CN106529126A - Processing method for inheriting monitoring image information after continuous monitoring interruption in brain dynamic electrical impedance imaging - Google Patents

Abstract

The invention discloses a processing method for inheriting monitoring image information after the continuous monitoring of craniocerebral dynamic electrical impedance imaging is interrupted, which is used for restoring the lost monitoring image information after the clinical continuous monitoring of craniocerebral dynamic electrical impedance imaging is interrupted. This method analyzes the measurement data before and after the interruption of continuous monitoring. First, the method of spline differential fitting is used to process the measurement data to suppress the image artifacts caused by the baseline change of the measurement data caused by the inconsistency of the electrode contact state before and after the interruption of continuous monitoring. The reconstruction matrix is augmented, and the prior information of the electrode displacement is introduced into the reconstruction matrix to reduce the influence of the electrodes on the monitoring image due to the inconsistent placement of the electrodes before and after the interruption of monitoring. After testing, this method can effectively suppress the image artifacts caused by the change of electrode contact state and placement position before and after monitoring interruption, restore the normal monitoring target, and improve the clinical practicability of electrical impedance imaging monitoring.

Description

Inheritance of monitoring image information after interruption of continuous monitoring of brain dynamic electrical impedance imaging processing method

technical field

The invention belongs to the technical field of dynamic electrical impedance imaging, and in particular relates to a processing method for inheriting monitoring image information after the continuous monitoring of cranial dynamic electrical impedance imaging is interrupted.

Background technique

Brain dynamic electrical impedance tomography technology uses electrodes evenly distributed around the head to continuously apply safe current excitation to the brain in real time and measure the boundary voltage, using data at two different times and taking the data at the earlier time as a reference frame, and the data at another moment is the foreground frame. After the difference between the two frames of data, combined with a certain imaging method, the impedance changes inside the brain at these two different moments are reconstructed. Therefore, continuous and stable data acquisition is the primary condition for the system to perform normal dynamic impedance imaging.

During actual clinical use, it often happens that monitoring has to be temporarily interrupted. For example, if a patient needs to undergo imaging examinations such as CT, the electrodes attached to the head need to be removed, and the electrodes need to be reattached if the monitoring is continued after the examination. After the electrodes are reattached, compared with the initial state of monitoring, the position and contact state of the electrodes are likely to change, and this change will directly affect the collected data. When restarting monitoring, if the reference frame at the beginning of monitoring is still used, the change of electrode-scalp contact status and electrode position may produce artifacts on the reconstructed image, obliterating the normal impedance change information. If the data at the restart of monitoring is selected as the reference frame, although the influence caused by the electrode-scalp contact state and electrode position change can be eliminated, the image information before the restart of monitoring cannot be directly reflected on the new monitoring image, resulting in the previous impedance Change information is lost. The interruption of continuous monitoring seriously affects the observation and diagnosis of the disease, which is not conducive to the clinical promotion and application of dynamic brain electrical impedance imaging.

Therefore, in order to eliminate the influence of electrode position and contact state changes on the image, and to preserve the previous image monitoring information, there is an urgent need for a method that can process the data after the continuous monitoring is interrupted.

Contents of the invention

The purpose of the present invention is to provide a processing method for inheriting monitoring image information after the continuous monitoring of craniocerebral dynamic electrical impedance imaging is interrupted. Improving the clinical applicability of dynamic electrical impedance imaging.

The present invention is achieved through the following technical solutions:

A processing method for inheriting monitoring image information after the continuous monitoring of craniocerebral dynamic electrical impedance imaging is interrupted. The image artifacts caused by the baseline change of the measurement data caused by the inconsistency of the electrode contact state before and after the interruption of monitoring; then, the image reconstruction algorithm is augmented, and the prior information of the electrode displacement is introduced into the reconstruction matrix to reduce the risk of electrode placement before and after the interruption of monitoring. The impact of inconsistencies in position on monitoring images.

