A method, device, terminal and storage medium for classifying functional magnetic resonance imaging of the brain

Abstract

The present invention discloses a method, device, terminal and storage medium for classifying functional magnetic resonance imaging of the brain. The method includes: respectively obtaining resting-state functional magnetic resonance images of patients with disorders of consciousness and healthy subjects; preprocessing the above data, and performing brain region segmentation on the preprocessed data based on a preset brain atlas template to obtain a number of brain regions, and extracting the average time series signal of each brain region; based on the method of a sliding time window, calculating the correlation coefficient of the average time series signals between brain regions, and constructing a dynamic functional connection matrix according to the correlation coefficient, and further calculating the dynamic coefficient of variation of each brain region under cross-time windows, and then combining with a classification model constructed based on a large sample data set to obtain the category of the functional magnetic resonance imaging of the brain. The embodiments of the present invention calculate the dynamic characteristics of local brain regions, making the detection method comprehensive. The classification model is constructed based on a large sample data set, and a more accurate image classification result can be achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical imaging, and particularly to a method, device, terminal and storage medium for classifying functional magnetic resonance imaging of the brain. Background Art

[0002] The state of impaired consciousness is mainly caused by traumatic brain injury, cerebral hemorrhage and cerebral hypoxia, and is mainly manifested as impaired consciousness to varying degrees. The state of impaired consciousness mainly includes coma, vegetative state and minimally conscious state. How to accurately diagnose the current consciousness state of patients with impaired consciousness has always been the focus and difficulty of research. The research on this issue has important social significance and helps to reduce the burden on the families and society of patients. Since the diagnostic method based on the behavioral feedback of patients still has a misdiagnosis rate of about 40%, more and more researchers have begun to use technologies such as brain imaging to study the neural markers that can reflect the current consciousness level of patients with impaired consciousness. Among them, resting-state functional magnetic resonance, as a new type of neuroimaging design, has the advantages of high spatial resolution, convenient and simple implementation, and non-invasiveness, and has been favored by researchers. However, for the detection of patients with impaired consciousness, most of the existing technologies adopt static functional connectivity analysis methods, ignoring the dynamic activity characteristics of the patients' brains in the time dimension. Although a few studies have adopted dynamic functional connectivity analysis methods, they all use k-means clustering to reduce the dimension of the brain activity patterns in the time dimension, and the final representative features obtained are also the arithmetic average of the clustering results, and then perform between-group analysis. Therefore, the deficiencies of the above methods are manifested in three aspects: First, the static analysis method ignores the dynamic activity characteristic information of the brain and is insufficient to comprehensively explore the brain activity basis supporting the normal consciousness level; Second, the previous methods mostly use clustering algorithms to detect the dynamic patterns of the brain, and the number of clusters is often determined by the researchers, which is difficult to avoid subjectivity; Finally, in the current dynamic functional connectivity research on impaired consciousness, there is a situation of small sample size, and verification based on large sample data is required.

[0003] Therefore, the existing technology still needs to be improved and developed. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method, device, terminal and storage medium for classifying functional magnetic resonance imaging of the brain, aiming at solving the problems that the method for classifying functional magnetic resonance imaging of the brain in the existing technology is not comprehensive enough and the detection result is not accurate enough.

[0005] The technical solution adopted by the present invention to solve the problem is as follows:

[0006] In the first aspect, an embodiment of the present invention provides a method for classifying functional magnetic resonance imaging of the brain, wherein the method includes:

[0007] Obtain the first resting-state functional magnetic resonance image of patients with consciousness disorders and the second resting-state functional magnetic resonance image of healthy subjects respectively;

[0008] Preprocess the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image, and perform brain region segmentation on the preprocessed first resting-state functional magnetic resonance image and second resting-state functional magnetic resonance image based on a preset brain atlas template to obtain a number of brain regions, and extract the average time series signal of each brain region;

[0009] Based on the sliding time window method, calculate the correlation coefficient of the average time series signals between brain regions, and construct a dynamic functional connectivity matrix according to the correlation coefficient;

[0010] Based on the dynamic functional connectivity matrix, calculate the dynamic coefficient of variation of each brain region under cross-time windows, and obtain the category of the brain functional magnetic resonance image according to the dynamic coefficient of variation of each brain region under cross-time windows and the classification model constructed based on a large sample data set.

[0011] In one implementation, before preprocessing the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image, it includes:

[0012] Screen the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image.

[0013] In one implementation, the preprocessing of the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image includes:

[0014] Perform removal of preset time points, inter-slice time correction, head motion correction, image registration, image segmentation, spatial normalization, denoising processing, and signal normalization processing on both the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image.

[0015] In one implementation, the extraction of the average time series signal of each brain region includes:

[0016] Extract the time series of each voxel in each brain region;

[0017] Average the time series of all voxels in each brain region to obtain the average time series signal of each brain region.

[0018] In one implementation, the calculation of the correlation coefficient of the average time series signals between brain regions based on the sliding time window method includes:

[0019] Obtain the window size and window step of the sliding time window;

[0020] Based on the window size and the window step, divide the average time series signal of each brain region into a number of time windows;

[0021] For each time window, calculate the Pearson correlation coefficient between brain regions in each time window.

[0022] In one implementation, the constructing a dynamic functional connectivity matrix according to the correlation coefficient includes:

[0023] Take the Pearson correlation coefficient between brain regions in each time window as the first matrix;

[0024] Form a dynamic functional connectivity matrix by all the first matrices corresponding to all time windows.

[0025] In one implementation, after the constructing a dynamic functional connectivity matrix according to the correlation coefficient includes:

[0026] Calculate the similarity degree of the row vectors of the first matrix in the dynamic functional connectivity matrix under different windows to obtain a similarity coefficient;

[0027] Based on the similarity coefficient, calculate the dynamic coefficient of variation of each brain region.

