CN102509123A - Brain functional magnetic resonance image classification method based on complex network - Google Patents

Abstract

The invention relates to a brain functional magnetic resonance image classification method based on a complex network, which comprises the following steps: pre-processing training sample images and test sample images, carrying out region segmentation, and extracting an average time sequence from each region; calculating the partial correlation coefficient among the average time sequences, carrying out matrix binarization on the partial correlation coefficient to obtain a complex network model, and calculating the feature path length, cost and clustering degree of the complex network model to respectively obtain network features of the training sample images and the test sample images; training to obtain an adaboost classifier; and by using the adaboost classifier obtained by training, classifying the test sample images. By using information in the brain functional magnetic resonance images as much as possible, the method can accurately classify the brain functional magnetic resonance images.

Description

A Classification Method of Brain Functional Magnetic Resonance Image Based on Complex Network

technical field

The invention belongs to the technical field of image processing, and in particular relates to a method for classifying brain functional magnetic resonance images based on a complex network.

Background technique

Functional Magnetic Resonance Imaging (fMRI) has been widely used in the diagnosis and treatment of neurological diseases due to its high spatial and temporal resolution and non-invasive features. fMRI generally refers to magnetic resonance imaging based on blood oxygen level-dependent (BOLD), which reflects changes in the brain by measuring changes in magnetic resonance signals caused by changes in cerebral blood flow and cerebral blood oxygen caused by neural activity. Activity. The brain is a complex system, and magnetic resonance images of the brain respond to stimulating conditions or lesions. It is an important application of computer-aided analysis to use image classification methods to calculate the possibility of certain attributes of brain fMRI images, or to automatically distinguish the category attributes of images.

Traditional fMRI image classification methods mainly include region of interest (ROI) method and voxel method. The classification method of the region of interest method divides the sample and the target into multiple target regions based on the prior knowledge of the target structure, and classifies the target accordingly; the classification method of the voxel method uses complex nonlinear registration to maximize To maximize the precise correspondence between individuals, and then use each spatial unit (voxel) of the image as the basis for classification. Both approaches assume a one-to-one correspondence between targets and the internal organization of samples. The former assumes that the prior image region exists in each target image and can be segmented accurately; the latter assumes that the voxels after nonlinear registration are one-to-one correspondence. However, such an assumption is unreasonable in many cases. Functional magnetic resonance images of people in different states will be disturbed by many factors. Traditional classification methods do not classify functional magnetic resonance images of brain according to the inherent properties of the brain, which will lead to a decline in classification performance.

Contents of the invention

(1) Technical problems to be solved

In order to overcome the deficiencies of the prior art, the technical problem to be solved by the present invention is to design a method for classifying brain functional magnetic resonance images with high classification accuracy and strong generalization performance.

(2) Technical solutions

In order to achieve the above object, the present invention proposes a method for classifying brain fMRI images based on a complex network, comprising the following steps:

Step Sa: Preprocessing the training sample image and the test sample image, then performing brain region segmentation, and extracting the average time series of each brain region;

Step Sb: Calculating the partial correlation coefficients between each average time series to obtain a partial correlation coefficient matrix;

Step Sc: Binarize the partial correlation coefficient matrix to obtain a complex network model;

Step Sd: calculating the characteristic path length, cost and clustering degree of the complex network model as the features of the fMRI image;

Step Se: using the network parameters of the training sample image as the feature of the training sample image in the feature of the functional magnetic resonance image, to train an adaptive improvement (adaboost) classifier;

Step Sf: classify the test sample image by using the trained adaptive boosting (adaboost) classifier.

(3) Beneficial effects

Aiming at the problem of brain functional magnetic resonance image classification, the present invention effectively improves the accuracy and stability of image classification by constructing a brain network model, calculating network characteristic parameters, training an adaptive boost (adaboost) classifier and the like.

The present invention can utilize as much information as possible in the functional magnetic resonance image of the brain, and the parameters of the brain network can reflect the activity of the brain in essence, making up for the deficiency that the traditional classification method cannot reflect the inherent attributes of the brain activity, and can accurately analyze the functional magnetic resonance of the brain. Images are classified.

Description of drawings

Fig. 1 is the method flowchart of the functional magnetic resonance image classification based on complex network provided by the present invention;

Fig. 2 is a comparison curve of classification receiver operating characteristics (ROC) using the classification method of the present invention (method A) compared with the existing classification method based on local features (method B) according to an embodiment of the present invention.

Detailed ways

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be described in further detail below in conjunction with specific embodiments and with reference to the accompanying drawings.

Functional magnetic resonance image classification based on complex networks is a new classification method for functional magnetic resonance images of brain. The method first establishes a complex brain network model, and calculates the characteristic path length and clustering degree of the brain network to represent different image patterns; then uses the characteristic path length and clustering degree to train an adaptive boost (adaboost) classifier; finally The test sample image is classified by using the trained adaptive boosting (adaboost) classifier.

With reference to Fig. 1, according to a kind of human brain fMRI image classification method of the present invention, can determine the category of test sample image according to training sample image, concrete implementation steps are as follows:

Step Sa, preprocessing the training sample image and the test sample image, then performing brain region segmentation, and extracting the average time series of each brain region;

1. Preprocessing of brain fMRI images

Due to the influence of various noises during the MRI scanning process, there are differences in the scale and position of the individual subjects themselves, so it is very necessary to preprocess the data before analyzing the data. In the data acquisition of the whole experiment, the main sources of noise information are: (1) physical head movement; (2) difference in scanning time between layers in the image; (3) inhomogeneity of the external magnetic field, etc. The common steps of brain fMRI image preprocessing include: slice scan time alignment, image sequence alignment, co-registration, standardization (or homogenization), spatial smoothing filter and temporal smoothing filter, etc.

2. Segmentation of brain fMRI images

Using the internationally accepted structural labeling template (AAL), the whole brain was divided into 90 brain regions. Structural labeling template is the most widely used brain structural template in the field of brain fMRI image research.

3. Extract the average time series of each brain region

According to the data of the preprocessed functional magnetic resonance image of the brain, extract the time series Y (matrix dimension D×N) of the activation values of each voxel contained in the corresponding brain area at different time points, where D is the The number of voxels inside, N is the number of time points. The activation value refers to the blood oxygen level dependence (BOLD) intensity of each voxel at different time points.

Step Sb: Calculating the average partial correlation coefficient between each time series. This step Sb specifically includes the following steps:

1. Calculate the covariance coefficient between the average time series

According to the time series of each brain region extracted in step Sa, calculate the covariance matrix S between each average time series, and each element _{si, j} of S is the covariance coefficient between the i-th and j-th time series ,

${\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; m</span>}}\underset{\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}<span\; class="notranslate">\; )</span>$

为第i个时间序列的平均值，为第j个时间序列的平均值。Among them, M is the number of time points, x _i (t) (i=1,...,M) is the i-th time series,

is the mean value of the i-th time series, is the average value of the jth time series.

2. Calculate the partial correlation coefficient between the average time series

According to the covariance coefficient matrix S between time series (matrix dimension is 90×90), calculate the partial correlation coefficient matrix R between time series (matrix dimension is 90×90), each element r _{i, j} of R is:

${\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}}{\sqrt{{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}}}$

为协方差矩阵S(矩阵维度为90×90)的逆矩阵的第{i，j}个元素。in,

is the {i, j}th element of the inverse matrix of the covariance matrix S (the matrix dimension is 90×90).

3. Perform Fisher transform on the partial correlation coefficient

According to the partial correlation coefficient matrix R (the matrix dimension is 90×90), calculate the partial correlation coefficient matrix F (the matrix dimension is 90×90) after Fisher transformation, and each element f _ij of F is:

${\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}}{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>$

Among them, f _ij is the {i, j}th element of the partial correlation coefficient matrix F (the matrix dimension is 90×90) after Fisher transformation, and r _ij is the element of the partial correlation coefficient matrix R (the matrix dimension is 90×90). The {i, j}th element.

Step Sc: Binarize the partial correlation coefficient matrix to obtain a complex network model;

其中N为网络中节点的数目)的十分之一。二值化的过程可描述为令Set the threshold T', and make the value greater than or equal to T' in the partial correlation coefficient matrix F (the matrix dimension is 90×90) after Fisher transformation be 1, and the value smaller than T' be 0, so as to obtain a complex network model. In the binarized matrix, 1 indicates that there is a connection between two brain regions, that is, there is an edge between two nodes in the network, and 0 indicates that there is no connection between two brain regions, that is, there is an edge between two nodes in the network. There is no edge. The method of threshold selection is: the number of edges that actually exist in the network is the number of edges that may exist in the network

where N is one-tenth of the number of nodes in the network). The process of binarization can be described as

${\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; \&Greater\; Equal;</span>{\mathrm{<span\; class="notranslate">\; T</span>}}^{<span\; class="notranslate">\; \&prime;</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; T</span>}}^{<span\; class="notranslate">\; \&prime;</span>}\end{array}\right.<span\; class="notranslate">\; ,</span>$

Among them, w _ij is the {i, j}th element of the binarized network, f _ij is the {i, j}th element of the partial correlation coefficient matrix F (the matrix dimension is 90×90) after Fisher transformation , T' is the selected threshold, |·| is the sign of absolute value calculation.

Step Sd: calculating the characteristic path length, cost and clustering degree of the complex network model as the features of the fMRI image;

According to the complex network model, the characteristic path length, cost and cluster degree of the complex network model are calculated as the features of the functional magnetic resonance image.

The characteristic path length provides the optimal path for information from one node to another node in the network. We can use the characteristic path length matrix to describe the characteristic path length l _ij of any two nodes i and j in the network. The average characteristic path length L of the network describes the average characteristic path length of any two nodes in the network, namely

$\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&NotEqual;</span>\mathrm{<span\; class="notranslate">\; j</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; l</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}$

Among them, N is the number of nodes in the network, that is, the number of divided brain regions is 90; l _ij is the characteristic path length between nodes i and j, and V is the set of all nodes in the network.

Cost is an important parameter to measure the nature of the network, and it is used to measure the overall cost of building the network. The calculation method is to compare the number of edges that actually exist in the network with the maximum number of edges that may exist in the network, namely:

$\mathrm{<span\; class="notranslate">\; K</span>}<span\; class="notranslate">\; =</span>\frac{\frac{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}{\mathrm{<span\; class="notranslate">\; K</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\frac{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}{\mathrm{<span\; class="notranslate">\; K</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span>$

Among them, N is the number of nodes in the network, K _i is the number of edges connected to node i in the network, and K is the cost of the network.

The degree of clustering is another important feature to measure the nature of the network, and it is used to measure the possibility that the adjacent nodes of a certain node are neighbors to each other. The value of the cluster degree C _i of a certain node i is equal to the ratio of the number of edges existing between its adjacent nodes to the number of all possible edges between them, that is

${\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{\frac{{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}$

就表示节点i的邻点之间可能存在的边数。Among them, e _i represents the number of edges existing between the neighbors of node i, k _i represents the number of neighbors of node i,

It means the number of edges that may exist between the neighbors of node i.

Step Se: using the network parameters of the training sample image as the features of the training sample image among the features of the fMRI image to train an adaptive boosting (adaboost) classifier.

(m为样本图像的数目)，自适应提高(adaboost)分类器在训练过程中逐渐调整三个线性分类器的权重，最后得到一个最优的自适应提高(adaboost)分类器。具体实施步骤如下：After obtaining the characteristics of the training sample image, firstly use the feature path length, cost and cluster degree as three linear classifiers, and use the weighted sum of these three linear classifiers to form a new adaptive boosting (adaboost) classifier. Initially, each The weight of each classifier is set to

(m is the number of sample images), the adaptive boosting (adaboost) classifier gradually adjusts the weights of the three linear classifiers during the training process, and finally an optimal adaptive boosting (adaboost) classifier is obtained. The specific implementation steps are as follows:

For a given sample (x ₁ , y ₁ ), ..., (x _m , y _m ), where x _i ∈ X, y _i ∈ Y = (-1, 1), X is the network of training sample images feature, Y is the image category, first set the weight of the initialization classifier as Then perform T iterations, the iteration process is as follows:

The variable t increases from 1 to T, each iteration first calculates the classification error ε _t obtained by classifying the training sample image for each feature h _t , and then calculates the new sample weight,

${\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}\mathrm{<span\; class="notranslate">\; ln</span>}<span\; class="notranslate">\; (</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}}{{\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>$

Finally, update the weights of each linear classifier,

${\mathrm{<span\; class="notranslate">\; D.</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; D.</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; Z</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}}\left\{\begin{array}{c}{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\\ {\mathrm{<span\; class="notranslate">\; e</span>}}^{{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&NotEqual;</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\end{array}\right.<span\; class="notranslate">\; ,</span>$

where Z _t is the normalization factor.

After the loop ends, the optimal adaptive boost (adaboost) classifier is obtained:

$\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; sign</span>}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; .</span>$

Step Sf: classify the test sample image by using the optimal adaptive boosting (adaboost) classifier obtained through training.

Input the test sample into the optimal adaptive boosting (adaboost) classifier obtained in the above steps to classify the test sample image, and the classification result is output through classification correct rate, true positive rate and false positive rate.

The effect of the functional magnetic resonance image classification method based on the complex network of the present invention can be illustrated by real functional magnetic resonance imaging data of the brain:

(1) Real data experiment process

In order to demonstrate the effect of the present invention, a real data set is used for testing in the embodiment, and a total of 39 subjects participated in the experiment, 20 males and 19 females. See Table 1 for the age group and clinical dementia classification information of the subjects. In the experiment, T2* weighted gradient echo planar imaging (Echo-Planar Imaging, EPI) sequence was used to obtain BOLD fMRI resting data after acupuncture stimulation.

Statistical Parametric Mapping (SPM) software ( http://www.fil.ion.ucl.ac.uk/spm/ ) was used to preprocess the data, including slice scan time alignment, image sequence alignment, co-registration, and standardization ( Or called homogenization), spatial smoothing filter. Using the method of the present invention (method A) to compare the existing classification method (method B) based on local features, obtain the receiver operating characteristic (ROC) curve and the area under the curve (AUC) of the classification method, and use the ROC Curves and AUC as measures of classifier performance.

Form 1 Subject Information

(2) Experimental results

The classification ROC curves of the two methods on the real experimental data set are shown in Figure 2, where the true positive rate in Figure 2 refers to the percentage that is actually positive and is correctly judged as positive according to the standard of the screening test , the false positive rate refers to the percentage that is actually negative but is wrongly judged positive according to the standard of the screening test. As shown in Figure 2, the ROC curve of method A is higher than that of method B in most threshold ranges; AUC value comparison: the AUC value of method A is 0.85, and the AUC value of method B is 0.78. The area under the curve (AUC) can measure the overall classification performance, posterior probability and sorting performance. The larger the AUC value, the better the overall performance of the classification method. Therefore, method A works better than method B.

Experimental results show that the complex network-based classification method for functional magnetic resonance images of the brain effectively improves the classification performance of functional magnetic resonance images of the brain.

The above is only a specific implementation mode in the present invention, but the scope of protection of the present invention is not limited thereto. Anyone familiar with the technology can understand the conceivable transformation or replacement within the technical scope disclosed in the present invention. All should be covered within the scope of protection of the claims of the present invention.

Claims (8)

1. a method for classifying functional magnetic resonance images of the brain based on a complex network, characterized in that, comprising the following steps: Step Sa: Preprocessing the training sample image and the test sample image, then performing brain region segmentation, and extracting the average time series of each brain region; Step Sb: Calculating the partial correlation coefficients between each average time series to obtain a partial correlation coefficient matrix; Step Sc: Binarize the partial correlation coefficient matrix to obtain a complex network model; Step Sd: calculating the characteristic path length, cost and clustering degree of the complex network model as the features of the fMRI image; Step Se: using the network parameters of the training sample image as the feature of the training sample image in the feature of the functional magnetic resonance image, to train an adaptive improvement (adaboost) classifier; Step Sf: classify the test sample image by using the trained adaptive boosting (adaboost) classifier. 2. the functional brain magnetic resonance image classification method based on complex network as claimed in claim 1, is characterized in that, described training sample image and test sample image are carried out preprocessing, is while retaining the brain function image detail, using The brain function image and the standard template are preprocessed by affine registration transformation, and the signal-to-noise ratio of the brain function image is improved. 3. the method for classifying functional magnetic resonance images of brain based on complex network as claimed in claim 1, is characterized in that, the step of described extraction brain area mean time series is: First, divide the whole brain into 90 brain regions according to the standard brain structure template, extract the activation values of each voxel in each brain region at different time points, and then average the activation values of each voxel to obtain the average brain region sequentially. 4. the fMRI image classification method based on complex network as claimed in claim 1, is characterized in that, the method for calculating the partial correlation coefficient between each mean time series is: First calculate the covariance matrix S between the average time series, the covariance matrix dimension is 90×90, each element _si,j of S is the covariance coefficient between the i-th and j-th time series, ${\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; m</span>}}\underset{\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>$ Among them, M is the number of time points, x _i (t) (i=1,...,M) is the i-th time series,

is the mean value of the i-th time series,

is the average value of the jth time series. Then, calculate the partial correlation coefficient matrix R between the average time series, the dimension of the partial correlation coefficient matrix R is 90×90, and each element r _{i, j} of R is: ${\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}}{\sqrt{{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}}}<span\; class="notranslate">\; ;</span>$ in,

is the {i, j}th element of the inverse matrix of the covariance matrix S; Finally, the Fisher transform is performed on the partial correlation coefficients to obtain the partial correlation coefficient matrix F after the Fisher transform. The dimension of the transformed partial correlation coefficient matrix is 90×90. 5. the functional brain functional magnetic resonance image classification method based on complex network as claimed in claim 1, is characterized in that, described partial correlation coefficient matrix binarization obtains the step of complex network model, comprises: Select the threshold to binarize the partial correlation coefficient matrix F after Fisher transformation. The dimension of the transformed partial correlation coefficient matrix is 90×90. After binarization, 1 indicates that there is a connection between two brain regions, that is, two There is an edge between nodes, and 0 means that there is no connection between the two brain regions, that is, there is no edge between two nodes in the network; The method for selecting the threshold is as follows: the number of edges actually existing in the network after the threshold is selected for binarization is one-tenth of the number of possible edges in the network. 6. the functional magnetic resonance image classification method based on complex network as claimed in claim 1, wherein said step of calculating the characteristic path length of this complex network model is: Use the characteristic path length matrix to describe the characteristic path length l _ij of any two nodes i and j in the network, and the average characteristic path length L of the network describes the average value of the characteristic path lengths of any two nodes in the network, namely $\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&NotEqual;</span>\mathrm{<span\; class="notranslate">\; j</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; l</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}$ Among them, N is the number of nodes in the network, that is, the number of divided brain regions is 90; l _ij is the characteristic path length between nodes i and j, and V is the set of all nodes in the network. 7. the functional magnetic resonance image classification method based on complex network as claimed in claim 1, wherein said step of calculating the cost of this complex network model is: Use the number of all edges that actually exist in the network to compare the number of edges that may exist in the network at most, that is: $\mathrm{<span\; class="notranslate">\; K</span>}<span\; class="notranslate">\; =</span>\frac{\frac{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}{\mathrm{<span\; class="notranslate">\; K</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\frac{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}{\mathrm{<span\; class="notranslate">\; K</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span>$ Among them, N is the number of nodes in the network, and K _i is the number of edges connected to node i in the network. 8. the functional magnetic resonance image classification method based on complex network as claimed in claim 1, wherein the step of calculating the degree of clustering of this complex network model is: The value of the cluster degree C _i of a certain node i is equal to the ratio of the number of edges existing between its adjacent nodes to the number of all possible edges between them, that is ${\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{\frac{{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}$ Among them, e _i represents the number of edges existing between the neighbors of node i, k _i represents the number of neighbors of node i,

It means the number of edges that may exist between the neighbors of node i.

Images