The above processing method specifically includes the following steps:

1) Obtain the baseline correction value of the measurement data

Select the last n frames of data before continuous monitoring is interrupted and the first n frames of data before restarting monitoring, connect the data together, and use splines for the i-th valid data channel data x _i (l) for data x(l) function fit, get the fitted sequence Among them, i∈N, N is the total number of measurement channels;

Then, detect Maximum value on the rising side at the baseline jump and the minima on the descending side Calculate baseline correction

2) Measure baseline correction for the restarted monitoring data

For the i-th measurement channel data sequence y _i (l) after restarting monitoring, use Baseline correction is performed on y _i (l). Compared with the data before the interruption of monitoring, if y _i (l) is on the side where the baseline rises, the corrected If y _i (l) is on the side of the baseline decline, the corrected

3) Augment the Gauss-Newton image reconstruction formula

For Δρ=-[J ^t J+λR] ^-1 Jz, Δρ is the electrical impedance change distribution vector at different times, and z is the boundary voltage difference vector at different times;

the Jacobian matrix expands to the constraint matrix expands to And use the prior information to determine the coefficient λ and the unit term of the constraint matrix R _aug ;

4) Reconstruct the impedance variation of the target field using the rectified data and the augmented reconstruction formula

which is in is the measured voltage data at time t1, It is the measured voltage data at time t2.

In step 1), the i-th effective data channel data x _i (l) is fitted using a spline function, specifically as follows:

Among them, S(x) is the spline fitting function of the data sequence x _i (l), find S(x) to minimize the value of L; among them, p∈[0,1] reflects the fitting function and the measured The proximity of the data; S(x) is a piecewise spline fitting function, written in a general form:

S _j (l)=a _j (ll _j ) ³ +b _j (ll _j ) ² +c _j (ll _j )+d _j ;

Add boundary conditions and use the least squares method to solve the coefficients of each segment fitting function of S(l). After obtaining the specific expression of S(l), find the fitting reconstruction sequence detection Maximum value on the rising side at the baseline jump and the minima on the descending side Calculate baseline correction

Step 3) The specific operation is:

The Gauss-Newton image reconstruction formula is as follows:

Δρ=-[J ^t J+λR] ^-1 Jz;

Among them, J is the Jacobian matrix, λR is the regularization constraint item, Δρ is the distribution vector of electrical impedance change at different times, and z is the boundary voltage difference vector at different times;

In order to suppress the influence of electrode position changes on imaging results, the matrix in the reconstruction formula is augmented: for the Jacobian matrix expands to n _meas is the number of measured voltages contained in one frame of data, n _elem is the number of finite element model units used for imaging reconstruction, n _dim is the imaging dimension, usually two-dimensional imaging, so take n _dim = 2, n _e is the number of electrodes, and the expansion part of the Jacobian matrix is filled with the disturbance due to the change of the measurement electrode position, namely:

Among them, A is the current excitation vector, H is the forward calculation matrix, It is a forward calculation matrix recalculated after the electrode position changes in the x or y direction, and the displacement is usually uniformly set as a priori constant;

Then, augment the constraint matrix R:

Will expands to The extension part is populated with prior estimates of the noise of the reconstructed data and a prior estimate of the change in the reconstructed conductivity, namely:

Among them, R _extra is a Laplacian filter, and the regularization parameter λ and the Laplacian template of R _extra are determined according to the prior information of the imaging field, then:

Among them, ave _noise is the average prior amplitude of the noise relative to the measured data, ave _conduct is the average prior amplitude of the conductivity change relative to the initial conductivity distribution, and ave _move is the average prior amplitude of the electrode displacement relative to the field radius amplitude, is the Laplacian operator.

Compared with the prior art, the present invention has the following beneficial technical effects:

The invention discloses a processing method for inheriting monitoring image information after the continuous monitoring of craniocerebral dynamic electrical impedance imaging is interrupted, which is used for recovering the lost monitoring image information after the interruption of clinical continuous monitoring of craniocerebral dynamic electrical impedance imaging. This method analyzes the measurement data before and after the interruption of continuous monitoring. First, the method of spline differential fitting is used to process the measurement data to suppress the image artifacts caused by the baseline change of the measurement data caused by the inconsistency of the electrode contact state before and after the interruption of continuous monitoring. The reconstruction matrix is augmented, and the prior information of the electrode displacement is introduced into the reconstruction matrix to reduce the influence of the electrodes on the monitoring image due to the inconsistent placement of the electrodes before and after the interruption of monitoring. After testing, this method can effectively suppress the image artifacts caused by the change of electrode contact state and placement position before and after monitoring interruption, restore the normal monitoring target, and improve the clinical practicability of electrical impedance imaging monitoring.

