CN113786185A - Static brain network feature extraction method and system based on convolutional neural network - Google Patents
Static brain network feature extraction method and system based on convolutional neural network Download PDFInfo
- Publication number
- CN113786185A CN113786185A CN202111097531.3A CN202111097531A CN113786185A CN 113786185 A CN113786185 A CN 113786185A CN 202111097531 A CN202111097531 A CN 202111097531A CN 113786185 A CN113786185 A CN 113786185A
- Authority
- CN
- China
- Prior art keywords
- brain
- convolutional
- neural network
- convolutional neural
- network
- 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.)
- Granted
Links
- 210000004556 brain Anatomy 0.000 title claims abstract description 122
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 50
- 230000003068 static effect Effects 0.000 title claims abstract description 45
- 238000000605 extraction Methods 0.000 title claims abstract description 27
- 208000014644 Brain disease Diseases 0.000 claims abstract description 18
- 238000012545 processing Methods 0.000 claims abstract description 13
- 238000007477 logistic regression Methods 0.000 claims abstract description 11
- 238000003745 diagnosis Methods 0.000 claims abstract description 6
- 239000000284 extract Substances 0.000 claims abstract description 6
- 238000000034 method Methods 0.000 claims description 30
- 230000006870 function Effects 0.000 claims description 16
- 238000002599 functional magnetic resonance imaging Methods 0.000 claims description 6
- 230000003993 interaction Effects 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical group [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 claims description 3
- 239000008280 blood Substances 0.000 claims description 3
- 238000002790 cross-validation Methods 0.000 claims description 3
- 229910052760 oxygen Inorganic materials 0.000 claims description 3
- 239000001301 oxygen Substances 0.000 claims description 3
- 230000000284 resting effect Effects 0.000 claims description 2
- 238000013135 deep learning Methods 0.000 description 9
- 239000000523 sample Substances 0.000 description 8
- 238000012549 training Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 208000024827 Alzheimer disease Diseases 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 238000002610 neuroimaging Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 206010003805 Autism Diseases 0.000 description 1
- 208000020706 Autistic disease Diseases 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 230000001490 effect on brain Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000002595 magnetic resonance imaging Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003012 network analysis Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 239000013610 patient sample Substances 0.000 description 1
- 238000002379 ultrasonic velocimetry Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 210000004885 white matter Anatomy 0.000 description 1
Images
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
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Animal Behavior & Ethology (AREA)
- Public Health (AREA)
- Pathology (AREA)
- Radiology & Medical Imaging (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Veterinary Medicine (AREA)
- High Energy & Nuclear Physics (AREA)
- Neurology (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Psychiatry (AREA)
- Signal Processing (AREA)
- Image Analysis (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention discloses a static brain network feature extraction method based on a convolutional neural network, which comprises the following steps: constructing a static brain network by calculating a pearson coefficient between each brain region; processing the static brain network by adopting a convolutional neural network with parameter size of [32,32,64,32] to extract corresponding brain region characteristics; connecting two convolutional layers behind the convolutional neural network, wherein the sizes of convolutional cores of the two convolutional layers are 32 and 16 respectively, and the step length is 116; and (3) sending the brain area features with the dimension of 1x32 after the convolutional layer processing into two full-connection layers with the sizes of 64 and 32 respectively to continuously extract the features, and then adopting a SoftMax logistic regression function to diagnose and classify the brain diseases. The invention can learn more characteristics with discriminative power and interpretability, can obviously improve the brain disease classification performance and has better classification performance for brain disease diagnosis.
Description
Technical Field
The invention relates to the technical field of machine learning and medical image processing, in particular to a static brain network feature extraction method and system based on a convolutional neural network.
Background
With the rapid development of the current biotechnology, brain imaging technologies, such as the modern Magnetic Resonance Imaging (MRI) technology including functional MRI (fMRI), provide a non-invasive way to explore the human brain, revealing the mechanism of brain structure and function that was not previously known. Brain network analysis can depict the interaction of different brain regions of the brain on a connection level, and becomes a new research hotspot in medical image analysis and neuroimaging.
