CN117122303B - Brain network prediction method, system, equipment and storage medium - Google Patents
Brain network prediction method, system, equipment and storage medium Download PDFInfo
- Publication number
- CN117122303B CN117122303B CN202311394088.5A CN202311394088A CN117122303B CN 117122303 B CN117122303 B CN 117122303B CN 202311394088 A CN202311394088 A CN 202311394088A CN 117122303 B CN117122303 B CN 117122303B
- Authority
- CN
- China
- Prior art keywords
- processing
- linear
- network
- brain
- subjected
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 210000004556 brain Anatomy 0.000 title claims abstract description 122
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000012545 processing Methods 0.000 claims abstract description 151
- 239000011159 matrix material Substances 0.000 claims abstract description 67
- 230000006870 function Effects 0.000 claims abstract description 66
- 230000000877 morphologic effect Effects 0.000 claims abstract description 50
- 238000000605 extraction Methods 0.000 claims abstract description 38
- 238000002599 functional magnetic resonance imaging Methods 0.000 claims abstract description 34
- 238000003384 imaging method Methods 0.000 claims abstract description 31
- 230000008859 change Effects 0.000 claims abstract description 13
- 230000014509 gene expression Effects 0.000 claims description 16
- 238000012937 correction Methods 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 9
- 239000000126 substance Substances 0.000 claims description 7
- 230000009466 transformation Effects 0.000 claims description 6
- 238000009499 grossing Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 3
- 238000011160 research Methods 0.000 abstract description 6
- 230000003925 brain function Effects 0.000 abstract description 3
- 238000013421 nuclear magnetic resonance imaging Methods 0.000 abstract description 3
- 238000013528 artificial neural network Methods 0.000 description 12
- 238000005516 engineering process Methods 0.000 description 4
- 238000002610 neuroimaging Methods 0.000 description 4
- 238000002595 magnetic resonance imaging Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000009792 diffusion process Methods 0.000 description 2
- 201000010099 disease Diseases 0.000 description 2
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 2
- 239000003814 drug Substances 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 230000008451 emotion Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000013140 knowledge distillation Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/05—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves
- A61B5/055—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/0033—Features or image-related aspects of imaging apparatus classified in A61B5/00, e.g. for MRI, optical tomography or impedance tomography apparatus; arrangements of imaging apparatus in a room
- A61B5/004—Features or image-related aspects of imaging apparatus classified in A61B5/00, e.g. for MRI, optical tomography or impedance tomography apparatus; arrangements of imaging apparatus in a room adapted for image acquisition of a particular organ or body part
- A61B5/0042—Features or image-related aspects of imaging apparatus classified in A61B5/00, e.g. for MRI, optical tomography or impedance tomography apparatus; arrangements of imaging apparatus in a room adapted for image acquisition of a particular organ or body part for the brain
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/20—ICT specially adapted for the handling or processing of medical images for handling medical images, e.g. DICOM, HL7 or PACS
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B2576/00—Medical imaging apparatus involving image processing or analysis
- A61B2576/02—Medical imaging apparatus involving image processing or analysis specially adapted for a particular organ or body part
- A61B2576/026—Medical imaging apparatus involving image processing or analysis specially adapted for a particular organ or body part for the brain
Abstract
The invention relates to the technical field of image processing, in particular to a prediction method, a system, equipment and a storage medium of a brain network, wherein the prediction method is characterized in that imaging group characteristics and morphological brain networks of T1 weighted imaging are subjected to linear processing, characteristic extraction, residual error connection processing and the like, and after the obtained depth characteristic matrix is subjected to pearson correlation processing, the depth characteristic matrix and the initial morphological brain networks are subjected to linear processing, characteristic extraction, residual error connection, multi-line change processing and the like, so that a prediction function connectivity network with rich node characteristics and network topology information is obtained, the topology structure and connection strength value of the brain function connectivity network are predicted and learned better, the accurate prediction from the T1 weighted imaging morphological brain network to the functional magnetic resonance imaging function connection network is realized, the prediction result with higher reality is obtained, and the application of structural nuclear magnetic resonance imaging and functional magnetic resonance imaging in clinical practice is enhanced, so that the prediction method has important significance to brain science research.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a brain network prediction method, a brain network prediction system, brain network prediction equipment and a brain network storage medium.
Background
Functional brain network research serves as an important branch in the field of neuroscience, and aims to deeply explore the interrelationship between different areas of the brain and the roles of the interrelationship in cognition, emotion, behavior and the like. In the last decades, brain network research has made great progress as neuroimaging technology has evolved. Among them, functional magnetic resonance imaging (functional magnetic resonance imaging, fMRI) has become an important tool for studying brain functional connectivity networks (Functional Connectivity Network, FCN) as a non-invasive neuroimaging technique. The brain network technology based on fMRI has high application value or potential in various researches such as excavating brain working mechanism, disease analysis and the like.
