CEREBRAL FUNCTION STATE EVALUATION DEVICE BASED ON BRAIN HEMOGLOBIN INFORMATION

Abstract

The present invention relates to a cerebral function state evaluation device, which comprises a brain oxyhemoglobin concentration variation acquiring component, acquiring brain oxyhemoglobin concentrations of a stroke patient, who is in a phase of completing finger-nose and heel-knee-tibia tests under instruction, by applying near-infrared spectroscopic brain imaging technology; a brain functional network constructing component; a typical feature acquiring component; and an evaluation model establishing component. The cerebral function state evaluation device evaluates a patient's motor ability based on brain hemoglobin information. By using the proposed evaluation device, an evaluation result can be given only if a patient completes several required actions. The device is inventive and simple to operate, and subjective factors in the process of the scale scoring can be avoided.

Claims

1. A cerebral function state evaluation device, comprising: a brain oxyhemoglobin concentration variation acquiring component, acquiring brain oxyhemoglobin concentrations of a stroke patient, who is in a phase of completing finger-nose and heel-knee-tibia tests under instruction by applying near-infrared spectroscopic brain imaging technology; a brain functional network constructing component, evaluating functional connection of the brain by analyzing oxyhemoglobin concentrations that acquired by the brain oxyhemoglobin concentration variation acquiring component, and constructing a brain functional network therewith; a typical feature acquiring component, calculating network topology parameters of the brain functional network constructed by the brain functional network constructing component; these network topology parameters, combined with the wavelet coherence coefficients between brain regions, are considered as an original feature set; the original feature set is screened by filtering and cooperative wrapper-based feature selection methods, and the final typical features are obtained, and an evaluation model establishing component, fitting the final typical features acquired by the typical feature acquiring component, and establishing an evaluation model of recovery level of the stroke patient by using a machine learning algorithm of a support vector regression machine.

2. The cerebral function state evaluation device according to claim 1, wherein in the completing finger-nose and heel-knee-tibia tests under instruction, the upper limbs perform the finger-nose action task, and the lower limbs perform the heel-knee-tibia task, and the upper and lower limbs on both the healthy and affected side respectively perform respective task for 4 times, where the rest time between every two tasks is 30 seconds.

3. The cerebral function state evaluation device according to claim 1, wherein in the a brain functional network constructing component, evaluating functional connection of the brain by analyzing oxyhemoglobin concentrations that acquired by the brain oxyhemoglobin concentration variation acquiring component, and constructing a brain functional network therewith, when evaluating the functional connection of the brain, wavelet coherence analysis method is used to calculate the coherence of each brain functional region, and the coherence coefficients are used to evaluate the functional connection of the brain.

4. The cerebral function state evaluation device according to claim 1, wherein in the a brain functional network constructing component, evaluating functional connection of the brain by analyzing oxyhemoglobin concentrations that acquired by the brain oxyhemoglobin concentration variation acquiring component, and constructing a brain functional network therewith, when the brain functional network is constructed, network parameters of the functional network are calculated, including an average node degree, a network density, and a clustering coefficient.

5. The cerebral function state evaluation device according to claim 3, wherein in the a typical feature acquiring component, calculating network topology parameters of the brain functional network constructed by the brain functional network constructing component; these network topology parameters, combined with the wavelet coherence coefficients between brain regions, are considered as an original feature set; the original feature set is screened by filtering and cooperative wrapper-based feature selection methods, and the final typical features are obtained, network parameters of different brain regions are compared respectively, and digital feature values of the network parameters are calculated, and the digital feature values include a covariance, a mean square error and a mean; and a corresponding mean, variance and coefficient of variation are calculated based on the coherence coefficient between brain regions calculated in the a brain functional network constructing component, evaluating functional connection of the brain by analyzing oxyhemoglobin concentrations that acquired by the brain oxyhemoglobin concentration variation acquiring component, and constructing a brain functional network therewith, when evaluating the functional connection of the brain, wavelet coherence analysis method is used to calculate the coherence of each brain functional region, and the coherence coefficients are used to evaluate the functional connection of the brain, all the above values corresponding to network parameters and coherence coefficients are combined as the original feature set.

6. The cerebral function state evaluation device according to claim 1, wherein in the a typical feature acquiring component, calculating network topology parameters of the brain functional network constructed by the brain functional network constructing component; these network topology parameters, combined with the wavelet coherence coefficients between brain regions, are considered as an original feature set; the original feature set is screened by filtering and cooperative wrapper-based feature selection methods, and the final typical features are obtained, when the original feature set is screened by using a feature selection method, the feature set is firstly preliminarily screened by a filter-based feature selection method; and then typical features are further selected from those preliminarily screened as the final typical features, by a wrapper-based feature selection method.

