CN114287950A - Heart disease computer-assisted classification method based on continuous coherence and Fourier transform - Google Patents
Heart disease computer-assisted classification method based on continuous coherence and Fourier transform Download PDFInfo
- Publication number
- CN114287950A CN114287950A CN202111610398.7A CN202111610398A CN114287950A CN 114287950 A CN114287950 A CN 114287950A CN 202111610398 A CN202111610398 A CN 202111610398A CN 114287950 A CN114287950 A CN 114287950A
- Authority
- CN
- China
- Prior art keywords
- fourier transform
- computer
- continuous
- coherence
- classification method
- 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.)
- Pending
Links
Images
Landscapes
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention relates to a heart disease computer-aided classification method based on continuous coherence and Fourier transform, belonging to the technical field of computer-aided diagnosis and signal processing. The invention takes the electrocardiogram as a one-dimensional time sequence signal, converts the signal into point cloud by utilizing Fourier transform, and researches the integral structural difference of the point cloud by continuously and homomorphically extracting topological features, thereby classifying the original signal, ensuring the application effect, avoiding the defects of information loss, easy interference by fine factors and the like. The Fourier transform is a powerful tool for analyzing periodic signals, and due to the completeness of the orthonormal base, the information of the signals can be completely converted into the information of point clouds. The method can keep accuracy, can greatly reduce calculated amount, is based on the topological characteristic of the signal, does not depend on adjustment parameters, and has remarkable advantages. The method has wide application prospect in the fields of the classification of the electrocardiosignals, biomedicine, signal processing and the like.
Description
Technical Field
The invention relates to a computer-aided classification method for pre-classifying heart disease abnormal conditions based on continuous coherence and Fourier transform, belonging to the technical field of computer-aided diagnosis and signal processing.
Background
Heart disease, one of the most fatal diseases in the world, poses a great threat to human health and safety. The method has great significance for protecting life and realizing personalized medical treatment by judging, predicting and preventing heart diseases in time in the face of growing patients. The conventional method for diagnosing heart diseases is mainly to observe an electrocardiogram for recording the electrical activity of heart contraction. However, interpretation of an electrocardiogram requires professional medical staff and detailed medical background knowledge, and if the interpretation is performed manually by the medical staff, the life health of a patient cannot be sufficiently guaranteed in the case of unbalanced medical resource distribution and a lack of conditions.
With the rapid development of computer technology, particularly artificial intelligence technology, the utilization of a computer system for electrocardiogram pre-classification prediction becomes a new technical means, effectively relieves the uneven distribution of artificial medical resources, and provides powerful help for doctors to further formulate specific treatment schemes.
However, when performing heart disease abnormality prediction classification on an electrocardiogram, the conventional computer-aided diagnosis system mainly performs feature extraction on a single-lead electrocardiogram and a standard 12-lead electrocardiogram from the viewpoints of signal processing, a dynamic system, statistical analysis, machine learning, and the like, and performs heart disease condition prediction classification according to the features. This approach is not ideal for pre-classifying heart disease abnormalities due to the fact that the detail factors are too large and the interpretability is poor.
Furthermore, currently, computer-aided systems rely mostly on the selection of manual parameters. While a sufficiently experienced code operator is required to make the reference for different patients, improper selection of parameters will substantially reduce the accuracy of the system diagnosis.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, and creatively provides a heart disease computer-aided classification method based on continuous coherence and Fourier transform to solve the technical problems of damaged signal structural information, poor interpretability, strong parameter dependence and the like in the process of carrying out electrocardiogram classification by a computer-aided diagnosis system.
The method has the innovation points that: the electrocardiogram is taken as a one-dimensional time sequence signal, the signal is converted into point cloud by utilizing Fourier transform for the first time, and the overall structural difference of the point cloud is researched by continuously and homologically extracting topological features, so that the original signals are classified, the application effect is ensured, the defects of information loss, easiness in interference of subtle factors and the like are avoided. The Fourier transform is a powerful tool for analyzing periodic signals, and meanwhile, due to the completeness of the orthonormal base, the information of the signals can be completely converted into the information of point clouds.