Description of drawings

Fig. 1 is a schematic flow chart of the method of the present invention;

Figure 2 is the reconstructed image at the moment before the interruption of continuous monitoring.

Fig. 3 is not processed by the method of the present invention, the one-dimensional data curve (a) after connecting the data of a period of time before the interruption of continuous monitoring and a section of data after restarting the monitoring, and reselecting the reference frame (c) and not reselecting The two-dimensional reconstructed image in the case of the reference frame (b);

Figure 4 is the one-dimensional data curve (a) after the correction step using this method, the two-dimensional reconstructed image (c) without using the improved algorithm (b) and using the improved imaging algorithm of this method.

detailed description

The present invention will be further described in detail below in conjunction with specific embodiments, which are explanations of the present invention rather than limitations.

A processing method for inheriting monitoring image information after continuous monitoring of craniocerebral dynamic electrical impedance imaging is interrupted according to the present invention, comprising the following steps:

(1) Obtain the baseline correction value of the measurement data. Select the last n frames of data before continuous monitoring is interrupted and the first n frames of data before restarting monitoring, and connect the data together as data x(l), for the ith (i∈N, N is the total number of measurement channels) valid data The channel data x _i (l) is fitted using a spline function:

Among them, S(x) is the spline fitting function of the data sequence x _i (l), and S(x) minimizes the value of L. where p∈[0,1] reflects the closeness of the fitting function to the measured data. S(x) is a piecewise spline fitting function, which can be written in general form:

S _j (l)＝a _j (ll _j ) ³ +b _j (ll _j ) ² +c _j (ll _j )+d _j

Add boundary conditions and use the least squares method to solve the coefficients of each segment fitting function of S(l). After obtaining the specific expression of S(l), find the fitting reconstruction sequence detection Maximum value on the rising side at the baseline jump and the minima on the descending side Calculate baseline correction

(2) Perform measurement baseline correction on the data of restarting monitoring. For the i-th measurement channel data sequence y _i (l) after restarting monitoring, use Baseline correction is performed on y _i (l). Compared with the data before the interruption of monitoring, if y _i (l) is on the side where the baseline rises, the corrected If y _i (l) is on the side of the baseline decline, the corrected

Repeat steps (1) (2) to perform the above processing on all measurement channels.

(3) Improving the imaging algorithm and augmenting the Gauss-Newton image reconstruction formula. For the Gauss-Newton imaging algorithm formula:

Δρ＝-[J ^t J+λR] ^-1 Jz

where is the J reconstruction matrix, λR is the regularization constraint item, z is the boundary voltage difference vector at different times, and Δρ is the distribution vector of electrical impedance changes at different times.

In order to suppress the influence of electrode position changes on imaging results, the matrix in the reconstruction formula should be augmented. For the Jacobian matrix expands to n _meas is the number of measured voltages contained in one frame of data, n _elem is the number of finite element model units used for imaging reconstruction, n _dim is the imaging dimension, usually two-dimensional imaging, so take n _dim = 2, n _e is the number of electrodes, and the expansion part of the Jacobian matrix is filled with the disturbance due to the change of the measurement electrode position, that is,

in

A is the current excitation vector, H is the forward calculation matrix, It is a forward calculation matrix that is recalculated after the electrode position changes in the x or y direction, and the displacement is usually uniformly set as a priori constant.

Then the constraint matrix R is augmented, and the expands to The extension part is filled with the noise prior estimate of the reconstructed data and the prior estimate of the reconstructed conductivity change, namely

Where R _extra is a Laplacian filter. Determining the regularization parameter λ and the Laplacian template of R _extra according to the prior information of the imaging field, we have

where ave _noise is the average prior amplitude of the noise relative to the measured data, ave _conduct is the average prior amplitude of the conductivity change relative to the initial conductivity distribution, and ave _move is the average prior amplitude of the electrode displacement relative to the field radius value.

