CN108520283B - Medical image classification method constructed based on brain region indirect relation network - Google Patents

Medical image classification method constructed based on brain region indirect relation network Download PDF

Info

Publication number
CN108520283B
CN108520283B CN201810331096.8A CN201810331096A CN108520283B CN 108520283 B CN108520283 B CN 108520283B CN 201810331096 A CN201810331096 A CN 201810331096A CN 108520283 B CN108520283 B CN 108520283B
Authority
CN
China
Prior art keywords
brain
relation
indirect
network
areas
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
Application number
CN201810331096.8A
Other languages
Chinese (zh)
Other versions
CN108520283A (en
Inventor
李颖
李静
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shandong Management University
Original Assignee
Shandong Management University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shandong Management University filed Critical Shandong Management University
Priority to CN201810331096.8A priority Critical patent/CN108520283B/en
Publication of CN108520283A publication Critical patent/CN108520283A/en
Application granted granted Critical
Publication of CN108520283B publication Critical patent/CN108520283B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Epidemiology (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Primary Health Care (AREA)
  • Public Health (AREA)
  • Image Analysis (AREA)

Abstract

The invention provides a medical image classification method constructed based on an indirect brain area relationship network, which is used for auxiliary diagnosis of mild cognitive impairment. The method comprises the steps of firstly establishing correlation vectors for two brain areas by utilizing the correlation between the two brain areas and the common neighborhood of the two brain areas, then obtaining correlation values between the two brain areas by calculating the lattice closeness of the correlation vectors of the two brain areas, obtaining the correlation values between any two brain areas according to the method, thus establishing a brain area indirect relation network, and finally classifying medical images based on the brain area indirect relation network. The method of the invention fully utilizes rich relation information between the brain areas and the common neighborhoods thereof to describe the correlation between the brain areas, and makes up the problem of insufficient information utilization in the process of constructing the direct relation network. In addition, the indirect relationship network has stronger distinguishability and robustness. The invention obviously improves the classification accuracy of normal people and patients with mild cognitive impairment, and has higher sensitivity and specificity.

Description

Medical image classification method constructed based on brain region indirect relation network
Technical Field
The invention relates to the field of medical image processing, in particular to a medical image classification method constructed based on an indirect brain area relationship network.
Background
Alzheimer's disease is a progressive, irreversible neurodegenerative disease that is usually accompanied by changes in brain structure and function. As one of the most common diseases, alzheimer disease is characterized by a decline in cognition and function, which seriously affects the normal life of people. Mild cognitive impairment is generally considered to be a transitional state between normal aging and alzheimer's disease, characterized by patients with mild cognitive impairment without affecting most of the daily activities. It has been found that approximately 10-15% of patients with mild cognitive impairment convert to alzheimer's disease annually, but only 1-2% of normal elderly suffer from alzheimer's disease. Therefore, early diagnosis and treatment of mild cognitive impairment are of great importance for delaying and preventing the conversion of mild cognitive impairment to alzheimer's disease.
Positron Emission Tomography (PET) is a standard non-invasive three-dimensional functional imaging modality. It uses positron nuclide to mark human body metabolites such as glucose and the like as an imaging agent, and provides biological metabolic information of diseases for clinic by measuring metabolic change of a focus part. In real life, PET has become an effective tool for diagnosing alzheimer's disease and mild cognitive impairment. For PET images, one can classify them using various pattern recognition methods, thereby enabling assisted diagnosis of diseases.
The human brain is highly complex, different areas of the brain have different functions, and the normal life of human beings requires the cooperation of different brain areas. If the brain areas of the brain are regarded as nodes in the network, and the cooperation between the brain areas is regarded as the connection between the nodes, the whole brain structure can be abstracted into a network model. Research finds that the change of the brain network structure is closely related to a plurality of diseases, so that a plurality of scholars at home and abroad construct the brain network by using different methods, and basis is provided for disease diagnosis and prediction. However, in the existing method, only limited information between two brain regions is used to depict the correlation between the brain regions (referred to as a direct relation network in the invention) when the network is constructed, and the correlation between the brain regions and the common neighborhood thereof is not considered, so that the correlation between every two brain regions can be better described, that is, the brain network cannot be constructed by utilizing more abundant deep information between a plurality of brain regions, so that the accuracy of distinguishing the brain network characteristics from patients and normal people is low. In addition, when a certain brain area is seriously influenced by noise, the direct relation network is also influenced correspondingly, and the classification effect is further influenced.
Disclosure of Invention
The invention provides a medical image classification method constructed based on a brain region indirect relation network for overcoming the defects of the prior art, and the method is used for auxiliary diagnosis of mild cognitive impairment. The method describes the correlation of two brain areas by using the correlation of the relationship between the two brain areas and the common neighborhood of the two brain areas, constructs a brain area indirect relationship network, and makes up the shortage of information of only using the direct relationship between every two brain areas. Therefore, the method has stronger distinguishability and robustness. Aiming at the problems of high network characteristic dimension and redundant information, the invention adopts a two-step characteristic dimension reduction method of double-sample t inspection and principal component analysis, and retains effective characteristics. And finally, classifying by using a support vector machine classifier. The method provided by the invention obviously improves the classification accuracy of normal people and patients with mild cognitive impairment, and has higher sensitivity and specificity.
The technical scheme of the invention is as follows:
a medical image classification method based on brain area indirect relation network construction firstly utilizes the correlation between two brain areas and their common neighborhoods to respectively establish correlation vectors for the two brain areas, then obtains the correlation value between the two brain areas by calculating the lattice closeness of the correlation vectors of the two brain areas, obtains the correlation value between any two brain areas according to the method, thereby constructing the brain area indirect relation network, and finally classifies the medical image based on the brain area indirect relation network, the method comprises the following steps:
(1) image pre-processing
Collecting PET image data, preprocessing with statistical parameter mapping software SPM, extracting metabolic intensity, and recording each sample as xi，
Figure BDA0001627966630000021
Wherein l1,l2,…,lnRespectively representing the size of each dimension of the sample, wherein N is the dimension of a single sample, and N represents the number of the samples;
(2) direct relationship network construction
For each sample xiPartitioning according to an anatomical template to obtain D brain areas as network nodes, calculating the correlation of the average signal intensity of any two brain areas as the direct relation value of the two brain areas, obtaining a D multiplied by D symmetric matrix as a direct relation network, wherein the ith row and jth column elements on the symmetric matrix are the direct relation values of the ith brain area and the jth brain area, and the diagonal elements are all set to be 0;
(3) indirect relational network construction
Using a D multiplied by D symmetric matrix as an indirect relation network, wherein the elements of the ith row and the jth column on the symmetric matrix are indirect relation values of the ith brain area and the jth brain area, diagonal elements are 0, and the indirect relation values of the two brain areas a and b are calculated according to the following modes:
firstly, according to the direct relation network, selecting the direct relation between the brain area a and the brain area b for the brain area a and the brain area b respectivelyThe K brain regions with the largest value are taken as their K neighbors, and then a common neighbor of brain region a and brain region b is found, denoted as the set U ═ n1,n2,…nM},M∈[0,D-2]Constructing a correlation vector of the brain region a by using the direct relation values of the brain region a and the brain region b and the common neighbor
Figure BDA0001627966630000031
And the correlation vector of brain region b
Figure BDA0001627966630000032
r(a,nM) Representing the brain region a and its neighbors nMA direct relation value of r (b, n)M) Representing the brain region b and its neighbors nMThen computing a correlation vector
Figure BDA0001627966630000033
And
Figure BDA0001627966630000034
as the indirect relationship value between brain region a and brain region b;
(4) feature dimension reduction
The method comprises two steps of t-test feature selection and principal component analysis dimension reduction, wherein in the first step, unpaired double-sample t-test is applied to find out features with significant difference in normal population and diseased population, and in the second step, principal component analysis is applied to further reduce dimensions to obtain optimal features for classification;
(5) support vector machine classification
And (4) training a support vector machine classifier by using the training sample based on the features selected in the step (4) to classify the test sample.
Specifically, in step (2), the direct relation value r (a, b) of any two brain regions a and b is calculated by:
Figure BDA0001627966630000035
wherein d (a, b) ═ t (a) -t (b)2T (a) and t (b) represent the mean signal intensity, σ, of brain region a and brain region b, respectivelyaAnd σbAre respectively provided withThe standard deviation of the mean signal intensity of brain region a and brain region b is represented.
In particular, in step (3), the relationship vector
Figure BDA0001627966630000036
And
Figure BDA0001627966630000037
fuzzy sets for correlations of
Figure BDA0001627966630000038
And
Figure BDA0001627966630000039
by lattice closeness of, i.e.
Figure BDA00016279666300000310
Wherein the content of the first and second substances,
Figure BDA00016279666300000311
Figure BDA00016279666300000312
wherein
Figure BDA00016279666300000313
Represents the inner product of the reaction mixture,
Figure BDA00016279666300000314
the external product is represented, the A is represented by the V, and the V is represented by the V.
The invention has the beneficial effects that: the indirect relation network is constructed by utilizing the closeness in fuzzy mathematics, the correlation between brain areas is described by fully utilizing rich relation information between the brain areas and the common neighborhoods of the brain areas, and the problem of insufficient information utilization in the direct relation network construction process is solved. In addition, the indirect relationship network has stronger distinguishability and robustness. The invention obviously improves the classification accuracy of normal people and patients with mild cognitive impairment, and has higher sensitivity and specificity.
Drawings
FIG. 1 is a schematic block diagram of a medical image classification method based on brain region indirect relationship network construction according to the present invention;
FIG. 2 is a schematic diagram of a PET image;
fig. 3 is a schematic block diagram of a process for solving for common neighbors of two brain regions.
Detailed Description
The invention is further illustrated with reference to the following figures and examples.
As shown in fig. 1, the method for diagnosing mild cognitive impairment based on the indirect relationship network of the present invention comprises the following specific steps:
1. image pre-processing
PET images of 154 normal and 195 patients with mild cognitive impairment were randomly selected from the ADNI database (as shown in fig. 2), pre-processed using a statistical parameter mapping Software Package (SPM), and metabolic intensities were extracted. Each sample is denoted xi，
Figure BDA0001627966630000041
Wherein l1,l2,…,lnRespectively, the size of each dimension of the sample, N is the dimension of a single sample, and N represents the number of samples.
2. Direct relationship network construction
For each sample xiAccording to the method, the brain is partitioned according to an Automatic Anatomical Labeling (AAL) template, cerebellum tissues are removed, and finally 90 brain areas are obtained and serve as network nodes. For each brain region, the mean and standard deviation of the signal intensity were calculated. Then, the correlation between any two brain regions is calculated, resulting in a 90 × 90 symmetric matrix.
The specific method is as follows, for any two brain regions a and b, the difference is calculated by using Euclidean distance:
d(a,b)＝(t(a)-t(b))2
where t (a) and t (b) represent the average signal intensity of brain region a and brain region b, respectively. The correlation of the two regions is defined by an exponential function:
Figure BDA0001627966630000042
wherein sigmaaAnd σbThe standard deviation of the mean signal intensity of brain region a and brain region b is represented, respectively.
By the above calculation, each sample xiA 90 x 90 correlation matrix is obtained that is symmetric about the diagonal and has all 1 diagonal elements. Each element in the matrix (excluding the diagonal elements) represents the correlation of the average signal strength of a pair of brain regions. The diagonal elements are set to 0, indicating that the region itself has no correlation. The matrix is a direct relationship network.
3. Indirect relational network construction
Direct relationship networks rely only on limited information between two brain regions, which roughly reflects the correlation between brain regions. The correlation of the relationship between two brain regions and their common neighborhood, referred to herein as an indirect relationship, provides more information and more accurately reflects the relationship of the two brain regions.
For any two brain regions a and b, the indirect relationship is calculated as follows. The direct relationship values of brain region a to other regions are first arranged in descending order. The larger the value, the more closely the direct relationship between the two brain regions is, so that the front K (K e [1,89 ]) with the closest relationship to the brain region a can be found]) Brain regions, which are considered to be K neighbors of brain region a. The same operation is performed on brain region b to find K neighbors of brain region b. In this case, K was set to 7 by the experiment. Then a common neighbor of brain region a and brain region b is found, denoted as the set U ═ n1,n2,…nM},M∈[0,88]. The direct relations between the brain region a and the brain region b and the common neighbor are respectively expressed as relation vectors
Figure BDA0001627966630000051
Figure BDA0001627966630000052
And
Figure BDA0001627966630000053
fig. 3 is a schematic diagram of a process of finding the mutual neighbor of brain region a and brain region b. Calculating a relationship vector
Figure BDA0001627966630000054
And
Figure BDA0001627966630000055
the correlation of (a) indicates an indirect relationship between the brain region a and the brain region b. The present invention uses the lattice closeness of the fuzzy sets to describe the correlation of two relationship vectors.
Figure BDA0001627966630000056
Figure BDA0001627966630000057
Wherein
Figure BDA0001627966630000058
Represents the inner product of the reaction mixture,
Figure BDA0001627966630000059
the external product is represented, the A is represented by the V, and the V is represented by the V.
Figure BDA00016279666300000510
The larger the size of the tube is,
Figure BDA00016279666300000511
the smaller the two fuzzy sets are closer together. Lattice closeness is defined as follows:
Figure BDA00016279666300000512
it is clear that,
Figure BDA00016279666300000513
the larger the correlation between the two relationship vectors. In particular, if two brain regions do not have a common neighbor,
Figure BDA00016279666300000514
is 0. For each individual, the indirect relationship of any two brain regions is calculated, resulting in a 90 × 90 symmetric matrix. The diagonal line is set to 0, indicating that the brain region itself has no indirect relationship. The invention extracts the upper triangular elements of the matrix and splices into an indirect relation vector, and the length of the indirect relation vector is 90 multiplied by (90-1)/2 which is 4005.
4. Feature dimension reduction
The feature dimensionality reduction is used for eliminating redundant and noise features and effectively avoiding the influence of dimensionality disasters. The characteristic dimension reduction of the invention comprises two steps: and (5) carrying out t-test feature selection and principal component analysis dimensionality reduction. First, a non-paired double sample t-test was applied to find features with significant differences (p <0.05, uncorrected) in normal and diseased populations, and retained, with the remaining features removed. Although the subset of features selected by the t-test is discriminative, it is not the optimal feature for classification. Since the selected features have correlation and are much larger in number than the number of samples. In order to solve the above problem, principal component analysis is applied to further reduce dimensions. The principal component analysis comprises the following steps:
(1) normalizing the raw data by using a zero-mean normalization method;
(2) calculating a correlation coefficient matrix of the samples;
(3) calculating an eigenvalue (λ) of a correlation coefficient matrix1,λ2,...,λp) And corresponding feature vectors vi＝(vi1,vi2,...,vip)，i＝1,2,3,...,p；
(4) The principal component analysis can obtain p principal components, but the variance of each principal component is decreased, the contained information amount is also decreased, and the first q principal components are generally selected according to the magnitude of the cumulative contribution rate of each principal component. The contribution ratio is a ratio of the variance of a principal component to the total variance, that is, a ratio of a feature value to the total of all feature values.
Figure BDA0001627966630000061
The larger the contribution rate is, the stronger the information indicating the original variables included in the principal component is. The number q of principal components is determined mainly by the cumulative contribution rate of the principal components, i.e.
Figure BDA0001627966630000062
th was set to 0.95 in this experiment.
(5) And respectively substituting the normalized original data into the principal component expressions to obtain new data of each sample under each principal component, namely the optimal characteristics for classification.
5. Support vector machine classification
And (4) training a support vector machine classifier by using the training sample based on the features selected in the step 4, and classifying the test sample. The goal of the support vector machine is to find the hyperplane with the largest separation to represent the largest separation of the two types of samples in the feature space. The boundaries of the hyperplane are represented by support vectors, i.e., training samples on the boundaries of the maximum interval. The method uses a support vector machine with a radial basis function kernel and uses grid search to obtain the optimal parameters of the support vector machine.
The effects of the present invention can be further illustrated by experimental results.
The experimental data were PET image data of 154 normal and 195 patients with mild cognitive impairment randomly selected from the ADNI database. And dividing the data set into a training set and a testing set, and performing experiments by adopting a ten-fold cross validation method. The method is compared with a classification method based on a direct relationship network, and the results of the experiment are recorded in table 1. As can be seen from table 1, after the indirect relationship features of the present invention are adopted, the accuracy, sensitivity and specificity of algorithm identification are effectively improved, which indicates that the indirect relationship network-based method utilizes richer multi-region relationship information than the direct relationship-based method, thereby facilitating the improvement of identification performance.
TABLE 1 comparison of Classification Performance between direct and Indirect relational features
Figure BDA0001627966630000071

Claims (1)

1. A medical image classification method based on brain area indirect relation network construction firstly utilizes the correlation between two brain areas and their common neighborhoods to respectively establish relation vectors for the two brain areas, then obtains the correlation value between the two brain areas by calculating the lattice closeness of the two brain area relation vectors, obtains the correlation value between any two brain areas according to the method, thereby constructing the brain area indirect relation network, and finally classifies the medical image based on the brain area indirect relation network, the method comprises the following steps:
(1) image pre-processing
Collecting PET image data, preprocessing with statistical parameter mapping software SPM, extracting metabolic intensity, and recording each sample as xi，
Figure FDA0002954010350000011
i∈[1,N]Wherein l is1，l2，…，lnRespectively representing the size of each dimension of the sample, wherein N is the dimension of a single sample, and N represents the number of the samples;
(2) direct relationship network construction
For each sample xiPartitioning according to an anatomical template to obtain D brain areas as network nodes, calculating the correlation of the average signal intensity of any two brain areas as the direct relation value of the D brain areas, obtaining a D multiplied by D symmetric matrix as a direct relation network, wherein the ith row and jth column elements on the symmetric matrix are the direct relation values of the ith brain area and the jth brain area, the diagonal elements are 0, and the direct relation values r (a, b) of any two brain areas a and bThe calculation method comprises the following steps:
Figure FDA0002954010350000012
wherein d (a, b) ═ t (a) -t (b)2T (a) and t (b) represent the mean signal intensity, σ, of brain region a and brain region b, respectivelyaAnd σbThe standard deviation of the mean signal intensity of brain region a and brain region b, respectively;
(3) indirect relational network construction
Using a D multiplied by D symmetric matrix as an indirect relation network, wherein the elements of the ith row and the jth column on the symmetric matrix are indirect relation values of the ith brain area and the jth brain area, diagonal elements are 0, and the indirect relation values of the two brain areas a and b are calculated according to the following modes:
firstly, according to a direct relation network, selecting K brain areas with the maximum direct relation value for the brain area a and the brain area b as K neighbors of the brain areas respectively, and then finding out the common neighbor of the brain area a and the brain area b, wherein the common neighbor is expressed as a set U ═ { n ═ n1,n2,…nM},M∈[0,D-2]Constructing a relation vector of the brain region a by using direct relation values of the brain region a and the brain region b and common neighbors
Figure FDA0002954010350000013
Relation vector with brain region b
Figure FDA0002954010350000014
r(a,nM) Representing the brain region a and its neighbors nMA direct relation value of r (b, n)M) Representing the brain region b and its neighbors nMThen computing a relationship vector
Figure FDA0002954010350000015
And
Figure FDA0002954010350000016
is used as an indirect relation value of the brain region a and the brain region b, and a relation vector
Figure FDA0002954010350000017
And
Figure FDA0002954010350000018
fuzzy sets for correlations of
Figure FDA0002954010350000019
And
Figure FDA00029540103500000110
by lattice closeness of, i.e.
Figure FDA00029540103500000111
Wherein the content of the first and second substances,
Figure FDA00029540103500000112
Figure FDA0002954010350000021
wherein
Figure FDA0002954010350000022
Represents the inner product of the reaction mixture,
Figure FDA0002954010350000023
representing the outer product, combining the A-type representation and extracting the V-type representation;
(4) feature dimension reduction
The method comprises two steps of t-test feature selection and principal component analysis dimension reduction, wherein in the first step, unpaired double-sample t-test is applied to find out features with significant difference in normal population and diseased population, and in the second step, principal component analysis is applied to further reduce dimensions to obtain optimal features for classification;
(5) support vector machine classification
And (4) training a support vector machine classifier by using the training sample based on the features selected in the step (4) to classify the test sample.
CN201810331096.8A 2018-04-13 2018-04-13 Medical image classification method constructed based on brain region indirect relation network Active CN108520283B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810331096.8A CN108520283B (en) 2018-04-13 2018-04-13 Medical image classification method constructed based on brain region indirect relation network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810331096.8A CN108520283B (en) 2018-04-13 2018-04-13 Medical image classification method constructed based on brain region indirect relation network

Publications (2)

Publication Number Publication Date
CN108520283A CN108520283A (en) 2018-09-11
CN108520283B true CN108520283B (en) 2021-04-27

Family

ID=63431100

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810331096.8A Active CN108520283B (en) 2018-04-13 2018-04-13 Medical image classification method constructed based on brain region indirect relation network

Country Status (1)

Country Link
CN (1) CN108520283B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111402198B (en) * 2020-02-11 2022-12-27 山东师范大学 Alzheimer disease classification method and system based on anatomical landmarks and residual error network
CN111814917B (en) * 2020-08-28 2020-11-24 成都千嘉科技有限公司 Character wheel image digital identification method with fuzzy state
CN112700876A (en) * 2020-12-28 2021-04-23 聊城大学 Brain function network construction method containing high-order information

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110053889A1 (en) * 2008-09-02 2011-03-03 Joanne Szabo Prenatal and postnatal screening and treatment of critical monosaccharide deficiencies for neurologic and immunologic function
CN103886328A (en) * 2014-03-19 2014-06-25 太原理工大学 Functional magnetic resonance image data classification method based on brain network modular structure characteristics
CN106667490A (en) * 2017-01-09 2017-05-17 北京师范大学 Magnetic resonance brain image based subject individual difference data relation analysis method

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110053889A1 (en) * 2008-09-02 2011-03-03 Joanne Szabo Prenatal and postnatal screening and treatment of critical monosaccharide deficiencies for neurologic and immunologic function
CN103886328A (en) * 2014-03-19 2014-06-25 太原理工大学 Functional magnetic resonance image data classification method based on brain network modular structure characteristics
CN106667490A (en) * 2017-01-09 2017-05-17 北京师范大学 Magnetic resonance brain image based subject individual difference data relation analysis method

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
"正态云模型研究回顾与展望";杨洁等;《计算机学报》;20180317;第41卷(第3期);第724-744页 *

Also Published As

Publication number Publication date
CN108520283A (en) 2018-09-11

Similar Documents

Publication Publication Date Title
WO2023077603A1 (en) Prediction system, method and apparatus for abnormal brain connectivity, and readable storage medium
Ben Ahmed et al. Classification of Alzheimer’s disease subjects from MRI using hippocampal visual features
CN114999629B (en) AD early prediction method, system and device based on multi-feature fusion
Segovia et al. Early diagnosis of Alzheimer’s disease based on partial least squares and support vector machine
CN108511055B (en) Ventricular premature beat recognition system and method based on classifier fusion and diagnosis rules
US20060104519A1 (en) System and method for a contiguous support vector machine
CN108520283B (en) Medical image classification method constructed based on brain region indirect relation network
Brahim et al. Graph Fourier transform of fMRI temporal signals based on an averaged structural connectome for the classification of neuroimaging
Merjulah et al. Classification of myocardial ischemia in delayed contrast enhancement using machine learning
US20230100365A1 (en) Automatic classification of healthy and disease conditions from images or digital standard 12-lead ECGs
Mozhdehfarahbakhsh et al. An MRI-based deep learning model to predict Parkinson’s disease stages
Yang et al. A novel method for automated congestive heart failure and coronary artery disease recognition using THC-Net
Ahammed et al. Bag-of-features model for asd fmri classification using svm
Buvaneswari et al. Detection and Classification of Alzheimer’s disease from cognitive impairment with resting-state fMRI
CN113052800B (en) Alzheimer disease image analysis method and device
KR20230070085A (en) Providing method of diagnostic information on Alzheimer&#39;s disease using brain network
CN111329467A (en) Heart disease auxiliary detection method based on artificial intelligence
Pallawi et al. Study of Alzheimer’s disease brain impairment and methods for its early diagnosis: a comprehensive survey
CN113723485B (en) Hypergraph processing method for brain image of mild hepatic encephalopathy
CN114983341A (en) Multi-modal feature fusion based multi-classification prediction system for Alzheimer&#39;s disease
CN113171105A (en) Myocardial ischemia identification and classification method based on integrated CNN
Dinu et al. A novel modelling technique for early recognition and classification of Alzheimer’s disease
Roland et al. An automated system for arrhythmia detection using ECG records from MITDB
Gahiwad et al. Brain Stroke Detection Using CNN Algorithm
Seraj et al. Instantaneous fMRI based cerebral parameters for automatic Alzheimer, mild cognitive impairment and healthy subject classification

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