CN115778402B - Artifact identification method and system for dynamic electrocardiosignal - Google Patents
Artifact identification method and system for dynamic electrocardiosignal Download PDFInfo
- Publication number
- CN115778402B CN115778402B CN202211540499.6A CN202211540499A CN115778402B CN 115778402 B CN115778402 B CN 115778402B CN 202211540499 A CN202211540499 A CN 202211540499A CN 115778402 B CN115778402 B CN 115778402B
- Authority
- CN
- China
- Prior art keywords
- heart beat
- data
- template
- marked
- heart
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention discloses a method and a system for identifying the artifact of a dynamic electrocardiosignal, wherein the method comprises the following steps: acquiring marked dynamic electrocardiograph data, wherein the marked dynamic electrocardiograph data comprises heart beat data, performing first preprocessing operation on the marked heart beat data, extracting and normalizing a characteristic value of each preprocessed heart beat to obtain a sample characteristic vector; inputting the sample feature vector into a pre-constructed support vector machine for training, and generating an artifact identification model; performing second preprocessing operation on the acquired dynamic electrocardiograph data, and extracting and normalizing the feature value of each preprocessed heart beat to obtain heart beat feature vectors; the heart beat feature vector is input into the artifact identification model, and the output result is used as an artifact identification result. According to the embodiment of the invention, the characteristics of the data to be analyzed are extracted through deep mining, and the support vector machine is adopted for training to obtain the identification model, so that the identification model is used for artifact identification of dynamic electrocardiosignals, and the accuracy of identification and the robustness of an algorithm are improved.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to a method and a system for identifying the artifact of a dynamic electrocardiosignal.
Background
The dynamic electrocardiograph is used for recording the whole process of the electrocardiograph's electrocardiographic activity for 24 hours or longer in the daily life state of the patient, inputting the electrocardiograph information obtained by the recorder into a computer for processing, and manually editing and printing out a dynamic electrocardiograph report, thereby providing the basis for diagnosing and treating diseases clinically.
The dynamic electrocardiogram is acquired in the daily life state of the patient, so that the dynamic electrocardiogram is inevitably interfered by different degrees in the recording process, including noise such as baseline drift, power frequency interference, myoelectric interference and the like. Although part of the interference can be filtered by designing a related filter, the filter cannot completely eliminate the interference introduced by reasons such as movement and the like due to the similarity with the shape of a real heart beat, and the interference is called as 'artifact', and is easily misjudged as ventricular premature beat or atrial premature beat during automatic program analysis, so that analysis and diagnosis of doctors are affected.
In the prior related art, the methods of template matching clustering, high-order statistics, wavelet analysis, independent component analysis and the like are mainly adopted for artifact identification, but the methods mainly utilize the characteristics of the data to be analyzed, so that the identification accuracy is low, the stability is poor, and the effect is not ideal in practical clinical application.
The prior art is therefore still in need of further development.
Disclosure of Invention
Aiming at the technical problems, the embodiment of the invention provides a method and a system for identifying the artifact of a dynamic electrocardiosignal, which can solve the technical problems that in the prior art, the artifact is identified mainly by adopting methods such as template matching clustering, high-order statistics, wavelet analysis, independent component analysis and the like, but the methods mainly use the characteristics of the data to be analyzed, so that the identification accuracy is lower and the stability is poor.
A first aspect of an embodiment of the present invention provides a method for identifying artifacts in a dynamic electrocardiographic signal, including:
Acquiring marked dynamic electrocardiographic data, wherein the marked dynamic electrocardiographic data comprises heart beat data, and the heart beat data type comprises true solid beats and artifacts;
Performing first preprocessing operation on the marked heart beat data, extracting a characteristic value from each preprocessed heart beat, and performing normalization processing to obtain a sample characteristic vector;
Inputting the sample feature vector into a pre-constructed support vector machine for training, and generating an artifact identification model;
acquiring acquired dynamic electrocardiographic data, performing second preprocessing operation on the acquired dynamic electrocardiographic data, extracting characteristic values of each preprocessed heart beat, and performing normalization processing to obtain heart beat characteristic vectors;
And inputting the heart beat feature vector into an artifact identification model, obtaining an output result of the artifact identification model, and taking the output result as the artifact identification result.
Optionally, performing a first preprocessing operation on the labeled beat data, including:
And carrying out template matching on the real heart beats and the artifacts in the marked heart beat data according to the positions in sequence, and counting the number of the heart beats of each template after all the marked data are matched.
Optionally, extracting a feature value from each preprocessed heart beat and performing normalization processing to obtain a sample feature vector, including:
Extracting characteristic values of each preprocessed heart beat, wherein the characteristic values comprise wavelet coefficients of the heart beats, the total number of heart beats of a template, the interval ratio of three intervals before and after the current heart beat, the matching degree of the current heart beat and the front and back heart beats, heart beat differential data information and low-frequency and high-frequency noise interference information;
And carrying out normalization processing on the extracted characteristic values to obtain sample characteristic vectors.
