阅读说明：本技术 基于Tucker分解和ripple时间窗的脑磁图源定位方法和装置 (Magnetoencephalogram source positioning method and device based on Tucker decomposition and ripple time window ) 是由 张冀聪 史丽娟 于 2020-12-22 设计创作，主要内容包括：提供一种基于Tucker分解和ripple时间窗的脑磁图源定位方法和装置,该装置包括：脑磁图传感器,用于获取用户的第一脑磁图信号；ripple检测单元,用于通过均方根方法检测所述第一脑磁图信号中的ripple时间窗作为源定位的时间窗口,获得所述ripple时间窗内的第一脑磁图信号作为第二脑磁图信号；基于高阶正交迭代的Tucker分解单元,用于对所述第二脑磁图信号的原始张量进行Tucker分解来计算所述原始张量的估计值；以及源定位单元,用于对所述估计值计算协方差矩阵,且通过波束成形方法中的LCMV逆问题求解方法计算所述第二脑磁图信号对应的源位置。本发明消除了噪声信号的影响,降低了计算的复杂度,保证了每次计算结果的一致性,提高了致痫区域定位的准确度。(A magnetoencephalography source positioning method and device based on Tucker decomposition and ripple time window are provided, the device comprises: the magnetoencephalogram sensor is used for acquiring a first magnetoencephalogram signal of a user; the ripple detection unit is used for detecting a ripple time window in the first magnetoencephalogram signal as a source-positioned time window through a root mean square method, and obtaining the first magnetoencephalogram signal in the ripple time window as a second magnetoencephalogram signal; a Tucker decomposition unit based on high-order orthogonal iteration, configured to perform Tucker decomposition on an original tensor of the second magnetoencephalogram signal to calculate an estimated value of the original tensor; and the source positioning unit is used for calculating a covariance matrix for the estimated value and calculating the source position corresponding to the second magnetoencephalogram signal by an LCMV inverse problem solving method in a beam forming method. The method eliminates the influence of noise signals, reduces the complexity of calculation, ensures the consistency of each calculation result and improves the accuracy of positioning the seizure area.)

1. A magnetoencephalography source positioning device based on Tucker decomposition and ripple time window is characterized by comprising:

the magnetoencephalogram sensor is used for acquiring a first magnetoencephalogram signal of a user;

a ripple detection unit, configured to detect a ripple time window in the first magnetoencephalogram signal as a source-located time window by a root mean square method, and obtain the first magnetoencephalogram signal in the ripple time window as a second magnetoencephalogram signal, where at least four continuous oscillation signals having a frequency of 80-250Hz and an amplitude higher than a background signal are defined as ripple;

a Tucker decomposition unit based on high-order orthogonal iteration, configured to perform Tucker decomposition on an original tensor of the second magnetoencephalogram signal to calculate an estimated value of the original tensor, where the estimated value calculated by the Tucker decomposition is constrained by the high-order orthogonal iteration, so that the estimated value remains unique, and a noise factor tensor in the original tensor is removed from the estimated value; and

a source positioning unit, configured to calculate a covariance matrix for the estimated value, and calculate a source position corresponding to the second magnetoencephalogram signal by an LCMV inverse problem solution method in beamforming;

preferably, the calculation process of the higher-order orthogonal iteration is as follows:

1) calculating factor matrix U of original tensor X by using high-order singular value decomposition algorithm⁽ⁿ⁾And a core tensor G; let k equal to 0;

2) let k be k +1 and N be 1,2, … N, the following operations are performed:

B^(k)←X×₁U^(1)T…×_n-1U^(n-1)T×_NU^NT；

and performs the core tensor B^(k)Determining the number Rn of main singular values of the singular value decomposition of the n-mode expansion, wherein B^(k)For the core tensor obtained for the kth iteration, then the operations are performed:

U^(N)←U(:,1:R_n)；

3) calculating a core tensor B obtained by the k iteration^(k)Meter G^(k)＝B^(k)And throughThe following formula determines whether to converge:

||G^(k)-G^(k-1)||_F<e; wherein E is a minimum value;

if the convergence condition is met, executing the next step, otherwise, returning to the step 2) to continue iteration until convergence;

4) output factor matrix U^(N)And the core tensor G^(k)(ii) a According to the factor matrix U^(N)And the core tensor G^(k)Reconstructing the estimate of the original tensor.