(4) Reconstruct the impedance variation of the target field using the corrected data and the augmented reconstruction formula. which is in For the reference frame voltage data after restarting monitoring, It can be abbreviated as Δρ=S _aug z for the foreground frame voltage data after restarting monitoring.

The specific implementation is as follows:

Put the electrical impedance measurement electrodes coated with conductive paste on the subject's head, and wrap the head with a bandage to fix the electrodes. After all the electrodes are in normal contact, start data collection and image monitoring. The image of normal continuous monitoring is shown in Figure 2, and the image contains obvious impedance change targets. In order to simulate the interruption of monitoring, all the measuring electrodes were removed from the subject's head, and the area where the electrodes were attached to the subject's head was wiped clean with gauze, and then all the electrical impedance measuring electrodes were reattached. Restricted by the actual situation of the operation, compared with the initial state, the re-attached electrodes will change in the distribution position and the contact impedance with the scalp, which will affect the boundary voltage and make the measurement data baseline inconsistent before and after restarting monitoring, as shown in Figure 3(a ) shown. If no processing is performed and the reference frame before the monitoring interruption is still used, strong artifacts will appear on the monitoring image, obliterating the original target, as shown in Figure 3(b); If the data is a background frame, there is no previous target information on the image, as shown in Figure 3(c). Therefore, a certain processing method is needed to suppress image artifacts and restore the original target information on the image without reselecting the imaging reference frame.

According to the process shown in Figure 1, when monitoring is restarted, the data and imaging algorithms are processed in the following steps:

Step 1: Obtain the baseline change of the measurement data caused by the inconsistency of the electrode scalp contact state. Read the last n frames of data of the monitoring process before the monitoring was interrupted, and the first n frames of data of the restarted monitoring, and combine them into a measurement data sequence x(l). The composition of the data sequence is expressed as a matrix:

x _meas (l) represents the measured data of the meas-th measurement channel of the l-th frame data, and l=2n. The number of measurement channels in the EIT measurement system is related to the number of electrodes and the excitation-measurement mode used. Select the data sequence x _i (l) of a single measurement channel for each processing, use the spline function to fit x _i (l) according to the following formula, and find the fitting function S(l):

The fitting function S(l) is a piecewise cubic polynomial, which can be written in the following general form:

S _j (l)＝a _j (ll _j ) ³ +b _j (ll _j ) ² +c _j (ll _j )+d _j

Adding the natural boundary condition b ₀ =b _l =0, and substituting each segmental boundary condition and derivative continuity condition, the matrix equation is obtained:

h _i =x _i+1 −x _i , q _i =2(h _i−1 +h _i ),

And get the expression of coefficient a _j , c _j with respect to b _j , d _j :

Let L'=0, obtain the corresponding cubic polynomial coefficients. After obtaining the expression of the spline fitting function S(l), calculate the fitting value sequence of the corresponding position detection Maximum value on the rising side at the baseline jump and the minima on the descending side Calculate baseline correction

Step 2: Correct the baseline of the measured data for changes. For the ith measurement channel data sequence y _i (l) of monitoring process 2, use Baseline correction is performed on y _i (l). As shown in Figure 3(a), compared with the data before monitoring interruption, y _i (l) is on the side of the baseline increase, and the corrected The rectification result is shown in Fig. 4(a). As shown in Figure 4(b), if imaging is performed directly, monitoring image artifacts caused by electrode position changes will appear on the monitoring images, which will affect the identification and judgment of monitoring targets, and image artifacts need to be suppressed.

Step 3: Change to an imaging algorithm to suppress image artifacts caused by electrode position changes. The Gauss-Newton image reconstruction formula used in dynamic imaging is augmented. There is a Gauss-Newton imaging algorithm formula:

Δρ＝-[J ^t J+λR] ^-1 Jz

where is the J reconstruction matrix, λR is the regularization constraint item, z is the boundary voltage difference vector at different times, and Δρ is the distribution vector of electrical impedance changes at different times.

For the Jacobian matrix expands to n _meas is the number of measured voltages contained in one frame of data, n _elem is the number of finite element model units used for reconstruction for imaging, n _e is the number of measurement electrodes, and the expansion part of the Jacobian matrix is filled with the precedence of the position change of the measurement electrodes. Check the disturbance compensation value, namely:

In order to obtain the filling elements, the forward calculation formula of electrical impedance imaging is used

z=HA

Among them, A is the current excitation vector, H is the forward calculation matrix, which is related to the coordinates of the finite element model used for imaging, and z is the boundary voltage vector. to calculate First set a priori quantity θ, modify the x-axis coordinate u _x of electrode 1 to u _x = u _x + θ, and recalculate the forward reconstruction matrix to obtain a new boundary voltage vector:

z _move = H _move A

then there is

In the same way, calculate the J matrix augmentation elements when other electrodes and y-axis coordinates are displaced, and use θ for the prior quantities.

Then the constraint matrix R is augmented, and the expands to The extension part is filled with the noise prior estimate of the reconstructed data and the prior estimate of the reconstructed conductivity change, namely

where R _extra is in the form of a discrete Laplacian filter, which can be written as:

The specific elements are: R _i,j|(extra) =2.1·δ ² , R _i,j|(extra) =-1·δ ² (unit i and unit j are adjacent), and other elements are 0. in ave _conduct is the proportional coefficient of the target conductivity change relative to the initial conductivity distribution, and ave _move is the proportional coefficient of the average electrode displacement relative to the radius of the field.

The constraint coefficient λ is:

Among them, ave _noise is the proportional coefficient of the noise signal relative to the measured data.

The corrected imaging result is shown in Fig. 4(c). Compared with the imaging results without algorithm processing and reselecting the imaging reference frame, the processed imaging results can effectively reflect the imaging target before the interruption of monitoring. Compared with the imaging results without algorithm processing and without reselecting the imaging reference frame, the processed imaging results can effectively suppress the image artifacts caused by the baseline change of the data measurement. After the algorithm compensation for the electrode position change and the baseline correction of the measurement data, compared with Figure 2, the imaging result processed by the algorithm in this paper effectively restores the original image information. Refer to the color scale bar on the right side of the image to ensure the restoration The image result is consistent with the original image in the value distribution. Therefore, the monitoring image information inheritance processing method can effectively deal with the loss of the original monitoring information after the continuous monitoring of the craniocerebral dynamic electrical impedance imaging is interrupted and the monitoring is restarted.

Claims (4)

1. the processing method that monitoring image information of having no progeny in a kind of cranium brain dynamic electric impedance imaging on-line monitor is inherited, its feature exist In the processing method is analyzed to the measurement data before and after on-line monitor interruption：First, the method being fitted using batten difference Measurement data is processed, measurement data baseline change caused by electrode contact state is inconsistent before and after suppressing on-line monitor to interrupt causes Image artifacts；Then, augmentation process is carried out to image reconstruction algorithm, the prior information of electrode displacement is introduced into reconstruction matrix, Before and after reducing monitoring interruption, electrode is because of the inconsistent impact to guarding image of riding position.

2. the place that monitoring image information of having no progeny in cranium brain dynamic electric impedance imaging on-line monitor according to claim 1 is inherited Reason method, it is characterised in that comprise the following steps：

1) obtain the Baseline wander value of measurement data

Selected on-line monitor interrupt before last n frame data and restart the front n frame data guarded, by data cube computation one Rise, be data x (l), to i-th valid data channel data x_iL () uses Spline-Fitting, obtain fitting sequence Wherein, i ∈ N, N are Measurement channel sum；

It is then detected thatThe maximum of uplifted side at the baseline transitionWith the minimum for declining sideCalculate base Line correction value

2) Baseline wander is measured to restarting the data guarded

To restarting the ith measurement channel data sequence y after guarding_iL (), usesTo y_iL () carries out Baseline wander, Compared with the data before monitoring interruption, if y_iL () is in the elevated side of baseline, then after correcting If y_iL side that () declines in baseline, then after correcting

3) augmentation process is carried out to Gauss-Newton image reconstruction formula

To Δ ρ=- [J^tJ+λR]^-1Jz, Δ ρ be not in the same time between electrical impedance change profile vector, z is border not in the same time Voltage difference vector；

By Jacobin matrixExpand toBy constraint matrixExpand toAnd coefficient lambda and constraint matrix R is determined using prior information_augUnit item；

4) impedance variations of target field domain are rebuild using the reconstruction formula of data and augmentation after correction

I.e.WhereinFor the measurement voltage data at t1 moment,For the measurement at t2 moment Voltage data.