Optionally, the acquiring the acquired dynamic electrocardiographic data performs a second preprocessing operation on the acquired dynamic electrocardiographic data, including:
Acquiring acquired dynamic electrocardiographic data, and performing filtering processing on the acquired dynamic electrocardiographic data to generate filtering data;
Performing heart beat detection on the filtered data based on a differential threshold algorithm, and recording the position information of each heart beat;
and carrying out clustering statistics on each heart beat, and counting the number of heart beats of each template.
Optionally, the matching the template of the real heart beat and the artifact in the marked heart beat data according to the positions sequentially includes:
sequentially carrying out similarity calculation on the marked heart beat data and the existing template according to the positions;
if the similarity is greater than or equal to a threshold value, judging that the heart beat data belongs to the template, and updating the data in the template;
if the similarity is smaller than the threshold value, matching calculation is carried out with the next template;
If the marked heart beat data are not matched with all the templates, a corresponding template is newly built based on the heart beat data.
A second aspect of an embodiment of the present invention provides a system for identifying artifacts in dynamic electrocardiographic signals, the system including: a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program when executed by the processor implementing the steps of:
Acquiring marked dynamic electrocardiographic data, wherein the marked dynamic electrocardiographic data comprises heart beat data, and the heart beat data type comprises true solid beats and artifacts;
Performing first preprocessing operation on the marked heart beat data, extracting a characteristic value from each preprocessed heart beat, and performing normalization processing to obtain a sample characteristic vector;
Inputting the sample feature vector into a pre-constructed support vector machine for training, and generating an artifact identification model;
acquiring acquired dynamic electrocardiographic data, performing second preprocessing operation on the acquired dynamic electrocardiographic data, extracting characteristic values of each preprocessed heart beat, and performing normalization processing to obtain heart beat characteristic vectors;
And inputting the heart beat feature vector into an artifact identification model, obtaining an output result of the artifact identification model, and taking the output result as the artifact identification result.
Optionally, the computer program when executed by the processor implements the steps of:
And carrying out template matching on the real heart beats and the artifacts in the marked heart beat data according to the positions in sequence, and counting the number of the heart beats of each template after all the marked data are matched.
Optionally, the computer program when executed by the processor further implements the steps of:
Extracting characteristic values of each preprocessed heart beat, wherein the characteristic values comprise wavelet coefficients of the heart beats, the total number of heart beats of a template, the interval ratio of three intervals before and after the current heart beat, the matching degree of the current heart beat and the front and back heart beats, heart beat differential data information and low-frequency and high-frequency noise interference information;
And carrying out normalization processing on the extracted characteristic values to obtain sample characteristic vectors.
Optionally, the computer program when executed by the processor further implements the steps of:
Acquiring acquired dynamic electrocardiographic data, and performing filtering processing on the acquired dynamic electrocardiographic data to generate filtering data;
Performing heart beat detection on the filtered data based on a differential threshold algorithm, and recording the position information of each heart beat;
and carrying out clustering statistics on each heart beat, and counting the number of heart beats of each template.
A third aspect of the embodiment of the present invention provides a non-volatile computer readable storage medium, where the non-volatile computer readable storage medium stores computer executable instructions that, when executed by one or more processors, cause the one or more processors to perform the above-mentioned artifact identification method for dynamic electrocardiographic signals.
In the technical scheme provided by the embodiment of the invention, marked dynamic electrocardiograph data is obtained, the marked dynamic electrocardiograph data comprises heart beat data, first preprocessing operation is carried out on the marked heart beat data, and feature values are extracted and normalized for each heart beat after preprocessing to obtain sample feature vectors; inputting the sample feature vector into a pre-constructed support vector machine for training, and generating an artifact identification model; performing second preprocessing operation on the acquired dynamic electrocardiograph data, and extracting and normalizing the feature value of each preprocessed heart beat to obtain heart beat feature vectors; the heart beat feature vector is input into the artifact identification model, and the output result is used as an artifact identification result. According to the embodiment of the invention, the characteristics of the data to be analyzed are extracted through deep mining, and the support vector machine is adopted for training to obtain the identification model, so that the identification model is used for artifact identification of dynamic electrocardiosignals, and the accuracy of identification and the robustness of an algorithm are improved.
Drawings
FIG. 1 is a flowchart of an embodiment of a method for identifying artifacts in a dynamic electrocardiographic signal according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of real heart beat and artifact of an embodiment of an artifact identification method for dynamic electrocardiographic signals according to an embodiment of the present invention;
FIG. 3a is a schematic diagram illustrating an interval of an embodiment of a method for identifying artifacts in a dynamic electrocardiographic signal according to an embodiment of the present invention;
FIG. 3b is a schematic diagram illustrating calculation of the interval correspondence of FIG. 3a according to an embodiment of the present invention;
FIG. 4a is a schematic diagram of an electrocardiographic waveform of an embodiment of a method for artifact identification of dynamic electrocardiographic signals according to an embodiment of the present invention;
FIG. 4b is a schematic diagram of differential signals of an embodiment of a method for identifying artifacts in dynamic electrocardiographic signals according to an embodiment of the present invention;
FIG. 5 is a schematic diagram illustrating a high-frequency noise calculation range of an embodiment of a method for identifying artifacts of a dynamic electrocardiographic signal according to an embodiment of the present invention;
FIG. 6 is a flowchart illustrating an embodiment of an identification preprocessing method for dynamic electrocardiographic signal artifact identification according to an embodiment of the present invention;
fig. 7 is a QRS wave detection flowchart of an embodiment of a method for identifying artifacts of a dynamic electrocardiographic signal according to an embodiment of the present invention;
Fig. 8 is a schematic hardware structure diagram of another embodiment of an artifact identification system for dynamic electrocardiographic signals according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to fall within the scope of the invention.