2. The tile decomposition and ripple time window based magnetoencephalogram source localization apparatus of claim 1, further comprising a source display unit for displaying the source location.

3. The magnetocaloric source localization device according to claim 2, wherein the source location is the seizure area of the brain.

4. The tile decomposition and ripple time window based magnetoencephalogram source localization device according to any one of claims 1-3, further comprising a preprocessing unit for performing filtering, linear trend elimination and independent component analysis on the first magnetoencephalogram signal to obtain a preprocessed first magnetoencephalogram signal, and the preprocessed first magnetogram signal is input to the ripple detection unit for subsequent processing.

5. The apparatus according to any of claims 1-3, wherein the ripple time window is a plurality of ripple time windows, and for each second magnetoencephalogram signal corresponding to the ripple time window, the source localization unit calculates a plurality of source locations corresponding to each second magnetoencephalogram signal, and uses a region having a first concentration corresponding to the plurality of source locations as a localization result.

6. A magnetoencephalogram source positioning method based on Tucker decomposition and ripple time window is characterized by comprising the following steps:

acquiring a first magnetoencephalogram signal of a user through a magnetoencephalogram sensor;

detecting a ripple time window in the first magnetoencephalogram signal as a source-positioned time window through a root mean square method, and obtaining the first magnetoencephalogram signal in the ripple time window as a second magnetoencephalogram signal, wherein at least four continuous oscillation signals with the frequency of 80-250Hz and the amplitude higher than that of a background signal are defined as ripple;

performing Tucker decomposition on an original tensor of the second magnetoencephalogram signal to calculate an estimated value of the original tensor, wherein the estimated value calculated by the Tucker decomposition is constrained by high-order orthogonal iteration to keep the estimated value unique, and a noise factor tensor in the original tensor is removed from the estimated value;

and calculating a covariance matrix according to the estimated value, and calculating a source position corresponding to the second magnetoencephalogram signal by an LCMV inverse problem solving method in beam forming.

7. The method of claim 6, wherein the source location is displayed by a source display unit.

8. The method of claim 7, wherein said source location is a seizure region of the brain.

9. The method according to any one of claims 6-8, wherein the pre-processed first magnetoencephalogram signal is obtained by filtering, linear trending and independent component analysis of the first magnetogram signal by a pre-processing unit, and is used for subsequent ripple time window detection.

10. The method according to any one of claims 6-8, wherein the ripple time window is plural, and for the second magnetoencephalogram signal corresponding to each ripple time window, a corresponding plurality of source positions are calculated, and a corresponding region is concentrated according to the plurality of source positions as a positioning result.

Technical Field

The invention relates to a magnetoencephalogram source positioning method and device based on Tucker decomposition and ripple time window, belonging to the field of biomedicine.

Background

Clinically, up to one third of epileptic patients are refractory epilepsy. While drug therapy with antiepileptic drugs is difficult to control effectively. Epileptic foci can affect cortical excitability, resulting in abnormal discharges, not only directly affecting focal or surrounding tissues, but also affecting distant brain regions. Therefore, these patients need to surgically excise epileptogenic foci (epileptogenic zones) to achieve seizure-free, and whether or not an epileptogenic focus (epileptogenic focus) is postoperatively dependent on the localization of the epileptogenic focus. However, it is difficult to determine the area due to the lack of a tool to directly measure the area.