The invention is realized by adopting the following technical scheme.
An electrocardiogram is considered as a one-dimensional time series signal. The continuous time series signal is first preprocessed by interpolation and filtering.
The signal is then embedded into a point cloud using a sliding window and a fourier transform.
And then, continuously carrying out coherence on the point cloud application to obtain a corresponding coherence map and a bar graph.
And finally, extracting topological features, constructing a feature space and segmenting by using a support vector machine.
Advantageous effects
Compared with the prior art, the method has the following advantages:
the method is a novel method based on topology, and discloses the interpretability of electrocardiosignal classification. Compared with the prior art, the method can keep accuracy, can greatly reduce calculated amount, is based on the topological characteristic of the signal, does not depend on adjustment parameters, and has remarkable advantages.
The method has wide application prospect in the fields of the classification of the electrocardiosignals, biomedicine, signal processing and the like.
Drawings
FIG. 1 is an overall flow diagram of the method of the present invention;
FIG. 2 is an example of four electrocardiograms of the method;
FIG. 3 is a schematic diagram of a noise reduction process in the present invention;
FIGS. 4 to 6 are schematic views of a normal electrocardiogram, a dot cloud chart and a continuous tone chart, respectively, according to the present invention;
FIGS. 7 to 9 are schematic diagrams of an electrocardiogram of ventricular premature beats, a dot cloud chart and a sustained concoction chart, respectively, according to the present invention;
FIGS. 10 to 12 are schematic diagrams of a left bundle branch block electrocardiogram, a dot cloud chart and a continuous tone chart, respectively, according to the present invention;
FIGS. 13 to 15 are schematic views of an electrocardiogram, a cloud point diagram and a continuous tone diagram of ventricular flutter in accordance with the present invention;
FIG. 16 is a graph showing the effect of the process of constructing the complex of Vietoris-Rips according to the present invention;
FIG. 17 is an exemplary graph of a bar graph and the creation of Betty numbers in the present invention;
FIG. 18 is a schematic diagram of the distribution of the raw ECG data set in the feature space according to the present invention;
FIG. 19 is a plane of the support vector machine SVM pair feature space segmentation in accordance with the present invention.
Detailed Description
The method of the present invention will be described in further detail with reference to the accompanying drawings.
As shown in fig. 1, a computer-aided classification method for heart disease based on persistent coherence and fourier transform comprises the following steps:
step 1: as shown in FIG. 2, the method of the present invention mainly assists in classifying the cardiac signals, such as health, left bundle branch block, ventricular flutter, and ventricular premature beat.
And preprocessing the continuous electrocardiosignals through interpolation and a filter.
An electrocardiogram is considered as a one-dimensional time series signal. Because the frequency of the alternating current used by power supply equipment in China is 50Hz, and the power frequency interference generated by the power supply equipment has a large influence on an electrocardiogram, the filter can be used for filtering and reducing noise of the electrocardio frequency above 50 Hz.
Then, the position of the R wave of each heartbeat is detected, and points with a fixed proportion (wave front 1/3, wave rear 2/3) behind the R wave front are selected to form a single heartbeat. According to the length of the R wave and different heartbeat lengths, the expansion transformation is uniformly carried out by utilizing interpolation, and the information content of a single heartbeat (for example, a single heartbeat with 300 points) is maintained. As shown in fig. 3.
And finally, segmenting continuous heartbeats according to the position of the R wave.
Step 2: the heartbeat signals are embedded into a point cloud using a sliding window and a fourier transform.
Specifically, the length of the sliding window W is set to M points for each heartbeat H. Wherein M is not less than 50. This is because, up to 50 points, it is possible to realize sufficiently long information including QRS complex, which is an important factor for judging the electrocardiogram, QRS complex reflects the change of left and right ventricular depolarization potential and time of the human body, first downward wave is Q wave, upward wave is R wave, and then downward wave is S wave.