However, fMRI is not a clinically routine acquisition sequence, and there is no such data in many hospitals or many disease diagnoses, resulting in a long-felt clinical popularization of fMRI sequences. With the rise of graph roll-up networks, predictions of different brain networks are made possible. At present, a lot of work is done from diffusion magnetic resonance imaging (diffusion magnetic resonance imaging, dMRI) to FCN prediction, but the accuracy of the prediction result of the method is lower, and the error between the prediction result and the real FCN is larger, so that the method is not beneficial to the clinical popularization and application of fMRI brain network technology.
Disclosure of Invention
The invention aims to provide a prediction method, a prediction system, prediction equipment and a storage medium of a brain network.
The technical scheme of the invention is as follows:
a method of predicting a brain network, comprising the operations of:
s1, acquiring image histology characteristics and morphological brain network of brain through T1 weighted imaging, wherein the image histology characteristics are subjected to linear processing to obtain linear image histology characteristics; the linear image histology characteristics, the linear image histology characteristics subjected to the first characteristic extraction processing and the morphological brain network are subjected to residual connection processing to obtain a depth characteristic matrix;
s2, performing pearson correlation processing on the depth feature matrix to obtain depth feature connection; the depth feature connection is subjected to linear processing to obtain linear depth feature connection; the linear depth feature connection, the linear depth feature connection subjected to the second feature extraction processing and the morphological brain network are subjected to residual connection processing and multi-linear change processing to obtain a prediction function connectivity network;
s3, acquiring functional magnetic resonance imaging of the brain, and acquiring a target functional connectivity network based on the functional magnetic resonance imaging; and the prediction function connectivity network learns the characteristics of the target function connectivity network, and when the characteristic loss entropy between the prediction function connectivity network and the target function connectivity network is smaller than a threshold value, a prediction result is output.
The operation of the first feature extraction in S1 specifically includes: constructing a graph structure based on the linear image histology characteristics and the morphological brain network; and acquiring important edge expression and a high-dimensional feature matrix in the graph structure, and executing the residual error connection processing operation in the S1 after parameter loss and nonlinear processing.
After the operation of S1, the method further includes: the linear depth feature matrix obtained after the linear processing of the depth feature matrix is subjected to residual connection processing with the linear depth feature matrix subjected to the third feature extraction processing and the morphological brain network to obtain an optimized depth feature matrix; the optimized depth feature matrix is used for executing the operation in S2.
The operation of acquiring the image histology characteristics and morphological brain network of the brain in the S1 and T1 weighted imaging specifically comprises the following steps: after the T1 weighted imaging is subjected to linear registration and nonlinear registration treatment, extracting texture features and strength features of different brain regions to obtain the image histology features; and the morphological brain network is obtained by the image group chemical characteristics through pearson correlation processing and L2, 1-mode processing.
The operation of the multi-linear change processing in S2 specifically includes: and sequentially performing first linear processing, first nonlinear processing, first parameter losing processing, second nonlinear processing, second parameter losing processing and second linear processing on the input subjected to residual connection processing to obtain the prediction function connectivity network.
The operation of obtaining the target function connectivity network based on the function magnetic resonance imaging in the step S3 specifically comprises the following steps: and performing time layer correction, head motion correction, spatial registration normalization and smoothing on the functional magnetic resonance imaging, extracting time sequence features of different brain regions, and performing pearson correlation processing and L2, 1-mode processing to obtain the target functional connectivity network.
The feature loss entropy in the S3 is obtained based on the mean square error loss, the average absolute value error loss, the global loss and the regional average pearson loss of the prediction function connectivity network and the target function connectivity network.
A prediction system for a brain network, comprising:
the depth feature matrix generation module is used for acquiring image histology features of the brain and morphological brain network of T1 weighted imaging, and the image histology features are subjected to linear processing to obtain linear image histology features; the linear image histology characteristics, the linear image histology characteristics subjected to the first characteristic extraction processing and the morphological brain network are subjected to residual connection processing to obtain a depth characteristic matrix;
the prediction function connectivity network generation module is used for obtaining depth feature connection through the pearson correlation processing of the depth feature matrix; the depth feature connection is subjected to linear processing to obtain linear depth feature connection; the linear depth feature connection, the linear depth feature connection subjected to the second feature extraction processing and the morphological brain network are subjected to residual connection processing and multi-linear change processing to obtain a prediction function connectivity network;
the learning module and the prediction result generation module are used for acquiring functional magnetic resonance imaging of the brain and obtaining a target functional connectivity network based on the functional magnetic resonance imaging; and the prediction function connectivity network learns the characteristics of the target function connectivity network, and when the characteristic loss entropy between the prediction function connectivity network and the target function connectivity network is smaller than a threshold value, a prediction result is output.