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
Referring to fig. 1, fig. 1 is a flowchart illustrating an embodiment of a method for identifying artifacts of a dynamic electrocardiographic signal according to an embodiment of the present invention. As shown in fig. 1, includes:
step S100, marked dynamic electrocardiographic data are obtained, wherein the marked dynamic electrocardiographic data comprise heart beat data, and the heart beat data type comprises true solid beats and artifacts;
Step 200, performing first preprocessing operation on the marked heart beat data, extracting a characteristic value from each preprocessed heart beat, and performing normalization processing to obtain a sample characteristic vector;
step S300, inputting the sample feature vector into a pre-constructed support vector machine for training, and generating an artifact identification model;
Step S400, acquiring acquired dynamic electrocardiographic data, performing second preprocessing operation on the acquired dynamic electrocardiographic data, extracting characteristic values of each preprocessed heart beat, and performing normalization processing to obtain heart beat characteristic vectors;
And S500, inputting the heart beat feature vector into an artifact identification model, obtaining an output result of the artifact identification model, and taking the output result as an artifact identification result.
In specific implementation, the technical scheme of the invention is used for acquiring a training module of the identification model and an identification module applied to dynamic electrocardiosignal artifact identification. The training module extracts characteristic values after preprocessing by using marked electrocardiographic data, takes a plurality of characteristic values as input, carries out model training by using a support vector machine, and finally obtains an artifact identification model. And in the identification module, preprocessing the acquired dynamic electrocardiographic data, extracting features by adopting a method consistent with the training module, and finally carrying out artifact identification by utilizing a model obtained by the training module to output an artifact identification result.
Marking a true heart beat and an artifact mark in dynamic electrocardiograph data; the real beats (including normal beats, ventricular beats, atrial beats, etc.) and artifacts in the dynamic electrocardiographic data are marked by the professional electrocardiographist by means of software, and the obtained results are used as training samples, as shown in fig. 2, wherein the real beats are marked as N (sinus beats) and V (ventricular premature beats), and the artifacts are marked as X (noise interference).
True-solid beats are relative to artifacts, true-heart beats include normal heart beats, ventricular heart beats, atrial heart beats, and the like. Wherein the normal heart beat comprises sinus heart beat, left bundle branch, right bundle, etc.; atrial beats include atrial premature beat, atrial escape beat, and the like; ventricular beats include ventricular premature beats, ventricular escape beats, and the like. The above description of N (sinus beat) and V (ventricular premature beat) is a specific description with respect to fig. 2.
Carrying out clustering statistics on the marked real heart beats and the marked artifacts, carrying out template matching on the dynamic electrocardiograph data real heart beats and the marked artifacts according to positions in sequence, and counting the number of the center beats of each template after all data matching is completed;
Extracting features of each heart beat to generate a training sample, wherein the extracted feature values comprise wavelet coefficients of the heart beats, the total number of heart beats of a template, the interval ratio of three intervals before and after the current heart beat, the matching degree of the current heart beat and the front and back heart beats, the information of heart beat differential data and low-frequency and high-frequency noise interference information;
Normalizing the extracted features to obtain feature vectors; after the feature extraction is completed, the support vector machine is used as a classifier to train the obtained feature vector, the normalized feature vector is used as input, and the radial basis function kernel is used as a kernel function to obtain an artifact recognition model for later artifact recognition.
Filtering the acquired dynamic electrocardiograph data, performing heart beat detection on the filtered data by adopting a differential threshold method, recording the position information of each heart beat, performing clustering statistics on each heart beat, and counting the number of heart beats of each template;
after the pretreatment is finished, extracting the characteristics of each heart beat (real heart beat and pseudo-difference), wherein the extracted characteristics and methods are consistent with those of a training module, and the extracted characteristic values are subjected to normalization treatment.
And inputting the normalized characteristic values into the recognition model obtained by the training module to obtain a final recognition result (real heart beat or artifact), and outputting the result to a user.
Further, performing a first preprocessing operation on the marked beat data, including:
And carrying out template matching on the real heart beats and the artifacts in the marked heart beat data according to the positions in sequence, and counting the number of the heart beats of each template after all the marked data are matched.
In the specific implementation, because the positions of the true heart beats and the artifacts in the electrocardiographic data are marked, the training pretreatment mainly carries out clustering statistics on the true heart beats and the artifacts. The method comprises the steps of sequentially carrying out template matching on the true and solid beats and the artifacts of the dynamic electrocardiographic data according to positions, and counting the number of center beats (true and artifacts) of each template after all data matching is completed.