The Magnetoencephalogram (MEG) is a method for monitoring brain functions in a non-invasive and real-time manner, and in the past decades, several brain source localization methods based on electroencephalogram or electroencephalogram, such as spike-based dipole-fitting methods, have been proposed, but the localization accuracy of these methods still cannot meet the requirement of preoperative accurate localization, the calculation complexity is high, the localization result is susceptible to noise signals, and for the same magnetoencephalogram signal, the results of multiple localizations have deviation and poor consistency.

In view of the above, the present invention is directed to a method and an apparatus for positioning a magnetoencephalogram source based on a Tucker decomposition and a ripple time window to solve one or more of the above technical problems.

Disclosure of Invention

In order to solve one or more technical problems in the prior art, the applicant finds, through long-term research, that accuracy of location of an epileptogenic focus (epileptogenic region) can be improved by extracting a ripple time window of an original magnetoencephalogram signal and combining with Tucker decomposition, and reference and basis can be provided for subsequent operations. Based on the research result, according to an aspect of the present invention, there is provided a magnetoencephalography source positioning device based on a Tucker decomposition and a ripple time window, comprising:

the magnetoencephalogram sensor is used for acquiring a first magnetoencephalogram signal of a user;

a ripple detection unit, configured to detect a ripple time window in the first magnetoencephalogram signal as a source-located time window by a root mean square method, and obtain the first magnetoencephalogram signal in the ripple time window as a second magnetoencephalogram signal, where at least four continuous oscillation signals having a frequency of 80-250Hz and an amplitude higher than a background signal are defined as ripple;

a Tucker decomposition unit based on high-order orthogonal iteration, configured to perform Tucker decomposition on an original tensor of the second magnetoencephalogram signal to calculate an estimated value of the original tensor, where the estimated value calculated by the Tucker decomposition is constrained by the high-order orthogonal iteration, so that the estimated value remains unique, and a noise factor tensor in the original tensor is removed from the estimated value; and

and the source positioning unit is used for calculating a covariance matrix for the estimated value and calculating a source position corresponding to the second magnetoencephalogram signal by an LCMV inverse problem solving method in beam forming.

According to another aspect of the invention, the calculation process of the higher order orthogonal iteration is as follows:

1) calculating factor matrix U of original tensor X by using high-order singular value decomposition algorithm⁽ⁿ⁾And a core tensor G; let k equal to 0;

2) let k be k +1 and N be 1,2, … N, the following operations are performed:

B^(k)←X×₁U^(1)T…×_n-1U^(n-1)T×_NU^NT；

and performs the core tensor B^(k)Determining the number Rn of main singular values of the singular value decomposition of the n-mode expansion, wherein B^(k)For the core tensor obtained for the kth iteration, then the operations are performed:

U^(N)←U(:,1:R_n)；

3) calculating a core tensor B obtained by the k iteration^(k)Meter G^(k)＝B^(k)And judging whether convergence occurs by the following formula:

||G^(k)-G^(k-1)||_F<e; wherein E is a minimum value;

if the convergence condition is met, executing the next step, otherwise, returning to the step 2) to continue iteration until convergence;

4) output factor matrix U^(N)And the core tensor G^(k)(ii) a According to the factor matrix U^(N)And the core tensor G^(k)Reconstructing the estimate of the original tensor.

According to another aspect of the present invention, the magnetoencephalogram source positioning device based on the Tucker decomposition and ripple time window further comprises a source display unit for displaying the source position.

According to yet another aspect of the invention, the source location is a seizure region of the brain.

According to another aspect of the present invention, the apparatus for source localization of magnetoencephalogram based on Tucker decomposition and ripple time window further includes a preprocessing unit, configured to perform filtering, linear trend elimination and independent component analysis on the first magnetoencephalogram signal to obtain a preprocessed first magnetoencephalogram signal, and the preprocessed first magnetogram signal is input to the ripple detecting unit for subsequent processing.

According to another aspect of the present invention, the ripple time window is a plurality of windows, and for the second magnetoencephalogram signal corresponding to each ripple time window, the source localization unit calculates a plurality of corresponding source locations, and concentrates a corresponding region according to the plurality of source locations as a localization result.