7. The cerebral function state evaluation device according to claim 6, wherein the filter-based feature selection method is a correlation coefficient method.

8. The cerebral function state evaluation device according to claim 6, wherein the wrapper-based feature selection method is a genetic algorithm.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0020] FIG. 1 is a schematic structural diagram of a cerebral function state evaluation device according to an embodiment of the present application.

[0021] FIG. 2 is a flow chart of a genetic algorithm in a cerebral function state evaluation device according to an embodiment of the present application.

DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] To make the objects, technical solutions, and advantages of the present invention clearer, the present invention is described in further detail with reference to accompanying drawings and examples. It should be understood that the specific examples described herein are merely provided for illustrating, instead of limiting the present invention.

[0023] Referring to FIG. 1, a cerebral function state evaluation device comprises:

[0024] a brain oxyhemoglobin concentration variation acquiring component 100, acquiring brain oxyhemoglobin concentrations of a stroke patient who is in a phase of completing finger-nose and heel-knee-tibia tests under instruction by applying near-infrared spectroscopic brain imaging technology;

[0025] a brain functional network constructing component 200, evaluating functional connection of the brain by analyzing oxyhemoglobin concentrations acquired by the brain oxyhemoglobin concentration variation acquiring component, and constructing a brain functional network therewith;

[0026] a typical feature acquiring component 300, calculating network topology parameters of the brain functional network constructed by the brain functional network constructing component; these network topology parameters, combined with the wavelet coherence coefficients between brain regions, are considered as an original feature set; the original feature set is screened by filtering and cooperative wrapper-based feature selection methods, and the final typical features are obtained; and an evaluation model establishing component 400, fitting the final typical features acquired by the typical feature acquiring component and establishing an evaluation model of recovery level of the stroke patient by using a machine learning algorithm of a support vector regression machine.

[0027] When a machine learning algorithm is used, the machine learning algorithm of a support vector regression machine is used to learn and fit the features obtained by the typical feature acquiring component, and establish an evaluation model.

[0028] Specifically, for a given training sample D={(x.sub.1,y.sub.1), (x.sub.2,y.sub.2), . . . , (x.sub.m,y.sub.m)}, the support vector regression machine is intended to obtain a formula below to fit the sample:

f(x)=.sup.Tx+b

[0029] The above problem can be converted into the following problem:

[00001]$\underset{,b,{}_i,{\hat{}}_i}{\min }.Math.\frac{1}{2}.Math.{.Math..Math.}^2+C.Math.\underset{i=1}{\overset{m}{.Math.}}.Math.\left({}_i+{\hat{}}_i\right)$$s.t.\{\begin{array}{c}f\left(x_i\right)-y_i+{}_i\\ y_i-f\left(x_i\right)+{}_i\\ {}_{i}0,{\hat{}}_{i}0,i=1,2,\mathrm{.Math.}.Math.,m\end{array}$

[0030] The Lagrangian multiplier is introduced, and the dual problem obtained is:

[00002]$\underset{,\hat{}}{\max }.Math.\underset{i=1}{\overset{m}{.Math.}}.Math.y_i\left({\hat{}}_i-{}_i\right)-\left({\hat{}}_i+{}_i\right)-\frac{1}{2}.Math.\underset{i=1}{\overset{m}{.Math.}}.Math.\underset{j=1}{\overset{m}{.Math.}}.Math.\left({\hat{}}_i-{}_i\right).Math.\left({\hat{}}_j-{}_j\right).Math.K\left(x_1.Math.x_j\right)$$s.t.\{\begin{array}{c}\underset{i=1}{\overset{m}{.Math.}}.Math.\left({\hat{}}_i-{}_i\right)\\ 0{}_i.Math.{\widehat{,}}_{i}C\end{array}$

[0031] The Lagrange multiplier is solved to obtain the offset:

[00003]$b=y_i+\mathrm{.Math.}-\underset{i=1}{\overset{m}{.Math.}}.Math.\left({\hat{}}_i-{}_i\right).Math.K\left(x_i,x_j\right)$

[0032] Then, the final fitting curve is:

[00004]$f\left(x\right)=\underset{i=1}{\overset{m}{.Math.}}.Math.\left({\hat{}}_i-{}_i\right).Math.K\left(x_i,x_j\right)+b$

[0033] It is to be understood that the components such as the brain oxyhemoglobin concentration variation acquiring component, the brain functional network constructing component, the typical feature acquiring component, the evaluation model establishing component, and the like can be implemented with hardware. Those skilled in the art should understand how to implement the above components through hardware (for example, discrete hardware elements, integrated circuits, digital circuits based on gate devices, analog circuit components, programmable hardware devices (such as microcontrollers, and FPGAs, etc.) and circuit systems composed of any combination of the above).