Further, performing template matching on the true heart beat and the artifact in the marked heart beat data according to positions in sequence, wherein the template matching comprises the following steps:
sequentially carrying out similarity calculation on the marked heart beat data and the existing template according to the positions;
if the similarity is greater than or equal to a threshold value, judging that the heart beat data belongs to the template, and updating the data in the template;
if the similarity is smaller than the threshold value, matching calculation is carried out with the next template;
If the marked heart beat data are not matched with all the templates, a corresponding template is newly built based on the heart beat data.
When the method is implemented, inputting real heart beat or artifact data, sequentially carrying out similarity calculation with the existing template, if the result is greater than or equal to a threshold value, considering the data to belong to the template, updating the data in the template, and otherwise, continuing to carry out matching calculation with the next template; if the data is not matched with all the templates, a template is newly built based on the data.
Further, extracting a characteristic value of each preprocessed heart beat, wherein the characteristic value comprises a wavelet coefficient of the heart beat, the total number of heart beats of a template, an interval ratio of three intervals before and after the current heart beat, the matching degree of the current heart beat and the front and rear heart beats, heart beat differential data information and low-frequency and high-frequency noise interference information;
And carrying out normalization processing on the extracted characteristic values to obtain sample characteristic vectors.
In specific implementation, after pretreatment, feature extraction is carried out on each heart beat (real heart beat and artifact), and the following feature values are mainly extracted:
1) Wavelet coefficient of heart beat
Selecting data of total 1 second of 400ms before and 600ms after the position of the heart beat to carry out 5-order wavelet decomposition, and taking the obtained wavelet coefficient as a characteristic value;
The wavelet transformation is a signal analysis method, can simultaneously analyze time domain and frequency domain, and has the characteristics of time-frequency localization and multi-resolution. The multi-resolution analysis characteristic of wavelet transformation is utilized to perform multi-resolution decomposition on signals under different scales. The basic theory of wavelet transformation is quite complex, involving a large number of difficult mathematical concepts and operations. From an application point of view we can consider the wavelet transform as a cascade of low-pass and high-pass filters, followed by a downsampling operation. Wavelet transforms are signals obtained after a series of filtering and downsampling operations, which are also known as wavelet coefficients. In this example, the db6 of the parent wavelet Daubechies substrate is used for wavelet decomposition, and after 5-scale wavelet decomposition, the obtained signal (wavelet coefficient) is used as a feature value.
2) Total number of heart beats of the template
Taking the total number of heart beats under the template where each heart beat is located as a characteristic value;
3) Interval ratio of 3 intervals before and after current heart beat
As shown in fig. 3a and 3b, R3 is the current beat, R0 to R2 are the first three beats, R4 to R6 are the last three beats, RR1 is the interval between two beats R0 and R1, the value is the position of R1 minus the position of R0, and other RR interval algorithms are similar. Obtaining 5 interval ratios as characteristic values by calculating the ratio between two adjacent intervals in 6 intervals;
4) Matching degree of current heart beat and front and back heart beat
Adopting a similarity calculation method to calculate the similarity between the current heart beat and the previous heart beat and the similarity between the current heart beat and the next heart beat respectively, and taking two calculated results as characteristic values;
5) Information of heart beat differential data
As shown in fig. 4a and 4b, the electrocardiograph signal obtains differential data through differential operation, searches for a maximum differential position (point P1 in the figure) and a minimum differential position (point P0 in the figure) on the differential data, and calculates three values as characteristic values respectively: the time interval between P0 and P1, the magnitudes of P0 and P1;
in this example, a backward differential operation is adopted, and the formula is as follows. Wherein D (n) is the difference result of the current point, X (n) is the signal value of the current point, and X (n-10 ms) is the signal value before 10 ms.
D(n)＝X(n)–X(n-10ms)
6) Low frequency and high frequency noise interference information
And selecting data of 400ms before and 600ms after the position of the heart beat for total 1 second to perform median filtering to extract low-frequency interference information, specifically searching the maximum value and the minimum value of the signal in the filtered signal obtained after median filtering, and taking the result of subtracting the minimum value from the maximum value as the interference value of low-frequency noise. As shown in fig. 5, the interference value calculation range of the high-frequency noise is calculated by using the current heart beat as the reference position, and signal noise values of a length of 100ms other than 150ms before and after the reference position are calculated. The calculation method is to perform second-order difference on the signals within 100ms, sequentially take absolute value addition for the difference result, and take the final addition result as a noise value. The interference value of the high frequency noise is the maximum of these two front-to-back noise values. The low-frequency and high-frequency noise interference values are taken as characteristic values.
The time in this example is the length of time that the current experiment gives the ideal result, but other time lengths may be chosen, but generally, it is not too short or too long.
In this example, a second-order backward difference is adopted, and a specific formula is as follows. Wherein D2 (n) is a second-order differential result of the current point, X (n) is a signal value of the current point, and X (n-5 ms) and X (n-10 ms) are signal values before 5ms and 10ms respectively.
D2(n)＝X(n)–2*X(n-5ms)+X(n-10ms)
And normalizing the characteristic values, mapping all the characteristic values into [0,1] to obtain a final characteristic vector, wherein the normalization can effectively improve the convergence rate of the algorithm. The method is to traverse all characteristic values of the current heart beat and search a maximum characteristic value FMax and a minimum characteristic value FMin. Each eigenvalue is then transformed using the following formula, where FIn and FOut are the input eigenvalue and the transformed eigenvalue, respectively.
