CN111046918A - ICA-CNN classified fMRI data space pre-smoothing and broadening method - Google Patents

ICA-CNN classified fMRI data space pre-smoothing and broadening method Download PDF

Info

Publication number
CN111046918A
CN111046918A CN201911144803.3A CN201911144803A CN111046918A CN 111046918 A CN111046918 A CN 111046918A CN 201911144803 A CN201911144803 A CN 201911144803A CN 111046918 A CN111046918 A CN 111046918A
Authority
CN
China
Prior art keywords
cnn
data
fmri
ica
smoothing
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
Application number
CN201911144803.3A
Other languages
Chinese (zh)
Other versions
CN111046918B (en
Inventor
林秋华
牛妍炜
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.)
Dalian University of Technology
Original Assignee
Dalian University of Technology
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 Dalian University of Technology filed Critical Dalian University of Technology
Priority to CN201911144803.3A priority Critical patent/CN111046918B/en
Publication of CN111046918A publication Critical patent/CN111046918A/en
Application granted granted Critical
Publication of CN111046918B publication Critical patent/CN111046918B/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
    • G06F18/21—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
    • G06F18/2134—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on separation criteria, e.g. independent component analysis
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/40—Detecting, measuring or recording for evaluating the nervous system
    • A61B5/4076—Diagnosing or monitoring particular conditions of the nervous system
    • A61B5/4088—Diagnosing of monitoring cognitive diseases, e.g. Alzheimer, prion diseases or dementia
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • 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
    • 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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Neurology (AREA)
  • Evolutionary Biology (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Psychiatry (AREA)
  • Pathology (AREA)
  • Veterinary Medicine (AREA)
  • Public Health (AREA)
  • Animal Behavior & Ethology (AREA)
  • Surgery (AREA)
  • Mathematical Physics (AREA)
  • Medical Informatics (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Physiology (AREA)
  • Neurosurgery (AREA)
  • Computational Linguistics (AREA)
  • Psychology (AREA)
  • Computing Systems (AREA)
  • Hospice & Palliative Care (AREA)
  • Developmental Disabilities (AREA)
  • Child & Adolescent Psychology (AREA)
  • Software Systems (AREA)
  • Fuzzy Systems (AREA)
  • Signal Processing (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

An ICA-CNN classified fMRI data space pre-smoothing and broadening method belongs to the field of biomedical signal processing. Firstly, applying spatial smoothing with different FWHM to fMRI observation data, and generating a new fMRI data set in an augmented mode; then the ICA-CNN framework is sent in, and the improvement of the classification performance of the patient and the healthy person is realized. The invention is adopted to classify 82 tested resting state fMRI data into patients and healthy people, three Gaussian filters with different FWHM are adopted to spatially smooth observed data, three groups of fMRI data sets are generated, and then the three groups of fMRI data sets are sent to an ICA-CNN framework for classification. Compared with the existing multi-model order data augmentation method, the method can improve the classification accuracy by 2%; if the two are combined, the classification accuracy can be improved by 12.71%. Therefore, the invention can independently improve the network classification performance, and is easy to combine with other augmentation methods, thereby obviously improving the classification accuracy.

Description

ICA-CNN classified fMRI data space pre-smoothing and broadening method
Technical Field
The invention belongs to the field of biomedical signal processing, and particularly relates to an ICA-CNN classified fMRI data space pre-smoothing and amplifying method.
Background
Convolutional Neural Networks (CNNs) have the advantages of local sensing, weight sharing, and the like, are prominent in tasks such as recognition, detection, classification, and the like, and are also useful in smart medicine. The resting-state fMRI (rs-fMRI) data has the advantages of non-invasiveness, high spatial resolution, easy acquisition on the test of a patient and the like, and is commonly used for analyzing and diagnosing neurological disorders such as schizophrenia. Therefore, CNNs with rs-fMRI data as training data would represent a unique advantage in the task of classifying healthy persons from patients.
In view of the problem that fMRI of patients is difficult to acquire and the data size is not large, linkawa et al propose an ICA-CNN classification framework for complex fMRI data of patients and healthy persons (patent application No. 201910350137.2). In this framework, two-dimensional slices of spatial components of interest are first separated using Independent Component Analysis (ICA), and then fed into two-dimensional CNN to extract features and classify. Compared with a method of directly utilizing observation data or using three-dimensional CNN classification, the framework reduces the requirement on fMRI data volume, and achieves better classification performance under the condition of limited fMRI data volume.
In fact, in the case where the fMRI data amount is constant, the data expansion can serve to further increase the data amount and improve the ICA-CNN classification effect. In the field of image classification, widely used data augmentation methods include rotation, flipping, noise addition, scaling, texture color transformation, and the like. However, these methods are not suitable for rs-fMRI analysis where the spatial structure cannot be varied arbitrarily. The patent application number 201910350137.2 only provides a data augmentation method with multiple model orders, and has important practical value in exploring other methods suitable for fMRI data augmentation under the ICA-CNN classification framework.
Disclosure of Invention
The invention provides an fMRI data amplification method suitable for an ICA-CNN classification framework, which can effectively expand fMRI sample size and further improve the classification performance of an ICA-CNN network. The specific scheme is that the fMRI observation data is subjected to spatial smoothing, a new fMRI data set is generated in an amplification mode, and then the fMRI data set is sent to an ICA-CNN framework, so that the classification performance of patients and healthy people is improved, and the method is shown in figure 1. The method comprises the following concrete steps:
the first step is as follows: inputting the four-dimensional fMRI observation amplitude data of the tested k
Figure BDA0002281880630000021
K is the total number of trials, T is the number of scan time points, X, Y, Z is the three dimensions of brain space.
The second step is that: the four-dimensional fMRI data of each tested object
Figure BDA0002281880630000022
Decomposition into a series of three-dimensional spatial signals x along a time axis(k)(1,x,y,z),…,x(k)(t,x,y,z),…,x(k)(T, x, y, z), wherein x(k)(T, X, Y, Z) is the three-dimensional scan at time T, T is 1, …, T, X is 1, …, X, Y is 1, …, Y, Z is 1, …, Z, K is 1, …, K.
The third step: the space is smooth. Spatial data x of a tested k at a time point t(k)(t, x, y, z), K1, …, K, convolved with a three-dimensional gaussian filter to achieve spatial smoothing as shown by:
Figure BDA0002281880630000023
in the formula
Figure BDA0002281880630000024
Representing a three-dimensional convolution calculation, g (x, y, z) is a three-dimensional zero-mean gaussian function defined as:
Figure BDA0002281880630000025
where σ is the standard deviation, the width of the filter is determined, and the larger σ is, the larger the smoothing degree is. Full width at Half maximum fwhm (fullbidth at Half maximum) is a common parameter for measuring the size of a gaussian kernel, having a linear relationship with sigma as follows,
Figure BDA0002281880630000026
in the pre-processing of fMRI data, the FWHM is typically taken to be the voxel size2-3 times of the total weight of the powder. Smoothing all the tested observation data by using G Gaussian filters with different FWHM sizes can expand each tested data set from 1 group to G group. For the tested k, the G groups of augmented data are recorded
Figure BDA0002281880630000027
As shown in fig. 2.
The fourth step: will be provided with
Figure BDA0002281880630000028
Is one-dimensional, i.e. the size of the space dimension is equal to X × Y × Z, i is 1, …, G, then the extrabrain voxels are removed, and only the intracerebral voxels are taken, obtaining
Figure BDA0002281880630000031
V is the number of endosomes, V<X×Y×Z。
The fifth step: principal Component Analysis (PCA) (principal Component analysis) pair
Figure BDA0002281880630000032
Figure BDA0002281880630000033
Reducing the vitamin content to obtain
Figure BDA0002281880630000034
N is the model order and is less than or equal to T.
And a sixth step: ICA isolation and extraction of the component of interest. Using Infmax algorithm pairs
Figure BDA0002281880630000035
Figure BDA0002281880630000036
Carrying out ICA separation; selecting a spatial activation map of the interested component from N ICA separation components based on the principle of maximum correlation coefficient with a spatial component template
Figure BDA0002281880630000037
Spatial composition template selected from textDocuments (S.M.Smith, P.T.Fox et al, "chromatography of the woven's functional and technical details activation and rest," Proceedings of the National academy of the United States of America, vol.106, No.31, pp.13040-13045,2009).
The seventh step: to pair
Figure BDA0002281880630000038
Performing extrabrain element zero filling, and restoring to three-dimensional space activation map, i.e.
Figure BDA0002281880630000039
Eighth step: will be provided with
Figure BDA00022818806300000310
Expand along the Z-axis into Z two-dimensional stacks of slices of size X Y, noted
Figure BDA00022818806300000311
And selecting W slices with large activation information according to the effective activation positions of the interested components, wherein W is less than or equal to Z and is marked as L, … and L + W-1.
The ninth step: cutting W pieces of two-dimensional slices
Figure BDA00022818806300000312
Normalized to [0,1 ] values]Is marked as
Figure BDA00022818806300000313
Then attaching a corresponding label to each slice
Figure BDA00022818806300000314
Figure BDA00022818806300000315
It is indicative of a healthy person,
Figure BDA00022818806300000316
representing patients with schizophrenia, establishing an augmented CNN sample set
Figure BDA00022818806300000317
The sample set of each sample to be tested is expanded from original W to W × G, i.e. the expansion is G times of original. The total K sample sets tested were expanded to K × W × G.
The tenth step: disorganizing the sequence of the tested sample, and proportionally dividing the sample set into a training set, a verification set and a test set; such as: 60% of the sample set is taken as a training set, 20% is taken as a verification set, and 20% is taken as a test set.
The eleventh step: and sending the training set into a CNN network for training. The CNN network structure used in the present invention is consistent with patent application No. 201910350137.2, and as shown in fig. 3, includes 2 convolutional layers, 2 max pooling layers, 1 full connection layer, and an output layer. Wherein, the size of the convolution layer core is 3 multiplied by 3, and the number of the convolution layer core is respectively 8 and 16; the size of the pooling layer core is 2 x 2, and the number of the pooling layer cores is 8 and 16 respectively; the full connection layer has 64 nodes, modified linear unit (ReLU) is used as an activation function, and the output layer uses Softmax to give the category to which each slice belongs. The batch processing size is 64, parameter updating is carried out by adopting an Adam algorithm, a two-class cross entropy function is used as a loss function of the network, and meanwhile, L is used2And (5) regularizing control weight and training a CNN model to have R rounds. At each round, calculating an accuracy ACC using the validation set; and saving the CNN model with the highest verification accuracy in the R round as the CNN model obtained by training. Let CNN verify as
Figure BDA0002281880630000041
Accuracy ACC is defined as follows:
Figure BDA0002281880630000042
wherein TP represents true positive, i.e.
Figure BDA0002281880630000043
TN indicates true negatives, i.e.
Figure BDA0002281880630000044
FP represents a false positive, i.e
Figure BDA0002281880630000045
FN indicates false negatives, i.e.
Figure BDA0002281880630000046
The twelfth step: and inputting the test set into the CNN model obtained by the eleventh training step to obtain the test accuracy ACC.
The thirteenth step: and 5-fold cross validation is adopted, namely the tenth step and the twelfth step are repeated for 5 times to obtain the accuracy ACC of the 5 times of tests, and the average result is calculated and recorded as the final classification accuracy of the CNN.
The invention has the beneficial effects that: the invention focuses on the classification problem in the field of computer-aided diagnosis, and provides a data augmentation method which aims at the fMRI data characteristics and is suitable for an ICA-CNN classification framework, and is used for training a CNN network and improving the classification result. For example, 82 rs-fMRI data tested were classified by the present invention as schizophrenic and healthy using FWHM of 6mm each3、8mm3、10mm3The three Gaussian filters perform spatial smoothing on all tested fMRI observation data to generate three groups of fMRI data sets, adopt Infmax to perform ICA separation, then select a Default Mode Network (DMN) to establish a sample set, and finally send the sample set to a CNN for training and testing. Compared with the multi-model order data amplification method provided by the patent 201910350137.2, the method can improve the classification accuracy by 2% under the condition of the same sample number. If the method is combined with a multi-model order augmentation method, three model orders of N-20, 60 and 100 are set for fMRI data obtained under each spatial smoothing kernel, spatial components are respectively extracted, a sample set is established, and then all the model orders are sent to an ICA-CNN network for classification. Compared with the method only using the multiple model order augmentation, the combination method can improve the classification accuracy by 12.71%. Therefore, the invention not only can independently improve the network classification performance, but also can be easily combined with other augmentation methods, thereby obviously improving the classification accuracy.
Drawings
FIG. 1 illustrates the location and usage steps of the present invention in the ICA-CNN classification framework.
FIG. 2 is a schematic diagram of the present invention for spatially smoothing and augmenting fMRI observations.
Fig. 3 is a CNN network architecture used in the present invention.
Detailed Description
An embodiment of the present invention will be described in detail below with reference to the accompanying drawings. There were 82 rs-fMRI magnitude data tested, including 42 schizophrenic patients and 40 healthy people. Each test contained 146 scans, each scan sharing whole-brain data of voxels X × Y × Z53 × 63 × 46 153594, with an intra-brain voxel V62336 and voxel size of 3 × 3 × 3mm3。
The first step is as follows: four-dimensional fMRI observation amplitude data of all 82 tested k are input
Figure BDA0002281880630000051
Figure BDA0002281880630000052
The second step is that: the four-dimensional fMRI data of each tested object
Figure BDA0002281880630000053
Decomposition into a series of three-dimensional spatial signals x along a time axis(k)(1,x,y,z),…,x(k)(t,x,y,z),…,x(k)(146, x, y, z), wherein x(k)(t, x, y, z) is the three-dimensional scan at time t, t is 1, …,146, x is 1, …,53, y is 1, …,63, z is 1, …,46, k is 1, …, 82.
The third step: the space is smooth. Spatial data x of a tested k at a time point t(k)(t, x, y, z), k is 1, …,82, and spatial smoothing is achieved by convolution with a three-dimensional gaussian filter, as shown in equation (1). According to the definition of the formula (2), 3 three-dimensional Gaussian filters are generated, and the FWHM is 6mm3、8mm3、10mm3. All the tested observation data are smoothed by using the 3 Gaussian filters, and each tested data set is expanded from 1 group to 3 groups. For the tested k, 3 sets of augmentation data were recordedDo it
Figure BDA0002281880630000061
Figure BDA0002281880630000062
The fourth step: will be provided with
Figure BDA0002281880630000063
Is one-dimensional, i.e. the size of the space dimension is 53 × 63 × 46-153594, i-1, 2,3, then off-brain voxels are removed and only intra-brain voxels are taken, resulting in
Figure BDA0002281880630000064
The fifth step: using PCA pairs
Figure BDA0002281880630000065
Reducing dimension, selecting model order N as 50 to obtain
Figure BDA0002281880630000066
And a sixth step: ICA isolation and extraction of the component of interest. Using Infmax algorithm pairs
Figure BDA0002281880630000067
ICA separation is carried out with k being 1, …,82, i being 1,2, 3; based on the principle of the largest correlation coefficient of DMN component template given by the literature (S.M. Smith, P.T. Fox et al, ' Correspondence of the branched's functional architecture and activation and ' Proceedings of the National Academy of Sciences of the United States of America, vol.106, No.31, pp.13040-13045,2009), the spatial activation map of DMN component is selected from 50 ICA separated components
Figure BDA0002281880630000068
The seventh step: to pair
Figure BDA0002281880630000069
Performing extrabrain element zero filling, and restoring to three-dimensional space activation map, i.e.
Figure BDA00022818806300000610
Eighth step: will be provided with
Figure BDA00022818806300000611
Expand along the z-axis into 46 two-dimensional stacks of slices of size 53 × 63, noted
Figure BDA00022818806300000612
According to the effective activation position of the interested component, selecting 25 slices with large activation information, specifically L12, z 12, … and 36.
The ninth step: 25 two-dimensional slices
Figure BDA00022818806300000613
Normalized to [0,1 ] values]Is marked as
Figure BDA0002281880630000071
Then attaching a corresponding label to each slice
Figure BDA0002281880630000072
Figure BDA0002281880630000073
It is indicative of a healthy person,
Figure BDA0002281880630000074
representing patients with schizophrenia, establishing an augmented CNN sample set
Figure BDA0002281880630000075
The number of samples tested is increased from W25 to W × G25 × 3 to 75, i.e., 3 times the original number. The number of samples of all 82 samples tested was increased to 82 × W × G, 6150.
The tenth step: the sequence of the tested samples is disturbed, 60% of sample sets are taken as training sets, 20% are taken as verification sets, and 20% are taken as testing sets.
The eleventh step: and (5) sending the training set into a CNN network for training, and training the CNN model R for 50 rounds. At each round, calculating the accuracy rate ACC by using the verification set according to the formula (3); and saving the CNN model with the highest verification accuracy in 50 rounds.
The twelfth step: and (4) inputting the test set into the CNN model obtained by the eleventh training step, and calculating according to the formula (3) to obtain the test accuracy ACC.
The thirteenth step: and 5-fold cross validation is adopted, namely the tenth step to the twelfth step are repeated for 5 times to obtain the test accuracy ACC for 5 times, and the average result is calculated to obtain the final classification accuracy of 0.7148.

Claims (3)

1. An ICA-CNN classified fMRI data space pre-smoothing augmentation method, which is characterized in that a new fMRI data set is generated by applying space smoothing to fMRI observation data and augmented, and then the fMRI data set is sent to an ICA-CNN frame to realize the improvement of the classification performance of schizophrenic patients and healthy people; the method is characterized by comprising the following steps:
the first step is as follows: inputting the four-dimensional fMRI observation amplitude data of the tested k
Figure FDA0002281880620000011
K is the total number of trials, T is the number of scan time points, X, Y, Z is the three dimensions of brain space;
the second step is that: the four-dimensional fMRI data of each tested object
Figure FDA0002281880620000012
Decomposition into a series of three-dimensional spatial signals x along a time axis(k)(1,x,y,z),…,x(k)(t,x,y,z),…,x(k)(T, x, y, z), wherein x(k)(T, X, Y, Z) is a three-dimensional scan at time T, T is 1, …, T, X is 1, …, X, Y is 1, …, Y, Z is 1, …, Z, K is 1, …, K;
the third step: smoothing the space, namely smoothing the space data x of the tested k at the time point t(k)(t, x, y, z), K1, …, K, convolved with a three-dimensional gaussian filter to achieve spatial smoothing as shown by:
Figure FDA0002281880620000013
in the formula
Figure FDA0002281880620000014
Representing a three-dimensional convolution calculation, g (x, y, z) is a three-dimensional zero-mean gaussian function defined as:
Figure FDA0002281880620000015
wherein, σ is standard deviation, determines the width of the filter, the larger σ is, the larger smoothing degree is, the full width at half maximum FWHM is a common parameter for measuring the size of Gaussian kernel, and has a linear relation with σ as follows,
Figure FDA0002281880620000016
smoothing the observation data of all the tested data by using G Gaussian filters with different FWHM sizes, expanding each tested data set from an original group to a G group, and recording the G group of expanded data as tested k
Figure FDA0002281880620000017
The fourth step: will be provided with
Figure FDA0002281880620000018
Is one-dimensional, i.e. the size of the space dimension is equal to X × Y × Z, i is 1, …, G, then the extrabrain voxels are removed, and only the intracerebral voxels are taken, obtaining
Figure FDA0002281880620000019
V is the number of endosomes, V<X×Y×Z；
The fifth step: PCA pairs Using principal component analysis
Figure FDA0002281880620000021
Reducing the vitamin content to obtain
Figure FDA0002281880620000022
N is the model order, and N is less than or equal to T;
and a sixth step: ICA separation and interesting component extraction, and Infmax algorithm pair
Figure FDA0002281880620000023
Figure FDA0002281880620000024
Carrying out ICA separation; selecting a spatial activation map of the interested component from N ICA separation components based on the principle of maximum correlation coefficient with a spatial component template
Figure FDA0002281880620000025
The seventh step: to pair
Figure FDA0002281880620000026
Performing extrabrain element zero filling, and restoring to three-dimensional space activation map, i.e.
Figure FDA0002281880620000027
Eighth step: will be provided with
Figure FDA0002281880620000028
Expand along the Z-axis into Z two-dimensional stacks of slices of size X Y, noted
Figure FDA0002281880620000029
Selecting W slices with large activation information according to the effective activation positions of the interested components, wherein W is less than or equal to Z and is recorded as L, … and L + W-1;
the ninth step: cutting W pieces of two-dimensional slices
Figure FDA00022818806200000210
Normalized to [0,1 ] values]Is marked as
Figure FDA00022818806200000211
Then attaching a corresponding label to each slice
Figure FDA00022818806200000212
Figure FDA00022818806200000213
It is indicative of a healthy person,
Figure FDA00022818806200000214
representing patients with schizophrenia, establishing an augmented CNN sample set
Figure FDA00022818806200000215
Expanding each tested sample set from original W to W multiplied by G, namely expanding the W multiplied by G, and expanding all K tested sample sets to K multiplied by W multiplied by G;
the tenth step: disorganizing the sequence of the tested sample, and proportionally dividing the sample set into a training set, a verification set and a test set;
the eleventh step: sending the training set into a CNN network for training; the CNN network structure comprises 2 convolutional layers, 2 maximum pooling layers, 1 full-connection layer and an output layer, wherein the size of each convolutional layer core is 3 multiplied by 3, and the number of the convolutional layers is 8 and 16 respectively; the size of the pooling layer core is 2 x 2, and the number of the pooling layer cores is 8 and 16 respectively; the method comprises the steps that a full-connection layer comprises 64 nodes, a modified linear unit ReLU is used as an activation function, an output layer gives the category of each slice by using Softmax, the batch processing size is 64, parameter updating is carried out by using an Adam algorithm, a two-class cross entropy function is used as a loss function of a network, and L is used simultaneously2Regularizing control weight, training a CNN model in R rounds, and calculating an accuracy ACC by using a verification set in each round; the CNN model with the highest verification accuracy in the R round is stored as the CNN model obtained by training, and the CNN verification result is made to be
Figure FDA0002281880620000031
Accuracy ACC is defined as follows:
Figure FDA0002281880620000032
wherein TP represents true positive, i.e.
Figure FDA0002281880620000033
TN indicates true negatives, i.e.
Figure FDA0002281880620000034
FP represents a false positive, i.e
Figure FDA0002281880620000035
FN indicates false negatives, i.e.
Figure FDA0002281880620000036
The twelfth step: inputting the test set into the CNN model obtained by the eleventh training step to obtain a test accuracy ACC;
the thirteenth step: and 5-fold cross validation is adopted, namely the tenth step and the twelfth step are repeated for 5 times to obtain the accuracy ACC of the 5 times of tests, and the average result is calculated and recorded as the final classification accuracy of the CNN.
2. The ICA-CNN classified fMRI data space pre-smoothing augmentation method as claimed in claim 1, wherein FWHM is 2-3 times of fMRI voxel size.
3. The ICA-CNN classified fMRI data space pre-smoothing augmentation method as claimed in claim 1 or 2, wherein in the tenth step, 60% of the sample set is used as the training set, 20% is used as the verification set, and 20% is used as the test set.
CN201911144803.3A 2019-11-21 2019-11-21 ICA-CNN classified fMRI data space pre-smoothing and broadening method Active CN111046918B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911144803.3A CN111046918B (en) 2019-11-21 2019-11-21 ICA-CNN classified fMRI data space pre-smoothing and broadening method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911144803.3A CN111046918B (en) 2019-11-21 2019-11-21 ICA-CNN classified fMRI data space pre-smoothing and broadening method

Publications (2)

Publication Number Publication Date
CN111046918A true CN111046918A (en) 2020-04-21
CN111046918B CN111046918B (en) 2022-09-20

Family

ID=70231787

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911144803.3A Active CN111046918B (en) 2019-11-21 2019-11-21 ICA-CNN classified fMRI data space pre-smoothing and broadening method

Country Status (1)

Country Link
CN (1) CN111046918B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117257312A (en) * 2023-11-20 2023-12-22 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) Method for augmenting magnetoencephalography data in machine learning

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108903942A (en) * 2018-07-09 2018-11-30 大连理工大学 A method of utilizing plural number fMRI spatial source phase identification spatial diversity
CN109222972A (en) * 2018-09-11 2019-01-18 华南理工大学 A kind of full brain data classification method of fMRI based on deep learning
CN110110776A (en) * 2019-04-28 2019-08-09 大连理工大学 A kind of tranquillization state plural number fMRI data ICA-CNN taxonomy model of patient and Healthy People

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108903942A (en) * 2018-07-09 2018-11-30 大连理工大学 A method of utilizing plural number fMRI spatial source phase identification spatial diversity
CN109222972A (en) * 2018-09-11 2019-01-18 华南理工大学 A kind of full brain data classification method of fMRI based on deep learning
CN110110776A (en) * 2019-04-28 2019-08-09 大连理工大学 A kind of tranquillization state plural number fMRI data ICA-CNN taxonomy model of patient and Healthy People

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
马士林等: "fMRI动态功能网络构建及其在脑部疾病识别中的应用", 《计算机科学》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117257312A (en) * 2023-11-20 2023-12-22 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) Method for augmenting magnetoencephalography data in machine learning
CN117257312B (en) * 2023-11-20 2024-01-26 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) Method for augmenting magnetoencephalography data in machine learning

Also Published As

Publication number Publication date
CN111046918B (en) 2022-09-20

Similar Documents

Publication Publication Date Title
CN108491849B (en) Hyperspectral image classification method based on three-dimensional dense connection convolution neural network
CN104537647B (en) A kind of object detection method and device
CN107944442A (en) Based on the object test equipment and method for improving convolutional neural networks
CN104299232B (en) SAR image segmentation method based on self-adaptive window directionlet domain and improved FCM
CN108830243A (en) Hyperspectral image classification method based on capsule network
CN107909109A (en) SAR image sorting technique based on conspicuousness and multiple dimensioned depth network model
CN110110776B (en) Method for constructing resting state complex fMRI data ICA-CNN classification framework of patient and healthy person
CN107491793B (en) Polarized SAR image classification method based on sparse scattering complete convolution
CN113095149A (en) Full-head texture network structure based on single face image and generation method
CN111832431A (en) Emotional electroencephalogram classification method based on CNN
CN104268561B (en) High spectrum image solution mixing method based on structure priori low-rank representation
CN111046918B (en) ICA-CNN classified fMRI data space pre-smoothing and broadening method
CN110870770B (en) ICA-CNN classified fMRI space activation map smoothing and broadening method
CN102567997A (en) Target detection method based on sparse representation and visual cortex attention mechanism
Grigorescu et al. Interpretable convolutional neural networks for preterm birth classification
CN107818567A (en) Brain local morphological feature description method based on cortical top point cloud
Ghasemzadeh et al. Hyperspectral face recognition using 3D discrete wavelet transform
CN110916661B (en) ICA-CNN classified fMRI intracerebral data time pre-filtering and amplifying method
CN116421200A (en) Brain electricity emotion analysis method of multi-task mixed model based on parallel training
CN114187475A (en) Method for explaining CNN classification result of multi-test complex fMRI data based on thermodynamic diagram
Hu et al. Hyperspectral image superresolution via deep structure and texture interfusion
Srivatsa et al. Application of least square denoising to improve admm based hyperspectral image classification
Seibert et al. Separable cosparse analysis operator learning
CN108197640A (en) High spectrum image fast filtering method based on three-dimensional Gabor filter
CN114176518B (en) Complex fMRI data space component phase anti-correction method for improving CNN classification performance

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