CN103383716A - Senile dementia computer-aided diagnosis method based on 18F-FDG PET image - Google Patents
Senile dementia computer-aided diagnosis method based on 18F-FDG PET image Download PDFInfo
- Publication number
- CN103383716A CN103383716A CN 201310259764 CN201310259764A CN103383716A CN 103383716 A CN103383716 A CN 103383716A CN 201310259764 CN201310259764 CN 201310259764 CN 201310259764 A CN201310259764 A CN 201310259764A CN 103383716 A CN103383716 A CN 103383716A
- Authority
- CN
- China
- Prior art keywords
- pet image
- fdg pet
- registration
- senile dementia
- training
- 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.)
- Pending
Links
Images
Landscapes
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention provides a senile dementia computer-aided diagnosis method based on an 18F-FDG PET image, which particularly comprises the following steps: step 1, standard maps of healthy men and women at different ages are built; step 2, training samples are selected; step 3, dimension reduction processing is performed to the training samples by adopting a third order principal element analytic method and a linear discriminate analysis method, and training characteristics of the training samples are extracted; step 4, for the situation of combination of each age group and gender, classifiers are built according to the corresponding training characteristics; step 5, training characteristics of the 18F-FDG PET image of a social visitor seeking medical treatment are acquired according to the method in the step 3, and the classifier corresponding to the age and gender of the social visitor seeking medical treatment is utilized to perform auxiliary diagnosis to senile dementia according to the training characteristics. The accurate diagnosis of senile dementia can be realized by utilizing the method provided by the invention, and the diagnosis efficiency is higher.
Description
Technical field
The present invention relates to technical field of medical image processing, be specifically related to a kind of senile dementia computer aided diagnosing method based on 18F-FDG PET image.
Background technology
Senile dementia, refer to a kind of continuation higher nerve functional activity obstacle, namely do not having under the state of the disturbance of consciousness, the obstacle of the aspects such as memory, thinking, analysis judgement, visual space identification, mood, along with the increase year by year of world population mean lifetime, population in the world is tending towards aging, and the incidence of disease of senile dementia also rises year by year, become into a large hidden danger of side of body people health in old age, also do not have at present a kind of effective method that it is treated.In a single day the mankind suffer from this illness, and As time goes on, it is more and more serious that the state of an illness can become, and can't reverse.Therefore, just seeming for the early diagnosis of this illness is even more important, and this provides good chance for the development that delays the state of an illness.
Relevant studies show that, the patient who has suffered from senile dementia, before the senile dementia illness occurred, its cerebral glucose metabolism will occur extremely.Therefore by judging the situation of cerebral glucose metabolism, just can diagnose this illness.
Some large hospitals have been bought positron imaging (PET at present, position emission tomography) imager, use 18F-FDG (fluorodeoxyglucose) so that 18F-FDG PET clear picture reflects the situation of cerebral glucose metabolism, to be widely used in the detection diagnosis of senile dementia.Therefore by computing machine, the 18F-FDG PET image of brain is processed, a kind of aid of automatic diagnosis can be provided for the doctor.
Summary of the invention
In view of this, the present invention proposes a kind of senile dementia computer aided diagnosing method based on 18F-FDG PET image, the method has very high diagnostic accuracy.
Realize that technical scheme of the present invention is as follows:
A kind of senile dementia computer aided diagnosing method based on 18F-FDG PET image, detailed process is:
At all age group, select respectively 18F-FDG PET image and the three-dimensional NMR structural images of N healthy male and N healthy women, and select the 18F-FDG PET image of N senile dementia male patient and N senile dementia female patient and three-dimensional NMR structural images as training sample;
The standard diagram that three-dimensional NMR structural images in training sample is corresponding with its age bracket of living in and sex carries out non-rigid registration, obtains corresponding deformation field; Adopt the rigid registration method, 18F-FDG PET image correspondence is registrated on the nuclear magnetic resonance structural images, obtain the 18F-FDG PET image after registration; Described deformation field is acted on 18F-FDG PET image after registration, obtain the 18F-FDG PET image with the respective standard atlas registration;
For 18F-FDG PET image Y each age bracket different sexes and the standard diagram registration
i C, adopt three rank pca methods and linear discriminant analysis method to carry out dimension-reduction treatment to it, extract the training characteristics of training sample, wherein, and i=1,2,3 ... N, C=1,2, C=1 represents health, C=2 represents ill;
Described dimension-reduction treatment, the detailed process of extracting training characteristics is:
Step 101, with the 18F-FDG PET image Y of standard diagram registration
i 1Consist of 3 rank tensor A
i, find the solution described tensor A
iAverage
And order
Step 102, general
Estimated value
With a low-dimensional 3 rank core tensors
With 3 basis matrix U
(1), U
(2), U
(3)Expression, namely
Calculate the 18F-FDG PET image Y with the standard diagram registration
i 1Corresponding low-dimensional 3 rank core tensors
And it is designated as
Step 104, for the 18F-FDG PET image Y of standard diagram registration
i 2, according to the method for step 101-103, the 18F-FDG PET image Y after the calculating registration
i 2Corresponding low-dimensional 3 rank core tensors
Open it is designated as
Step 105, employing linear discriminant analysis method are to low-dimensional 3 rank core tensors
With
Carry out dimension-reduction treatment, obtain respectively Y
i 1And Y
i 2Corresponding training characteristics;
Step 4, for the combined situation of each age bracket and sex, the training characteristics corresponding according to it set up sorter;
Step 5, obtain the training characteristics of the 18F-FDG PET image of asking the doctor according to the method for step 3, utilize the sorter corresponding with asking doctor's age and sex training characteristics to be carried out the auxiliary diagnosis of senile dementia.
Further, sorter of the present invention is to adopt the method for support vector machine to set up.
Beneficial effect
At first, the present invention is by using 18F-FDG PET image, adopt three rank pca methods and linear discriminant analysis method to carry out dimension-reduction treatment to it, extract the training characteristics of training sample, and set up sorter based on training characteristics, the sorter that utilization is set up is realized the auxiliary diagnosis to senile dementia; The present invention is carrying out dimension-reduction treatment to training sample, can improve diagnosis efficiency of the present invention.
Secondly, the present invention is based on training characteristics, utilize the sorting technique of support vector machine, can obtain very high classification results, thereby effectively play the effect of auxiliary diagnosis.
Again, the present invention is when 18F-FDG PET image registration is to the standard diagram, utilize the nuclear magnetic resonance structural images as medium, because the nuclear magnetic resonance structural images has high resolution, than directly with 18F-FDG PET image registration to standard diagram, adopt method for registering of the present invention to have higher accuracy.
Description of drawings
Fig. 1 is the process flow diagram of senile dementia computer aided diagnosing method of the present invention;
Fig. 2 is the schematic diagram that 3 dimension tensors are launched into each mode matrix.
Embodiment
The present invention is described in detail below in conjunction with the drawings and specific embodiments.