FOut＝(FIn-FMin)/(FMax-FMin)。
Further, acquiring the acquired dynamic electrocardiographic data, and performing a second preprocessing operation on the acquired dynamic electrocardiographic data, including:
Acquiring acquired dynamic electrocardiographic data, and performing filtering processing on the acquired dynamic electrocardiographic data to generate filtering data;
Performing heart beat detection on the filtered data based on a differential threshold algorithm, and recording the position information of each heart beat;
and carrying out clustering statistics on each heart beat, and counting the number of heart beats of each template.
In specific implementation, as shown in fig. 6, the collected electrocardiographic data is filtered to filter out the interference of noise such as baseline drift, power frequency interference, myoelectric interference and the like. And secondly, performing heart beat detection on the filtered data by adopting a differential threshold value method, and recording the position information of each heart beat (real heart beat and false heart beat). And finally, carrying out clustering statistics by adopting a method consistent with training pretreatment, and counting the number of center beats (real heart beats and artifacts) of each template.
And filtering the baseline drift, the power frequency interference and the myoelectricity interference by adopting different filtering methods.
Baseline drift. Baseline drift is generally due to respiration of the subject, electrode drift, etc. during electrocardiographic measurements. The characteristic is that the frequency is lower (generally below 0.7 Hz), the change is gentle but the change amplitude is larger. In this example, median filtering is used to filter baseline drift interference, a window with a length of 800ms is used to obtain the median value of signals in the window, and the original signal minus the median value signal is used as the final filtering result. Specifically, each sampling point in the electrocardiosignal is sequentially processed as follows: and calculating the median value of the signals in the total 800ms window of 400ms before and 400ms after the current point, and subtracting the median value from the value of the current point to obtain the result of median filtering of the current point.
And (5) power frequency interference. The power frequency interference is mainly caused by the electromagnetic field action generated by the alternating current power supply, the annular circuit between the electrocardio acquisition instrument and the human body and other factors, and the frequency is 50Hz or 60Hz. The power frequency interference mainly consists of sinusoidal signals, and appears as regularly distributed moire on the electrocardiogram. In this example, a trap is used to eliminate power frequency interference noise, and the filtering formula is as follows. Wherein Y (n) is a filtering result, Y (n-20 ms) is a filtering result of the first 20ms, X (n) is a signal value of the current point, and X (n-20 ms) is a signal value of the first 20 ms.
Y(n)＝0.86*Y(n-20ms)+0.93*(X(n)–X(n-20ms))
Myoelectric interference. Myoelectric interference is caused by diseases such as thyroid gland, muscle excitation contraction, and human body tension and cold stimulus, and the frequency is between 5 and 2000 Hz. The myoelectric interference belongs to a high-frequency interference signal relative to the electrocardiosignal. In this example, a low-pass filter is used to eliminate myoelectric interference noise, and the filtering formula is as follows. Wherein Y (n) is a filtering result, Y (n-1) and Y (n-2) are filtering results of the first two points, X (n) is a signal value of the current point, and X (n-10 ms) and X (n-20 ms) are signal values before 10ms and 20ms respectively.
Y(n)＝2*Y(n-1)–Y(n-2)+X(n)–2*X(n-10ms)+X(n-20ms)
Differential thresholding. As shown in fig. 7, QRS wave (i.e., heart beat) detection is performed using a differential thresholding method. Specifically, after filtering the electrocardiographic data, sequentially performing differential, square and moving average processing, then searching for a peak value, comparing the peak value with a threshold value after obtaining the peak value, if the peak value is greater than the threshold value, considering that a QRS wave is detected, otherwise treating the detected QRS wave as noise, and finally updating the threshold value, for example, making the updated threshold value be half of the previous threshold value, then pointing to the differential, square and moving average processing again, searching for a new peak value and comparing again. It should be noted that, the QRS wave detection in this embodiment belongs to the prior art, and specifically, the related description about the processes of difference, square, moving average processing, peak searching, threshold comparison, and the like in the method and system for detecting an electrocardiograph signal real-time heart rate (CN 104586384B) of the chinese patent of the invention may refer to the present invention.
The embodiment of the invention provides a false difference identification method of a dynamic electrocardiosignal, wherein in the selection of a characteristic value, the characteristic parameter comprises heart beat data self information (wavelet coefficient, differential information), related information (the number of matched heart beats, interval ratio and front and back heart beat matching degree) and noise interference information (low-frequency and high-frequency noise interference values) of other data. The characteristics of the data to be analyzed are extracted through deep mining, and a support vector machine is adopted for training to obtain an identification model for artifact identification of dynamic electrocardiosignals. The characteristics for artifact identification not only comprise the data to be analyzed, but also comprise the similarity of the data to be analyzed and the front and back data, the interval ratio of the data sequence, the noise of the data to be analyzed and other information, so that the influence of factors such as individual difference on the identification result is avoided, and the identification accuracy and the algorithm robustness are improved.
It should be noted that, there is not necessarily a certain sequence between the steps, and those skilled in the art will understand that, in different embodiments, the steps may be performed in different orders, that is, may be performed in parallel, may be performed interchangeably, or the like.