The brain network prediction device comprises a processor and a memory, wherein the processor executes a computer program stored in the memory to realize the brain network prediction method.
A computer readable storage medium storing a computer program, wherein the computer program when executed by a processor implements the above-described brain network prediction method.
The invention has the beneficial effects that:
the invention provides a brain network prediction method, which is characterized in that the image group chemical characteristics and morphological brain network of T1 weighted imaging are subjected to linear processing, characteristic extraction, residual error connection and the like, the obtained depth characteristic matrix is subjected to Pearson correlation processing, and then is subjected to linear processing, characteristic extraction, residual error connection, multi-linear change processing and the like with the initial morphological brain network, so that a prediction function connectivity network with rich node characteristics and network topology information is obtained, the topology structure and connection strength value of the brain function connectivity network are predicted and learned better, the accurate prediction from the T1 weighted imaging morphological brain network to the functional magnetic resonance imaging functional connection network is realized, the prediction result with higher reality is obtained, the application of structural nuclear magnetic resonance imaging and functional magnetic resonance imaging in clinical practice is enhanced, and the brain network prediction method has important significance to brain science research.
Drawings
The aspects and advantages of the present application will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention.
In the drawings:
FIG. 1 is a flow chart of a prediction method in an embodiment.
Detailed Description
Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings.
The present embodiment provides a brain network prediction method, referring to fig. 1, including the following operations:
s1, acquiring image histology characteristics and morphological brain network of brain through T1 weighted imaging, wherein the image histology characteristics are subjected to linear processing to obtain linear image histology characteristics; the linear image histology characteristics, the linear image histology characteristics subjected to the first characteristic extraction processing and the morphological brain network are subjected to residual connection processing to obtain a depth characteristic matrix;
s2, performing pearson correlation processing on the depth feature matrix to obtain depth feature connection; the depth feature connection is subjected to linear processing to obtain linear depth feature connection; the linear depth feature connection, the linear depth feature connection subjected to the second feature extraction processing and the morphological brain network are subjected to residual connection processing and multi-linear change processing to obtain a prediction function connectivity network;
s3, acquiring functional magnetic resonance imaging of the brain, and acquiring a target functional connectivity network based on the functional magnetic resonance imaging; and the prediction function connectivity network learns the characteristics of the target function connectivity network, and when the characteristic loss entropy between the prediction function connectivity network and the target function connectivity network is smaller than a threshold value, a prediction result is output.
S1, acquiring image histology characteristics and morphological brain network of brain through T1 weighted imaging, and obtaining linear image histology characteristics after linear processing of the image histology characteristics; and carrying out residual connection processing on the linear image histology characteristics, the linear image histology characteristics subjected to the first characteristic extraction processing and the morphological brain network to obtain a depth characteristic matrix.
T1 weighted imaging is acquired. First, T1 weighted imaging (T1 WI) of the brain in digital imaging and communications in medicine (Digital Imaging and Communications in Medicine, DICOM) is acquired, then the T1 weighted imaging in DICOM format is converted to a three-dimensional image in neuroimaging informatics initiative (Neuroimaging Informatics Technology Initiative, nifi) format, and patient privacy data is deleted for subsequent processing analysis.
The operation of acquiring the image histology characteristics and morphological brain network of the T1 weighted imaging is as follows: after T1 weighted imaging is subjected to linear registration and nonlinear registration treatment, texture features and strength features of different brain regions are extracted, and image histology features are obtained; the morphological brain network is obtained by the image group characteristics through the Pearson correlation processing and the L2 and 1 paradigm processing.
The T1 weighted imaging also comprises N4 deviation correction processing before the linear registration processing. The method comprises the following steps: after N4 deviation correction is carried out on the T1 weighted imaging, linear registration and nonlinear registration are carried out on the T1 weighted imaging to an MNI152 standard space, and the texture characteristics and the intensity characteristics of 90 brain areas and 25-dimensional brain areas corresponding to each T1 weighted imaging after registration are extracted by utilizing the brain area position information in the AAL map, so that image histology characteristics are obtained; then, the image group is subjected to pearson correlation processing and L2,1 paradigm processing to obtain a morphological brain network.
The first feature extraction may be performed by constructing a graph structure based on the linear image histology features and the morphological brain network; and after the graph structure is subjected to node information aggregation processing, the operation of residual connection processing in S1 is executed.