More recently, traditional methods of machine learning have been used in the analysis and classification of brain networks. For example, researchers have taken advantage of the brain network for early brain disease diagnosis and classification, and have achieved very good performance. In these studies, it is typical to extract local measures of brain regions (e.g., clustering coefficients, etc.) from the brain network as features for disease classification. The feature selection is to select an optimal subset of an original feature set from data by using a certain strategy, namely, according to certain criteria and methods, a k-dimensional feature subset which can provide more information for people is selected from the original d-dimensional feature set, actually, features which are useful for a specific task are selected from high-dimensional data, and redundant items in the data are discarded, so that the purpose of reducing the sample dimension is achieved, the occurrence of 'dimension disaster' is avoided, the learning of other subsequent target tasks is promoted, and the classification performance is improved. For example, Chen et al use the weights of edges as features for the classification of AD (Alzheimer's) and MCI (millicognititive impact). For example, the existing research utilizes white matter to construct a structural connection network from DTI images, and extracts 6 physiological parameters to construct 6 different structural networks for MCI classification, thereby obtaining good effect. Some studies have proposed the use of the LASSO model to construct a functionally connected network and are used for the study of autism in children. There are also studies that propose the use of the grouplsoso method to construct MCI and NC functionally connected networks and for classification. However, the previous methods of et al only analyze brain network connectivity in a patient sample set and are therefore not suitable for classification. Some previous methods have constructed connected networks with the same topology on the MCI and NC, so that the difference in topology between the patient network and the normal human network is ignored during the classification process. Since the locality measures only the characteristics of the local structure of the network, classification performance may be affected.
With the gradual fire heat of the deep learning method, the convolution neural network method is abnormally exploded in the image field. With ImageNet games, various improved representative of deep networks, such as AlexNet, VGG, GoogleNet, ResNet, etc., emerge. The models have excellent effects in image classification, and the classification precision exceeds the human discrimination precision once. Recently, related personnel propose a static brain network feature extraction method based on deep learning by combining with the characteristics of medical image data, and the method is applied to medical images.
Although the convolutional neural network is mostly used for classifying pictures, based on the characteristics of the pictures, the essence of the convolutional neural network is a matrix formed by a plurality of pixels, in the research on the brain based on rs-fMRI data, a common method is to abstract image type data into numerical type data and construct a brain network, which is similar to a functional brain network formed in a medical image, and in addition, the characteristics between a normal brain region and a patient brain region may be difficult to distinguish by extracting the characteristics through a traditional machine learning method, but the classification performance may be influenced because of local measurement.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a static brain network feature extraction method and a system based on a convolutional neural network. The invention overcomes the defects of the existing method, and further learns more discriminative and explanatory characteristics. The method can be used for the analysis and research of medical images and can also obviously improve the performance of brain disease classification theoretically. Finally, verification is carried out on a real brain disease data set, and experimental results show that compared with the traditional learning method, the method has better classification performance on brain disease diagnosis.
In order to achieve the purpose, the invention adopts the following technical scheme:
a static brain network feature extraction method based on a convolutional neural network comprises the following steps:
s1, dividing the brain space of each sample into 116 interested areas according to the automatic anatomical marking template, and forming a static brain network by calculating a Pearson coefficient between each brain area;
s2, processing the static brain network by adopting a convolutional neural network with the parameter size of [32,32,64,32] to extract corresponding brain area characteristics, and respectively extracting and obtaining brain area space characteristics and brain area marginalization characteristics after convolution and full connection; wherein, the number of channels of the convolutional neural network is [32,32], the parameters of the full connection layer are [64,32], and the convolution kernel is 116;
s3, connecting two convolutional layers behind the convolutional neural network, wherein the sizes of convolutional cores of the two convolutional layers are 32 and 16 respectively, the step length is 116, the parameter of the first convolutional layer is 1x116x32, and the parameter of the second convolutional layer is 1x1x 32; the two layers of convolution layers continue to carry out convolution operation on the similar features of different positions of the brain area space extracted in the step S2 for deep extraction;
and S4, sending the spatial features of the brain area with the dimension of 1x32 after the convolution layer processing in the step S3 into two full-connection layers with the sizes of 64 and 32 respectively to continuously extract the marginalized features of the brain area, transmitting the marginalized features of the brain area to the full-connection layers to contain the same features as the learning target, and then adopting a SoftMax logistic regression function to diagnose and classify the brain diseases.