[0034] The cerebral function state evaluation device evaluates a patient's motor ability based on brain hemoglobin information. By using the proposed evaluation device, an evaluation result can be given only if the patient completes several required actions. The device is inventive and simple to operate, and subjective factors in the process of the scale scoring can be avoided.

[0035] In another embodiment, in the completing finger-nose and heel-knee-tibia tests under instruction, the upper limbs perform the finger-nose task, the lower limbs perform the heel-knee-tibia task, and the upper and lower limbs on both the healthy and affected side respectively perform the respective task for 4 times, where the rest time between two tests is 30 seconds.

[0036] In another embodiment, in the a brain functional network constructing component, evaluating functional connection of the brain by analyzing oxyhemoglobin concentrations acquired by the brain oxyhemoglobin concentration variation acquiring component, and constructing a brain functional network therewith, when evaluating the functional connection of the brain, wavelet coherence analysis method is used to calculate the coherence of each brain functional region, and the coherence coefficients are used to evaluate the functional connection of the brain.

[0037] Specifically, in the pre-processing, a method of mathematical morphology filtering is used to perform baseline correction on an original signal, and then a moving average smoothing method is used to remove high-frequency components in the signal:

[0038] Input sequence f(n) and structural element k(m) are defined.

[0039] Erosion operation is defined as:

(fk)(n)=min.sub.m=0, . . . ,M-1{f(n+m)k(m)}n=0,1, . . . ,NM

[0040] Dilation operation is defined as:

(fk)(n)=max.sub.m=0, . . . ,M-1{f(nm)+k(m)}n=0,1, . . . ,NM

[0041] Morphological opening operation is defined as:

(fk)(n)=[(fk)k](n)

[0042] Morphological closing operation is defined as:

(f.Math.k)(n)=[(fk)k](n)

[0043] Signal after baseline correction f.sub.correction:

f.sub.correction=f.sub.0(f.sub.0k+f.sub.0.Math.k)/2

[0044] in which f.sub.0 is the original signal.

[0045] Then the signal f.sub.correction is smoothed to obtain a preprocessed signal f.sub.preprocess:

f.sub.preprocess=smooth(f.sub.correction)

[0046] in which smooth() is a moving averaging operator.

[0047] When evaluating the functional connection of the brain, wavelet coherence analysis method is used to calculate the coherence of each brain functional region at a center frequency of 0.04 Hz, and the functional connection of the brain is evaluated with the coherence coefficients.

[0048] Morlet wavelet is defined as:

.sub.morlet()=.sup.1/4e.sup.i.sup.0.sup.e.sup..sup.2.sup./2

[0049] Continuous wavelet transform is defined as:

[00005]$W_n\left(s\right)=\underset{n=0}{\overset{N-1}{.Math.}}.Math.x_n.Math.\left[\frac{\left(n^{}-n\right).Math.{}_t}{s}\right]$

[0050] Discrete Fourier transform is performed on x.sub.n, and according to the convolution theory, it is obtained that:

[00006]$W_n\left(s\right)=\underset{k=0}{\overset{N-1}{.Math.}}.Math.x_k.Math.\left(s.Math..Math.{}_k\right).Math.e^{i.Math..Math.{}_k.Math.n.Math..Math.{}_t}$

[0051] in which the angular frequency is defined as:

[00007]${}_k=\{\begin{array}{c}\frac{2.Math..Math..Math.k}{N.Math..Math.{}_t}:k\frac{N}{2}\\ -\frac{2.Math..Math..Math.k}{N.Math..Math.{}_t}:k>\frac{N}{2}:\end{array}$

[0052] The smoothing operation over time is defined as S.sub.time:

[00008]${\stackrel{\_}{W}}_n^2\left(s\right)=\frac{1}{n_a}.Math.\underset{n=n_1}{\overset{n_2}{.Math.}}.Math.{.Math.W_n\left(s\right).Math.}^2$

[0053] The smoothing operation over scale is defined as S.sub.scale: scale:

[00009]${\stackrel{\_}{W}}_n^2=\frac{{}_j.Math.{}_t}{C_{}}.Math.\underset{j=j_1}{\overset{j_2}{.Math.}}.Math.\frac{{.Math.W_n\left(s_j\right).Math.}^2}{s_j}$

[0054] The smoother is defined as:

S(W)=S.sub.scale(S.sub.time(W.sub.n(s)))

[0055] The cross spectrum is defined as:

W.sub.n.sup.XY(s)=|W.sub.n.sup.X(s)W.sub.n.sup.Y*(s)|

[0056] Wavelet coherence coefficient:

[00010]$R_n^2\left(s\right)=\frac{{.Math.S\left(s^{-1}.Math.W_n^{\mathrm{XY}}\left(s\right)\right).Math.}^2}{S\left(s^{-1}.Math.{.Math.W_n^X\left(s\right).Math.}^2\right).Math.S\left(s^{-1}.Math.{.Math.W_n^Y\left(s\right).Math.}^2\right)}$

[0057] In another embodiment, in the a brain functional network constructing component, evaluating functional connection of the brain by analyzing oxyhemoglobin concentrations that acquired by the brain oxyhemoglobin concentration variation acquiring component, and constructing a brain functional network therewith, when the brain functional network is constructed, network parameters of the functional network are calculated, including an average node degree, a network density, and a clustering coefficient.

[0058] Specifically, when the brain functional network is constructed, a threshold is set according to the strength of functional connection of the brain, an adjacency matrix is obtained based on the threshold, and network parameters such as an average node degree, a network density, and a clustering coefficient of the functional network are calculated based on the value of the adjacency matrix.

[0059] The adjacency matrix is calculated:

[00011]$M_{\mathrm{adj}}\left(i,j\right)=\{\begin{array}{cc}1& R\left(i,j\right)T\\ 0& R\left(i,j\right)<T\end{array}$

[0060] in which R(i,j) is a wavelet coherence value of a channel i and a channel j, which are calculated with the following equation

[00012]$R_n^2\left(s\right)=\frac{{.Math.S\left(s^{-1}.Math.W_n^{\mathrm{XY}}\left(s\right)\right).Math.}^2}{S\left(s^{-1}.Math.{.Math.W_n^X\left(s\right).Math.}^2\right).Math.S\left(s^{-1}.Math.{.Math.W_n^Y\left(s\right).Math.}^2\right)}$

[0061] and T is a set threshold.

[0062] The brain functional network is constructed according to the adjacency matrix, and then the following three brain functional network parameters are calculated:

[0063] N is defined as the number of nodes in the network, and when T is given, they are calculated as follows.

[0064] Average node degree:

[00013]$K={}_{\mathrm{degree}}\left(M_{\mathrm{adj}}\right)=\frac{1}{N}.Math.\underset{i=1}{\overset{N}{.Math.}}.Math.K_i$

[0065] K.sub.i represents a degree value at each node i.

[00014]$K_i=\underset{j=1,ji}{\overset{N}{.Math.}}.Math.M_{\mathrm{adj}}\left(i,j\right)$

[0066] Network density:

[00015]$D={}_{\mathrm{density}}\left(M_{\mathrm{adj}}\right)=\frac{\underset{i=1,ij}{\overset{N}{.Math.}}.Math.\underset{j=1,ji}{\overset{N}{.Math.}}.Math.M_{\mathrm{adj}}\left(i,j\right)}{2.Math.N\left(N-1\right)}$

[0067] Clustering coefficient of a node i:

[00016]$C_i=\frac{2.Math.e_i}{K_i\left(K_i-1\right)}$

[0068] e.sub.i represents the number of nodes adjacent to the node i

[0069] Clustering coefficient of the network:

[00017]$C={}_{\mathrm{clustercoff}}\left(M_{\mathrm{adj}}\right)=\frac{1}{N}.Math.\underset{i=1}{\overset{N}{.Math.}}.Math.C_i$

[0070] In another embodiment, in the a typical feature acquiring component, calculating network topology parameters of the brain functional network constructed by the brain functional network constructing component; these network topology parameters, combined with the wavelet coherence coefficients between brain regions, are considered as the original feature set; the original feature set is screened by filtering and cooperative wrapper-based feature selection methods, and the final typical features are obtained, network parameters of different brain regions are compared respectively, and digital feature values of the network parameters are calculated, and the digital feature values include a covariance, a mean square error and a mean; and a corresponding mean, variance and coefficient of variation are calculated based on the coherence coefficient between brain regions calculated in the a brain functional network constructing component, evaluating functional connection of the brain by analyzing oxyhemoglobin concentrations acquired by the brain oxyhemoglobin concentration variation acquiring component, and constructing a brain functional network therewith, when evaluating the functional connection of the brain, wavelet coherence analysis method is used to calculate the coherence of each brain functional region, and the coherence coefficients are used to evaluate the functional connection of the brain, all the above values corresponding to network parameters and coherence coefficients are combined as the original feature set.