According to another aspect of the present invention, there is also provided a method for positioning a magnetoencephalogram source based on a Tucker decomposition and a ripple time window, comprising the steps of:

acquiring a first magnetoencephalogram signal of a user through a magnetoencephalogram sensor;

detecting a ripple time window in the first magnetoencephalogram signal as a source-positioned time window through a root mean square method, and obtaining the first magnetoencephalogram signal in the ripple time window as a second magnetoencephalogram signal, wherein at least four continuous oscillation signals with the frequency of 80-250Hz and the amplitude higher than that of a background signal are defined as ripple;

performing Tucker decomposition on an original tensor of the second magnetoencephalogram signal to calculate an estimated value of the original tensor, wherein the estimated value calculated by the Tucker decomposition is constrained by high-order orthogonal iteration to keep the estimated value unique, and a noise factor tensor in the original tensor is removed from the estimated value;

and calculating a covariance matrix for the estimated value, and calculating a source position corresponding to the second magnetoencephalogram signal by an LCMV inverse problem solving method.

According to still another aspect of the present invention, the source location is displayed by a source display unit.

According to yet another aspect of the invention, the source location is taken as a seizure region of the brain.

According to another aspect of the invention, the preprocessing unit filters, trends linearly and analyzes independent components of the first magnetoencephalogram signal to obtain a preprocessed first magnetoencephalogram signal, and the preprocessed first magnetogram signal is used for subsequent ripple time window detection.

According to another aspect of the present invention, the ripple time window is a plurality of windows, for each second magnetoencephalogram signal corresponding to the ripple time window, a plurality of corresponding source positions are calculated, and a region having the first concentration corresponding to the plurality of source positions is used as a positioning result.

Compared with the prior art, the invention has one or more of the following technical effects:

(1) in the magnetoencephalogram source positioning, high-frequency noise and low-frequency noise in magnetoencephalogram signals are effectively removed through a Tucker decomposition method, the influence of noise signals on the positioning result of epileptogenic foci (epileptogenic regions) is reduced, and meanwhile, the complexity of calculation is reduced.

(2) The result of the Tucker decomposition is subjected to uniqueness constraint by using a high-order orthogonal iteration (HOOI) method, and the consistency of the calculation result every time is ensured.

(3) The accuracy of the location of the epileptic focus (epileptogenic region) is further improved by taking ripple as a time window for source location selection and combining with Tucker decomposition, and reference and basis are provided for subsequent operations.

(4) By detecting a plurality of ripple time windows and obtaining a plurality of corresponding source positions, the final positioning result is determined according to the corresponding areas in the source position sets, and the positioning precision is further improved; combining multiple source localization results, taking the brain region with the first concentration (45% -85%) of more than 10 brain regions (source localization regions) as the localization result can further reduce the workload of doctors and eliminate potential interference.

Drawings

So that the manner in which the above recited features of the present invention can be understood in detail, a more particular description of the invention, briefly summarized above, may be had by reference to embodiments. The drawings relate to preferred embodiments of the invention and are described below:

FIG. 1 is a schematic diagram of a magnetoencephalography source positioning device based on a Tucker decomposition and ripple time window according to a preferred embodiment of the present invention;

FIG. 2 is a diagram illustrating the detection result of a ripple time window according to a preferred embodiment of the present invention;

FIG. 3 is a schematic diagram showing the comparison between the positioning accuracy of the tracker decomposition based on the ripple time window and the positioning accuracy of the dipole-fitting based on the spike.

DETAILED DESCRIPTION OF EMBODIMENT (S) OF INVENTION

Reference will now be made in detail to the various embodiments, one or more examples of which are illustrated in each figure. The examples are provided by way of explanation and are not meant as limitations. For example, features illustrated or described as part of one embodiment can be used on or in conjunction with any other embodiment to yield yet a further embodiment. It is intended that the present invention encompass such modifications and variations.