The operation of the first feature extraction may further be: constructing a graph structure based on linear image histology characteristics and morphological brain network; and acquiring important edge expression and a high-dimensional feature matrix in the graph structure, and executing the operation of residual connection processing in S1 after parameter loss and nonlinear processing. The method comprises the following steps: taking a 90×25 matrix of the linear image histology feature and a 90×90 matrix of the morphological brain network as node features and connectivity features respectively, constructing a graph structure G= (A, X), wherein A is the node feature, and X is the connectivity feature; the important edge expression of the adjacency matrix in the graph structure G is obtained by the following calculation formula:，/>for important edge expression, D is an angle matrix; meanwhile, the connectivity characteristics in the graph structure G are subjected to high-dimensional linear transformation to obtain a high-dimensional characteristic matrix, and a calculation formula is as follows: />，/>For the high-dimensional feature matrix, linear () is a high-dimensional Linear layer, R is a set, n is the dimension of the input high-dimensional Linear layer, +.>Is vector dimension after linear transformation; then, the important edge expression and the high-dimensional feature matrix are subjected to parameter loss (which can be realized by a Dropout layer in the neural network) and nonlinearityAfter processing (which can be realized by a ReLU layer in a neural network), residual connection processing is carried out on the linear image group characteristics, and then a depth characteristic matrix is obtained.
S2, performing pearson correlation processing on the depth feature matrix to obtain depth feature connection; the depth feature connection is subjected to linear processing to obtain linear depth feature connection; and performing residual connection processing and multi-linear change processing on the linear depth feature connection, the linear depth feature connection subjected to the second feature extraction processing and the morphological brain network to obtain a prediction function connectivity network.
To further enrich the information of the depth feature matrix, the operation of S1 further includes: performing residual connection processing on the linear depth feature matrix obtained after linear processing of the depth feature matrix, the linear depth feature matrix subjected to third feature extraction processing and a morphological brain network to obtain an optimized depth feature matrix; the optimized depth feature matrix is used to perform the operation in S2.
The operation of the multiple linear variation process may be: and (3) carrying out first linear processing (which can be realized by a linear layer in the neural network), nonlinear processing (which can be realized by a ReLU layer in the neural network), parameter loss processing (which can be realized by a Dropout layer in the neural network) and second linear processing (which can be realized by a linear layer in the neural network) on the input subjected to residual connection processing in sequence to obtain the prediction function connectivity network.
The operation of the multiple linear variation process may also be: and (3) carrying out first linear processing (which can be realized by a linear layer in a neural network), first nonlinear processing (which can be realized by a ReLU layer in the neural network), first parameter loss processing (which can be realized by a Dropout layer in the neural network), second nonlinear processing (which can be realized by a ReLU layer in the neural network), second parameter loss processing (which can be realized by a Dropout layer in the neural network) and second linear processing (which can be realized by a linear layer in the neural network) on the input subjected to residual connection processing in sequence to obtain the prediction function connectivity network.
In addition, after the operation of the residual connection processing in S2, further includes: performing linear processing on the input subjected to residual connection processing to obtain linear optimized depth characteristic connection; and performing residual connection processing and multi-linear change processing on the linear optimized depth feature connection, the linear optimized depth feature connection subjected to the second feature extraction processing and the morphological brain network, wherein the obtained optimized prediction function connectivity network is used for executing the operation of S3.
The second feature extraction and the third feature extraction are the same as the first feature extraction, and include constructing a graph structure, acquiring important edge expressions and high-dimensional feature matrices, losing parameters and nonlinear processing, which are not repeated here for the sake of space.
S3, acquiring functional magnetic resonance imaging of the brain, and acquiring a target functional connectivity network based on the functional magnetic resonance imaging; the prediction function connectivity network learns the characteristics of the target function connectivity network, and when the characteristic loss entropy between the prediction function connectivity network and the target function connectivity network is smaller than a threshold value, a prediction result is output.
Based on the functional magnetic resonance imaging, the operation of obtaining the target functional connectivity network is specifically as follows: performing time layer correction, head motion correction, spatial registration normalization and smoothing on the functional magnetic resonance imaging, extracting time sequence features of different brain regions, and performing pearson correlation processing and L2, 1-range processing to obtain a target functional connectivity network. Specifically, after temporal correction, head motion correction, spatial registration normalization and smoothing are carried out on the acquired functional magnetic resonance imaging by utilizing an spm12 tool, the brain region positions of the AAL map are utilized to extract the time sequence features of 90 brain region positions corresponding to the smoothed functional magnetic resonance imaging, and then pearson correlation processing and L2 and 1-range processing are carried out to obtain the target functional connectivity network.