As shown in Figure 1, the present invention is based on the senile dementia computer aided diagnosing method of 18F-FDG PET image, concrete process is:
According to selected age bracket (usually greater than 40 years old, 5-10 year an age bracket), choose respectively the three-dimensional NMR structural images of some (quantity is greater than 100) healthy males and healthy women, adopt method (the groupwise B-spline registration of non-rigid body, Dartel, ANTS etc.), set up male sex's standard diagram and women's standard diagram of all ages and classes section based on above-mentioned three-dimensional NMR structural images, the size of this standard diagram is N
x* N
y* N
z
At all age group, select respectively 18F-FDG PET image and the three-dimensional NMR structural images of N healthy male and N healthy women, and select N by having of making a definite diagnosis of the doctor 18F-FDG PET image with senile dementia female patient that the senile dementia male sex and N made a definite diagnosis by the doctor and three-dimensional NMR structural images (quantity is greater than 30) as training sample;
The standard diagram that three-dimensional NMR structural images in training sample is corresponding with its age bracket of living in and sex carries out non-rigid registration (B-spline registration, Dartel, ANTS etc.), obtains corresponding deformation field; Adopt the rigid registration method, 18F-FDG PET image correspondence is registrated on the three-dimensional NMR structural images, and described deformation field is applied on 18F-FDG PET image after corresponding registration, obtain the 18F-FDG PET image with the respective standard atlas registration;
the present invention is because the pixel of 18F-FDG PET image is not high, directly may there be the not high situation of registration accuracy in 18F-FDG PET image registration to standard diagram, the present invention with the higher three-dimensional NMR structural images of pixel value as medium, first the standard diagram that the nuclear magnetic resonance structural images is corresponding with it carries out registration, obtain the corresponding deformation field, again 18F-FDG PET image and nuclear magnetic resonance structural images are carried out registration (because this two image is all to derive from same people, therefore the precision of its registration is higher) obtain the 18F-FDGPET image after registration, then described deformation field is acted on the 18F-FDG PET image after registration, namely realized 18F-FDG PET image registration to the standard diagram corresponding with it, thereby guaranteed registration accuracy.
For 18F-FDG PET image Y each age bracket different sexes and the standard diagram registration
i C, adopt three rank pca methods and linear discriminant analysis method to carry out dimension-reduction treatment to it, extract the training characteristics of training sample, wherein, and i=1,2,3 ... N, C=1,2, when C=1 represents health, when C=2 represents ill.
Suppose in step 1 since 40 years old, take 5 years old as step-length, 10 age brackets have been chosen, comprise masculinity and femininity in each age bracket, there are 20 kinds of situations in the combination of age bracket and sex, and each in 20 kinds of situations all includes 18F-FDG PET image ill and healthy and the standard diagram registration, and the method for the training characteristics that each situation is obtained is identical, and the below describes arbitrary situation.
Dimension-reduction treatment, the detailed process of extracting training characteristics is:
Because each situation comprises healthy and the 18F-FDG PET image Y standard diagram registration
i 1With 18F-FDG PET image Y ill and the standard diagram registration
i 2, following steps 101-step 103 is to Y
i 1And Y
i 2Carry out respectively;
Step 101, with the 18F-FDG PET image Y of standard diagram registration
i 1Consist of an I
1* I
2* I
33 rank tensor A
i, I wherein
1=N
x, I
2=N
y, I
3=N
z, find the solution described tensor A
iAverage
And order
Step 102, by adopting the Turkcy model, will
Estimated value
With a low-dimensional 3 rank core tensors
(dimension is J
1* J
2* J
3, J
1<<I
1, J
2<<I
2, J
3<<I
3, this dimension is artificially determined) and 3 basis matrix U
(1), U
(2), U
(3)Expression, namely
The below describes this expression principle:
Use the multiplication of tensor here, for 3 rank tensor I
1* I
2* I
3Can expand into a size is I
1* (I
2I
3) matrix, be called pattern 1 (mode-1), perhaps expanding into a size is I
2* (I
3I
1) matrix, be called pattern 2 (mode-2), perhaps expanding into a size is I
3* (I
1I
2) matrix, be called mode 3 (mode-3).
Fig. 2 is that one 3 dimension tensor is launched into illustrating of 3 mode matrix.Mode-1maxtrixF wherein
(1)Be expressed as pattern 1, the vector that wherein launches to obtain comprises I
1Row and I
2* I
3Row, Mode-2maxtrixF
(2)And Mode-3maxtrixF
(3)With Mode-1maxtrixF
(1)Similar.Simultaneously, tensor
Pattern 1 matrix I
1* (I
2I
3) can be J with any size
u* I
1Matrix U multiply each other, the representation of multiplied result is
The expression tensor
Pattern 1 multiplication of matrices.Tensor
Pattern 2 matrix I
2* (I
3I
1) can be J with any size
n* I
2Matrix U multiply each other, the representation of multiplied result is
The expression tensor
Pattern 2 multiplications of matrices.Tensor
Mode 3 matrix I
3* (I
1I
2) can be J with any size
n* I
3Matrix U multiply each other, the representation of multiplied result is
The expression tensor
The mode 3 multiplication of matrices.