In the following description of the drawings, like reference numerals designate identical or similar structures. Generally, only the differences between the individual embodiments will be described. Descriptions of parts or aspects in one embodiment can also be applied to corresponding parts or aspects in another embodiment, unless explicitly stated otherwise.

Example 1

Referring to fig. 1-3, fig. 1 is a schematic diagram of a positioning apparatus for a magnetoencephalogram source based on a Tucker decomposition and ripple time window according to a preferred embodiment of the present invention; FIG. 2 is a diagram illustrating the detection result of a ripple time window according to a preferred embodiment of the present invention; FIG. 3 is a schematic diagram showing the comparison between the positioning accuracy of the tracker decomposition based on the ripple time window and the positioning accuracy of the dipole-fitting based on the spike. In accordance with a preferred embodiment of the present invention, referring to fig. 1-2, there is provided a magnetoencephalography source localization apparatus based on Tucker decomposition and ripple time window, comprising:

the magnetoencephalogram sensor is used for acquiring a first magnetoencephalogram signal of a user;

a ripple detection unit, configured to detect a ripple time window in the first magnetoencephalogram signal as a source-located time window by a root mean square method, and obtain the first magnetoencephalogram signal in the ripple time window as a second magnetoencephalogram signal, where at least four continuous oscillation signals having a frequency of 80-250Hz and an amplitude higher than a background signal are defined as ripple;

a Tucker decomposition unit based on high-order orthogonal iteration, configured to perform Tucker decomposition on an original tensor of the second magnetoencephalogram signal to calculate an estimated value of the original tensor, where the estimated value calculated by the Tucker decomposition is constrained by the high-order orthogonal iteration, so that the estimated value remains unique, and a noise factor tensor in the original tensor is removed from the estimated value; and

and the source positioning unit is used for calculating a covariance matrix for the estimated value and calculating a source position corresponding to the second magnetoencephalogram signal by an LCMV (linear constrained minimum variance) inverse problem solving method in a beam forming method.

Advantageously, the method improves the detection precision of the source position by the Tucker decomposition based on the high-order orthogonal iteration (HOOI) and taking the ripple time period of the MEG signal as the positioning time window, can be widely applied to deep source and shallow source positioning, and improves the reference value of clinical brain source positioning in preoperative evaluation.

According to another preferred embodiment of the present invention, the calculation process of the higher order orthogonal iteration is as follows:

1) calculating factor matrix U of original tensor X by using high-order singular value decomposition algorithm⁽ⁿ⁾And a core tensor G; let k equal to 0;

2) let k be k +1 and N be 1,2, … N, the following operations are performed:

B^(k)←X×₁U^(1)T…×_n-1U^(n-1)T×_NU^NT；

and performs the core tensor B^(k)Determining the number Rn of main singular values of the singular value decomposition of the n-mode expansion, wherein B^(k)For the core tensor obtained for the kth iteration, then the operations are performed:

U^(N)←U(:,1:R_n)；

3) calculating a core tensor B obtained by the k iteration^(k)Meter G^(k)＝B^(k)And judging whether convergence occurs by the following formula:

||G^(k)-G^(k-1)||_F<e; wherein E is a minimum value;

if the convergence condition is met, executing the next step, otherwise, returning to the step 2) to continue iteration until convergence;

4) output factor matrix U^(N)And the core tensor G^(k)(ii) a According to the factor matrix U^(N)And the core tensor G^(k)Reconstructing the estimate of the original tensor.

According to another preferred embodiment of the present invention, the apparatus for source localization of magnetoencephalogram based on Tucker decomposition and ripple time window further comprises a source display unit for displaying the source location.

According to yet another preferred embodiment of the invention, the source location is an epileptogenic zone (epileptogenic zone) of the brain.

According to another preferred embodiment of the present invention, the apparatus for source localization of magnetoencephalogram based on Tucker decomposition and ripple time window further comprises a preprocessing unit, configured to perform filtering, linear trend elimination and Independent Component Analysis (ICA) on the first magnetoencephalogram signal to obtain a preprocessed first magnetoencephalogram signal, and the preprocessed first magnetogram signal is input to the ripple detecting unit for subsequent processing.