And then, taking the target function connectivity network as a teacher network, taking the prediction function connectivity network as a student network, and commanding the prediction function connectivity network to learn the characteristics of the target function connectivity network by using a knowledge distillation method, and outputting a prediction result, namely the trained prediction function connectivity network, when the characteristic loss entropy between the prediction function connectivity network and the target function connectivity network is smaller than a threshold value.
The feature loss entropy is obtained based on the mean square error loss, the average absolute value error loss, the global loss and the regional average pearson loss of the prediction function connectivity network and the target function connectivity network.
The method can be obtained by the following calculation formula:
，
L NS the entropy is lost for the features,L MSE in order to be a mean square error loss,L L1 in order to average the absolute value error loss,L WPCC in the event of a global loss,L MRPCC is the regional average pearson loss. Wherein the method comprises the steps ofs i Represents the firstiIndividual predictive FCNs, i.e. predictive functional connectivity networks,y i first, theiThe true FCN, i.e. the target functional connectivity network,andrepresenting true and predicted individualsrFCN connections for individual brain regions,Cov()andσ()which represent covariance calculation and standard deviation calculation, respectively.
To demonstrate the accuracy of the prediction method provided in this example, the following experiment was performed.
Experimental setup. The dataset was at 6:2: the ratio of 2 is randomly divided into training, validation and test sets. The experimental development environment used was pytorch1.9.0, 200 runs were performed on the NVIDIA RTX 2080 graphics processor during the model training period, an Adam optimizer was used, the initial learning rate was batch size set to 4, the proposed model was on the NVIDIA RTX 2080 graphics processor, the model was trained for about 30 minutes, and the experimental environment and specific settings are shown in table 1.
Table 1 summary of experimental parameter settings
。
In the experiment, the mean square error, the pearson correlation and the peak signal to noise ratio are used as evaluation indexes, and the effect of the prediction method provided by the embodiment is evaluated, and the effect of the prediction method provided by the embodiment is better as shown in table 2.
TABLE 2 prediction result parameter Table of the present embodiment
。
The embodiment also provides a prediction system of a brain network, including:
the depth feature matrix generation module is used for acquiring image histology features and morphological brain networks of the brain through T1 weighted imaging, and obtaining linear image histology features after the image histology features are subjected to linear processing; performing residual connection processing on the linear image histology characteristics, the linear image histology characteristics subjected to the first characteristic extraction processing and the morphological brain network to obtain a depth characteristic matrix;
the prediction function connectivity network generation module is used for obtaining depth feature connection through the pearson correlation processing of the depth feature matrix; the depth feature connection is subjected to linear processing to obtain linear depth feature connection; the linear depth feature connection, the linear depth feature connection subjected to the second feature extraction processing and the morphological brain network are subjected to residual connection processing and multi-linear change processing to obtain a prediction function connectivity network;
the learning module and the prediction result generation module are used for acquiring functional magnetic resonance imaging of the brain and obtaining a target functional connectivity network based on the functional magnetic resonance imaging; the prediction function connectivity network learns the characteristics of the target function connectivity network, and when the characteristic loss entropy between the prediction function connectivity network and the target function connectivity network is smaller than a threshold value, a prediction result is output.
The embodiment also provides a prediction device of the brain network, which comprises a processor and a memory, wherein the prediction method of the brain network is realized when the processor executes the computer program stored in the memory.
The present embodiment also provides a computer-readable storage medium storing a computer program, wherein the computer program when executed by a processor implements the above-described brain network prediction method.
The embodiment provides a prediction method of a brain network, which is characterized in that the image group chemical characteristics and morphological brain network of T1 weighted imaging are subjected to linear processing, characteristic extraction, residual connection and the like, the obtained depth characteristic matrix is subjected to pearson correlation processing, and then is subjected to linear processing, characteristic extraction, residual connection, multi-linear change processing and the like with the initial morphological brain network, so that a prediction function connectivity network with rich node characteristics and network topology information is obtained, the topology structure and connection strength value of the brain function connectivity network are predicted and learned better, the accurate prediction from the T1 weighted imaging morphological brain network to the functional magnetic resonance imaging functional connection network is realized, the prediction result with higher reality is obtained, the application of structural nuclear magnetic resonance imaging and functional magnetic resonance imaging in clinical practice is enhanced, and the prediction method has important significance to brain science research.
Claims (7)
1. A method of predicting a brain network, comprising the operations of:
s1, acquiring image histology characteristics and morphological brain network of brain through T1 weighted imaging, wherein the image histology characteristics are subjected to linear processing to obtain linear image histology characteristics; the linear image histology characteristics, the linear image histology characteristics subjected to the first characteristic extraction processing and the morphological brain network are subjected to residual connection processing to obtain a depth characteristic matrix;
the first feature extraction operation is as follows: constructing a graph structure based on the linear image histology characteristics and the morphological brain network; acquiring important edge expression and a high-dimensional feature matrix in the graph structure, and executing the operation of residual error connection processing in the S1 after parameter loss and nonlinear processing; the method comprises the following steps: matrix of linear image group morphology feature, and morphology brain netThe matrix of the complex is respectively used as node characteristics and connectivity characteristics, a graph structure G= (A, X) is constructed, A is the node characteristics, and X is the connectivity characteristics; the important edge expression of the adjacent matrix in the graph structure G is obtained by the following calculation formula:，/>for important edge expression, D is an angle matrix; meanwhile, the connectivity features in the graph structure G are subjected to high-dimensional linear transformation to obtain a high-dimensional feature matrix, and a calculation formula is as follows: />，/>For the high-dimensional feature matrix, linear () is a high-dimensional Linear layer, R is a set, n is the dimension of the input high-dimensional Linear layer, +.>Is vector dimension after linear transformation; then, carrying out parameter loss and nonlinear processing on the important edge expression and the high-dimensional feature matrix, and carrying out residual connection processing on the important edge expression and the high-dimensional feature matrix and the linear image group chemical feature to obtain the depth feature matrix;
the operation of acquiring the image histology characteristics and morphological brain network of the brain through T1 weighted imaging is specifically as follows: after the T1 weighted imaging is subjected to linear registration and nonlinear registration treatment, extracting texture features and strength features of different brain regions to obtain the image histology features; the morphological brain network is obtained through the Pearson correlation processing and L2,1 paradigm processing of the image group chemical characteristics;
s2, performing pearson correlation processing on the depth feature matrix to obtain depth feature connection; the depth feature connection is subjected to linear processing to obtain linear depth feature connection; the linear depth feature connection, the linear depth feature connection subjected to the second feature extraction processing and the morphological brain network are subjected to residual connection processing and multi-linear change processing to obtain a prediction function connectivity network;
the operation of the second feature extraction process is the same as the operation of the first feature extraction process;
s3, acquiring functional magnetic resonance imaging of the brain, and acquiring a target functional connectivity network based on the functional magnetic resonance imaging; the prediction function connectivity network learns the characteristics of the target function connectivity network, and when the characteristic loss entropy between the prediction function connectivity network and the target function connectivity network is smaller than a threshold value, a prediction result is output; the prediction result is a prediction function connectivity network;
the feature loss entropy is obtained based on a mean square error loss, an average absolute value error loss, a global loss and a regional average pearson loss of the prediction function connectivity network and the target function connectivity network.
2. The method for predicting a brain network according to claim 1, further comprising, after the operation of S1:
the linear depth feature matrix obtained after the linear processing of the depth feature matrix is subjected to residual connection processing with the linear depth feature matrix subjected to the third feature extraction processing and the morphological brain network to obtain an optimized depth feature matrix;
the optimized depth feature matrix is used for executing the operation in the step S2;
the operation of the third feature extraction process is the same as the operation of the first feature extraction process.
3. The method for predicting a brain network according to claim 1, wherein the operation of the multilingual change process in S2 is specifically: and sequentially performing first linear processing, first nonlinear processing, first parameter losing processing, second nonlinear processing, second parameter losing processing and second linear processing on the input subjected to residual connection processing to obtain the prediction function connectivity network.
4. The method for predicting a brain network according to claim 1, wherein the operation of obtaining the target functional connectivity network based on the functional magnetic resonance imaging in S3 is specifically:
and performing time layer correction, head motion correction, spatial registration normalization and smoothing on the functional magnetic resonance imaging, extracting time sequence features of different brain regions, and performing pearson correlation processing and L2, 1-mode processing to obtain the target functional connectivity network.