In order to optimize the technical scheme, the specific measures adopted further comprise:
further, the convolutional neural network based static brain network feature extraction method according to claim 1, wherein in step S1, the pearson coefficient between each brain region is calculated according to the following formula:
F(i,j)＝pcc(xi,xj)
where pc denotes the correlation between two brain regions, xiAnd xjBlood oxygen signals representing brain region i and brain region j, respectively, and F (i, j) describes the interaction between brain region i and brain region j.
Further, the process of using SoftMax logistic regression function for diagnosis and classification of brain diseases comprises the following steps:
perform 3 classification tasks: emci, ncvs. l mci, and a 4 classification hcvs. emcivs. l mcivs.ad;
adopting a 5-fold cross validation strategy for each classification, and classifying by using a SoftMax classifier;
the precision of each fold is averaged to obtain the final precision.
Further, in step S4, the SoftMax logistic regression function is:
Sa＝ea/∑beb
wherein S isaThe probability of the a-th category among all the b categories is shown, wherein a represents the a-th category, and b represents a total b categories.
Further, the sample employs a resting state functional magnetic resonance imaging dataset.
The invention also provides a static brain network feature extraction system based on the convolutional neural network, which comprises the following steps:
the static brain network generation module is used for dividing the brain space of each sample into 116 interested areas according to the automatic anatomical marking template and forming a static brain network by calculating a Pearson coefficient between every two brain areas;
a convolutional neural network with parameter size [32,32,64,32] for processing the static brain network to extract corresponding brain region features; wherein, the number of channels of the convolutional neural network is [32,32], the parameters of the full connection layer are [64,32], and the convolution kernel is 116;
two convolutional layers connected behind the convolutional neural network, the sizes of convolutional cores of the two convolutional layers are respectively 32 and 16, the step length is 116, the parameter of the first convolutional layer is 1x116x32, and the parameter of the second convolutional layer is 1x1x 32; after convolution, spatial features of the brain areas are respectively extracted and used for identifying similar features of the brain areas at different spatial positions; the two layers of convolution layers continue to carry out convolution operation on the extracted similar features at different positions of the brain area space;
two fully-connected layers with the sizes of 64 and 32 respectively are used for continuously extracting the edge features of the brain area from the characteristics with the dimension of 1x32 after the convolution layer processing;
and the SoftMax classifier adopts a SoftMax logistic regression function to diagnose and classify the brain diseases.
The invention firstly constructs a static functional brain network, then extracts deeper features by using a deep learning convolutional neural network method, then performs learning training on the learned features in the convolutional neural network, and performs test training results on a test set to obtain better effect on brain disease classification.
Drawings
Fig. 1 is a flowchart of a static brain network feature extraction method based on a convolutional neural network according to an embodiment of the present invention.
Fig. 2 is a structural diagram of a convolutional neural network-based static brain network feature extraction system according to an embodiment of the present invention.
Fig. 3 is an emcivs. hc2 classification training diagram of an embodiment of the present invention.
Fig. 4 is an advs.hc2 classification training diagram of an embodiment of the present invention.
Fig. 5 is an advs.l mcivs.emcivs.hc4 classification training diagram according to an embodiment of the present invention.
Detailed Description
The present invention will now be described in further detail with reference to the accompanying drawings.
It should be noted that the terms "upper", "lower", "left", "right", "front", "back", etc. used in the present invention are for clarity of description only, and are not intended to limit the scope of the present invention, and the relative relationship between the terms and the terms is not limited by the technical contents of the essential changes.
Example one
Fig. 1 is a flowchart of a convolutional neural network-based static brain network feature extraction method according to an embodiment of the present invention, in this embodiment, for a scanned image in a medical data set, a static functional brain network is formed by correlations between different brain regions through Pearson coefficients, then the static functional brain network is trained in a constructed convolutional neural network, and finally, a SoftMax function is used on a test set for brain disease classification by using a trained model. Referring to fig. 1, the extraction method specifically includes the following steps:
step one, for all samples, a static brain network (SFCN) is formed by calculating a correlation coefficient between each brain region through a pearson coefficient (PCC).
In the first step, a plurality of samples of fMRI scanning are obtained, for each scanned sample, a brain space of each sample is divided into 116 regions of interest (ROIs) according to an Automatic Anatomical Labeling (AAL) template, and a static functional brain network is constructed by using a pearson correlation coefficient, which is used for measuring the correlation between two brain regions. The constructed static brain network is a weighted fully connected network. The formula is as follows:
F(i,j)＝pcc(xi,xj)
where pc denotes the correlation between two brain regions (PCC is used in this study), xiAnd xjBlood Oxygen (BOLD) signals representing brain regions i and j, respectively, F (i, j) describe the interaction between each brain region. Thus, given a time sequence of scans over a certain time, resulting in a set of FC networks, F (i, j) may implicitly describe the static nature of the network.
And step two, extracting the brain region characteristics from the constructed SFCN through a Convolutional Neural Network (CNN) constructed in the deep learning.
In the second step, for a complete static functional brain network extracted in the first step, the static functional brain network is a three-dimensional matrix (sample number 13456 137), and a convolutional neural network with parameter size of [32,32,64,32] is constructed in the second step, wherein [32,32] is the number of channels, [64,32] is the size of a full connection layer, and 116 (representing a brain region) is the size of a convolution kernel, and is used for the step size of each step during convolution. After the SFCN is subjected to convolution operation through the built first layer, the size of the output feature is 1x116x32, wherein 32 is the number of channels of the first layer, and then the feature extracted from the first layer is input into a second layer of convolution for operation, wherein the size of the dimension after the two layers of convolution is 1x 32.
And step three, building two layers of convolution layers in the step two, wherein the size of each layer of convolution kernel is 32 and 16 respectively, the step length of each step is 116, the parameter size of the first layer is 1x116x32, the size of the second layer after convolution operation is 1x1x16, and the features automatically extracted by the CNN built by the two layers are sent into the built two layers of full-connection layers to continue convolution operation to extract the related brain area features.
In the third step, based on the convolutional neural network with convolutional kernels of [32,32,64 and 32] in the deep learning, the brain region features with the dimensionality of 1x32 after two layers of convolution are input into the fully-connected layer, the size of the fully-connected output dimensionality of the first layer is 64, the fully-connected output dimensionality of the second layer is 32, and then the fully-connected operation is continued.
And step four, sending the features after the convolutional layers in the step three into two full-connection layers with the sizes of 64 and 32 respectively to continuously extract the features, and then diagnosing and classifying the brain diseases by using a SoftMax logistic regression function.
And in the fourth step, after the convolution of the two layers in the third step, the one-dimensional vector after the full connection of the two layers is continuously carried out is sent into a SoftMax function for classification. The SoftMax function is actually the gradient logarithm normalization of finite discrete probability distribution, and is a generalization of the logistic function. The SoftMax function is used in a multi-classification process, and maps the outputs of a plurality of neurons into a (0,1) interval, which can be understood as a probability, so as to perform multi-classification, and the formula is as follows:
Si＝ei/∑jej
wherein S isiIndicating the probability that the ith class occupies all the j classes. i represents the ith class and j represents a total of j classes.
And fifthly, training a built convolutional neural network model, and then evaluating the performance of the model in a test set.
In this example, the published resting-state functional magnetic resonance imaging (rest-statefMRI) dataset, i.e., the alzheimer neuroimaging initiative (ADNI), was mainly used. For the ADNI dataset, the ADNI data set included a total of 174 subjects, including 48 normal subjects, 50 early MCI patients (eMCI),45 late MCI patients (lMCI), and 31 patients with Alzheimer's Disease (AD). Specifically, the imaging matrix size was 49 × 58, the slice thickness was 3mm, the flip angle (flip angle) was 90 °, the axial slices (axialslices) were 47, the TE and TR were 30ms and 2.2s, respectively, and the voxel size was 3 × 3 × 4mm 3.
After going through the feature extraction of the above several steps, 3 classification tasks are then performed: emci, ncvs. l mci and a 4-class hcvs. emcivs. l mcivs.ad where for each class, 5-fold cross-validation strategy was used for the experiment, and finally a SoftMax classifier was used for classification, and then the average of the accuracy of each fold was taken as the final accuracy. In order to evaluate the effect of the different methods, the present embodiment employs two indexes, i.e., the accuracy of all categories, and the accuracy of classification for each category. The proposed method (CNN-FCN) was first compared with a Baseline method (Baseline) using the clustering coefficients of static FC networks as features, and with a method for SVM classification, the results of which are given in tables 1 and 2.
Table 1: 2 comparison of methods in classification tasks (eMCIVs. HC and ADVs. HC)
ACC＝Accuracy
Table 2: comparison results of methods in 4 classification task (ADvs. lMCIvs. eMCIVs. HC)
Fig. 3 to 5 show the training times of the convolutional neural network built in the deep learning in each compromise of every 200 rounds in different classification tasks and the accuracy in the training. As can be seen from tables 1 and 2, the proposed CNN method based on deep learning has better classification performance than other 2 comparison methods, and it can be observed from the results that the performance of the CNN method based on deep learning in two-group comparison is always better than that of the method for classifying brain disease diagnosis based on SVM, which indicates that the proposed new method is more advantageous than the traditional learning method for extracting features for classification. These results indicate that the classification performance will be further improved by the proposed deep learning based extraction of features of brain networks.
Example two
Correspondingly, on the basis of the extraction method, the embodiment also provides a static brain network feature extraction system based on the convolutional neural network. Referring to fig. 2, the extraction system comprises a static brain network generation module, a convolutional neural network with parameter size [32,32,64,32], two convolutional layers, two fully-connected layers with size of 64 and 32 respectively, and a SoftMax classifier which are connected in sequence.
And the static brain network generation module is used for dividing the brain space of each sample into 116 interested areas according to the automatic anatomical marking template, and forming the static brain network by calculating a Pearson coefficient between each brain area.
A convolutional neural network with parameter size [32,32,64,32] for processing the static brain network to extract corresponding brain region features; the number of channels of the convolutional neural network is [32,32], the parameters of the full-link layer are [64,32], and the convolution kernel is 116.
Two convolutional layers connected behind the convolutional neural network, the sizes of convolutional cores of the two convolutional layers are respectively 32 and 16, the step length is 116, the parameter of the first convolutional layer is 1x116x32, and the parameter of the second convolutional layer is 1x1x 32; these two convolutional layers continue the convolution operation on the corresponding brain region features extracted in step S2.
And two fully-connected layers with the sizes of 64 and 32 respectively are used for continuously extracting the features of the brain area with the dimension of 1x32 after the convolution layer processing.
And the SoftMax classifier adopts a SoftMax logistic regression function to diagnose and classify the brain diseases.
The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above-mentioned embodiments, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that modifications and embellishments within the scope of the invention may be made by those skilled in the art without departing from the principle of the invention.
Claims (6)
1. A static brain network feature extraction method based on a convolutional neural network is characterized by comprising the following steps:
s1, dividing the brain space of each sample into 116 interested areas according to the automatic anatomical marking template, and forming a static brain network by calculating a Pearson coefficient between each brain area;
s2, processing the static brain network by adopting a convolutional neural network with the parameter size of [32,32,64,32] to extract corresponding brain area characteristics, and respectively extracting and obtaining brain area space characteristics and brain area marginalization characteristics after convolution and full connection; wherein, the number of channels of the convolutional neural network is [32,32], the parameters of the full connection layer are [64,32], and the convolution kernel is 116;
s3, connecting two convolutional layers behind the convolutional neural network, wherein the sizes of convolutional cores of the two convolutional layers are 32 and 16 respectively, the step length is 116, the parameter of the first convolutional layer is 1x116x32, and the parameter of the second convolutional layer is 1x1x 32; the two layers of convolution layers continue to carry out convolution operation on the similar features of different positions of the brain area space extracted in the step S2 for deep extraction;
and S4, sending the spatial features of the brain area with the dimension of 1x32 after the convolution layer processing in the step S3 into two full-connection layers with the sizes of 64 and 32 respectively to continuously extract the marginalized features of the brain area, transmitting the marginalized features of the brain area to the full-connection layers to contain the same features as the learning target, and then adopting a SoftMax logistic regression function to diagnose and classify the brain diseases.
2. The convolutional neural network based static brain network feature extraction method of claim 1, wherein in step S1, the pearson coefficient between each brain region is calculated according to the following formula:
F(i,j)＝pcc(xi,xj)
where pc denotes the correlation between two brain regions, xiAnd xjBlood oxygen signals representing brain region i and brain region j, respectively, and F (i, j) describes the interaction between brain region i and brain region j.
3. The convolutional neural network based static brain network feature extraction method of claim 1, wherein the process of using SoftMax logistic regression function for diagnosis and classification of brain diseases comprises the following steps:
perform 3 classification tasks: emci, ncvs. l mci, and a 4 classification hcvs. emcivs. l mcivs.ad;
adopting a 5-fold cross validation strategy for each classification, and classifying by using a SoftMax classifier;
the precision of each fold is averaged to obtain the final precision.
4. The convolutional neural network-based static brain network feature extraction method according to claim 1 or 3, wherein in step S4, the SoftMax logistic regression function is:
Sa＝ea/∑beb
wherein S isaThe probability of the a-th category among all the b categories is shown, wherein a represents the a-th category, and b represents a total b categories.
5. The convolutional neural network based static brain network feature extraction method of claim 1, wherein the sample adopts a resting state functional magnetic resonance imaging dataset.
6. A convolutional neural network based static brain network feature extraction system, the extraction system comprising:
the static brain network generation module is used for dividing the brain space of each sample into 116 interested areas according to the automatic anatomical marking template and forming a static brain network by calculating a Pearson coefficient between every two brain areas;
a convolutional neural network with parameter size [32,32,64,32] for processing the static brain network to extract corresponding brain region features; wherein, the number of channels of the convolutional neural network is [32,32], the parameters of the full connection layer are [64,32], and the convolution kernel is 116;
two convolutional layers connected behind the convolutional neural network, the sizes of convolutional cores of the two convolutional layers are respectively 32 and 16, the step length is 116, the parameter of the first convolutional layer is 1x116x32, and the parameter of the second convolutional layer is 1x1x 32; after convolution, spatial features of the brain areas are respectively extracted and used for identifying similar features of the brain areas at different spatial positions; the two layers of convolution layers continue to carry out convolution operation on the extracted similar features at different positions of the brain area space;
two fully-connected layers with the sizes of 64 and 32 respectively are used for continuously extracting the edge features of the brain area from the characteristics with the dimension of 1x32 after the convolution layer processing;
and the SoftMax classifier adopts a SoftMax logistic regression function to diagnose and classify the brain diseases.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111097531.3A CN113786185B (en) | 2021-09-18 | 2021-09-18 | Static brain network feature extraction method and system based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111097531.3A CN113786185B (en) | 2021-09-18 | 2021-09-18 | Static brain network feature extraction method and system based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113786185A true CN113786185A (en) | 2021-12-14 |
| CN113786185B CN113786185B (en) | 2024-05-07 |
Family
ID=79183989
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111097531.3A Active CN113786185B (en) | 2021-09-18 | 2021-09-18 | Static brain network feature extraction method and system based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113786185B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114360068A (en) * | 2022-01-12 | 2022-04-15 | 安徽师范大学 | Bone action identification method based on multi-stream packet shuffling graph convolution neural network |
| CN116051849A (en) * | 2023-04-03 | 2023-05-02 | 之江实验室 | Brain network data feature extraction method and device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107909117A (en) * | 2017-09-26 | 2018-04-13 | 电子科技大学 | A kind of sorting technique and device based on brain function network characterization to early late period mild cognitive impairment |
| CN110522448A (en) * | 2019-07-12 | 2019-12-03 | 东南大学 | A kind of brain network class method based on figure convolutional neural networks |
| CN110797123A (en) * | 2019-10-28 | 2020-02-14 | 大连海事大学 | Graph convolution neural network evolution method of dynamic brain structure |
| CN111009324A (en) * | 2019-11-25 | 2020-04-14 | 东北大学 | Mild cognitive impairment auxiliary diagnosis system and method based on brain network multi-feature analysis |
| AU2020102977A4 (en) * | 2020-10-23 | 2020-12-24 | Bhima, Ravi Teja DR | A Deep learning technique to recognise brain activity by fMRI and DTI image fusion |
| CN113040715A (en) * | 2021-03-09 | 2021-06-29 | 北京工业大学 | Human brain function network classification method based on convolutional neural network |
-
2021
- 2021-09-18 CN CN202111097531.3A patent/CN113786185B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107909117A (en) * | 2017-09-26 | 2018-04-13 | 电子科技大学 | A kind of sorting technique and device based on brain function network characterization to early late period mild cognitive impairment |
| CN110522448A (en) * | 2019-07-12 | 2019-12-03 | 东南大学 | A kind of brain network class method based on figure convolutional neural networks |