3. the place that monitoring image information of having no progeny in cranium brain dynamic electric impedance imaging on-line monitor according to claim 2 is inherited Reason method, it is characterised in that step 1) in, to i-th valid data channel data x_iL () uses Spline-Fitting, specifically Carry out as the following formula：

$L=p\underset{j=0}{\overset{n}{\Σ}}{(x_{i,j}-S(l_j\left)\right)}^2+(1-p){\∫}_m^n{\left(S^{\′\′}\right(l\left)\right)}^{2}dl;$

Wherein, S (x) is data sequence x_iL the spline-fit function of (), asks S (x) to make the value of L get minimum；Wherein, p ∈ [0,1] Reflect the degree of closeness of fitting function and measured data；S (x) is segmentation batten fitting function, is write as general type：

S_j(l)=a_j(l-l_j)³+b_j(l-l_j)²+c_j(l-l_j)+d_j；

Add boundary condition to solve each piecewise fitting function coefficients of S (l) using least square method, obtain S (l) expressions Afterwards, seek fitting reconfiguration sequenceDetectionThe maximum of uplifted side at the baseline transitionIt is minimum with decline side ValueCalculate Baseline wander value

4. the place that monitoring image information of having no progeny in cranium brain dynamic electric impedance imaging on-line monitor according to claim 2 is inherited Reason method, it is characterised in that step 3) concrete operations are：

Gauss -- Newtonian image reconstruction formula such as following formula：

Δ ρ=- [J^tJ+λR]^-1Jz；

Wherein, J is Jacobin matrix, and λ R are regularization constraint item, Δ ρ be not in the same time between electrical impedance change profile vector, z For boundary voltage difference vector not in the same time；

For suppressing electrode position change influence on RT, augmentation process is carried out to the matrix in reconstruction formula：For refined Respectively compare matrixExpand ton_measFor the measurement voltage number that frame data are included, n_elem For the FEM model unit number that imaging reconstruct used is used, n_dimFor being imaged dimension, two-dimensional imaging is generally carried out, so taking n_dim=2, n_eFor number of electrodes, the expansion of Jacobin matrix is filled to due to the disturbance of measuring electrode change in location, i.e.,：

${\mathrm{\Δv}}_{n_{electrode}|d_{move}}=AH-{\mathrm{AH}}_{n_{electrode}|d_{move}},d_{move}=xory;$

Wherein, A is current excitation vector, and H is positive calculating matrix,After changing for electrode position in an x or y direction The positive calculating matrix for recalculating, the displacement generally unification are set to a priori constant；

Then, augmentation process is carried out to constraint matrix R：

WillExpand toExpansion is filled to the noise prior estimate of reconstruct data With reconstruct conductivity variations prior estimate, i.e.,：

$R_{aug}=\left(\begin{array}{cc}R& \\ & R_{extra}\end{array}\right);$

Wherein, R_extraFor a Laplace filter, regularization parameter λ and R are determined according to the prior information of imaging field domain_extra Laplce's template, then have：

$\mathrm{\λ}=\frac{{\mathrm{ave}}_{noise}}{{\mathrm{ave}}_{conduct}},R_{extra}={\left(\frac{{\mathrm{ave}}_{conducl}}{{\mathrm{ave}}_{move}}\right)}^2{\▿}^2;$

Wherein, ave_noiseFor noise relative to measurement data average priori amplitude, ave_conductFor conductivity variations relative to The average priori amplitude of initial conductivity distribution, ave_moveIt is the average priori amplitude of electrode displacement relative to field domain radius, For Laplace operator.