CN109157210A - A kind of epicardial potential method for reconstructing based on ADMM and neural network - Google Patents
A kind of epicardial potential method for reconstructing based on ADMM and neural network Download PDFInfo
- Publication number
- CN109157210A CN109157210A CN201810759651.7A CN201810759651A CN109157210A CN 109157210 A CN109157210 A CN 109157210A CN 201810759651 A CN201810759651 A CN 201810759651A CN 109157210 A CN109157210 A CN 109157210A
- Authority
- CN
- China
- Prior art keywords
- admm
- model
- electrocardio
- neural network
- epicardial potential
- 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
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
Abstract
The invention discloses a kind of epicardial potential method for reconstructing based on ADMM and neural network, comprising the following steps: establish electrocardio reverse temperature intensity model；It is ADMM algorithm model by the electric reverse temperature intensity model conversation based on ADMM algorithm；ADMM recursive neural network model is converted by the ADMM algorithm model, calculates electrocardio inverse problem result.The present invention can more accurately solve electrocardio inverse problem, it is ensured that the accuracy of epicardial potential reconstructed results, it is with a high credibility.
Description
Technical field
The present invention relates to epicardial potential technical field, in particular to a kind of external membrane of heart based on ADMM and neural network
Current potential method for reconstructing.
Background technique
The main target of electrocardio inverse problem research is speculated in heart according to cardioelectric field Potential distribution caused by body surface
Electrical activity, to judge the health status of heart.
For electrocardio inverse problem, we finally need to obtain unique solution.However, due to electrocardio inverse problem be it is ill, make
Substantially unstable at its equation, so that electrocardio inverse problem is stablized the most common way is to carry out Regularization, a success
Regularization process potential problem can be made to be stablized and feasible solution.
Electrocardio inverse problem is to rebuild epicardial potential based on body surface potential, its maximum feature is its ill-posedness, canonical
Change method selects optimal solution by a series of limitations of prior information stablize inverse problem.When solving electrocardio inverse problem, such as
It is vital that, which chooses regularization method,.
Currently used regularization method has Tikhonov (Tikhonov regularization), TSVD (truncated singular value decomposition
Hair) and TTLS (truncated total least squares), these are all direct regularization methods, they are all using the think of for seeking pseudoinverse
Think, then apply some smoothness constraints, a final step can obtain the solution of inverse problem.The advantages of these methods is to calculate simply
And time complexity is low, but the accuracy of its result relies heavily on the quality of parameter, and for different numbers
According to it is also unstable for acquiring the solution of inverse problem.Some methods are fine for the calculating effect of a certain sets of data, but change into
Other data errors are larger, it is therefore desirable to find a kind of stable method to overcome these disadvantages.
Summary of the invention
The purpose of the present invention is to provide a kind of epicardial potential method for reconstructing based on ADMM and neural network, energy
It is enough more accurately to solve electrocardio inverse problem, it is ensured that the accuracy of epicardial potential reconstructed results, it is with a high credibility.
To achieve the above object, the invention adopts the following technical scheme:
A kind of epicardial potential method for reconstructing based on ADMM and neural network, comprising the following steps:
S1, electrocardio reverse temperature intensity model is established；
S2, it is based on ADMM algorithm, is ADMM algorithm model by the electric reverse temperature intensity model conversation；
S3, it converts the ADMM algorithm model on ADMM recursive neural network model, calculates electrocardio inverse problem result.
Preferably, the electrocardio reverse temperature intensity model are as follows:
Wherein, ΦEFor epicardial potential, ΦTFor body surface potential, A is the conversion extrapolated by Thorax volume conductor model
Matrix.
Preferably, the step S2 is realized especially by following methods:
Based on ADMM algorithm and Augmented Lagrangian Functions, the electrocardio reverse temperature intensity model is expressed as:
The ADMM algorithm model are as follows:
Wherein,
Preferably, the ADMM recursive neural network model in the step S3, using μ (0), ΦT(0), E (0) is used as net
Network input, ΦEIt is exported as network, emulates preset standard value in dataFor web tab.
After adopting the above technical scheme, compared with the background technology, the present invention, having the advantages that
The present invention can more accurately solve electrocardio inverse problem, it is ensured that the accuracy of epicardial potential reconstructed results, it can
Reliability is high.
Detailed description of the invention
Fig. 1 is flow diagram of the invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments,
The present invention will be described in further detail.It should be appreciated that specific embodiment described herein is only used to explain this hair
It is bright, it is not intended to limit the present invention.
Embodiment
Referring to Fig. 1, the invention discloses a kind of epicardial potential method for reconstructing based on ADMM and neural network, packet
Include following steps:
S1, electrocardio reverse temperature intensity model is established:
Wherein, ΦEFor epicardial potential, ΦTFor body surface potential, A is the conversion extrapolated by Thorax volume conductor model
Matrix.
The derivation process of electrocardio reverse temperature intensity model is as follows:
Torso model volume conductor model is the bridge for connecting cardiac electrical activity and body surface electrical activity, in order to simplify complexity
Torso model volume conductor model, facilitating electrocardio inverse problem is indicated and is solved with numerical method.People are by heart -- trunk
Region segmentation obtains the threedimensional model of Thorax volume conductor at triangle interconnected, and obtains mould using Element BEM
The numerical value of type is stated, wherein the current potential at each moment is stated by node potential.In this way, we can be by electrocardio just
The model of problem is expressed as:
ΦT=A ΦE (1)
Electrocardio inverse problem model can indicate are as follows:
ΦE=A-ΦT (2)
Due to the Ill-posed characteristic of A matrix, we can not be directly to A matrix inversion, this is but also electrocardio inverse problem becomes spine
The pathosis problem of hand.So converting electrocardio inverse problem to the general convex optimization problem that can be solved:
Wherein, λ represents penalty factor, (ΦE) represent to (ΦE) bound term taken.
Using λ | | ΦE||2As penalty term, and in order to optimize operation, required majorization of solutions problem model is rewritten
Are as follows:
Thus obtain above-mentioned electrocardio reverse temperature intensity model.
S2, it is based on ADMM algorithm, is ADMM algorithm model by electric reverse temperature intensity model conversation.
Above-mentioned optimization problem is solved using ADMM algorithm, then formula (4) can indicate are as follows:
According to ADMM algorithm principle, the Lagrangian formulation of formula (3) can be indicated are as follows:
According to ADMM algorithm, iterative process include the following three steps:
Formula (7) are unfolded, preset parameter E and μ, update ΦE:
Formula (8) are unfolded, preset parameter ΦEAnd μ, update E:
It enablesIt can obtain:
Formula (9) are unfolded, preset parameter ΦEAnd E, update μ:
S3, it converts ADMM algorithm model on ADMM recursive neural network model, calculates electrocardio inverse problem result.Using μ
(0)、ΦT(0), E (0) is used as network inputs, ΦEIt is exported as network, emulates preset standard value in dataFor network mark
Label.
Experimental evaluation
One, evaluation criterion
The accuracy for solving electrocardio inverse problem is quantitatively evaluated using two relative error RE, related coefficient CC aspects.
Relative error RE indicate absolute error shared by true value ratio, can definitely reflect the order of accuarcy of measurement, it and survey
Amount accuracy is positively correlated.Related coefficient CC illustrates the quantitative measurement of certain type of correlation and dependence.
RE is defined as:
CC is defined as:
Wherein, NkIndicate the node total number of heart geometric jacquard patterning unit surface, ΦHIndicate that the potential value of heart surface, subscript ' ∧ ' refer to
Reference value, subscriptRefer to average value.
Two, experiment condition
Using true torso model volume conductor model, with the heart surface electricity obtained by ECGsim software emulation
Position and body surface potential data, each heartbeat data 500ms, the cardiac potential data of generation are (257*500), body surface potential number
According to for (64*500), by adjusting heart surface potential source, we produce 19500 datas, and are selected at random using wherein
80%, i.e., 15600 are carried out training parameter, and remaining 20% (i.e. 3900) are as verifying collection to verify network operation result.
And the error evaluation of result is trained using quantitative RE and CC, then by the regularization of result and the other three mainstream
Method is compared.Result images are visualized by map3d, SCIRUN4.7, MATLABR2013a to realize.
Three, experimental result
Epicardial potential weight is carried out using epicardial potential method of the present invention and the analysis of Tikhonov, TTLS, ADMM method
It builds, RE the and CC value of obtained result is as shown in table 1:
Table 1
| Tikhonov | TTLS | ADMM | our method | |
| RE | 0.184395 | 0.235334 | 0.160781 | 0.140548 |
| CC | 0.961311 | 0.956793 | 0.958757 | 0.964171 |
It can be seen from Table 1 that epicardial potential method of the present invention is compared with other three kinds of methods, RE is smaller, and CC is more
Greatly, it can be deduced that the result obtained using the present embodiment epicardial potential method for reconstructing is more accurate.
More than, it is merely preferred embodiments of the present invention, but scope of protection of the present invention is not limited thereto, appoints
In the technical scope disclosed by the present invention, any changes or substitutions that can be easily thought of, all by what those familiar with the art
It is covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be with scope of protection of the claims
It is quasi-.
Claims (4)
1. a kind of epicardial potential method for reconstructing based on ADMM and neural network, which comprises the following steps:
S1, electrocardio reverse temperature intensity model is established；
S2, it is based on ADMM algorithm, is ADMM algorithm model by the electric reverse temperature intensity model conversation；
S3, it converts the ADMM algorithm model on ADMM recursive neural network model, calculates electrocardio inverse problem result.
2. a kind of epicardial potential method for reconstructing based on ADMM and neural network as described in claim 1, which is characterized in that
The electrocardio reverse temperature intensity model are as follows:
Wherein, ΦEFor epicardial potential, ΦTFor body surface potential, A is the transition matrix extrapolated by Thorax volume conductor model.
3. a kind of epicardial potential method for reconstructing as claimed in claim 2, which is characterized in that the step S2 especially by with
Lower method is realized:
Based on ADMM algorithm and Augmented Lagrangian Functions, the electrocardio reverse temperature intensity model is expressed as:
The ADMM algorithm model are as follows:
Wherein,
4. a kind of epicardial potential method for reconstructing as claimed in claim 3, which is characterized in that the ADMM in the step S3 changes
For neural network model, using μ (0), ΦT(0), E (0) is used as network inputs, ΦEIt is exported as network, emulates in data and preset
Standard valueFor web tab.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810759651.7A CN109157210A (en) | 2018-07-11 | 2018-07-11 | A kind of epicardial potential method for reconstructing based on ADMM and neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810759651.7A CN109157210A (en) | 2018-07-11 | 2018-07-11 | A kind of epicardial potential method for reconstructing based on ADMM and neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109157210A true CN109157210A (en) | 2019-01-08 |
Family
ID=64897693
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810759651.7A Pending CN109157210A (en) | 2018-07-11 | 2018-07-11 | A kind of epicardial potential method for reconstructing based on ADMM and neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109157210A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110575166A (en) * | 2019-09-30 | 2019-12-17 | 北京信息科技大学 | Method and device for time-frequency analysis of human electroencephalogram signals |
| CN111714124A (en) * | 2020-06-18 | 2020-09-29 | 中国科学院深圳先进技术研究院 | Magnetic resonance film imaging method, device, imaging equipment and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150234779A1 (en) * | 2014-02-20 | 2015-08-20 | Mitsubishi Electric Research Laboratories, Inc. | Method for Solving Quadratic Programs for Convex Sets with Linear Equalities by an Alternating Direction Method of Multipliers with Optimized Step Sizes |
| CN105224942A (en) * | 2015-07-09 | 2016-01-06 | 华南农业大学 | A kind of RGB-D image classification method and system |
| CN105378762A (en) * | 2014-04-08 | 2016-03-02 | 微软技术许可有限责任公司 | Deep learning using alternating direction method of multipliers |
| US20160275416A1 (en) * | 2015-03-20 | 2016-09-22 | Nec Laboratories America, Inc. | Fast Distributed Nonnegative Matrix Factorization and Completion for Big Data Analytics |
| CN106093814A (en) * | 2016-06-02 | 2016-11-09 | 浙江理工大学 | A kind of cardiac magnetic resonance imaging method based on multiple dimensioned low-rank model |
| CN107260159A (en) * | 2017-06-28 | 2017-10-20 | 浙江理工大学 | A kind of cardiac electric functional imaging method based on convolutional neural networks |
-
2018
- 2018-07-11 CN CN201810759651.7A patent/CN109157210A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150234779A1 (en) * | 2014-02-20 | 2015-08-20 | Mitsubishi Electric Research Laboratories, Inc. | Method for Solving Quadratic Programs for Convex Sets with Linear Equalities by an Alternating Direction Method of Multipliers with Optimized Step Sizes |
| CN105378762A (en) * | 2014-04-08 | 2016-03-02 | 微软技术许可有限责任公司 | Deep learning using alternating direction method of multipliers |
| US20160275416A1 (en) * | 2015-03-20 | 2016-09-22 | Nec Laboratories America, Inc. | Fast Distributed Nonnegative Matrix Factorization and Completion for Big Data Analytics |
| CN105224942A (en) * | 2015-07-09 | 2016-01-06 | 华南农业大学 | A kind of RGB-D image classification method and system |
| CN106093814A (en) * | 2016-06-02 | 2016-11-09 | 浙江理工大学 | A kind of cardiac magnetic resonance imaging method based on multiple dimensioned low-rank model |
| CN107260159A (en) * | 2017-06-28 | 2017-10-20 | 浙江理工大学 | A kind of cardiac electric functional imaging method based on convolutional neural networks |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110575166A (en) * | 2019-09-30 | 2019-12-17 | 北京信息科技大学 | Method and device for time-frequency analysis of human electroencephalogram signals |
| CN111714124A (en) * | 2020-06-18 | 2020-09-29 | 中国科学院深圳先进技术研究院 | Magnetic resonance film imaging method, device, imaging equipment and storage medium |
| WO2021253671A1 (en) * | 2020-06-18 | 2021-12-23 | 中国科学院深圳先进技术研究院 | Magnetic resonance cine imaging method and apparatus, and imaging device and storage medium |
| CN111714124B (en) * | 2020-06-18 | 2023-11-03 | 中国科学院深圳先进技术研究院 | Magnetic resonance film imaging method, device, imaging equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Pereira Coutinho et al. | Novel fiducial and non‐fiducial approaches to electrocardiogram‐based biometric systems | |
| Xue et al. | Multiuser gesture recognition using sEMG signals via canonical correlation analysis and optimal transport | |
| CN111513709A (en) | Non-local neural network myocardial transmembrane potential reconstruction method based on iterative contraction threshold algorithm | |
| CN114863225B (en) | Image processing model training method, image processing model generation device, image processing model equipment and image processing model medium | |
| CN109157210A (en) | A kind of epicardial potential method for reconstructing based on ADMM and neural network | |
| CN109961508A (en) | A kind of heart static three-dimensional model method for reconstructing | |
| Stenroos et al. | Boundary element computations in the forward and inverse problems of electrocardiography: comparison of collocation and Galerkin weightings | |
| CN110811596B (en) | Noninvasive cardiac potential reconstruction method based on low rank and sparse constraint and non-local total variation | |
| CN113876419A (en) | Mapping efficiency by suggesting locations of mapping points | |
| Karoui et al. | Cardiac activation maps reconstruction: a comparative study between data-driven and physics-based methods | |
| Meister et al. | Graph convolutional regression of cardiac depolarization from sparse endocardial maps | |
| CN112401905B (en) | Natural action electroencephalogram recognition method based on source localization and brain network | |
| US20230263450A1 (en) | A cardiac transmembrane potential reconstruction method based on graph convolutional neural network and iterative threshold contraction algorithm | |
| Xie et al. | Non‐invasive reconstruction of dynamic myocardial transmembrane potential with graph‐based total variation constraints | |
| Nallikuzhy et al. | Spatially enhanced ECG using patient-specific dictionary learning | |
| CN115500841A (en) | Ventricular premature beat positioning method for fusion of time domain and frequency domain feature deep learning | |
| Wang et al. | A New ISTA-net Solution to Inverse Problem of Electrocardiology | |
| Cluitmans et al. | Physiology-based regularization improves noninvasive reconstruction and localization of cardiac electrical activity | |
| Wang et al. | An ADMM-net solution to inverse problem of electrocardiology | |
| Wang et al. | ECGI with a deep neural network and 2D normalized body surface potential maps | |
| Xie et al. | Lnista-Net: learning nonlocal regularized iterative shrinkage-thresholding network for noninvasive cardiac transmembrane potential imaging | |
| CN113781600B (en) | Electrical impedance imaging shape reconstruction method based on deep learning | |
| Miri et al. | Applicability of body surface potential map in computerized optimization of biventricular pacing | |
| CN113052927B (en) | Imaging detection method for improving spatial resolution | |
| Sanromán-Junquera et al. | Effect of interpolation on electroanatomical mapping |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190108 |
|
| RJ01 | Rejection of invention patent application after publication |