According to another preferred embodiment of the present invention, the ripple time window is a plurality of windows, and the source localization unit calculates a plurality of corresponding source locations for the second magnetoencephalogram signal corresponding to each ripple time window, and concentrates a corresponding region as a localization result according to the plurality of source locations. Specifically, for example, a region corresponding to at least Y iterations of the Z source positions is output as a final result, and Y is preferably equal to or greater than 3. According to experimental verification, the repeatedly corresponding (indicating) areas for more than 3 times are basically close to the real epileptogenic areas, so that the positioning accuracy is further improved, and reliable reference is provided for later operations of doctors.

Preferably, the ripple time windows are Z, and for the second magnetoencephalogram signal corresponding to each ripple time window, the source localization unit calculates Z source locations corresponding to each ripple time window, determines Z brain regions corresponding to the Z source locations respectively, and takes the brain region with the first concentration in the Z brain regions as a localization result. Research shows that the Z is more than or equal to 10, the first concentration ratio is 45% -85%, accurate positioning of a real seizure area can be achieved, and compared with single source positioning, the method can further reduce workload of doctors and eliminate potential interference.

According to another preferred embodiment of the present invention, there is also provided a method for positioning a magnetoencephalogram source based on a Tucker decomposition and a ripple time window, comprising the steps of:

acquiring a first magnetoencephalogram signal of a user through a magnetoencephalogram sensor;

detecting a ripple time window in the first magnetoencephalogram signal as a source-positioned time window through a root mean square method, and obtaining the first magnetoencephalogram signal in the ripple time window as a second magnetoencephalogram signal, wherein at least four continuous oscillation signals with the frequency of 80-250Hz and the amplitude higher than that of a background signal are defined as ripple;

performing Tucker decomposition on an original tensor of the second magnetoencephalogram signal to calculate an estimated value of the original tensor, wherein the estimated value calculated by the Tucker decomposition is constrained by high-order orthogonal iteration to keep the estimated value unique, and a noise factor tensor in the original tensor is removed from the estimated value;

and calculating a covariance matrix according to the estimated value, and calculating a source position corresponding to the second magnetoencephalogram signal by an LCMV inverse problem solving method in beam forming.

According to still another preferred embodiment of the present invention, the source position is displayed by a source display unit.

According to a further preferred embodiment of the invention, the source location is a seizure region of the brain.

According to another preferred embodiment of the present invention, the pre-processing unit filters, trends linearly, and analyzes the independent components of the first magnetoencephalogram signal to obtain a pre-processed first magnetoencephalogram signal, which is used for subsequent ripple time window detection.

According to another preferred embodiment of the present invention, the ripple time window is a plurality of windows, and for the second magnetoencephalogram signal corresponding to each ripple time window, a plurality of corresponding source positions are calculated, and a corresponding region in the plurality of source positions is determined as a positioning result.

According to another preferred embodiment of the present invention, there is also provided a novel source localization apparatus based on Tucker decomposition and ripple time window, including a ripple detection unit, a tensor decomposition unit, a source localization unit, and a verification unit. Preferably, the Ripple detection unit applies a root mean square method. The MEG (magnetoencephalography) signal adopts Tucker decomposition to calculate the estimation value of an original tensor, the Tucker decomposition result is kept unique based on the mode constraint of high-order orthogonal iteration, and the factor tensor of which the main information is noise is removed. And the source positioning unit calculates a covariance matrix for the estimated value of the tensor decomposed by the Tucker, so that the source position of the epileptic brain magnetic signal is calculated by an LCMV method. The validation unit includes validation on top of the real magnetoencephalography data, wherein the real data-based validation compares the ripple-based Tucker decomposition with spike-based clinical dipole methods. As shown in fig. 3, which shows the comparison result, wherein the present invention greatly improves the positioning accuracy of the seizure-causing region of the brain with respect to the spike-based clinical dipole method.