The method for identifying the artifacts of the dynamic electrocardiographic signal in the embodiment of the present invention is described above, and the system for identifying the artifacts of the dynamic electrocardiographic signal in the embodiment of the present invention is described below, referring to fig. 8, fig. 8 is a schematic hardware structure diagram of another embodiment of a system for identifying the artifacts of the dynamic electrocardiographic signal in the embodiment of the present invention, as shown in fig. 8, the system 10 includes: memory 101, processor 102, and a computer program stored on the memory and executable on the processor, which when executed by processor 101, performs the steps of:
Acquiring marked dynamic electrocardiographic data, wherein the marked dynamic electrocardiographic data comprises heart beat data, and the heart beat data type comprises true solid beats and artifacts;
Performing first preprocessing operation on the marked heart beat data, extracting a characteristic value from each preprocessed heart beat, and performing normalization processing to obtain a sample characteristic vector;
Inputting the sample feature vector into a pre-constructed support vector machine for training, and generating an artifact identification model;
acquiring acquired dynamic electrocardiographic data, performing second preprocessing operation on the acquired dynamic electrocardiographic data, extracting characteristic values of each preprocessed heart beat, and performing normalization processing to obtain heart beat characteristic vectors;
And inputting the heart beat feature vector into an artifact identification model, obtaining an output result of the artifact identification model, and taking the output result as the artifact identification result.
Specific implementation steps are the same as those of the method embodiment, and are not repeated here.
Optionally, the computer program when executed by the processor 101 also implements the steps of:
And carrying out template matching on the real heart beats and the artifacts in the marked heart beat data according to the positions in sequence, and counting the number of the heart beats of each template after all the marked data are matched.
Specific implementation steps are the same as those of the method embodiment, and are not repeated here.
Optionally, the computer program when executed by the processor 101 also implements the steps of:
Extracting characteristic values of each preprocessed heart beat, wherein the characteristic values comprise wavelet coefficients of the heart beats, the total number of heart beats of a template, the interval ratio of three intervals before and after the current heart beat, the matching degree of the current heart beat and the front and back heart beats, heart beat differential data information and low-frequency and high-frequency noise interference information;
And carrying out normalization processing on the extracted characteristic values to obtain sample characteristic vectors.
Specific implementation steps are the same as those of the method embodiment, and are not repeated here.
Optionally, the computer program when executed by the processor 101 also implements the steps of:
Acquiring acquired dynamic electrocardiographic data, and performing filtering processing on the acquired dynamic electrocardiographic data to generate filtering data;
Performing heart beat detection on the filtered data based on a differential threshold algorithm, and recording the position information of each heart beat;
and carrying out clustering statistics on each heart beat, and counting the number of heart beats of each template.
Specific implementation steps are the same as those of the method embodiment, and are not repeated here.
Optionally, the computer program when executed by the processor 101 also implements the steps of:
sequentially carrying out similarity calculation on the marked heart beat data and the existing template according to the positions;
if the similarity is greater than or equal to a threshold value, judging that the heart beat data belongs to the template, and updating the data in the template;
if the similarity is smaller than the threshold value, matching calculation is carried out with the next template;
If the marked heart beat data are not matched with all the templates, a corresponding template is newly built based on the heart beat data.
Specific implementation steps are the same as those of the method embodiment, and are not repeated here.
Embodiments of the present invention provide a non-transitory computer-readable storage medium storing computer-executable instructions for execution by one or more processors, e.g., to perform the method steps S100 through S500 of fig. 1 described above.
By way of example, nonvolatile storage media can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically erasable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM), which acts as external cache memory. By way of illustration and not limitation, RAM may be available in many forms such as Synchronous RAM (SRAM), dynamic RAM, (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), SYNCHLINK DRAM (SLDRAM), and Direct Rambus RAM (DRRAM). The disclosed memory components or memories of the operating environment described in embodiments of the present invention are intended to comprise one or more of these and/or any other suitable types of memory.
The above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (3)
1. A method for identifying artifacts in dynamic electrocardiographic signals, comprising:
Acquiring marked dynamic electrocardiographic data, wherein the marked dynamic electrocardiographic data comprises heart beat data, and the heart beat data type comprises true solid beats and artifacts;
Performing first preprocessing operation on the marked heart beat data, extracting a characteristic value from each preprocessed heart beat, and performing normalization processing to obtain a sample characteristic vector;
Inputting the sample feature vector into a pre-constructed support vector machine for training, and generating an artifact identification model;
acquiring acquired dynamic electrocardiographic data, performing second preprocessing operation on the acquired dynamic electrocardiographic data, extracting characteristic values of each preprocessed heart beat, and performing normalization processing to obtain heart beat characteristic vectors;
Inputting the heart beat feature vector into an artifact identification model, obtaining an output result of the artifact identification model, and taking the output result as an artifact identification result;
The first preprocessing operation is performed on the marked heart beat data, and the first preprocessing operation comprises the following steps:
sequentially performing template matching on the true heart beats and the artifacts in the marked heart beat data according to positions, and counting the number of the heart beats of each template after all marked data are matched;
Extracting characteristic values from each preprocessed heart beat and carrying out normalization processing to obtain sample characteristic vectors, wherein the method comprises the following steps:
Extracting characteristic values of each preprocessed heart beat, wherein the characteristic values comprise wavelet coefficients of the heart beats, the total number of heart beats of a template, the interval ratio of three intervals before and after the current heart beat, the matching degree of the current heart beat and the front and back heart beats, heart beat differential data information and low-frequency and high-frequency noise interference information;
Normalizing the extracted characteristic values to obtain sample characteristic vectors;
The acquiring the acquired dynamic electrocardiographic data, and performing a second preprocessing operation on the acquired dynamic electrocardiographic data, includes:
Acquiring acquired dynamic electrocardiographic data, and performing filtering processing on the acquired dynamic electrocardiographic data to generate filtering data;
Performing heart beat detection on the filtered data based on a differential threshold algorithm, and recording the position information of each heart beat;
carrying out clustering statistics on each heart beat, and counting the number of heart beats of each template;
the template matching is sequentially carried out on the true heart beat and the false heart beat in the marked heart beat data according to the positions, and the template matching comprises the following steps:
sequentially carrying out similarity calculation on the marked heart beat data and the existing template according to the positions;
if the similarity is greater than or equal to a threshold value, judging that the heart beat data belongs to the template, and updating the data in the template;
if the similarity is smaller than the threshold value, matching calculation is carried out with the next template;
If the marked heart beat data are not matched with all the templates, a corresponding template is newly built based on the heart beat data.
2. A system for artifact identification of dynamic electrocardiographic signals, the system comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program when executed by the processor implementing the steps of:
Acquiring marked dynamic electrocardiographic data, wherein the marked dynamic electrocardiographic data comprises heart beat data, and the heart beat data type comprises true solid beats and artifacts;
Performing first preprocessing operation on the marked heart beat data, extracting a characteristic value from each preprocessed heart beat, and performing normalization processing to obtain a sample characteristic vector;
Inputting the sample feature vector into a pre-constructed support vector machine for training, and generating an artifact identification model;
acquiring acquired dynamic electrocardiographic data, performing second preprocessing operation on the acquired dynamic electrocardiographic data, extracting characteristic values of each preprocessed heart beat, and performing normalization processing to obtain heart beat characteristic vectors;
Inputting the heart beat feature vector into an artifact identification model, obtaining an output result of the artifact identification model, and taking the output result as an artifact identification result;
The first preprocessing operation is performed on the marked heart beat data, and the first preprocessing operation comprises the following steps:
sequentially performing template matching on the true heart beats and the artifacts in the marked heart beat data according to positions, and counting the number of the heart beats of each template after all marked data are matched;
Extracting characteristic values from each preprocessed heart beat and carrying out normalization processing to obtain sample characteristic vectors, wherein the method comprises the following steps:
Extracting characteristic values of each preprocessed heart beat, wherein the characteristic values comprise wavelet coefficients of the heart beats, the total number of heart beats of a template, the interval ratio of three intervals before and after the current heart beat, the matching degree of the current heart beat and the front and back heart beats, heart beat differential data information and low-frequency and high-frequency noise interference information;
Normalizing the extracted characteristic values to obtain sample characteristic vectors;
The acquiring the acquired dynamic electrocardiographic data, and performing a second preprocessing operation on the acquired dynamic electrocardiographic data, includes:
Acquiring acquired dynamic electrocardiographic data, and performing filtering processing on the acquired dynamic electrocardiographic data to generate filtering data;
Performing heart beat detection on the filtered data based on a differential threshold algorithm, and recording the position information of each heart beat;
carrying out clustering statistics on each heart beat, and counting the number of heart beats of each template;
the template matching is sequentially carried out on the true heart beat and the false heart beat in the marked heart beat data according to the positions, and the template matching comprises the following steps:
sequentially carrying out similarity calculation on the marked heart beat data and the existing template according to the positions;
if the similarity is greater than or equal to a threshold value, judging that the heart beat data belongs to the template, and updating the data in the template;
if the similarity is smaller than the threshold value, matching calculation is carried out with the next template;
If the marked heart beat data are not matched with all the templates, a corresponding template is newly built based on the heart beat data.
3. A non-transitory computer-readable storage medium storing computer-executable instructions that, when executed by one or more processors, cause the one or more processors to perform the method of dynamic electrocardiographic signal artifact identification of claim 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211540499.6A CN115778402B (en) | 2022-12-02 | 2022-12-02 | Artifact identification method and system for dynamic electrocardiosignal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211540499.6A CN115778402B (en) | 2022-12-02 | 2022-12-02 | Artifact identification method and system for dynamic electrocardiosignal |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115778402A CN115778402A (en) | 2023-03-14 |
| CN115778402B true CN115778402B (en) | 2024-08-30 |
Family
ID=85445081
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211540499.6A Active CN115778402B (en) | 2022-12-02 | 2022-12-02 | Artifact identification method and system for dynamic electrocardiosignal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115778402B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118964907A (en) * | 2024-07-17 | 2024-11-15 | 南方医科大学南方医院 | Artifact recognition and correction system for vital sign data during surgery based on machine learning |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109303561A (en) * | 2018-11-01 | 2019-02-05 | 杭州质子科技有限公司 | It is a kind of to clap the recognition methods clapped with the abnormal heart based on the artifact heart of misclassification and supervised learning |
Family Cites Families (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8983586B2 (en) * | 2013-03-14 | 2015-03-17 | Medtronic, Inc. | Beat-morphology matching scheme for cardiac sensing and event detection |
| CN109645979A (en) * | 2017-10-10 | 2019-04-19 | 深圳市理邦精密仪器股份有限公司 | Ambulatory ecg signal artifact identification method and device |
| CN108937915B (en) * | 2018-07-24 | 2021-02-19 | 杭州质子科技有限公司 | Method for identifying premature beat in dynamic electrocardiogram |
| CN109171707A (en) * | 2018-10-24 | 2019-01-11 | 杭州电子科技大学 | A kind of intelligent cardiac figure classification method |
| CN111281378A (en) * | 2020-02-13 | 2020-06-16 | 苏州百慧华业精密仪器有限公司 | Method and device for screening suspected T-wave electricity alternate sections in dynamic electrocardiogram |
| CN111297351A (en) * | 2020-02-13 | 2020-06-19 | 苏州百慧华业精密仪器有限公司 | Motion artifact identification method and device in dynamic electrocardiogram |
| CN111419212A (en) * | 2020-02-27 | 2020-07-17 | 平安科技（深圳）有限公司 | Method and device for processing electrocardiogram data, storage medium and computer equipment |
| CN113712525B (en) * | 2020-05-21 | 2024-07-12 | 深圳市理邦精密仪器股份有限公司 | Physiological parameter processing method and device and medical equipment |
| CN113995420B (en) * | 2020-07-14 | 2025-01-24 | 深圳华清心仪医疗电子有限公司 | A dynamic electrocardiogram data processing method and processing device based on RR interval ratio scatter diagram |
| CN112528783B (en) * | 2020-11-30 | 2024-04-16 | 深圳邦健生物医疗设备股份有限公司 | Electrocardiogram and heart beat data clustering method, device, electronic equipment and medium |
| CN114494798B (en) * | 2022-01-28 | 2025-06-27 | 纳龙健康科技股份有限公司 | Electrocardiogram artifact confirmation method, terminal device and storage medium |
| CN114795235B (en) * | 2022-04-14 | 2023-04-07 | 中国人民解放军陆军第八十二集团军医院 | Single-lead electrocardiogram monitoring method and system based on morphological contour algorithm |
-
2022
- 2022-12-02 CN CN202211540499.6A patent/CN115778402B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109303561A (en) * | 2018-11-01 | 2019-02-05 | 杭州质子科技有限公司 | It is a kind of to clap the recognition methods clapped with the abnormal heart based on the artifact heart of misclassification and supervised learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115778402A (en) | 2023-03-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP6367442B2 (en) | Method and system for disease analysis based on conversion of diagnostic signals | |
| Saini et al. | QRS detection using K-Nearest Neighbor algorithm (KNN) and evaluation on standard ECG databases | |
| EP1802230B1 (en) | Monitoring physiological activity using partial state space reconstruction | |
| CA2979135C (en) | Systems, apparatus and methods for sensing fetal activity | |
| Sayadi et al. | A model-based Bayesian framework for ECG beat segmentation | |
| US9050007B1 (en) | Extraction of cardiac signal data | |
| CN111956203B (en) | Electrocardiosignal parameterization method, model training method, device, equipment and medium | |
| Lee et al. | Personal identification using a robust eigen ECG network based on time-frequency representations of ECG signals | |
| Bulusu et al. | Transient ST-segment episode detection for ECG beat classification | |
| CN112603325A (en) | Electrocardio abnormity discrimination system and method based on morphological filtering and wavelet threshold | |
| Kaur et al. | On the detection of cardiac arrhythmia with principal component analysis | |
| Rao et al. | P-and T-wave delineation in ECG signals using parametric mixture Gaussian and dynamic programming | |
| CN115778402B (en) | Artifact identification method and system for dynamic electrocardiosignal | |
| Boucheham et al. | Piecewise linear correction of ECG baseline wander: a curve simplification approach | |
| Rao et al. | Performance identification of different heart diseases based on neural network classification | |
| CN115881276B (en) | Time-frequency double-bar code characteristic image generation method of electrocardiosignal and storage medium | |
| Khandait et al. | ECG signal processing using classifier to analyses cardiovascular disease | |
| Soe et al. | ECG signal classification using discrete wavelet transform and pan tompkins algorithm | |
| Rodrigues et al. | The issue of automatic classification of heartbeats | |
| Banerjee et al. | A classification approach for myocardial infarction using voltage features extracted from four standard ECG leads | |
| Gómez-Herrero et al. | Relative estimation of the Karhunen-Loève transform basis functions for detection of ventricular ectopic beats | |
| Pander et al. | Fuzzy-based algorithm for QRS detection | |
| Davydov et al. | Myocardial infarction detection using wavelet analysis of ECG signal | |
| CN115040137B (en) | Electrocardiosignal parameterization method, model training method, device, equipment and medium | |
| Vysiya et al. | Automatic detection of cardiac arrhythmias in ECG signal for IoT application |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |