CN113723239A - Magnetic resonance image classification method and system based on causal relationship - Google Patents
Magnetic resonance image classification method and system based on causal relationship Download PDFInfo
- Publication number
- CN113723239A CN113723239A CN202110951379.4A CN202110951379A CN113723239A CN 113723239 A CN113723239 A CN 113723239A CN 202110951379 A CN202110951379 A CN 202110951379A CN 113723239 A CN113723239 A CN 113723239A
- Authority
- CN
- China
- Prior art keywords
- magnetic resonance
- causal
- resonance image
- brain
- representing
- 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
- 230000001364 causal effect Effects 0.000 title claims abstract description 171
- 238000000034 method Methods 0.000 title claims abstract description 45
- 210000004556 brain Anatomy 0.000 claims abstract description 161
- 238000012216 screening Methods 0.000 claims abstract description 61
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 47
- 238000010586 diagram Methods 0.000 claims abstract description 36
- 238000003745 diagnosis Methods 0.000 claims abstract description 8
- 238000012545 processing Methods 0.000 claims abstract description 8
- 230000000694 effects Effects 0.000 claims description 61
- 230000006870 function Effects 0.000 claims description 58
- 239000002245 particle Substances 0.000 claims description 25
- 238000005457 optimization Methods 0.000 claims description 23
- 238000012353 t test Methods 0.000 claims description 20
- 239000011159 matrix material Substances 0.000 claims description 17
- 238000001228 spectrum Methods 0.000 claims description 14
- 238000012795 verification Methods 0.000 claims description 11
- 238000010276 construction Methods 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 9
- 238000007635 classification algorithm Methods 0.000 claims description 6
- OAICVXFJPJFONN-UHFFFAOYSA-N Phosphorus Chemical compound [P] OAICVXFJPJFONN-UHFFFAOYSA-N 0.000 claims description 5
- 230000001105 regulatory effect Effects 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000002776 aggregation Effects 0.000 claims description 3
- 238000004220 aggregation Methods 0.000 claims description 3
- 201000010099 disease Diseases 0.000 claims description 3
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims description 3
- 230000036541 health Effects 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000012307 MRI technique Methods 0.000 claims description 2
- 238000010200 validation analysis Methods 0.000 claims 1
- 230000007246 mechanism Effects 0.000 abstract description 10
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 6
- 239000008280 blood Substances 0.000 description 6
- 210000004369 blood Anatomy 0.000 description 6
- 229910052760 oxygen Inorganic materials 0.000 description 6
- 239000001301 oxygen Substances 0.000 description 6
- 208000014644 Brain disease Diseases 0.000 description 4
- 238000010219 correlation analysis Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 230000003044 adaptive effect Effects 0.000 description 2
- 230000002490 cerebral effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000002790 cross-validation Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000002950 deficient Effects 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 238000011478 gradient descent method Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000013138 pruning Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011157 data evaluation Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002595 magnetic resonance imaging Methods 0.000 description 1
Images
Classifications
-
- 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
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention discloses a magnetic resonance image classification method and a system based on causal relationship, relating to the technical field of magnetic resonance image classification, wherein the method comprises the following steps: acquiring a magnetic resonance image with tag information; processing the magnetic resonance image to obtain BOLD signals of different brain areas; BOLD signals of different brain areas are used as variables to be input into a causal algorithm, and causal connection graphs of different brain areas are constructed; constructing characteristics by using the causal connection diagram, and screening the constructed characteristics to obtain optimal characteristics; and classifying the magnetic resonance image to be classified based on the optimal characteristics to obtain a classification result. The method can better fit with the actual working mechanism of the brain, reflect the difference of the working mechanisms of the brains of different subjects, has higher accuracy of classification results, and provides powerful suggestions for subsequent diagnosis and treatment.
Description
Technical Field
The invention relates to the technical field of magnetic resonance image classification, in particular to a magnetic resonance image classification method and system based on causal relationship.
Background
The existing brain disease diagnosis methods comprise a correlation analysis method, a nuclear method, a graph embedding method, a deep learning-based method and the like, the methods have certain effects in the field of brain disease diagnosis, but some defects still exist, the methods classify people on the basis of correlation, but the correlation analysis is greatly different from the real working mechanism of the brain, and the causal connection mode between brain areas cannot be correctly reflected; moreover, since the correlation analysis does not reflect the correct cause and effect relationship, the accuracy of the predicted diagnosis result cannot be ensured.
Chinese patent application CN111916204A published on 11/10/2020 provides a brain disease data evaluation method based on an adaptive sparse deep neural network, which includes: collecting documented brain disease data; summarizing each individual and the corresponding data characteristic and the change value thereof into a piece of unit data, and then constructing a data matrix by using the unit data corresponding to each individual; dividing a data matrix into a training set and a test set; training the deep neural network model based on the sparse enhancement BP algorithm by utilizing a training set and a testing set to obtain the trained deep neural network model based on the sparse enhancement BP algorithm; the principle of the method is still based on a deep learning method, each individual and the corresponding data characteristics and the change value thereof are summarized into a unit data, then the unit data corresponding to each individual is utilized to construct a data matrix, the data matrix is based on correlation, but correlation analysis is greatly different from the real working mechanism of the brain, the causal connection mode between brain areas cannot be correctly reflected, and the diagnosis result is inaccurate.
Disclosure of Invention
The invention provides a magnetic resonance image classification method and system based on causal relationship, aiming at overcoming the defects that the difference of brain working mechanisms is not considered and the classification result is inaccurate when the brain magnetic resonance image is classified in the prior art, the method and system can be more fit with the actual brain working mechanism, reflect the difference of the brain working mechanisms of different subjects, and have higher accuracy of the classification result.
In order to solve the technical problems, the technical scheme of the invention is as follows:
the invention provides a magnetic resonance image classification method based on causal relationship, which comprises the following steps:
s1: acquiring a magnetic resonance image with tag information;
s2: processing the magnetic resonance image to obtain BOLD signals of different brain areas;
s3: BOLD signals of different brain areas are used as variables to be input into a causal algorithm, and causal connection graphs of different brain areas are constructed;
s4: constructing features using a causal connection graph;
s5: screening the constructed features to obtain optimal features;
s6: and classifying the magnetic resonance image to be classified based on the optimal characteristics to obtain a classification result.
Preferably, the magnetic resonance image is acquired using a magnetic resonance imaging technique with tag information that is classified into a diseased tag and a normal tag according to a health condition. The label information is used as prior knowledge, and the magnetic resonance image is combined to be used as a data base, so that the subsequently constructed features have stronger correlation with the label information.
Preferably, the specific method of S2 is: and dividing the brain into different brain areas by using the AAL map, and performing aggregation calculation on the BOLD signal of each brain element in the brain areas to obtain the BOLD signals of the different brain areas. The BOLD signal is called as a blood oxygen level dependent signal, when the oxygen-deficient blood signal is increased, the signal of a certain brain area of the brain is weakened, when the oxygen-enriched blood signal is increased, the signal of the certain brain area of the brain is strengthened, and the BOLD signal reflects the activity of cerebral neurons.
Preferably, in step S3, the causal algorithm is a two-step causal discovery algorithm, and BOLD signals of different brain regions are input into the two-step causal algorithm as variables to construct a causal connection diagram of the different brain regions, specifically:
s31: setting a first-step objective function, optimizing the first-step objective function, and learning a skeleton diagram of the causal graph; the first step objective function is:
wherein ,
representing the edge weights of the skeleton map after optimization,
the BOLD signal representing the jth brain region,
representing the BOLD signal of the kth brain region, T being the time series length of the BOLD signal, n representing the number of magnetic resonance images,
weight, β, representing LASSO regressionjkExpressing regression coefficient, lambda expressing the first regulating parameter, | · | | non-woven phosphor2Represents L2 regularization, | - | represents L1 regularization;
the framework diagram is learned by adopting the self-adaptive Lasso regression, and compared with the Lasso regression, the self-adaptive Lasso regression has the function of pruning the framework diagram, can remove redundant edges and ensures the sparsity of results.
S32: setting a second-step objective function, optimizing the second-step objective function, and learning causal connection weight; the second step objective function is:
wherein ,
expressing causal connection weights of j and k brain intervals in the ith optimized magnetic resonance image, wherein L represents likelihood; λ represents a first regulation parameter, γ represents a second regulation parameter, and α represents a third regulation parameter;
representing the causal connection weight of the j and k brain intervals in the i, h and l magnetic resonance images; y ish、ylRepresenting h, l magnetic resonanceLabel information of the image; n represents the number of magnetic resonance images, m represents the number of brain regions, and T represents the time series length of the BOLD signal;
the function of the probability density is represented by,
causal connection weight matrix W representing brain regions in the ith magnetic resonance image(i)J (th) line of (1), x(i)(t) BOLD signal of brain region in ith magnetic resonance image at time t, W(i)A causal connection weight matrix for each brain region in the ith magnetic resonance image is represented.
In the second step, the effect of a first adjusting parameter lambda of a first regular term in the objective function is to prune the causal graph, remove redundant edges and ensure the sparsity of the causal graph; the second regulation parameter gamma of the second regularization term acts to reduce causal connection map differences, making the results more reliable; the third adjustment parameter α of the third regularization term has the effect of increasing the difference between different brain region causal graphs, making the causal graphs more capable of expressing different classes of brain region causal connection patterns, and incorporating as a priori knowledge tag information into the causal connection graphs.
S33: and constructing a cause and effect connection diagram according to the skeleton diagram of the cause and effect diagram and the cause and effect connection weight.
The first step of the two-step cause and effect discovery algorithm is to find a skeleton graph of a cause and effect graph, the skeleton graph is an undirected graph without weights and used for representing the dependency existing among brain regions, and the second step learns the cause and effect relation among the brain regions.
Preferably, the first-step objective function and the second-step objective function are optimized by using an adaptive gradient descent method.
Preferably, the specific method for screening the constructed features to obtain the optimal features comprises the following steps:
the constructed features include causal connection weight features, noise term variance features, and noise term variance power spectra;
s51: carrying out primary screening on the causal connection weight characteristics by using a double-sample t test, wherein the target function of the double-sample t test is as follows:
wherein ,tjkP-value, mu, representing causal connection weight characteristics of j-th and k-th brain intervalsa,jkMeans, mu, of causal connection weights of j-th and k-th brain regions in the magnetic resonance image group ab,jkMeans S representing the mean of the causal connection weights of the j-th and k-th brain regions in the magnetic resonance image group baRepresenting the set of magnetic resonance images ID, S in the magnetic resonance image group abRepresents a set of magnetic resonance images ID in the magnetic resonance image group b;
Da,jkrepresenting the variance of causal connection weights, D, for the j-th and k-th brain regions in the magnetic resonance image group ab,jkA variance of causal connection weights of j-th and k-th brain regions in the magnetic resonance image group b;
setting a threshold value, and screening tjkA corresponding causal connection weight characteristic when the threshold is greater than the threshold;
s52: performing secondary screening on the primarily screened features by utilizing a particle swarm optimization algorithm:
inputting the causal connection weight characteristic, the noise term variance characteristic and the noise term variance power spectrum after the double-sample t test and screening into a particle swarm optimization algorithm, updating the speed and the position until a preset iteration number is reached, and screening out an optimal solution as an optimal characteristic.
The causal connection weight characteristics represent structural information of a causal connection graph, the noise term variance characteristics represent noise with real data, and the noise term variance power spectrum represents dynamic connectivity frequency information.
The purpose of feature screening is to prevent overfitting, avoid dimension disasters and improve generalization capability; the double-sample t test is one of hypothesis tests, and is mainly used for determining whether the overall mean values of two independent group samples are equal or not so as to select the features with larger difference from the mean values in a disease label and a normal label; the particle swarm optimization algorithm can prevent the local optimal solution from being trapped, and a better effect is obtained.
Preferably, the S6 is specifically:
s61: training the classifier by using a classification algorithm based on the optimal characteristics, and classifying the magnetic resonance image to be classified by using the trained classifier to obtain a classification result;
s62: and verifying the classification result of the classifier by adopting a cross verification method.
The effect of the cross-validation method is to prevent overfitting.
The invention also provides a magnetic resonance image classification system based on causal relationship, which comprises:
the data acquisition module is used for acquiring a magnetic resonance image with label information;
the preprocessing module is used for processing the magnetic resonance image to obtain BOLD signals of different brain areas;
the causal graph building module is used for inputting BOLD signals of different brain areas as variables into a causal algorithm to build causal connection graphs of the different brain areas;
a feature construction module for constructing features using the causal connection graph;
the characteristic screening module is used for screening the constructed characteristics to obtain the optimal characteristics;
and the diagnosis module is used for classifying the magnetic resonance images to be classified based on the optimal characteristics to obtain a classification result.
Preferably, the cause and effect graph building module comprises a first step unit, a second step unit and a building unit;
the cause and effect algorithm in the cause and effect graph building module is a two-step cause and effect finding algorithm, BOLD signals of different brain areas are used as variables to be input into the two-step cause and effect algorithm, and cause and effect connection graphs of the different brain areas are built;
the first step unit is used for setting a first step objective function, optimizing the first step objective function and learning a skeleton diagram of the causal graph; the first step objective function is:
wherein ,
representing the edge weights of the skeleton map after optimization,
the BOLD signal representing the jth brain region,
representing the BOLD signal of the kth brain region, T being the time series length of the BOLD signal, n representing the number of magnetic resonance images,
weight, β, representing LASSO regressionjkExpressing regression coefficient, lambda expressing the first regulating parameter, | · | | non-woven phosphor2Represents L2 regularization, | - | represents L1 regularization;
the second-step unit is used for setting a second-step objective function, optimizing the second-step objective function and learning causal connection weight; the second step objective function is:
wherein ,
expressing causal connection weights of j and k brain intervals in the ith optimized magnetic resonance image, wherein L represents likelihood; λ represents a first regulation parameter, γ represents a second regulation parameter, and α represents a third regulation parameter;
denotes the ithH, causal connection weight of j and k brain intervals in l magnetic resonance images; y ish、ylTag information representing the h, l magnetic resonance image; n represents the number of magnetic resonance images, m represents the number of brain regions, and T represents the time series length of the BOLD signal;
the function of the probability density is represented by,
causal connection weight matrix W representing brain regions in the ith magnetic resonance image(i)J (th) line of (1), x(i)(t) BOLD signal of brain region in ith magnetic resonance image at time t, W(i)A causal connection weight matrix representing each brain region in the ith magnetic resonance image;
and the construction unit is used for constructing the cause and effect connection diagram according to the skeleton diagram and the cause and effect connection weight of the cause and effect diagram.
Preferably, the feature screening module comprises a primary screening unit and a secondary screening unit;
the constructed features include causal connection weight features, noise term variance features, and noise term variance power spectra;
the primary screening unit is used for carrying out primary screening on the causal connection weight characteristics by using a double-sample t test, and the target function of the double-sample t test is as follows:
wherein ,tjkP-value, mu, representing causal connection weight characteristics of j-th and k-th brain intervalsa,jkMeans, mu, of causal connection weights of j-th and k-th brain regions in the magnetic resonance image group ab,jkMeans S representing the mean of the causal connection weights of the j-th and k-th brain regions in the magnetic resonance image group baRepresenting the set of magnetic resonance images ID, S in the magnetic resonance image group abRepresents a set of magnetic resonance images ID in the magnetic resonance image group b;
Da,jkrepresenting the variance of causal connection weights, D, for the j-th and k-th brain regions in the magnetic resonance image group ab,jkA variance of causal connection weights of j-th and k-th brain regions in the magnetic resonance image group b;
setting a threshold value, and screening tjkA corresponding causal connection weight characteristic when the threshold is greater than the threshold;
and the secondary screening unit is used for carrying out secondary screening on the primarily screened features by utilizing a particle swarm optimization algorithm:
inputting the causal connection weight characteristic, the noise term variance characteristic and the noise term variance power spectrum after the double-sample t test and screening into a particle swarm optimization algorithm, updating the speed and the position until a preset iteration number is reached, and screening out an optimal solution as an optimal characteristic.
Preferably, the diagnostic module comprises a classification unit and a verification unit;
the classification unit is used for training the classifier by utilizing a classification algorithm based on the optimal characteristics, and classifying the magnetic resonance image to be classified by utilizing the trained classifier to obtain a classification result;
and the verification unit is used for verifying the classification result of the classifier by adopting a cross verification method.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
the data base of the invention is magnetic resonance images fused with label information, BOLD signals of different brain areas are obtained by processing the magnetic resonance images, and causal connection graphs of different brain areas are constructed; the causal graph can better reflect the real working mechanism of the brain; the specificity of the brain working mechanism of different subjects can be better reflected by utilizing the structural characteristics of a causal connection diagram in combination with prior knowledge, and the brain working mechanism is strongly explained; and screening the constructed features to screen out the optimal features, classifying the magnetic resonance images to be classified based on the optimal features, and obtaining a classification result with higher accuracy.
Drawings
Fig. 1 is a flowchart of a magnetic resonance image classification method based on causal relationship according to embodiment 1.
Fig. 2 is a block diagram of a magnetic resonance image classification system based on causal relationship as described in embodiment 2.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the patent;
for the purpose of better illustrating the embodiments, certain features of the drawings may be omitted, enlarged or reduced, and do not represent the size of an actual product;
it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The technical solution of the present invention is further described below with reference to the accompanying drawings and examples.
Example 1
The present embodiment provides a magnetic resonance image classification method based on causal relationship, as shown in fig. 1, the method includes:
s1: acquiring a magnetic resonance image with tag information;
s2: processing the magnetic resonance image to obtain BOLD signals of different brain areas;
s3: BOLD signals of different brain areas are used as variables to be input into a causal algorithm, and causal connection graphs of different brain areas are constructed;
s4: constructing features using a causal connection graph;
s5: screening the constructed features to obtain optimal features;
s6: and classifying the magnetic resonance image to be classified based on the optimal characteristics to obtain a classification result.
And acquiring a magnetic resonance image with label information by using a magnetic resonance imaging technology, wherein the label information is divided into a diseased label and a normal label according to the health condition. The label information is used as prior knowledge, and the magnetic resonance image is combined to be used as a data base, so that the subsequently constructed features have stronger correlation with the label information.
The specific method of S2 is as follows: and dividing the brain into different brain areas by using the AAL map, and performing aggregation calculation on the BOLD signal of each brain element in the brain areas to obtain the BOLD signals of the different brain areas. The BOLD signal is called as a blood oxygen level dependent signal, when the oxygen-deficient blood signal is increased, the signal of a certain brain area of the brain is weakened, when the oxygen-enriched blood signal is increased, the signal of the certain brain area of the brain is strengthened, and the BOLD signal reflects the activity of cerebral neurons. In this embodiment, the magnetic resonance image is input into the AAL atlas to obtain BOLD signals of 116 brain regions of the brain.
In step S3, the cause and effect algorithm is a two-step cause and effect discovery algorithm, and BOLD signals of different brain areas are input as variables into the two-step cause and effect algorithm to construct a cause and effect connection diagram of different brain areas, specifically:
s31: setting a first-step objective function, optimizing the first-step objective function, and learning a skeleton diagram of the causal graph; the first step objective function is:
wherein ,
representing the edge weights of the skeleton map after optimization,
the BOLD signal representing the jth brain region,
representing the BOLD signal of the kth brain region, T being the time series length of the BOLD signal, n representing the number of magnetic resonance images,
weight, β, representing LASSO regressionjkExpressing regression coefficient, lambda expressing the first regulating parameter, | · | | non-woven phosphor2Representing L2 regularization, | · | representing L1 regularization.
The framework diagram is learned by adopting the self-adaptive Lasso regression, and compared with the Lasso regression, the self-adaptive Lasso regression has the function of pruning the framework diagram, can remove redundant edges and ensures the sparsity of results.
S32: setting a second-step objective function, optimizing the second-step objective function, and learning causal connection weight; the second step objective function is:
wherein ,
expressing causal connection weights of j and k brain intervals in the ith optimized magnetic resonance image, wherein L represents likelihood; λ represents a first regulation parameter, γ represents a second regulation parameter, and α represents a third regulation parameter;
representing the causal connection weight of the j and k brain intervals in the i, h and l magnetic resonance images; y ish、ylTag information representing the h, l magnetic resonance image; n represents the number of magnetic resonance images, m represents the number of brain regions, and T represents the time series length of the BOLD signal;
the function of the probability density is represented by,
causal connection weight matrix W representing brain regions in the ith magnetic resonance image(i)J (th) line of (1), x(i)(t) BOLD signal of brain region in ith magnetic resonance image at time t, W(i)A causal connection weight matrix for each brain region in the ith magnetic resonance image is represented.
In the second step, the effect of a first adjusting parameter lambda of a first regular term in the objective function is to prune the causal graph, remove redundant edges and ensure the sparsity of the causal graph; the second regulation parameter gamma of the second regularization term acts to reduce causal connection map differences, making the results more reliable; the third adjustment parameter α of the third regularization term has the effect of increasing the difference between different brain region causal graphs, making the causal graphs more capable of expressing different classes of brain region causal connection patterns, and incorporating as a priori knowledge tag information into the causal connection graphs.
S33: and constructing a cause and effect connection diagram according to the skeleton diagram of the cause and effect diagram and the cause and effect connection weight.
The first step of the two-step cause and effect discovery algorithm is to find a skeleton graph of a cause and effect graph, the skeleton graph is an undirected graph without weights and used for representing the dependency existing among brain regions, and the second step learns the cause and effect relation among the brain regions.
In the embodiment, a cause and effect discovery model is established, and a two-step cause and effect discovery algorithm is applied to the cause and effect discovery model, so that the method has stronger feature extraction capability and higher robustness in classification application; the causal connection diagram is more suitable for classification by adding regularization terms for increasing the difference of different causal connection diagrams into an objective function based on a structural equation model and label information of prior knowledge.
And optimizing the first-step objective function and the second-step objective function by adopting a self-adaptive gradient descent method.
The specific method for screening the constructed features to obtain the optimal features comprises the following steps:
the characteristics of the causal connection diagram construction comprise a causal connection weight characteristic, a noise term variance characteristic and a noise term variance power spectrum; the causal connection weight characteristics represent structural information of a causal connection graph, the noise term variance characteristics represent noise with real data, and the noise term variance power spectrum represents dynamic connectivity frequency information.
S51: carrying out primary screening on the causal connection weight characteristics by using a double-sample t test, wherein the target function of the double-sample t test is as follows:
wherein ,tjkP-value, mu, representing causal connection weight characteristics of j-th and k-th brain intervalsa,jkMeans, mu, of causal connection weights of j-th and k-th brain regions in the magnetic resonance image group ab,jkMeans S representing the mean of the causal connection weights of the j-th and k-th brain regions in the magnetic resonance image group baRepresenting the set of magnetic resonance images ID, S in the magnetic resonance image group abRepresents a set of magnetic resonance images ID in the magnetic resonance image group b;
Da,jkrepresenting the variance of causal connection weights, D, for the j-th and k-th brain regions in the magnetic resonance image group ab,jkA variance of causal connection weights of j-th and k-th brain regions in the magnetic resonance image group b;
setting a threshold value, and screening tjkAnd if the weight is larger than the threshold value, the corresponding causal connection weight characteristic is obtained. In the present embodiment, the threshold is set to 0.05.
S52: performing secondary screening on the primarily screened features by utilizing a particle swarm optimization algorithm:
inputting the causal connection weight characteristic, the noise term variance characteristic and the noise term variance power spectrum after the double-sample t test and screening into a particle swarm optimization algorithm, updating the speed and the position until a preset iteration number is reached, and screening out an optimal solution as an optimal characteristic.
The calculation formula of the variance characteristic of the noise term is as follows:
U＝X-BX
wherein, U represents the noise vector of the BOLD signal of the brain area, X is the BOLD signal of the brain area, and B is a cause and effect matrix;
the particle swarm optimization algorithm characteristic screening is constructed based on the particle swarm optimization algorithm, the basic idea of the particle swarm optimization algorithm is that a plurality of particles are randomly distributed in different spatial positions in an assumed space, each particle has two attributes of position and speed, and the position of each particle is changed according to the two attributes. And (4) obtaining an optimal solution through iteration.
The velocity update formula is:
wherein w is a first weight parameter, a is a second weight parameter, b is a third weight parameter, rand is a random number, pbest is a local optimal solution of each particle, gbest is a global optimal solution of all particles,
for the velocity of the t +1 th iteration particle i,
for the speed of the particle i in the t-th iteration,
the position of the particle i is iterated for the t round;
the location update formula is:
wherein sigmoid (. cndot.) is an activation function,
when in use
When selecting the corresponding feature, when
When so, the corresponding feature is not selected.
The purpose of feature screening is to prevent overfitting, avoid dimension disasters and improve generalization capability; the double-sample t test is one of hypothesis tests, and is mainly used for determining whether the overall mean values of two independent group samples are equal or not so as to select the features with larger difference from the mean values in a disease label and a normal label; the particle swarm optimization algorithm can prevent the local optimal solution from being trapped, and a better effect is obtained.
The S6 specifically includes:
s61: training the classifier by using a classification algorithm based on the optimal characteristics, and classifying the magnetic resonance image to be classified by using the trained classifier to obtain a classification result;
s62: and verifying the classification result of the classifier by adopting a cross verification method. The effect of the cross-validation method is to prevent overfitting. The classifier is an SVM model, a gradient lifting decision tree model and a deep learning model.
Example 2
This embodiment provides a magnetic resonance image classification system based on causal relationship, which is used to implement the magnetic resonance image classification method based on causal relationship as described in embodiment 1, and as shown in fig. 2, the system includes:
the data acquisition module is used for acquiring a magnetic resonance image with label information;
the preprocessing module is used for processing the magnetic resonance image to obtain BOLD signals of different brain areas;
the causal graph building module is used for inputting BOLD signals of different brain areas as variables into a causal algorithm to build causal connection graphs of the different brain areas;
a feature construction module for constructing features using the causal connection graph;
the characteristic screening module is used for screening the constructed characteristics to obtain the optimal characteristics;
and the diagnosis module is used for classifying the magnetic resonance images to be classified based on the optimal characteristics to obtain a classification result.
The cause and effect graph building module comprises a first step unit, a second step unit and a building unit;
the cause and effect algorithm in the cause and effect graph building module is a two-step cause and effect finding algorithm, BOLD signals of different brain areas are used as variables to be input into the two-step cause and effect algorithm, and cause and effect connection graphs of the different brain areas are built;
the first step unit is used for setting a first step objective function, optimizing the first step objective function and learning a skeleton diagram of the causal graph; the first step objective function is:
wherein ,
representing the edge weights of the skeleton map after optimization,
the BOLD signal representing the jth brain region,
representing the BOLD signal of the kth brain region, T being the time series length of the BOLD signal, n representing the number of magnetic resonance images,
weight, β, representing LASSO regressionjkExpressing regression coefficient, lambda expressing the first regulating parameter, | · | | non-woven phosphor2Represents L2 regularization, | - | represents L1 regularization;
the second-step unit is used for setting a second-step objective function, optimizing the second-step objective function and learning causal connection weight; the second step objective function is:
wherein ,
expressing causal connection weights of j and k brain intervals in the ith optimized magnetic resonance image, wherein L represents likelihood; λ represents a first regulation parameter, γ represents a second regulation parameter, and α represents a third regulation parameter;
representing the causal connection weight of the j and k brain intervals in the i, h and l magnetic resonance images; y ish、ylTag information representing the h, l magnetic resonance image; n represents the number of magnetic resonance images, m represents the number of brain regions, and T represents the time series length of the BOLD signal;
the function of the probability density is represented by,
causal connection weight matrix W representing brain regions in the ith magnetic resonance image(i)J (th) line of (1), x(i)(t) BOLD signal of brain region in ith magnetic resonance image at time t, W(i)A causal connection weight matrix representing each brain region in the ith magnetic resonance image;
and the construction unit is used for constructing the cause and effect connection diagram according to the skeleton diagram and the cause and effect connection weight of the cause and effect diagram.
The characteristic screening module comprises a primary screening unit and a secondary screening unit;
the constructed features include causal connection weight features, noise term variance features, and noise term variance power spectra;
the primary screening unit is used for carrying out primary screening on the causal connection weight characteristics by using a double-sample t test, and the target function of the double-sample t test is as follows:
wherein ,tjkP-value, mu, representing causal connection weight characteristics of j-th and k-th brain intervalsa,jkMeans, mu, of causal connection weights of j-th and k-th brain regions in the magnetic resonance image group ab,jkMeans S representing the mean of the causal connection weights of the j-th and k-th brain regions in the magnetic resonance image group baRepresenting the set of magnetic resonance images ID, S in the magnetic resonance image group abRepresenting magnetic resonance images I in a magnetic resonance image group bD, collecting;
Da,jkrepresenting the variance of causal connection weights, D, for the j-th and k-th brain regions in the magnetic resonance image group ab,jkA variance of causal connection weights of j-th and k-th brain regions in the magnetic resonance image group b;
setting a threshold value, and screening tjkA corresponding causal connection weight characteristic when the threshold is greater than the threshold;
and the secondary screening unit is used for carrying out secondary screening on the primarily screened features by utilizing a particle swarm optimization algorithm:
inputting the causal connection weight characteristic, the noise term variance characteristic and the noise term variance power spectrum after the double-sample t test and screening into a particle swarm optimization algorithm, updating the speed and the position until a preset iteration number is reached, and screening out an optimal solution as an optimal characteristic.
The diagnostic module comprises a classification unit and a verification unit;
the classification unit is used for training the classifier by utilizing a classification algorithm based on the optimal characteristics, and classifying the magnetic resonance image to be classified by utilizing the trained classifier to obtain a classification result;
and the verification unit is used for verifying the classification result of the classifier by adopting a cross verification method.
It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.
Claims (10)
1. A method for causal relationship-based classification of magnetic resonance images, the method comprising:
s1: acquiring a magnetic resonance image with tag information;
s2: processing the magnetic resonance image to obtain BOLD signals of different brain areas;
s3: BOLD signals of different brain areas are used as variables to be input into a causal algorithm, and causal connection graphs of different brain areas are constructed;
s4: constructing features using a causal connection graph;
s5: screening the constructed features to obtain optimal features;
s6: and classifying the magnetic resonance image to be classified based on the optimal characteristics to obtain a classification result.
2. A causal relationship-based magnetic resonance image classification method as claimed in claim 1, wherein in S1, a magnetic resonance image with label information is acquired by using a magnetic resonance imaging technique, and the label information is classified into a disease label and a normal label according to a health condition.
3. The causal relationship-based magnetic resonance image classification method of claim 1, wherein the specific method of step S2 is: and dividing the brain into different brain areas by using the AAL map, and performing aggregation calculation on the BOLD signal of each brain element in the brain areas to obtain the BOLD signals of the different brain areas.
4. The causal relationship-based magnetic resonance image classification method according to claim 1, wherein in step S3, the causal algorithm is a two-step causal discovery algorithm, BOLD signals of different brain regions are input as variables into the two-step causal algorithm, and a causal connection map of the different brain regions is constructed, specifically:
s31: setting a first-step objective function, optimizing the first-step objective function, and learning a skeleton diagram of the causal graph; the first step objective function is:
wherein ,
representing the edge weights of the skeleton map after optimization,
the BOLD signal representing the jth brain region,
representing the BOLD signal of the kth brain region, T being the time series length of the BOLD signal, n representing the number of magnetic resonance images,
weight, β, representing LASSO regressionjkExpressing regression coefficient, lambda expressing the first regulating parameter, | · | | non-woven phosphor2Represents L2 regularization, | - | represents L1 regularization;
s32: setting a second-step objective function, optimizing the second-step objective function, and learning causal connection weight; the second step objective function is:
wherein ,
expressing causal connection weights of j and k brain intervals in the ith optimized magnetic resonance image, wherein L represents likelihood; λ represents a first regulation parameter, γ represents a second regulation parameter, and α represents a third regulation parameter;
represents the j (th) position in the i (th), h (l) magnetic resonance images,k causal connection weights between brain regions; y ish、ylTag information representing the h, l magnetic resonance image; n represents the number of magnetic resonance images, m represents the number of brain regions, and T represents the time series length of the BOLD signal;
the function of the probability density is represented by,
causal connection weight matrix W representing brain regions in the ith magnetic resonance image(i)J (th) line of (1), x(i)(t) BOLD signal of brain region in ith magnetic resonance image at time t, W(i)A causal connection weight matrix representing each brain region in the ith magnetic resonance image;
s33: and constructing a cause and effect connection diagram according to the skeleton diagram of the cause and effect diagram and the cause and effect connection weight.
5. The method for classifying magnetic resonance images based on causal relationship as claimed in claim 1, wherein in said step S5, the constructed features are screened, and the specific method for obtaining the optimal features is:
the constructed features include causal connection weight features, noise term variance features, and noise term variance power spectra;
s51: carrying out primary screening on the causal connection weight characteristics by using a double-sample t test, wherein the target function of the double-sample t test is as follows:
wherein ,tjkP-value, mu, representing causal connection weight characteristics of j-th and k-th brain intervalsa,jkMeans, mu, of causal connection weights of j-th and k-th brain regions in the magnetic resonance image group ab,jkMeans S representing the mean of the causal connection weights of the j-th and k-th brain regions in the magnetic resonance image group baRepresenting the set of magnetic resonance images ID, S in the magnetic resonance image group abRepresents a set of magnetic resonance images ID in the magnetic resonance image group b;
Da,jkrepresenting the variance of causal connection weights, D, for the j-th and k-th brain regions in the magnetic resonance image group ab,jkA variance of causal connection weights of j-th and k-th brain regions in the magnetic resonance image group b;
setting a threshold value, and screening tjkA corresponding causal connection weight characteristic when the threshold is greater than the threshold;
s52: performing secondary screening on the primarily screened features by utilizing a particle swarm optimization algorithm:
inputting the causal connection weight characteristic, the noise term variance characteristic and the noise term variance power spectrum after the double-sample t test and screening into a particle swarm optimization algorithm, updating the speed and the position until a preset iteration number is reached, and screening out an optimal solution as an optimal characteristic.
6. A causal relationship-based magnetic resonance image classification method according to claim 1, wherein said S6 is specifically:
s61: training the classifier by using a classification algorithm based on the optimal characteristics, and classifying the magnetic resonance image to be classified by using the trained classifier to obtain a classification result;
s62: and verifying the classification result of the classifier by adopting a cross verification method.
7. A causal relationship-based magnetic resonance image classification system, the system comprising:
the data acquisition module is used for acquiring a magnetic resonance image with label information;
the preprocessing module is used for processing the magnetic resonance image to obtain BOLD signals of different brain areas;
the causal graph building module is used for inputting BOLD signals of different brain areas as variables into a causal algorithm to build causal connection graphs of the different brain areas;
a feature construction module for constructing features using the causal connection graph;
the characteristic screening module is used for screening the constructed characteristics to obtain the optimal characteristics;
and the diagnosis module is used for classifying the magnetic resonance images to be classified based on the optimal characteristics to obtain a classification result.
8. The causal relationship-based magnetic resonance image classification system of claim 7, wherein the causal graph construction module comprises a first step unit, a second step unit, and a construction unit;
the cause and effect algorithm in the cause and effect graph building module is a two-step cause and effect finding algorithm, BOLD signals of different brain areas are used as variables to be input into the two-step cause and effect algorithm, and cause and effect connection graphs of the different brain areas are built;
the first step unit is used for setting a first step objective function, optimizing the first step objective function and learning a skeleton diagram of the causal graph; the first step objective function is:
wherein ,
representing the edge weights of the skeleton map after optimization,
the BOLD signal representing the jth brain region,
representing the BOLD signal of the kth brain region, T being the time series length of the BOLD signal, n representing the number of magnetic resonance images,
weight, β, representing LASSO regressionjkDenotes the regression coefficient, and λ denotes the first control parameter，||·||2Represents L2 regularization, | - | represents L1 regularization;
the second-step unit is used for setting a second-step objective function, optimizing the second-step objective function and learning causal connection weight; the second step objective function is:
wherein ,
expressing causal connection weights of j and k brain intervals in the ith optimized magnetic resonance image, wherein L represents likelihood; λ represents a first regulation parameter, γ represents a second regulation parameter, and α represents a third regulation parameter;
representing the causal connection weight of the j and k brain intervals in the i, h and l magnetic resonance images; y ish、ylTag information representing the h, l magnetic resonance image; n represents the number of magnetic resonance images, m represents the number of brain regions, and T represents the time series length of the BOLD signal;
the function of the probability density is represented by,
causal connection weight matrix W representing brain regions in the ith magnetic resonance image(i)J (th) line of (1), x(i)(t) BOLD signal of brain region in ith magnetic resonance image at time t, W(i)A causal connection weight matrix representing each brain region in the ith magnetic resonance image;
and the construction unit is used for constructing the cause and effect connection diagram according to the skeleton diagram and the cause and effect connection weight of the cause and effect diagram.
9. The causal relationship-based magnetic resonance image classification system of claim 7, wherein the feature screening module comprises a primary screening unit and a secondary screening unit;
the constructed features include causal connection weight features, noise term variance features, and noise term variance power spectra;
the primary screening unit is used for carrying out primary screening on the causal connection weight characteristics by using a double-sample t test, and the target function of the double-sample t test is as follows:
wherein ,tjkP-value, mu, representing causal connection weight characteristics of j-th and k-th brain intervalsa,jkMeans, mu, of causal connection weights of j-th and k-th brain regions in the magnetic resonance image group ab,jkMeans S representing the mean of the causal connection weights of the j-th and k-th brain regions in the magnetic resonance image group baRepresenting the set of magnetic resonance images ID, S in the magnetic resonance image group abRepresents a set of magnetic resonance images ID in the magnetic resonance image group b;
Da,jkrepresenting the variance of causal connection weights, D, for the j-th and k-th brain regions in the magnetic resonance image group ab,jkA variance of causal connection weights of j-th and k-th brain regions in the magnetic resonance image group b;
setting a threshold value, and screening tjkA corresponding causal connection weight characteristic when the threshold is greater than the threshold;
and the secondary screening unit is used for carrying out secondary screening on the primarily screened features by utilizing a particle swarm optimization algorithm:
inputting the causal connection weight characteristic, the noise term variance characteristic and the noise term variance power spectrum after the double-sample t test and screening into a particle swarm optimization algorithm, updating the speed and the position until a preset iteration number is reached, and screening out an optimal solution as an optimal characteristic.
10. The causal relationship-based magnetic resonance image classification system of claim 7, wherein the diagnostic module includes a classification unit and a validation unit;
the classification unit is used for training the classifier by utilizing a classification algorithm based on the optimal characteristics, and classifying the magnetic resonance image to be classified by utilizing the trained classifier to obtain a classification result;
and the verification unit is used for verifying the classification result of the classifier by adopting a cross verification method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110951379.4A CN113723239B (en) | 2021-08-18 | 2021-08-18 | Magnetic resonance image classification method and system based on causal relationship |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110951379.4A CN113723239B (en) | 2021-08-18 | 2021-08-18 | Magnetic resonance image classification method and system based on causal relationship |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113723239A true CN113723239A (en) | 2021-11-30 |
| CN113723239B CN113723239B (en) | 2023-10-20 |
Family
ID=78676864
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110951379.4A Active CN113723239B (en) | 2021-08-18 | 2021-08-18 | Magnetic resonance image classification method and system based on causal relationship |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113723239B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114820460A (en) * | 2022-04-02 | 2022-07-29 | 南京航空航天大学 | Method and device for analyzing correlation of single gene locus and time sequence brain image |
| CN114882273A (en) * | 2022-04-24 | 2022-08-09 | 电子科技大学 | Visual identification method, device, equipment and storage medium applied to narrow space |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106548206A (en) * | 2016-10-27 | 2017-03-29 | 太原理工大学 | Multi-modal nuclear magnetic resonance image data classification method based on minimum spanning tree |
| CN107133651A (en) * | 2017-05-12 | 2017-09-05 | 太原理工大学 | The functional magnetic resonance imaging data classification method of subgraph is differentiated based on super-network |
| CN109770903A (en) * | 2018-12-29 | 2019-05-21 | 中国科学院自动化研究所 | The classification prediction technique of functional magnetic resonance imaging, system, device |
| CN109993230A (en) * | 2019-04-04 | 2019-07-09 | 江南大学 | A kind of TSK Fuzzy System Modeling method towards brain function MRI classification |
-
2021
- 2021-08-18 CN CN202110951379.4A patent/CN113723239B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106548206A (en) * | 2016-10-27 | 2017-03-29 | 太原理工大学 | Multi-modal nuclear magnetic resonance image data classification method based on minimum spanning tree |
| CN107133651A (en) * | 2017-05-12 | 2017-09-05 | 太原理工大学 | The functional magnetic resonance imaging data classification method of subgraph is differentiated based on super-network |
| CN109770903A (en) * | 2018-12-29 | 2019-05-21 | 中国科学院自动化研究所 | The classification prediction technique of functional magnetic resonance imaging, system, device |
| CN109993230A (en) * | 2019-04-04 | 2019-07-09 | 江南大学 | A kind of TSK Fuzzy System Modeling method towards brain function MRI classification |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114820460A (en) * | 2022-04-02 | 2022-07-29 | 南京航空航天大学 | Method and device for analyzing correlation of single gene locus and time sequence brain image |
| CN114820460B (en) * | 2022-04-02 | 2023-09-29 | 南京航空航天大学 | Method and device for correlation analysis of single gene locus and time sequence brain image |
| CN114882273A (en) * | 2022-04-24 | 2022-08-09 | 电子科技大学 | Visual identification method, device, equipment and storage medium applied to narrow space |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113723239B (en) | 2023-10-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ghosh et al. | Automatic detection and classification of diabetic retinopathy stages using CNN | |
| Prasad et al. | Early detection of diabetic retinopathy from digital retinal fundus images | |
| JP2022551683A (en) | Methods and systems for non-invasive genetic testing using artificial intelligence (AI) models | |
| CN108648191A (en) | Pest image-recognizing method based on Bayes's width residual error neural network | |
| Karthikeyan et al. | Feature selection and parameters optimization of support vector machines based on hybrid glowworm swarm optimization for classification of diabetic retinopathy | |
| Dey et al. | Automatic detection of whitefly pest using statistical feature extraction and image classification methods | |
| KR101687217B1 (en) | Robust face recognition pattern classifying method using interval type-2 rbf neural networks based on cencus transform method and system for executing the same | |
| CN111863244B (en) | Functional connection mental disease classification method and system based on sparse pooling graph convolution | |
| CN113723239A (en) | Magnetic resonance image classification method and system based on causal relationship | |
| Yusuf et al. | Web-based cataract detection system using deep convolutional neural network | |
| Sharma et al. | Machine learning approach for detection of diabetic retinopathy with improved pre-processing | |
| CN114463843A (en) | Multi-feature fusion fish abnormal behavior detection method based on deep learning | |
| KR20210067815A (en) | Method for measuring health condition of user and apparatus therefor | |
| Rosli et al. | Development of CNN transfer learning for dyslexia handwriting recognition | |
| CN110569880A (en) | Method for decoding visual stimulation by using artificial neural network model | |
| Bhowmik et al. | Reducing false prediction on covid-19 detection using deep learning | |
| Yang et al. | Retinal vessel segmentation based on an improved deep forest | |
| Subbarao et al. | Detection of Retinal Degeneration via High-Resolution Fundus Images using Deep Neural Networks | |
| Swain et al. | Diabetic retinopathy using image processing and deep learning | |
| Chakravarthy et al. | DR-NET: A Stacked Convolutional Classifier Framework for Detection of Diabetic Retinopathy | |
| Parkavi et al. | Decision support system using CNN for detecting the type of disease in fruits and vegetables | |
| Rubini et al. | Genetic Optimized Stacked Auto Encoder Based Diabetic Retinopathy Classification. | |
| Ashok et al. | Detection of macular diseases from optical coherence tomography images: Ensemble learning approach using vgg-16 and inception-v3 | |
| Abou El-Maged et al. | Enhanced CT-Image for Covid-19 Classification using ResNet-50 | |
| Sivapriya et al. | Diagnostic Classification of Lung Disease with Chest X-Ray Images Using SNN |
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 |