5. A prediction system for a brain network, comprising:
the depth feature matrix generation module is used for acquiring image histology features of the brain and morphological brain network of T1 weighted imaging, and the image histology features are subjected to linear processing to obtain linear image histology features; the linear image histology characteristics, the linear image histology characteristics subjected to the first characteristic extraction processing and the morphological brain network are subjected to residual connection processing to obtain a depth characteristic matrix; the first feature extraction operation is as follows: constructing a graph structure based on the linear image histology characteristics and the morphological brain network; acquiring important edge expression and a high-dimensional feature matrix in the graph structure, and executing the operation of residual error connection processing in the S1 after parameter loss and nonlinear processing; the method comprises the following steps: taking a matrix of linear image histology characteristics and a matrix of morphological brain network as node characteristics and connectivity characteristics respectively, constructing a graph structure G= (A, X), wherein A is the node characteristics, and X is the connectivity characteristics; the important edge expression of the adjacent matrix in the graph structure G is obtained by the following calculation formula:，/>for important edge expression, D is an angle matrix; meanwhile, the connectivity characteristics in the graph structure G are subjected to high-dimensional linear transformation to obtain a high-dimensional characteristic matrix, and a calculation formula is obtainedThe method comprises the following steps: />，/>For the high-dimensional feature matrix, linear () is a high-dimensional Linear layer, R is a set, n is the dimension of the input high-dimensional Linear layer, +.>Is vector dimension after linear transformation; then, carrying out parameter loss and nonlinear processing on the important edge expression and the high-dimensional feature matrix, and carrying out residual connection processing on the important edge expression and the high-dimensional feature matrix and the linear image group chemical feature to obtain the depth feature matrix; the operation of acquiring the image histology characteristics and morphological brain network of the brain through T1 weighted imaging is specifically as follows: after the T1 weighted imaging is subjected to linear registration and nonlinear registration treatment, extracting texture features and strength features of different brain regions to obtain the image histology features; the morphological brain network is obtained through the Pearson correlation processing and L2,1 paradigm processing of the image group chemical characteristics;
the prediction function connectivity network generation module is used for obtaining depth feature connection through the pearson correlation processing of the depth feature matrix; the depth feature connection is subjected to linear processing to obtain linear depth feature connection; the linear depth feature connection, the linear depth feature connection subjected to the second feature extraction processing and the morphological brain network are subjected to residual connection processing and multi-linear change processing to obtain a prediction function connectivity network; the operation of the second feature extraction process is the same as the operation of the first feature extraction process;
the learning module and the prediction result generation module are used for acquiring functional magnetic resonance imaging of the brain and obtaining a target functional connectivity network based on the functional magnetic resonance imaging; the prediction function connectivity network learns the characteristics of the target function connectivity network, and when the characteristic loss entropy between the prediction function connectivity network and the target function connectivity network is smaller than a threshold value, a prediction result is output; the prediction result is a prediction function connectivity network; the feature loss entropy is obtained based on a mean square error loss, an average absolute value error loss, a global loss and a regional average pearson loss of the prediction function connectivity network and the target function connectivity network.
6. A prediction device of a brain network, comprising a processor and a memory, wherein the processor implements the prediction method of a brain network according to any one of claims 1-4 when executing a computer program stored in the memory.
7. A computer readable storage medium for storing a computer program, wherein the computer program when executed by a processor implements the method of predicting a brain network according to any one of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311394088.5A CN117122303B (en) | 2023-10-26 | 2023-10-26 | Brain network prediction method, system, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311394088.5A CN117122303B (en) | 2023-10-26 | 2023-10-26 | Brain network prediction method, system, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117122303A CN117122303A (en) | 2023-11-28 |
| CN117122303B true CN117122303B (en) | 2024-01-30 |
Family
ID=88858508
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311394088.5A Active CN117122303B (en) | 2023-10-26 | 2023-10-26 | Brain network prediction method, system, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117122303B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117350352B (en) * | 2023-12-06 | 2024-02-23 | 烟台大学 | Learning method, system and equipment from structural brain network to functional connectivity network |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106874680A (en) * | 2017-02-22 | 2017-06-20 | 天津天士力医药商业有限公司 | It is a kind of to detect the brain network analysis method that blood-nourishing and brain-refreshing ball improves cognitive function |
| CN110473635A (en) * | 2019-08-14 | 2019-11-19 | 电子科技大学 | A kind of analysis method of teenager's brain structural network and brain function cyberrelationship model |
| CN112690777A (en) * | 2021-01-13 | 2021-04-23 | 常州大学 | Neurological disorder diagnosis system based on state transition dynamic brain network algorithm |
| CN114255228A (en) * | 2021-12-23 | 2022-03-29 | 福州大学 | Brain function connection network analysis method based on resting state function magnetic resonance image |
| CN114391826A (en) * | 2022-01-24 | 2022-04-26 | 西北大学 | Human characterization prediction method and device based on edge-driven graph neural network |
| CN114418982A (en) * | 2022-01-14 | 2022-04-29 | 太原理工大学 | Method for constructing DTI multi-parameter fusion brain network |
| CN114984457A (en) * | 2022-05-17 | 2022-09-02 | 复旦大学 | Patient light stimulation parameter selection method and system based on brain function network |
| CN115082311A (en) * | 2022-06-17 | 2022-09-20 | 烟台大学 | Single image super-resolution reconstruction method based on depth attention residual error network |
| CN116312965A (en) * | 2022-09-07 | 2023-06-23 | 常州大学 | Cognitive function state prediction method based on multi-mode brain network |
| CN116807447A (en) * | 2023-08-31 | 2023-09-29 | 北京智精灵科技有限公司 | Brain age prediction modeling method, cognition improving method and system of dynamic brain network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114376558B (en) * | 2022-03-24 | 2022-07-19 | 之江实验室 | Brain atlas individuation method and system based on magnetic resonance and twin map neural network |
-
2023
- 2023-10-26 CN CN202311394088.5A patent/CN117122303B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106874680A (en) * | 2017-02-22 | 2017-06-20 | 天津天士力医药商业有限公司 | It is a kind of to detect the brain network analysis method that blood-nourishing and brain-refreshing ball improves cognitive function |
| CN110473635A (en) * | 2019-08-14 | 2019-11-19 | 电子科技大学 | A kind of analysis method of teenager's brain structural network and brain function cyberrelationship model |
| CN112690777A (en) * | 2021-01-13 | 2021-04-23 | 常州大学 | Neurological disorder diagnosis system based on state transition dynamic brain network algorithm |
| CN114255228A (en) * | 2021-12-23 | 2022-03-29 | 福州大学 | Brain function connection network analysis method based on resting state function magnetic resonance image |
| CN114418982A (en) * | 2022-01-14 | 2022-04-29 | 太原理工大学 | Method for constructing DTI multi-parameter fusion brain network |
| CN114391826A (en) * | 2022-01-24 | 2022-04-26 | 西北大学 | Human characterization prediction method and device based on edge-driven graph neural network |
| CN114984457A (en) * | 2022-05-17 | 2022-09-02 | 复旦大学 | Patient light stimulation parameter selection method and system based on brain function network |
| CN115082311A (en) * | 2022-06-17 | 2022-09-20 | 烟台大学 | Single image super-resolution reconstruction method based on depth attention residual error network |
| CN116312965A (en) * | 2022-09-07 | 2023-06-23 | 常州大学 | Cognitive function state prediction method based on multi-mode brain network |
| CN116807447A (en) * | 2023-08-31 | 2023-09-29 | 北京智精灵科技有限公司 | Brain age prediction modeling method, cognition improving method and system of dynamic brain network |
Non-Patent Citations (2)
| Title |
|---|
| Deep Reinforcement Learning Guided Graph Neural Networks for Brain Network Analysis;Xusheng Zhao et.al;Neural Networks;第154卷;56-67 * |
| 基于网络拓扑结构的脑功能网络重构技术研究;李艺茹;万方数据库;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117122303A (en) | 2023-11-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112150478B (en) | Method and system for constructing semi-supervised image segmentation framework | |
| CN106951499B (en) | A kind of knowledge mapping representation method based on translation model | |
| CN107833629A (en) | Aided diagnosis method and system based on deep learning | |
| JP7373042B2 (en) | Brain function registration method based on graph model | |
| CN117122303B (en) | Brain network prediction method, system, equipment and storage medium | |
| CN113314205B (en) | Efficient medical image labeling and learning system | |
| CN111325750B (en) | Medical image segmentation method based on multi-scale fusion U-shaped chain neural network | |
| CN111161314B (en) | Target object position area determination method and device, electronic equipment and storage medium | |
| CN112837741B (en) | Protein secondary structure prediction method based on cyclic neural network | |
| CN109448703A (en) | In conjunction with the audio scene recognition method and system of deep neural network and topic model | |
| CN117172294B (en) | Method, system, equipment and storage medium for constructing sparse brain network | |
| Qiang et al. | Learning brain representation using recurrent Wasserstein generative adversarial net | |
| CN113449878B (en) | Data distributed incremental learning method, system, equipment and storage medium | |
| CN114241240A (en) | Method and device for classifying brain images, electronic equipment and storage medium | |
| CN111695570B (en) | Small sample lower semantic segmentation method based on variational prototype reasoning | |
| CN115937129B (en) | Method and device for processing left and right half brain relations based on multi-mode magnetic resonance image | |
| CN116188428A (en) | Bridging multi-source domain self-adaptive cross-domain histopathological image recognition method | |
| Goutham et al. | Brain tumor classification using Efficientnet-B0 model | |
| Liu et al. | Spatial-temporal convolutional attention for mapping functional brain networks | |
| CN117350352B (en) | Learning method, system and equipment from structural brain network to functional connectivity network | |
| CN117292232B (en) | Method, system and equipment for acquiring multidimensional space characteristics of T1 weighted imaging | |
| CN113255789A (en) | Video quality evaluation method based on confrontation network and multi-tested electroencephalogram signals | |
| CN114190884A (en) | Longitudinal analysis method, system and device for brain disease data | |
| Wei et al. | Open-Vocabulary Skeleton Action Recognition with Diffusion Graph Convolutional Network and Pre-Trained Vision-Language Models | |
| CN115274092A (en) | Alzheimer's disease identification method based on CBAM-CNN |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |