed comprises standard lead electrocardiograph data and extended lead electrocardiograph data; the determining unit is used for determining a soft mask vector set corresponding to the electrocardiograph data of the preset leads to be analyzed, wherein the soft mask vector set comprises soft mask vectors corresponding to each lead in the electrocardiograph data of the preset leads to be analyzed one by one; The computing unit is used for carrying out weighted computation on the preset lead electrocardiograph data to be analyzed according to the soft mask vectors to obtain weighted preset lead electrocardiograph data; The first feature extraction unit is used for inputting the preset lead electrocardiograph data into a first branch to perform feature extraction to obtain microscopic morphological features; the second feature extraction unit is used for inputting RR interval sequences corresponding to the preset lead electrocardiograph data into a second branch for feature extraction to obtain macro rhythm features; The processing unit is used for calling a preset topological structure of a preset cross attention mechanism, respectively processing main view features corresponding to the standard lead electrocardiograph data and auxiliary view features corresponding to the extended lead electrocardiograph data in the microscopic morphological features, and fusing the processed results to obtain panoramic space features; And the fusion unit is used for fusing the panoramic space characteristics and the macro rhythm characteristics to obtain an analysis result of the electrocardio data of the preset lead number to be analyzed.
  9. 9. An analysis device for lead electrocardiographic data, comprising at least one processor and a memory coupled to the processor, wherein: the memory is used for storing a computer program; the processor is configured to execute the computer program to enable the analysis device of the lead electrocardiographic data to implement the method of analyzing lead electrocardiographic data according to any one of claims 1 to 7.
  10. 10. A computer program product comprising computer readable instructions which, when run on an electronic device, cause the electronic device to implement a method of analysing lead electrocardiographic data according to any one of claims 1 to 7.

Description

Analysis method and related device for lead electrocardiograph data Technical Field The application relates to the technical field of data processing, in particular to a method and a related device for analyzing lead electrocardiograph data. Background With the deep application of artificial intelligence technology in the field of medical diagnosis, an Electrocardiogram (ECG) automatic analysis technology based on a deep learning model has become a research hotspot, and the technology aims to assist doctors to rapidly and accurately identify cardiac diseases such as arrhythmia, myocardial ischemia and the like. However, despite significant advances made in standard datasets by existing methods, the accuracy of the analysis results of the lead electrocardiographic data in actual clinical settings is still low. Disclosure of Invention In view of the above problems, the present application provides a method and related apparatus for analyzing lead electrocardiographic data, so as to achieve the purpose of improving accuracy of an analysis result of electrocardiographic data. The specific scheme is as follows: the first aspect of the application provides a method for analyzing lead electrocardiographic data, comprising the following steps: Acquiring preset lead electrocardiographic data to be analyzed, wherein the preset lead electrocardiographic data to be analyzed comprises standard lead electrocardiographic data and extended lead electrocardiographic data; Determining a soft mask vector set corresponding to the electrocardiographic data of the preset leads to be analyzed, wherein the soft mask vector set comprises soft mask vectors corresponding to each lead in the electrocardiographic data of the preset leads to be analyzed one by one; weighting calculation is carried out on the preset lead electrocardiograph data to be analyzed according to each soft mask vector, and the weighted preset lead electrocardiograph data are obtained; Inputting electrocardiographic data of a preset lead number into a first branch for feature extraction to obtain microscopic morphological features; Inputting RR interval sequences corresponding to preset lead electrocardiograph data into a second branch for feature extraction to obtain macro rhythm features; Invoking a preset topological structure of a preset cross attention mechanism, respectively processing main view features corresponding to standard lead electrocardiograph data and auxiliary view features corresponding to expanded lead electrocardiograph data in microscopic morphological features, and fusing the processed results to obtain panoramic space features; And fusing the panoramic space characteristics and the macro rhythm characteristics to obtain an analysis result of the electrocardio data of the preset lead number to be analyzed. In one possible implementation, determining a soft mask vector set corresponding to the preset lead number electrocardiographic data to be analyzed includes: performing short-time Fourier transform on the electrocardiograph data of the preset leads to be analyzed to obtain a two-dimensional time-frequency representation set; calculating the power spectrum density of each lead according to each two-dimensional time-frequency representation in the two-dimensional time-frequency representation set to obtain a power spectrum density set; calculating the instantaneous signal-to-noise ratio of each lead according to each power spectrum density in the power spectrum density set to obtain an instantaneous signal-to-noise ratio set; And calculating a soft mask vector of each lead according to each instantaneous signal-to-noise ratio in the instantaneous signal-to-noise ratio set to obtain a soft mask vector set. In one possible implementation, weighting calculation is performed on each lead according to each soft mask vector to obtain weighted preset lead number electrocardiographic data, including: and carrying out Hadamard product on the electrocardiograph data corresponding to each lead in the electrocardiograph data of the preset lead number to be analyzed and the corresponding soft mask vector to obtain the electrocardiograph data of the preset lead number. In one possible implementation, inputting the preset lead number electrocardiographic data into the first branch for feature extraction to obtain the micro morphological feature, including: and inputting the electrocardio data of the preset lead number into a deep convolution neural network in the first branch to extract the characteristics, thereby obtaining the microcosmic morphological characteristics. In one possible implementation, inputting the RR interval sequence corresponding to the preset lead electrocardiograph data into the second branch for feature extraction to obtain the macro-rhythm feature, including: r wave detection is carried out on the electrocardio data of the preset lead number to obtain RR interval sequences; and calling a pre-trained time sequence feature extractio