Preferably, the ripple detection unit defines at least four continuous oscillations of 80-250Hz and amplitude above the background signal as ripple. Preferably, at least four oscillations of 80-250Hz and amplitudes above the first threshold of the background signal are defined as ripple, further improving the positioning accuracy. Preferably, in the present invention, a Root Mean Square (RMS) method is used for detecting ripples in MEG, and the conditions thereof can be specifically selected as follows: the waveform power value should be 3-9 times the background signal and the duration of the waveform should last at least 15ms (corresponding to at least four oscillations). Preferably, false positive ripples can be eliminated manually.

According toIn another preferred embodiment of the present invention, in the orthogonal iteration process, the original tensor X is input; outputting the factor matrix U after iteration is finished⁽ⁿ⁾And a core tensor G.

The calculation process of the orthogonal iteration is preferably as follows:

1) calculating a factor matrix U of an original tensor X by using a high-order singular value decomposition (HOSVD) algorithm⁽ⁿ⁾And a core tensor G. Let k equal to 0;

2) let k be k +1 and N be 1,2, … N, the following operations are performed:

B^(k)←X×₁U^(1)T…×_n-1U^(n-1)T×_NU^NT(ii) a Wherein, X is multiplied by nU^(n)TRepresenting tensor X and matrix U^{(n) of T}n-modulo product;

and performs the core tensor B^(k)Determining the number Rn of main singular values of the SVD, and then executing the operation:

U^(N)←U(:,1:R_n)

3) computing the core tensor B for the kth iteration^(k)Meter G^(k)＝B^(k)And whether to converge is judged by the following convergence formula:

||G^(k)-G^(k-1)||_F<E

e is a minimum value.

If the convergence condition is met, executing the next step, otherwise returning to the step 2 to continue iteration until convergence.

4) Output factor matrix U^(N)And the core tensor G^(k)(ii) a According to the factor matrix U^(N)And the core tensor G^(k)Reconstructing the estimate of the original tensor.

It is understood that, among other things, the core tensor G is initialized⁽⁰⁾Is the zero tensor (all elements are 0). I G^(k)-G^(k-1)||_F<E denotes determination of G^(k)-G^(k-1)Whether a convergence formula of a convergence condition is met.

Preferably, for the estimated value, an inverse problem is solved by using a beamforming (beamforming) analysis method to calculate the covariance matrix and the position of the estimated source, and the principle of beamforming is as follows: the signal is filtered through time domain information and frequency domain information to obtain information in a given direction while attenuating noise interference from other directions.

Compared with the prior art, the invention has one or more of the following technical effects:

(1) in the magnetoencephalogram source positioning, high-frequency noise and low-frequency noise in magnetoencephalogram signals are effectively removed through a Tucker decomposition method, the influence of noise signals on the positioning result of epileptogenic foci (epileptogenic regions) is reduced, and meanwhile, the complexity of calculation is reduced.

(2) The result of the Tucker decomposition is subjected to uniqueness constraint by using a high-order orthogonal iteration (HOOI) method, and the consistency of the calculation result every time is ensured.

(3) The accuracy of the location of the epileptic focus (epileptogenic region) is further improved by taking ripple as a time window for source location selection and combining with Tucker decomposition, and reference and basis are provided for subsequent operations.

(4) By detecting a plurality of ripple time windows and obtaining a plurality of corresponding source positions, the final positioning result is determined according to the corresponding areas in the source position sets, and the positioning precision is further improved; combining multiple source localization results, taking the brain region with the first concentration (45% -85%) of more than 10 brain regions (source localization regions) as the localization result can further reduce the workload of doctors and eliminate potential interference.

While the foregoing is directed to embodiments of the present invention, other and further embodiments of the invention may be devised without departing from the basic scope thereof, and the scope thereof is determined by the claims that follow.

The above embodiments are merely preferred embodiments of the present invention, which are not intended to limit the present invention, and the features of the embodiments that do not violate each other may be combined with each other. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.