Sliding on the heartbeat H by using a sliding window, and taking out a continuous signal S with the length of M every time of slidingi. At this time, fast Fourier transform is performed to generate a sequence of complex numbers of size M, the j-th complex number in the sequence including its amplitude
Angle of rotation
And phase
Three features. According to the characteristics of fast fourier, there are:
where Fs represents the sampling frequency and N represents the number of points of a single sample.
Thus, for any i, SiThe generated complex sequences all have the same orthogonal base, SiRepresenting a continuous signal. Extraction of
As a coordinate of the jth position, a point cloud of heartbeats H is thus generated. As shown in fig. 4 to 15, twelve diagrams respectively show images of four electrocardiograms, a point cloud diagram and a continuous tone diagram.
And step 3: and applying continuous coherence to the point cloud to obtain a coherence map and a bar graph corresponding to the point cloud.
Specifically, vietoris-rips complex (V-R complex for short) is constructed for point clouds, and the corresponding topology is simply complex according to different dimensions. The invention selects a one-dimensional simple complex to reflect the topological characteristics of the point cloud, as shown in fig. 16. As time and radius expand, the birth and death times of different generators (in the V-R manifold, one circle comes out as the radius increases, and this circle is called a generator) are recorded to form a coherence map (see fig. 6) and a bar map (see fig. 17).
And 4, step 4: extracting topological features as spatial dimensions to construct a feature space, and extracting a segmentation plane by using a Support Vector Machine (SVM) to segment.
Specifically, according to the concordance diagram and the bar graph, the continuous entropy, the maximum generation element survival time and the maximum Betty number survival time are extracted. The one-dimensional Betty number is a topological invariant, and can distinguish the scale of one-dimensional circulation in different topological spaces. Therefore, in constructing the replica, the maximum Betty number lifetime is chosen to describe the most stable case.
By examining the sustained coherence map, it was found that there was a significant difference in the distance of the different points to the diagonal. Thus, the most durable cycle is expressed in terms of the largest life. In addition to focusing on the details of the persistent graph, the entire property needs to be described, thus introducing entropy. The behavior of the original electrocardiogram in the feature space is shown in fig. 18.
And finally, segmenting the points in the feature space by using a support vector machine of the linear kernel to obtain the final segmentation accuracy and the final segmentation plane. The new electrocardiosignals can be classified by utilizing the cutting plane.
Example verification
The method is applied to a data set consisting of 250 heartbeats, and the data is derived from an MIT-BIH heartbeat database. The specific identified condition is shown in fig. 19, which results in an accuracy of more than 82% in distinguishing normal heartbeats from abnormal heartbeats, wherein an accuracy of 99% in distinguishing ventricular flutter from normal heartbeats is achieved.
Claims (8)
1. A heart disease computer-assisted classification method based on continuous coherence and Fourier transform is characterized by comprising the following steps:
step 1: preprocessing the continuous electrocardiosignals through interpolation and a filter;
step 2: embedding the heartbeat signals into a point cloud by using a sliding window and Fourier transform, wherein the method comprises the following steps:
setting the length of a sliding window W as M points for each heartbeat H, wherein M is not less than 50;
sliding on the heartbeat H by using a sliding window, and taking out a continuous signal S with the length of M every time of slidingi(ii) a At this time, fast Fourier transform is performed to generate a sequence of complex numbers of size M, the j-th complex number in the sequence including its amplitude
Angle of rotation
And phase
Three characteristics; according to the characteristics of fast fourier, there are:
wherein Fs represents the sampling frequency and N represents the number of points of a single sample;
for arbitrary i, SiThe generated complex sequences all have the same orthogonal base, SiRepresents a continuous signal; extraction of
As coordinates of the jth position, thereby producing a point cloud of heartbeats H;
and step 3: applying continuous coherence to the point cloud to obtain a coherence map and a bar graph corresponding to the point cloud;
and 4, step 4: extracting topological features as space dimensions to construct a feature space, extracting a segmentation plane by using a support vector machine to segment to obtain final segmentation accuracy and a segmentation plane, and classifying new electrocardiosignals by using the segmentation plane.
2. The computer-aided classification method of heart diseases based on persistent coherence and fourier transform as claimed in claim 1, characterized in that the cardiac electrical signals comprise: health, left bundle branch block, ventricular flutter and ventricular premature beat.
3. The computer-aided classification method of heart diseases based on continuous coherence and Fourier transform as claimed in claim 1, wherein the electrocardiogram is regarded as a one-dimensional time series signal, and the electrocardio frequency is first filtered and de-noised by a filter;
then, detecting the position of the R wave of each heartbeat, and selecting points with a fixed proportion behind the R wave front to form a single heartbeat; according to the length of the R wave and the different heartbeat lengths, carrying out unified telescopic transformation by utilizing interpolation to maintain the information content of a single heartbeat;
and finally, segmenting continuous heartbeats according to the position of the R wave.
4. The computer-aided classification method of heart diseases based on continuous harmony fourier transform as claimed in claim 3, wherein the fixed ratio after selecting the R wave front is: wavefront 1/3, wavefront 2/3.
5. The computer-aided classification method of heart diseases based on continuous harmony fourier transform as claimed in claim 3, wherein the information amount for maintaining a single heartbeat is 300 points.
6. The computer-aided classification method of heart diseases based on continuous harmony and Fourier transform as claimed in claim 3, characterized in that the cardioelectric frequency above 50Hz is filtered and de-noised by a filter.
7. The computer-aided classification method for heart diseases based on persistent coherence and fourier transform as claimed in claim 1, wherein in step 3, vietoris-rips complex is constructed for point clouds, and corresponding topology is simply complex according to different dimensions; and selecting a one-dimensional simple complex to reflect the topological characteristics of the point cloud, and recording the time of birth and death of different generators along with the expansion of time and radius to form a concoction map and a bar graph.
8. The computer-aided classification method for heart diseases based on persistent coherence and fourier transform as claimed in claim 1, wherein in step 4, persistent entropy, maximum generator lifetime and maximum beth number lifetime are extracted from coherence maps and bar maps.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111610398.7A CN114287950A (en) | 2021-12-27 | 2021-12-27 | Heart disease computer-assisted classification method based on continuous coherence and Fourier transform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111610398.7A CN114287950A (en) | 2021-12-27 | 2021-12-27 | Heart disease computer-assisted classification method based on continuous coherence and Fourier transform |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114287950A true CN114287950A (en) | 2022-04-08 |
Family
ID=80970132
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111610398.7A Pending CN114287950A (en) | 2021-12-27 | 2021-12-27 | Heart disease computer-assisted classification method based on continuous coherence and Fourier transform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114287950A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116416417A (en) * | 2023-06-09 | 2023-07-11 | 远洋装饰工程股份有限公司 | Toilet hidden drainage structure auxiliary overhaul system based on augmented reality |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20150009843A (en) * | 2013-07-17 | 2015-01-27 | 건국대학교 산학협력단 | System for classification electrocardiogram signals |
| US20200074281A1 (en) * | 2018-08-31 | 2020-03-05 | Fujitsu Limited | Computer-readable recording medium, abnormality determination method, and abnormality determination device |
| CN111753709A (en) * | 2020-06-19 | 2020-10-09 | 太原科技大学 | Fractional Fourier transform-based three-dimensional pucker-palate recognition method |
| CN112184869A (en) * | 2020-10-09 | 2021-01-05 | 北京理工大学 | Point cloud simplification method for keeping geometric features based on absolute Gaussian curvature estimation |
| CN113331845A (en) * | 2021-05-31 | 2021-09-03 | 浙江大学 | Electroencephalogram signal feature extraction and accuracy discrimination method based on continuous coherence |
-
2021
- 2021-12-27 CN CN202111610398.7A patent/CN114287950A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20150009843A (en) * | 2013-07-17 | 2015-01-27 | 건국대학교 산학협력단 | System for classification electrocardiogram signals |
| US20200074281A1 (en) * | 2018-08-31 | 2020-03-05 | Fujitsu Limited | Computer-readable recording medium, abnormality determination method, and abnormality determination device |
| CN111753709A (en) * | 2020-06-19 | 2020-10-09 | 太原科技大学 | Fractional Fourier transform-based three-dimensional pucker-palate recognition method |
| CN112184869A (en) * | 2020-10-09 | 2021-01-05 | 北京理工大学 | Point cloud simplification method for keeping geometric features based on absolute Gaussian curvature estimation |
| CN113331845A (en) * | 2021-05-31 | 2021-09-03 | 浙江大学 | Electroencephalogram signal feature extraction and accuracy discrimination method based on continuous coherence |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116416417A (en) * | 2023-06-09 | 2023-07-11 | 远洋装饰工程股份有限公司 | Toilet hidden drainage structure auxiliary overhaul system based on augmented reality |
| CN116416417B (en) * | 2023-06-09 | 2023-09-05 | 远洋装饰工程股份有限公司 | Toilet hidden drainage structure auxiliary overhaul system based on augmented reality |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Luo et al. | Patient-specific deep architectural model for ECG classification | |
| CN109117730B (en) | Real-time electrocardiogram atrial fibrillation judgment method, device and system and storage medium | |
| Karpagachelvi et al. | ECG feature extraction techniques-a survey approach | |
| Lu et al. | Feature selection for ECG signal processing using improved genetic algorithm and empirical mode decomposition | |
| CN108324263B (en) | Noninvasive cardiac electrophysiology inversion method based on low-rank sparse constraint | |
| CN111449644A (en) | Bioelectricity signal classification method based on time-frequency transformation and data enhancement technology | |
| Khan et al. | Electrocardiogram heartbeat classification using convolutional neural networks for the detection of cardiac Arrhythmia | |
| Guan et al. | Low-dimensional denoising embedding transformer for ECG classification | |
| CN112932498A (en) | T wave morphology classification system with strong generalization capability based on deep learning | |
| Zhang et al. | Detection of atrial fibrillation from variable-duration ECG signal based on time-adaptive densely network and feature enhancement strategy | |
| Huang et al. | A multiview feature fusion model for heartbeat classification | |
| CN114287950A (en) | Heart disease computer-assisted classification method based on continuous coherence and Fourier transform | |
| CN113180685B (en) | Electrocardio abnormity discrimination system and method based on morphological filtering and wavelet threshold | |
| Spicher et al. | ECG delineation using a piecewise Gaussian derivative model with parameters estimated from scale-dependent algebraic expressions | |
| Zeng et al. | Deep arrhythmia classification based on SENet and lightweight context transform | |
| Afkhami et al. | ECG based detection of left ventricular hypertrophy using higher order statistics | |
| Kishore et al. | An innovative machine learning approach for classifying ECG signals in healthcare devices | |
| Jabborov | New way to process signals digitally | |
| Rashkovska et al. | Clustering of heartbeats from ECG recordings obtained with wireless body sensors | |
| CN115813409A (en) | Ultra-low-delay moving image electroencephalogram decoding method | |
| Ayushi et al. | A survey of ECG classification for arrhythmia diagnoses using SVM | |
| CN114569116A (en) | Three-channel image and transfer learning-based ballistocardiogram ventricular fibrillation auxiliary diagnosis system | |
| Abeysekera et al. | Patterns in Hilbert transforms and Wigner-Ville distributions of electrocardiogram data | |
| Lu et al. | A New Multichannel Parallel Network Framework for the Special Structure of Multilead ECG | |
| Fraser et al. | Time-delay lifts for physiological signal exploration: An application to ECG analysis |
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 |