[0071] When the digital features between the parameter variation curves and the wavelet coherence values between brain regions are calculated, the network parameter variation curves of different brain regions and the network parameter variation curves of the healthy side and the affected side during the task phase are compared, and a covariance, a mean square error, a volatility, a mean, and other digital feature values between variation curves are calculated and used as a feature set. In addition, the mean, variance and coefficient of variation of the coherence values between the brain regions are also added to the feature set.

[0072] The following parameters between curves are calculated:

[0073] Covariance:

[00018]$\mathrm{Cov}\left(X,Y\right)=\frac{1}{N-1}.Math.\underset{T=t_1}{\overset{t_2}{.Math.}}.Math.\left({}_X\left(T\right)-\stackrel{\_}{{}_X}\right).Math.\left({}_Y\left(T\right)-\stackrel{\_}{{}_Y}\right)$

[0074] Root mean square error:

[00019]$\mathrm{RMS}.Math..Math.E\left(X,Y\right)=\sqrt{\frac{1}{N}.Math.\underset{T=t_1}{\overset{t_2}{.Math.}}.Math.{}_X\left(T\right)-{}_Y\left(T\right)}$

[0075] Mean:

[00020]$E\left(X\right)=\frac{1}{N}.Math.\underset{T=t_1}{\overset{t_2}{.Math.}}.Math.{}_X\left(T\right)$

[0076] Volatility:

[00021]$\mathrm{Flu}\left(X\right)=\frac{1}{N}.Math.\underset{T=t_1}{\overset{t_2}{.Math.}}.Math..Math.d\left({}_X\left(T\right)\right).Math.$

[0077] In another embodiment, in the a typical feature acquiring component, calculating network topology parameters of the brain functional network constructed by the brain functional network constructing component; these network topology parameters, combined with the wavelet coherence coefficients between brain regions, are considered as the original feature set; the original feature set is screened by filtering and cooperative wrapper-based feature selection methods, and the final typical features are obtained, when the original feature set is screened by using a feature selection method, the feature set is firstly preliminarily screened by a filter-based feature selection method; and then typical features are further selected from those preliminarily screened as the final typical features, by a wrapper-based feature selection method.

[0078] In another embodiment, the filter-based feature selection method is a correlation coefficient method.

[0079] Specifically, a sample set D={(x.sub.1,y.sub.1), (x.sub.2,y.sub.2), . . . , (x.sub.m,y.sub.m)} is given, in which x.sub.i is a feature set and y.sub.i is a true value.

[0080] A Pearson correlation coefficient operator Peason() is defined, and the square of the Pearson correlation coefficient of each column of features in the feature set with the true values r.sup.2 is calculated as follows:

r.sup.2(i)=Peason(x.sup.i,y).sup.2

[0081] r.sup.2 is sorted, and the features corresponding to the 25 highest r.sup.2 are used as preliminary features.

[0082] In another embodiment, the wrapper-based feature selection method is a genetic algorithm.

[0083] For specific steps of the genetic algorithm, reference may be made to FIG. 2.

[0084] In the present invention, wavelet coherence analysis method is used, and the correlation between nodes in a central frequency band can be analyzed, which is beneficial to monitoring the correlation between brain regions in each neural activity frequency band and physiologically active frequency band.

[0085] In the present invention, complex network analysis method is used, and the cooperation between various brain regions is analyzed from the perspective of the data transmission capability and the work efficiency between the nodes, which is beneficial to finding the parameter index reflecting the working state of the brain.

[0086] In the present invention, the algorithm of support vector regression machine is used, and an optimal regression model can be established according to the information of the feature parameters, thereby improving the accuracy of determining the state of the brain.

[0087] In the present invention, the patient's motor ability is evaluated based on the brain information. Based on the proposed evaluation device, an evaluation result can be given only if the patient completes several required actions. The device is inventive and simple to operate, and subjective factors during the evaluation can be avoided.

[0088] In the present invention, the near-infrared spectroscopic brain imaging technology is used for test, which is simple in operation. It has low requirements on the external environment, and low susceptibility to environmental noise. The subjects will not be negatively affected. Throughout the test, the patient completes the finger-nose and heel-knee-tibia tests in the natural environment, and the resulting analysis results are more reliable to assess the patient's recovery level.

[0089] The technical features of the above-described embodiments may be used in combination. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, where no contradiction exists, all the combinations of these technical features are contemplated in the scope of the present invention.

[0090] The above-described embodiments are merely illustrative of several implementations of the present invention, and the description is specific and particular, but is not to be construed as limiting the scope of the present invention. It should be pointed out that for those of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present invention, all of which fall within the protection scope of the present invention. Therefore, the protection scope of the present invention is defined by the appended claims.