| CN110797123A (en) * | 2019-10-28 | 2020-02-14 | 大连海事大学 | Graph convolution neural network evolution method of dynamic brain structure |
| CN111009324A (en) * | 2019-11-25 | 2020-04-14 | 东北大学 | Mild cognitive impairment auxiliary diagnosis system and method based on brain network multi-feature analysis |
| AU2020102977A4 (en) * | 2020-10-23 | 2020-12-24 | Bhima, Ravi Teja DR | A Deep learning technique to recognise brain activity by fMRI and DTI image fusion |
| CN113040715A (en) * | 2021-03-09 | 2021-06-29 | 北京工业大学 | Human brain function network classification method based on convolutional neural network |
Non-Patent Citations (3)
| Title |
|---|
| 接标等: "Designing weighted correlation kernels in convolutional neural networks for functional connectivity based brain disease diagnosis", 《MEDICAL IMAGE ANALYSIS》, 31 December 2020 (2020-12-31) * |
| 罗杰等: "基于三维卷积神经网络的大脑fMRI信号识别方法", 《中国体视学与图像分析》, vol. 24, no. 3, pages 191 - 197 * |
| 董鹏等: "Integration of Handcrafted and Embedded Features from Functional Connectivity Network with rs-fMRI for Brain Disease Classification", 《MACHINE LEARNING IN MEDICAL IMAGING》, 21 September 2021 (2021-09-21) * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114360068A (en) * | 2022-01-12 | 2022-04-15 | 安徽师范大学 | Bone action identification method based on multi-stream packet shuffling graph convolution neural network |
| CN116051849A (en) * | 2023-04-03 | 2023-05-02 | 之江实验室 | Brain network data feature extraction method and device |
| CN116051849B (en) * | 2023-04-03 | 2023-07-07 | 之江实验室 | Brain network data feature extraction method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113786185B (en) | 2024-05-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Singh et al. | Shallow 3D CNN for detecting acute brain hemorrhage from medical imaging sensors | |
| CN113040715B (en) | Human brain function network classification method based on convolutional neural network | |
| CN112101451B (en) | Breast cancer tissue pathological type classification method based on generation of antagonism network screening image block | |
| CN114926477B (en) | Brain tumor multi-mode MRI image segmentation method based on deep learning | |
| CN113616184A (en) | Brain network modeling and individual prediction method based on multi-mode magnetic resonance image | |
| Yue et al. | Hierarchical feature extraction for early Alzheimer’s disease diagnosis | |
| CN111009321A (en) | Application method of machine learning classification model in juvenile autism auxiliary diagnosis | |
| CN111090764B (en) | Image classification method and device based on multitask learning and graph convolution neural network | |
| Zhang et al. | Automatic assessment of full left ventricular coverage in cardiac cine magnetic resonance imaging with fisher-discriminative 3-D CNN | |
| CN112634211B (en) | MRI image segmentation method, device and equipment based on multiple neural networks | |
| CN113786185B (en) | Static brain network feature extraction method and system based on convolutional neural network | |
| CN111540467B (en) | Schizophrenia classification identification method, operation control device and medical equipment | |
| CN111009324A (en) | Mild cognitive impairment auxiliary diagnosis system and method based on brain network multi-feature analysis | |
| CN114038564B (en) | Noninvasive risk prediction method for diabetes | |
| Zhou et al. | Automatic optic disc detection using low-rank representation based semi-supervised extreme learning machine | |
| CN117952964B (en) | Fundus medical image analysis method based on computer vision technology | |
| SJ et al. | A Features Fusion Approach for Neonatal and Pediatrics Brain Tumor Image Analysis Using Genetic and Deep Learning Techniques. | |
| Tobin et al. | Using a patient image archive to diagnose retinopathy | |
| CN112863650A (en) | Cardiomyopathy identification system based on convolution and long-short term memory neural network | |
| Pallawi et al. | Study of Alzheimer’s disease brain impairment and methods for its early diagnosis: a comprehensive survey | |
| Tajammal et al. | Deep Learning‐Based Ensembling Technique to Classify Alzheimer’s Disease Stages Using Functional MRI | |
| CN113191413B (en) | Prostate multimode MR image classification method and system based on foveal residual error network | |
| CN112562819B (en) | Report generation method of ultrasonic multi-section data for congenital heart disease | |
| CN113283465B (en) | Diffusion tensor imaging data analysis method and device | |
| Lin et al. | Hybrid CNN-SVM for alzheimer’s disease classification from structural MRI and the alzheimer’s disease neuroimaging initiative (ADNI) |
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 |