[0028] In one implementation, the obtaining the brain functional magnetic resonance image category according to the dynamic coefficient of variation of each brain region across time windows and a classification model constructed based on a large sample data set includes:

[0029] Obtain data samples;

[0030] Based on the support vector machine classification algorithm, take the dynamic coefficient of variation of each brain region across time windows as a feature vector, classify the data samples through a classification model constructed based on a large sample data set to obtain the brain functional magnetic resonance image category, and use the cross - validation method to test the classification model constructed based on the large sample data set.

[0031] In a second aspect, an embodiment of the present invention further provides a brain functional magnetic resonance image classification device, wherein the device includes:

[0032] An image acquisition module, configured to respectively acquire a first resting - state functional magnetic resonance image of a patient with disorder of consciousness and a second resting - state functional magnetic resonance image of a healthy subject;

[0033] A segmentation and extraction module, configured to preprocess the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image, and perform brain region segmentation on the preprocessed first and second resting-state functional magnetic resonance images based on a preset brain atlas template to obtain a plurality of brain regions, and extract the average time series signal of each of the brain regions;

[0034] A correlation coefficient calculation and construction module, configured to calculate the correlation coefficient of the average time series signals between brain regions based on a sliding time window method, and construct a dynamic functional connectivity matrix according to the correlation coefficient;

[0035] A dynamic coefficient of variation calculation and image category obtaining module, configured to calculate the dynamic coefficient of variation of each of the brain regions across time windows based on the dynamic functional connectivity matrix, and obtain the category of the brain functional magnetic resonance image according to the dynamic coefficient of variation of each of the brain regions across time windows and a classification model constructed based on a large sample data set.

[0036] In a third aspect, an embodiment of the present invention further provides an intelligent terminal, including a memory, and one or more programs, where one or more programs are stored in the memory and are configured to be executed by one or more processors, and the one or more programs include instructions for executing the brain functional magnetic resonance image classification method as described in any one of the above.

[0037] In a fourth aspect, an embodiment of the present invention further provides a non-transitory computer-readable storage medium, when instructions in the storage medium are executed by a processor of an electronic device, enabling the electronic device to execute the brain functional magnetic resonance image classification method as described in any one of the above.

[0038] Advantages of the present invention: In the embodiments of the present invention, first, the first resting-state functional magnetic resonance image of a patient with disturbance of consciousness and the second resting-state functional magnetic resonance image of a healthy subject are obtained respectively; then, the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image are preprocessed, and brain region segmentation is performed on the preprocessed first resting-state functional magnetic resonance image and second resting-state functional magnetic resonance image based on a preset brain atlas template to obtain a plurality of brain regions, and the average time series signal of each brain region is extracted; then, based on the sliding time window method, the correlation coefficient of the average time series signals between brain regions is calculated, and a dynamic functional connection matrix is constructed according to the correlation coefficient; finally, based on the dynamic functional connection matrix, the dynamic coefficient of variation of each brain region under cross-time windows is calculated, and according to the dynamic coefficient of variation of each brain region under cross-time windows and a classification model constructed based on a large sample data set, the category of the brain functional magnetic resonance image is obtained; it can be seen that in the embodiments of the present invention, the dynamic characteristics of local brain regions are calculated, so that the detection method is comprehensive, and the classification model is constructed based on a large sample data set, and a more accurate image classification result can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0040] Figure 1 Schematic flowchart of a method for classifying brain functional magnetic resonance images provided by an embodiment of the present invention.

[0041] Figure 2 Schematic flowchart of a method for classifying brain functional magnetic resonance images in an implementation manner provided by an embodiment of the present invention.

[0042] Figure 3 Schematic diagram of the similarity matrix of the dynamic functional connection of healthy subjects provided by an embodiment of the present invention.

[0043] Figure 4 Schematic diagram of the similarity matrix of the dynamic functional connection of a patient with disturbance of consciousness provided by an embodiment of the present invention.

[0044] Figure 5 Schematic diagram of the V dfc mean difference between the healthy subject group and the patient group with disturbance of consciousness in a window provided by an embodiment of the present invention.

[0045] Figure 6The classification effect diagram of large sample data of the brain region dynamic coefficient of variation feature provided by an implementation manner of the embodiment of the present invention.

[0046] Figure 7 The principle block diagram of the brain functional magnetic resonance image classification device provided by the embodiment of the present invention.

[0047] Figure 8 The internal structure principle block diagram of the intelligent terminal provided by the embodiment of the present invention. Detailed implementation manners

[0048] The present invention discloses a method, device, terminal and storage medium for classifying brain functional magnetic resonance images. To make the objectives, technical solutions and effects of the present invention clearer and more definite, the following further describes the present invention in detail with reference to the accompanying drawings and by way of examples. It should be understood that the specific examples described herein are only used to explain the present invention and are not used to limit the present invention.

[0049] Those skilled in the art of the present technology can understand that unless specifically stated otherwise, the singular forms "a", "an", "the" and "said" used herein may also include the plural forms. It should be further understood that the term "including" used in the specification of the present invention means the presence of the described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or their groups. It should be understood that when we say that an element is "connected" or "coupled" to another element, it can be directly connected or coupled to other elements, or there may also be intermediate elements. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The phrase "and / or" used herein includes all or any unit and all combinations of one or more related listed items.

[0050] Those skilled in the art of the present technology can understand that unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as the general understanding of those of ordinary skill in the art to which the present invention belongs. It should also be understood that terms such as those defined in a general dictionary should be understood to have a meaning consistent with the meaning in the context of the prior art, and will not be interpreted in an idealized or overly formal sense unless specifically defined as here.

[0051] In the existing technology, the methods for analyzing resting-state functional magnetic resonance images are mainly divided into two categories, namely static and dynamic functional connectivity analysis. Among them, static functional connectivity analysis assumes that the activity state of the brain is relatively stable within several minutes, and measures the resonance degree between different brain regions through the Pearson correlation coefficient. Based on the static analysis method, according to different spatial scales, the activities of specific brain regions, specific functional networks, and the whole-brain functional network of the brain can be measured. In the study of disorders of consciousness, by establishing a linear classification model, some studies have found that the activity of the default network can distinguish different consciousness states (accuracy rate = 85.3%). In addition, in addition to specific functional networks (such as the default network), it has been found that other functional networks in the whole brain, especially the primary auditory and visual networks, can distinguish patients in the minimally conscious state and the "vegetative" state with an accuracy rate of 80%. However, the above studies are mainly based on static analysis methods, and compare the consciousness levels by calculating the correlations between brain regions over the entire period of time. With the in-depth study, researchers have found that the brain also exhibits specific activity patterns within several minutes, and this pattern can be detected by means of dynamic functional connectivity. The methods of dynamic functional connectivity mainly include sliding time window analysis, co-activation pattern analysis, etc. Through k-means clustering, the dynamic functional connectivity method reduces the dimensionality of the continuous brain spatial activity patterns in the time dimension, and finally obtains several representative brain activity patterns. Among them, the sliding time window analysis is based on a time window of a specific length, constructs the brain functional network connection matrix one by one, and uses the k-means method for clustering, and finally obtains several representative functional matrices. The co-activation pattern analysis directly performs k-means clustering based on each time sampling point, and finally obtains several representative brain activity patterns. Based on the above methods, researchers have found that the decline in the consciousness level is manifested as the breakdown of the dynamic balance of the brain functional network. At the same time, by means of dynamic connectivity, some studies have pointed out that the decline in the consciousness level is related to the decline in the functional diversity and information integration ability of the posterior midline brain regions of the brain. Defects of existing calculations: For the diagnosis of patients with disorders of consciousness, most of the existing technologies adopt static functional connectivity analysis methods, ignoring the dynamic activity characteristics of the patients' brains in the time dimension. Although a few studies have adopted dynamic functional connectivity analysis methods, they all use k-means clustering to reduce the dimensionality of the brain activity patterns in the time dimension, and the final representative features obtained are also the arithmetic average of the clustering results, and then perform between-group analysis.Therefore, the deficiencies of the above methods are manifested in three aspects: First, the static analysis method ignores the dynamic activity characteristic information of the brain and is insufficient to comprehensively explore the brain activity basis supporting the normal consciousness level; Second, the previous methods perform clustering based on k-means, and the number of clusters is often determined by the researcher, which is difficult to avoid subjectivity; Finally, in the current research on dynamic functional connectivity of disorders of consciousness, the sample size is small, and verification based on large-sample data is required.

[0052] To solve the problems of the existing technology, this embodiment provides a method for classifying brain functional magnetic resonance images. By calculating the dynamic coefficient of variation of local brain regions, the detection method is comprehensive, and the classification model is constructed based on a large-sample data set, which can achieve more accurate image classification results. Specifically, in implementation, first, the first resting-state functional magnetic resonance image of a patient with a disorder of consciousness and the second resting-state functional magnetic resonance image of a healthy subject are respectively obtained; then, the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image are preprocessed, and based on a preset brain atlas template, the preprocessed first resting-state functional magnetic resonance image and second resting-state functional magnetic resonance image are segmented into brain regions to obtain several brain regions, and the average time series signal of each brain region is extracted; then, based on the sliding time window method, the correlation coefficient of the average time series signals between brain regions is calculated, and based on the correlation coefficient, a dynamic functional connectivity matrix is constructed; finally, based on the dynamic functional connectivity matrix, the dynamic coefficient of variation of each brain region under the cross-time window is calculated, and based on the dynamic coefficient of variation of each brain region under the cross-time window and the classification model constructed based on the large-sample data set, the category of the brain functional magnetic resonance image is obtained.

[0053] Exemplary method

[0054] This embodiment provides a method for classifying brain functional magnetic resonance images, and this method can be applied to an intelligent terminal for medical imaging. Specifically, as Figure 1-2 shown, the method includes:

[0055] Step S100, respectively obtain the first resting-state functional magnetic resonance image of a patient with a disorder of consciousness and the second resting-state functional magnetic resonance image of a healthy subject;

[0056] Specifically, the resting-state functional magnetic resonance image can be obtained by using a nuclear magnetic resonance scanner. In order to classify the subsequent detection results, the present invention will obtain the resting-state functional magnetic resonance image of a patient with a disorder of consciousness and the resting-state functional magnetic resonance image of a healthy subject.

[0057] After obtaining the resting-state functional magnetic resonance images of the patient with a disorder of consciousness and the healthy subject, the following steps can be executed as Figure 1-2The following steps are shown: S200, preprocessing the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image, and performing brain region segmentation on the preprocessed first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image based on a preset brain map template to obtain multiple brain regions, and extracting the average time series signal of each brain region;

[0058] Specifically, the first and second resting-state fMRI images can be preprocessed using the SPM12 (Statistical Parametric Mapping) software package, and the signal-to-noise ratio of the preprocessed first and second resting-state fMRI images is improved. Because the first and second resting-state fMRI images represent whole-brain imaging data, brain region segmentation is required for subsequent, more detailed analysis. In this embodiment, the preset brain map template is a standard brain map derived from the ICA results of the HCP public dataset, which includes 32 brain regions. Therefore, the first and second resting-state fMRI images are segmented into several brain regions based on the 32 brain regions in the standard brain map, and the average time series signal of each brain region is extracted to prepare for the subsequent construction of the dynamic functional connectivity matrix.

[0059] In one implementation, the following step may be performed before step S200: screening the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image.

[0060] Specifically, in order to eliminate the impact of severe structural damage on signal quality and data analysis, data in which brain structures are obviously deformed, distorted, or missing in the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image are eliminated to complete the screening of the above data.

[0061] In step S200, the preprocessing of the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image includes the following steps: removing preset time points, inter-layer time correction, head motion correction, image registration, image segmentation, spatial normalization, denoising and signal normalization are performed on the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image.

[0062] Specifically, the preprocessing process includes: removing the first 5 time points, interslice time correction, head motion correction, structural-functional image registration, image segmentation, spatial normalization, removing the influence of physiological noise (average signals of cerebrospinal fluid and white matter), head motion (head motion parameters in 6 dimensions), band-pass (0.01 - 0.08 Hz) filtering, and signal normalization. After preprocessing both the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image, the influence caused by subject head motion, equipment noise, and physiological noise during data acquisition can be minimized, and the signal-to-noise ratio of the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image can be improved. Among them, removing the images of the first 5 time points is considered due to the uneven magnetic field when the magnetic resonance machine starts. Since magnetic resonance data is scanned slice by slice during acquisition, there are time differences between different slices, so interpolation method is needed for interslice correction. In addition, since the subject will have head motion during the magnetic resonance experiment, in order to exclude the influence of sampling points with large head motion, head motion correction needs to be performed based on the reference time point. Finally, the individual functional imaging data is projected into the standard MNI (Montreal Neurological Institute) space, that is, signal normalization is performed, preparing for subsequent between-group comparison and large-sample data classification.

[0063] In step S200, the extracting the average time series signal of each brain region includes the following steps: extracting the time series of each voxel in each brain region; averaging the time series of all voxels in each brain region to obtain the average time series signal of each brain region.

[0064] Specifically, a voxel is a volume element, which is the smallest unit for digital data segmentation in three-dimensional space. Therefore, a brain region can contain multiple voxels. Extract the time series of each voxel in each brain region of the first resting-state functional magnetic resonance image, and extract the time series of each voxel in each brain region of the second resting-state functional magnetic resonance image. It should be noted that the above time series is the residual series after preprocessing and regression of noise. Finally, average the time series of all voxels in each brain region of the first resting-state functional magnetic resonance image to obtain the average time series signal of each brain region of the first resting-state functional magnetic resonance image; average the time series of all voxels in each brain region of the second resting-state functional magnetic resonance image to obtain the average time series signal of each brain region of the second resting-state functional magnetic resonance image.

[0065] After obtaining the average time series signal of each brain region of the first resting-state functional magnetic resonance image and the average time series signal of each brain region of the second resting-state functional magnetic resonance image, the following can be performed as Figure 1-2The following steps are shown: S300. Based on the sliding time window method, calculate the correlation coefficient of the average time series signals between brain regions, and construct a dynamic functional connectivity matrix according to the correlation coefficient.

[0066] Specifically, the sliding time window method is a commonly used method for constructing a dynamic functional connectivity matrix. According to the window size (w) and the step size (s), the original time series is divided into several overlapping (w≥s) or non-overlapping (w≤s) time periods. In this embodiment, in each window (w), calculate the correlation coefficient of the average time series signals between pairwise brain regions in the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image. The correlation coefficient can be the Pearson correlation coefficient. Finally, the correlation coefficients can be constructed into a dynamic functional connectivity matrix.

[0067] In step S300, the method of calculating the correlation coefficient of the average time series signals between brain regions based on the sliding time window includes the following steps: Obtain the window size and window step of the sliding time window; Based on the window size and the window step, divide the average time series signal of each brain region into several time windows; For each time window, calculate the Pearson correlation coefficient between brain regions in each time window.

[0068] Specifically, the window size and window step of the sliding time window can be set manually and input into the system. The system obtains the window size w and window step s, and uses the rule of overlapping (w≥s) or non-overlapping (w≤s) to divide the average time series signal of each brain region into several time windows. It should be noted that the division method of the average time series signal of each brain region in the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image is the same. For each time window, calculate the Pearson correlation coefficient between pairwise brain regions in each time window. In this embodiment, with a window length of 22 TRs (44 seconds) and a step size of 1 TR (2 seconds), when the original time series length is 235 TRs, a total of 214 consecutive windows are obtained. After division by the sliding time window, the original time series matrix of 32 brain regions (32×235) is converted into (32×214) consecutive time windows. Based on the time series of each window, calculate the Pearson correlation coefficient between pairwise of the 32 brain regions. In any window, the Pearson correlation coefficient between pairwise brain regions is calculated by the following formula:

[0069]

[0070] where x and y are two brain regions respectively, n is the number of brain regions (such as 32), and the value of r ranges from -1 to 1. The closer the value is to -1, the more negative correlation is shown between the two brain regions. On the contrary, the closer the value is to 1, the more positive correlation is shown between the brain regions.

[0071] In step S300, constructing the dynamic functional connectivity matrix according to the correlation coefficient includes the following steps: using the Pearson correlation coefficient between brain regions in each time window as the first matrix; and forming the dynamic functional connectivity matrix by all the first matrices corresponding to all time windows.

[0072] Specifically, using the Pearson correlation coefficient between brain regions in each time window as the first matrix means that the calculated correlation coefficients between any two of the 32 brain regions form a 32×32 correlation matrix, which is called the first matrix. Then, forming the dynamic functional connectivity matrix by all the first matrices corresponding to all time windows. That is to say, in the time dimension, based on the sliding time window method, 214 32×32 first matrices, i.e., the dynamic functional connectivity matrix, can be obtained.

[0073] In one implementation, after constructing the dynamic functional connectivity matrix according to the correlation coefficient, the following steps are included: calculating the similarity between the upper triangular regions of the first matrices in the dynamic functional connectivity matrix to obtain the similarity coefficient; and calculating the dynamic stability value for each time window based on the similarity coefficient.

[0074] Specifically, to improve the computational efficiency, the dynamic functional connectivity matrix (214×32×32) is converted into a two-dimensional matrix (214×496), where 496 = 32×(32 - 1) / 2, that is, taking the upper triangular elements of the functional connectivity matrix for each window. Here, the first dimension of the matrix is the number of windows (214), and the second dimension is the correlation coefficients between any two of the 32 brain regions. To measure the similarity of the functional connectivity matrices of different windows, a similarity matrix of the dynamic functional connectivity two-dimensional matrix is further constructed based on the correlation coefficient. For the dynamic functional connectivity matrices of any two windows, their similarity coefficient (S) can be measured by the following formula:

[0075] S(t w1 ,t w2 ) = corr[TriUpper(dFC(t w1 )), TriUpper(dFC(t w2 ))]

[0076] where S(t w1 ,t w2 ) refers to the similarity coefficient between the upper triangular regions of the dynamic functional connectivity matrices of window w1 and window w2, which measures the similarity degree between the upper triangular regions of the dynamic functional connectivity matrices of any two windows. The larger the value, the higher the similarity. dFC(t w1 ) and dFC(t w2) refer to the dynamic functional connectivity matrices under windows w1 and w2 respectively, and the similarity degree between the two is realized by calculating the correlation coefficient of their upper triangular matrices. Therefore, after calculating the similarity coefficient, the two-dimensional dynamic functional connectivity matrix (214*496) based on the sliding time window is converted into a similarity matrix of dynamic functional connectivity (214*214). Applying the similarity matrix to the dataset of large-sample patients with disorders of consciousness, the results are as Figure 3-4 , where Figure 3 is the similarity matrix of the dynamic functional connectivity of healthy subjects, Figure 4 are the similarity matrices of the dynamic functional connectivity of patients with disorders of consciousness respectively. Based on the results of the similarity matrix, it can be obtained that for the brains of subjects with intact consciousness, their dynamic functional connectivity shows good continuity in the time dimension, that is, there is a high correlation coefficient between the functional connectivities of different windows. On the contrary, for the brains of patients with disorders of consciousness with impaired consciousness levels, their dynamic functional connectivity has poor continuity in time, that is, the correlation coefficient is low. Therefore, from the perspective of the whole brain, the differences in dynamic functional connectivity similarity at different consciousness levels can be intuitively seen.

[0077] Then, based on the similarity coefficient, the dynamic stability value of each time window is calculated. In practice, in order to further quantify the stability of the dynamic functional connectivity matrix of the whole brain at any time window, the dynamic stability value of any time window is calculated based on the following formula:

[0078] V dfc (t w1 ) = 1 - S(t w1 , t w2 )

[0079] where, for the dynamic functional connectivity matrix of any time window, its dynamic stability value V dfc (t w1 ) is defined as 1 minus the similarity coefficient between this time window and the adjacent time window. The larger the value of V dfc (t w1 ), the worse the dynamic stability of the time window t w1 , and its functional network structure will be converted in the next window; vice versa. As Figure 5 shows the inter-group differences in dynamic stability V dfc between different consciousness levels (healthy subjects and patients with disorders of consciousness). The results show that this index shows good discrimination at different consciousness levels, that is, the value of the whole-brain dynamic functional connectivity V dfc of patients with disorders of consciousness is significantly higher than that of healthy subjects. According to the definition, that is, the stability of the whole-brain dynamic functional network of patients with disorders of consciousness is poor. This result has been well verified in the large-sample data in the embodiments.

[0080] After obtaining the similarity coefficient, the following steps as shown in Figure 1-2 can be performed: S400. Based on the dynamic functional connectivity matrix, calculate the dynamic coefficient of variation of each brain region across time windows, and obtain the category of the brain functional magnetic resonance image according to the dynamic coefficient of variation of each brain region across time windows and the classification model constructed based on the large sample data set.

[0081] Specifically, in order to further calculate the degree of dynamic functional connectivity variation of the brain region, based on the constructed dynamic functional connectivity matrix (214 * 32 * 32), extract the correlation coefficients of brain region j with all other brain regions in each window, that is, the row vectors of the correlation matrix, to form the dynamic matrix (214 * 1 * 32) of the functional connectivity pattern of this brain region. Further, calculate the dynamic coefficient of variation of each brain region j across time windows based on the following formula:

[0082]

[0083] where TV j is the dynamic coefficient of variation of brain region j, which measures the stability of the dynamic connectivity pattern of the brain region across time windows. FC w1 (j, :) and FC w2 (j, :) are the row vectors of the correlation coefficients of brain region j in time window w1 and time window w2. Finally, TV j takes the mean of 1 minus the correlation coefficients of the row vectors of brain region j between any two windows. The larger the TV j value, the lower the similarity of the functional connectivity pattern of brain region j in the time dimension, that is, the higher the time variation degree, and vice versa. After obtaining the dynamic coefficient of variation of each brain region across time windows, the category of the brain functional magnetic resonance image can be obtained according to the dynamic coefficient of variation of each brain region across time windows and the classification model constructed based on the large sample data set. Correspondingly, the step of obtaining the category of the brain functional magnetic resonance image according to the dynamic coefficient of variation of each brain region across time windows and the classification model constructed based on the large sample data set includes the following steps: obtaining data samples; based on the support vector machine classification algorithm, taking the dynamic coefficient of variation of each brain region across time windows as the classification vector, classifying the data samples through the classification model constructed based on the large sample data set to obtain the category of the brain functional magnetic resonance image, and using the cross-validation method to test the classification model constructed based on the large sample data set.

[0084] Specifically, the classification model can be a binary classifier. First, data samples are obtained. Based on the sklearn toolkit, the support vector machine classification algorithm is used, and the dynamic coefficient of variation of each brain region j under the cross-time window is used as the feature vector to train the binary classifier so that it can distinguish between patients with disorders of consciousness and healthy subjects. Before training the binary classifier, the features are screened based on statistical comparison (independent samples t-test). During the classifier training process, based on the obtained data samples, the feature vectors of healthy subject samples are labeled as 1, and the feature vectors of patients with disorders of consciousness are labeled as 2. The linear support vector machine is used for model training and testing. Subsequently, the leave-one-out cross-validation method is used to verify the effect of the binary classifier. One data sample is taken out as the test set in turn, and the remaining data is used as the training set. If the classifier correctly judges the healthy subject sample as class 1 and the patient with disorders of consciousness sample as class 2, it is a correct distinction. Finally, the mean value of the correct rates of all cross-validation tests is used as the correct rate of the binary classifier classification. Further, the classifier in the present invention is applied to a large sample dataset of patients with disorders of consciousness (n = 61). Different levels of consciousness can be distinguished by this classifier. The results show that the correct rate of the classification model reaches 90%, that is, among 61 samples, 55 are correctly identified by the model, fully verifying that the correct rate of the classification model constructed based on the large sample dataset in the present invention is very high, as Figure 6 shown, and more accurate image classification results of patients with disorders of consciousness can be obtained.

[0085] Exemplary device

[0086] As Figure 7 shown in, an embodiment of the present invention provides a brain functional magnetic resonance image classification device, which includes an image acquisition module 501, a segmentation and extraction module 502, a correlation coefficient calculation and construction module 503, and a dynamic coefficient of variation calculation and image category obtaining module 504, wherein:

[0087] The image acquisition module 501 is configured to respectively acquire a first resting-state functional magnetic resonance image of a patient with a disorder of consciousness and a second resting-state functional magnetic resonance image of a healthy subject;

[0088] The segmentation and extraction module 502 is configured to preprocess the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image, and perform brain region segmentation on the preprocessed first resting-state functional magnetic resonance image and second resting-state functional magnetic resonance image based on a preset brain atlas template to obtain a plurality of brain regions, and extract the average time series signal of each brain region;

[0089] The correlation coefficient calculation and construction module 503 is configured to calculate the correlation coefficient of the average time series signals between brain regions based on the sliding time window method, and construct a dynamic functional connection matrix according to the correlation coefficient;

[0090] The dynamic coefficient of variation calculation and image category obtaining module 504 is configured to calculate the dynamic coefficient of variation of each of the brain regions across time windows based on the dynamic functional connectivity matrix, and obtain the category of the functional magnetic resonance image of the brain according to the dynamic coefficient of variation of each of the brain regions across time windows and a classification model constructed based on a large sample data set.

[0091] Based on the above embodiments, the present invention further provides an intelligent terminal, the principle block diagram of which can be as Figure 8 shown. The intelligent terminal includes a processor, a memory, a network interface, a display screen, and a temperature sensor connected through a system bus. Among them, the processor of the intelligent terminal is used to provide computing and control capabilities. The memory of the intelligent terminal includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the intelligent terminal is used to communicate with an external terminal through a network connection. The computer program, when executed by the processor, implements a method for classifying functional magnetic resonance images of the brain. The display screen of the intelligent terminal can be a liquid crystal display screen or an electronic ink display screen, and the temperature sensor of the intelligent terminal is pre-set inside the intelligent terminal to detect the operating temperature of internal devices.

[0092] Those skilled in the art can understand that Figure 8 the schematic diagram in

[0093] merely shows the block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the intelligent terminal to which the solution of the present invention is applied. The specific intelligent terminal may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0094] Obtain the first resting-state functional magnetic resonance image of a patient with consciousness disorder and the second resting-state functional magnetic resonance image of a healthy subject respectively;

[0095] Preprocess the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image, and perform brain region segmentation on the preprocessed first resting-state functional magnetic resonance image and second resting-state functional magnetic resonance image based on a preset brain atlas template to obtain a plurality of brain regions, and extract the average time series signal of each of the brain regions;

[0096] Based on the sliding time window method, calculate the correlation coefficient of the average time series signals between brain regions, and construct a dynamic functional connectivity matrix according to the correlation coefficient;

[0097] Based on the dynamic functional connectivity matrix, calculate the dynamic coefficient of variation of each brain region across time windows, and obtain the category of the brain functional magnetic resonance image according to the dynamic coefficient of variation of each brain region across time windows and the classification model constructed based on the large sample data set.

[0098] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided by the present invention can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0099] In summary, the present invention discloses a method, apparatus, terminal, and storage medium for classifying brain functional magnetic resonance images. The method includes: respectively obtaining a first resting-state functional magnetic resonance image of a patient with disorders of consciousness and a second resting-state functional magnetic resonance image of a healthy subject; preprocessing the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image, and performing brain region segmentation on the preprocessed first resting-state functional magnetic resonance image and second resting-state functional magnetic resonance image based on a preset brain atlas template to obtain a plurality of brain regions, and extracting the average time series signal of each brain region; calculating the correlation coefficient of the average time series signals between brain regions based on the sliding time window method, and constructing a dynamic functional connectivity matrix according to the correlation coefficient; calculating the dynamic coefficient of variation of each brain region under cross-time windows based on the dynamic functional connectivity matrix, and obtaining the category of the brain functional magnetic resonance image according to the dynamic coefficient of variation of each brain region under cross-time windows and a classification model constructed based on a large sample data set. The embodiments of the present invention calculate the dynamic characteristics of local brain regions, making the detection method comprehensive. The classification model is constructed based on a large sample data set, and a more accurate image classification result can be achieved.

[0100] Based on the above embodiments, the present invention discloses a method for classifying brain functional magnetic resonance images. It should be understood that the application of the present invention is not limited to the above examples. For those of ordinary skill in the art, improvements or transformations can be made according to the above description. All such improvements and transformations should fall within the protection scope of the appended claims of the present invention.

Claims

1. A method for classifying functional magnetic resonance imaging of the brain, characterized in that, The method includes: Obtaining a first resting-state functional magnetic resonance image of a patient with disorders of consciousness and a second resting-state functional magnetic resonance image of a healthy subject respectively; Preprocessing the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image, and performing brain region segmentation on the preprocessed first resting-state functional magnetic resonance image and second resting-state functional magnetic resonance image based on a preset brain atlas template to obtain a plurality of brain regions, and extracting the average time series signal of each brain region; Based on the sliding time window method, calculating the correlation coefficient of the average time series signals between brain regions, and constructing a dynamic functional connectivity matrix according to the correlation coefficient; Based on the dynamic functional connectivity matrix, calculating the dynamic coefficient of variation of each brain region under cross-time windows, and obtaining the category of the functional magnetic resonance image of the brain according to the dynamic coefficient of variation of each brain region under cross-time windows and a classification model constructed based on a large sample data set; The constructing a dynamic functional connectivity matrix according to the correlation coefficient includes: Taking the Pearson correlation coefficient between brain regions in each time window as the first matrix; Composing all the first matrices corresponding to all time windows into a dynamic functional connectivity matrix; After the constructing a dynamic functional connectivity matrix according to the correlation coefficient, it includes: converting the dynamic functional connectivity matrix into a two-dimensional matrix, the first dimension of the two-dimensional matrix is the number of windows, the second dimension is the correlation coefficient between every two brain regions, calculating the similarity between the upper triangular regions of the first matrices in the dynamic functional connectivity matrix to obtain a similarity coefficient; calculating the dynamic stability value of each time window based on the similarity coefficient; For the dynamic functional connectivity matrices of any two windows, their similarity coefficient can be measured by the following formula: S(t w1 ,t w2 ) = corr[TriUpper(dFC(t w1 )), TriUpper(dFC(t w2 ))], Among them, S(t w1 , t w2 ) refers to the similarity coefficient between the upper triangular region of the dynamic functional connectivity matrix of window w1 and window w2, which measures the similarity degree between the upper triangular regions of the dynamic functional connectivity matrices of any two windows. The larger the value, the higher the similarity; dFC(t w1 ) and dFC(t w2 ) respectively refer to the dynamic functional connectivity matrices under window w1 and window w2, and the similarity degree between the two is realized by calculating the correlation coefficient of their upper triangular matrices; Calculating the dynamic stability value of any time window based on the following formula: V dfc (t w1 ) = 1 - S(t w1 , t w2 ) Among them, for the dynamic functional connectivity matrix of any time window, its dynamic stability value V dfc (t w1 ) is defined as 1 minus the similarity coefficient between this time window and the adjacent time window; the larger the value of V dfc (t w1 ), the worse the dynamic stability of time window t w1 , and its functional network structure will be transformed in the next window; vice versa; the calculation formula for the dynamic coefficient of variation of each brain region across time windows is as follows: Among them, TV j That is, the dynamic coefficient of variation of brain region j, which measures the stability of the dynamic connection pattern of the brain region across time windows, FC w1 (j, :) and FC w2 (j, :) are the row vectors of the correlation coefficients of brain region j in time window w1 and time window w2. TV j It takes 1 minus the mean of the row vector correlation coefficients between any two windows of brain region j. TV j The larger the value, the lower the similarity of the functional connection pattern of brain region j in the time dimension, that is, the higher the degree of temporal variation, and vice versa; Before the preprocessing the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image, and performing brain region segmentation on the preprocessed first resting-state functional magnetic resonance image and second resting-state functional magnetic resonance image based on a preset brain atlas template to obtain a plurality of brain regions, and extracting the average time series signal of each brain region, it includes: Screening the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image, and removing the data in the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image in which obvious deformation, distortion or absence of brain structure occurs; The calculating the correlation coefficient of the average time series signals between brain regions based on the sliding time window method includes: Obtaining the window size and window step length of the sliding time window, and the window size and window step length of the sliding time window are set manually and input into the system; Based on the window size and the window step length, dividing the average time series signal of each brain region into a plurality of time windows in a manner of coincidence or non-coincidence; For each time window, calculating the Pearson correlation coefficient between brain regions in each time window; In any window, the Pearson correlation coefficient between every two brain regions is calculated by the following formula: Wherein, x and y are two brain regions respectively, n is the number of brain regions, and the value of r ranges from -1 to 1. The closer the value is to -1, the more negative correlation is shown between the two brain regions. On the contrary, the closer the value is to 1, the more positive correlation is shown between the brain regions. The categories of the functional magnetic resonance imaging obtained according to the dynamic coefficient of variation of each brain region under the cross-time window and the classification model constructed based on the large sample data set include: Obtain data samples; The classification model is a binary classifier. Based on the sklearn toolkit, using the support vector machine classification algorithm, the dynamic coefficient of variation of each brain region under the cross-time window is used as the classification feature vector to train the binary classifier so that it can distinguish between patients with disorders of consciousness and healthy subjects. Before training the binary classifier, the features are screened based on statistical comparison. During the training process of the binary classifier, based on the obtained data samples, the feature vectors of the healthy subject samples are labeled as 1, and the feature vectors of the patients with disorders of consciousness samples are labeled as 2. The linear support vector machine is used for model training and testing. The leave-one-out cross-validation method is used to verify the effect of the binary classifier. One data sample is taken out as the test set in turn, and the remaining data is used as the training set. If the binary classifier correctly judges the healthy subject samples as category 1 and the patients with disorders of consciousness samples as category 2, it is a correct distinction. The average value of the correct rates of all cross-validation tests is used as the correct rate of the binary classifier classification.

2. The brain functional magnetic resonance image classification method according to claim 1, wherein The preprocessing of the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image includes: Both the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image are subjected to removing preset time points, interlayer time correction, head motion correction, image registration, image segmentation, spatial normalization, denoising processing, and signal normalization processing.

3. The brain functional magnetic resonance image classification method according to claim 1, wherein The extraction of the average time series signal of each brain region includes: Extract the time series of each voxel in each brain region; Average the time series of all voxels in each brain region to obtain the average time series signal of each brain region.

4. A brain functional magnetic resonance image classification device, characterized in that, The device includes: An image acquisition module for respectively acquiring the first resting-state functional magnetic resonance image of a patient with disorders of consciousness and the second resting-state functional magnetic resonance image of a healthy subject; A segmentation and extraction module for preprocessing the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image, and performing brain region segmentation on the preprocessed first resting-state functional magnetic resonance image and second resting-state functional magnetic resonance image based on a preset brain atlas template to obtain a number of brain regions, and extracting the average time series signal of each brain region; A correlation coefficient calculation and construction module for calculating the correlation coefficient of the average time series signals between brain regions based on the sliding time window method, and constructing a dynamic functional connection matrix according to the correlation coefficient. A dynamic coefficient of variation calculation and image category obtaining module, which is used to calculate the dynamic coefficient of variation of each brain region under cross-time windows based on the dynamic functional connectivity matrix, and obtain the category of the brain functional magnetic resonance image according to the dynamic coefficient of variation of each brain region under cross-time windows and a classification model constructed based on a large sample data set; The constructing of the dynamic functional connectivity matrix according to the correlation coefficient includes: Taking the Pearson correlation coefficient between brain regions in each time window as the first matrix; Composing all the first matrices corresponding to all time windows into a dynamic functional connectivity matrix; After constructing the dynamic functional connectivity matrix according to the correlation coefficient, it includes: converting the dynamic functional connectivity matrix into a two-dimensional matrix, where the first dimension of the two-dimensional matrix is the number of windows, and the second dimension is the correlation coefficient between every two brain regions, calculating the similarity between the upper triangular regions of the first matrices in the dynamic functional connectivity matrix to obtain a similarity coefficient; calculating the dynamic stability value of each time window based on the similarity coefficient; For the dynamic functional connectivity matrices of any two windows, their similarity coefficient can be measured by the following formula: S(t w1 ,t w2 ) = corr[TriUpper(dFC(t w1 )), TriUpper(dFC(t w2 ))], Among them, S(t w1 , t w2 ) refers to the similarity coefficient between the upper triangular region of the dynamic functional connectivity matrix of window w1 and window w2, which measures the similarity degree between the upper triangular regions of the dynamic functional connectivity matrices of any two windows. The larger the value, the higher the similarity; dFC(t w1 ) and dFC(t w2 ) respectively refer to the dynamic functional connectivity matrices under window w1 and window w2, and their similarity degree is realized by calculating the correlation coefficient of their upper triangular matrices; Calculating the dynamic stability value of any time window based on the following formula: V dfc (t w1 ) = 1 - S(t w1 , t w2 ), Among them, for the dynamic functional connectivity matrix of any time window, its dynamic stability value V dfc (t w1 ) is defined as 1 minus the similarity coefficient between this time window and the adjacent time window; the larger the value of V dfc (t w1 ), the worse the dynamic stability of time window t w1 , and its functional network structure will be transformed in the next window; vice versa; The formula for calculating the dynamic coefficient of variation of each brain region under cross-time windows is: Among them, TV j That is, the dynamic coefficient of variation of brain region j, which measures the stability of the dynamic connection pattern of the brain region across time windows, FC w1 (j, :) and FC w2 (j, :) are the row vectors of the correlation coefficients of brain region j in time window w1 and time window w2, TV j It takes the mean of 1 minus the correlation coefficients of the row vectors between any two windows of brain region j, TV j The larger the value, the lower the similarity of the functional connection pattern of brain region j in the time dimension, that is, the higher the degree of temporal variation, and vice versa; Before the segmentation and extraction module, it includes: Screening the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image, and removing the data with obvious deformation, distortion or missing brain structures in the first resting-state functional magnetic resonance image and the second resting-state functional magnetic resonance image; The calculating of the correlation coefficient of the average time series signals between brain regions based on the sliding time window method includes: Obtaining the window size and window step of the sliding time window, and the window size and window step of the sliding time window are set manually and input into the system; Dividing the average time series signals of each brain region into several time windows based on the window size and the window step in a manner of overlapping or non-overlapping; For each time window, calculating the Pearson correlation coefficient between brain regions in each time window; In any window, the Pearson correlation coefficient between every two brain regions is calculated by the following formula: Where x and y are two brain regions respectively, n is the number of brain regions, and the value of r ranges from -1 to 1. The closer the value is to -1, the more negative correlation is shown between the two brain regions. On the contrary, the closer the value is to 1, the more positive correlation is shown between the brain regions; The obtaining of the category of the brain functional magnetic resonance image according to the dynamic coefficient of variation of each brain region under cross-time windows and a classification model constructed based on a large sample data set includes: Obtaining data samples; The classification model is a binary classifier. Based on the sklearn toolkit and using the support vector machine classification algorithm, taking the dynamic coefficient of variation of each brain region under cross-time windows as a classification feature vector to train the binary classifier so that it can distinguish between patients with disorders of consciousness and healthy subjects; Before the binary classifier training, features are screened based on statistical comparison; during the binary classifier training process, based on the obtained data samples, the feature vectors of healthy subject samples are labeled as 1, and the feature vectors of patients with disturbance of consciousness samples are labeled as 2, and a linear support vector machine is used for model training and testing; the leave-one-out cross-validation method is used to verify the effect of the binary classifier, and one data sample is taken out as the test set in turn, and the remaining data is used as the training set; if the binary classifier correctly judges the healthy subject sample as class 1 and the patient with disturbance of consciousness sample as class 2, it is a correct distinction; the average value of the correct rates of all cross-validation tests is used as the correct rate of the binary classifier classification.

5. An intelligent terminal, characterized in that, It includes a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors. The one or more programs include instructions for executing the method according to any one of claims 1-3.

6. A non - transitory computer - readable storage medium, characterized in that, When the instructions in the storage medium are executed by the processor of the electronic device, the electronic device is enabled to execute the method according to any one of claims 1-3.

Images