Calculate the 18F-FDG PET image Y with the standard diagram registration
i 1Corresponding low-dimensional 3 rank core tensors
And it is designated as
The process that the present invention finds the solution optimum solution is prior art, does not elaborate at this.
Simultaneously, arbitrarily and the tensor A of the 18F-FDG PET image construction of standard diagram registration
iCan pass through low-dimensional 3 rank core tensors
And optimum solution
Expression, that is:
Step 104, for the 18F-FDG PET image Y of standard diagram registration
i 2, according to the method for step 101-103, the 18F-FDG PET image Y after the calculating registration
i 2Corresponding low-dimensional 3 rank core tensors
Open it is designated as
Step 105, with described low-dimensional 3 rank core tensors
In element be in line in order
With described low-dimensional 3 rank core tensors
In element be in line in order
According to the classification under training sample, can adopt the method for linear discriminant analysis, to low-dimensional 3 rank core tensors
With
Carry out dimension-reduction treatment, obtain respectively Y
i 1And Y
i 2Corresponding training characteristics.
Concrete grammar is as follows:
Scatter matrix in scatter matrix and class between the model class
Wherein, C=2 is classification (health, ill), N
jThe sample number of corresponding classification, μ
jBe the average of corresponding sample, μ is the average of (in any situation) all samples.
The dimension of setting W equals n (n<10), and the criterion function below optimization finds the projection matrix that has one group of optimum discriminant vector to consist of
With the core tensor
In element be in line in order
Project to this matrix W
opt, its projection coefficient consists of final training characteristics; With the core tensor
In element be in line in order
Project to this matrix W
opt, its projection coefficient consists of final training characteristics.
Step 4, for the combined situation of each age bracket and sex, the training characteristics corresponding according to it set up sorter;
The training characteristics that utilization obtains adopts the method for support vector machine to set up sorter; Corresponding like this each age bracket and different sexes have been set up corresponding sorter.
Step 5, obtain the training characteristics of the 18F-FDG PET image of asking the doctor according to the method for step 3, utilize the sorter corresponding with asking doctor's age and sex training characteristics to be carried out the auxiliary diagnosis of senile dementia.Namely
At first, obtain 18F-FDG PET image and the three-dimensional NMR structural images of asking the doctor;
Secondly, according to the method for step 3, the 18F-FDG PET image of asking the doctor is processed, obtained training characteristics; That is:
Use rigid registration with 18F-FDG PET image registration on the three-dimensional NMR structural images, according to its sex and age section, with its nuclear magnetic resonance structural images and corresponding standard diagram registration (B-spline registration, Dartel, ANTS etc.), obtain corresponding deformation field, this deformation field is applied on 18F-FDG PET image after rigid registration, obtain and the 18F-FDG PET image of corresponding standard diagram registration.Adopt the method for three rank pivot analysis and the method for linear discriminant analysis to carry out dimension-reduction treatment, obtain training characteristics;
Again, according to the age of asking the doctor and sex, select corresponding sorter that training characteristics is classified, diagnose out and ask whether the doctor is senile dementia.
In sum, these are only preferred embodiment of the present invention, is not for limiting protection scope of the present invention.Within the spirit and principles in the present invention all, any modification of doing, be equal to replacement, improvement etc., within all should being included in protection scope of the present invention.
Claims (3)
1. the senile dementia computer aided diagnosing method based on 18F-FDG PET image, is characterized in that, detailed process is:
Step 1, based on the three-dimensional NMR structural images of healthy male and the healthy women of all ages and classes section, Criterion collection of illustrative plates;
Step 2, selection training sample;
At all age group, select respectively 18F-FDG PET image and the three-dimensional NMR structural images of N healthy male and N healthy women, and select the 18F-FDG PET image of N senile dementia male patient and N senile dementia female patient and three-dimensional NMR structural images as training sample;
Step 3, based on standard diagram, extract the training characteristics of training sample;
The standard diagram that three-dimensional NMR structural images in training sample is corresponding with its age bracket of living in and sex carries out non-rigid registration, obtains corresponding deformation field; Adopt the rigid registration method, 18F-FDG PET image correspondence is registrated on the three-dimensional NMR structural images, obtain the 18F-FDG PET image after registration; Described deformation field is acted on 18F-FDG PET image after registration, obtain the 18F-FDG PET image with the respective standard atlas registration;
For 18F-FDG PET image Y each age bracket different sexes and the respective standard atlas registration
i C, adopt three rank pca methods and linear discriminant analysis method to carry out dimension-reduction treatment to it, extract the training characteristics of training sample, wherein, and i=1,2,3 ... N, C=1,2, C=1 represents health, C=2 represents ill;
Described dimension-reduction treatment, the detailed process of extracting training characteristics is:
Step 101, with the 18F-FDG PET image Y of standard diagram registration
i 1Consist of 3 rank tensor A
i, find the solution described tensor A
iAverage
And order
Step 102, general
Estimated value
With a low-dimensional 3 rank core tensors
With 3 basis matrix U
(1), U
(2), U
(3)Expression, namely
Calculate the 18F-FDG PET image Y with the standard diagram registration
i 1Corresponding low-dimensional 3 rank core tensors
And it is designated as
Step 104, for the 18F-FDGPET image Y of standard diagram registration
i 2, according to the method for step 101-103, the 18F-FDG PET image Y after the calculating registration
i 2Corresponding low-dimensional 3 rank core tensors
Open it is designated as
Step 105, employing linear discriminant analysis method are to low-dimensional 3 rank core tensors
With
Carry out dimension-reduction treatment, obtain respectively Y
i 1And Y
i 2Corresponding training characteristics;
Step 4, for the combined situation of each age bracket and sex, the training characteristics corresponding according to it set up sorter;
Step 5, obtain the training characteristics of the 18F-FDG PET image of asking the doctor according to the method for step 3, utilize the sorter corresponding with asking doctor's age and sex training characteristics to be carried out the auxiliary diagnosis of senile dementia.
2. senile dementia computer aided diagnosing method according to claim 1, is characterized in that described N 〉=30.
3. senile dementia computer aided diagnosing method according to claim 1, is characterized in that, described sorter is to adopt the method for support vector machine to set up.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201310259764 CN103383716A (en) | 2013-06-26 | 2013-06-26 | Senile dementia computer-aided diagnosis method based on 18F-FDG PET image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201310259764 CN103383716A (en) | 2013-06-26 | 2013-06-26 | Senile dementia computer-aided diagnosis method based on 18F-FDG PET image |
Publications (1)
Publication Number | Publication Date |
---|---|
CN103383716A true CN103383716A (en) | 2013-11-06 |
Family
ID=49491504
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 201310259764 Pending CN103383716A (en) | 2013-06-26 | 2013-06-26 | Senile dementia computer-aided diagnosis method based on 18F-FDG PET image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103383716A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104361318A (en) * | 2014-11-10 | 2015-02-18 | 中国科学院深圳先进技术研究院 | Disease diagnosis auxiliary system and disease diagnosis auxiliary method both based on diffusion tensor imaging technology |
-
2013
- 2013-06-26 CN CN 201310259764 patent/CN103383716A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104361318A (en) * | 2014-11-10 | 2015-02-18 | 中国科学院深圳先进技术研究院 | Disease diagnosis auxiliary system and disease diagnosis auxiliary method both based on diffusion tensor imaging technology |
CN104361318B (en) * | 2014-11-10 | 2018-02-06 | 中国科学院深圳先进技术研究院 | A kind of medical diagnosis on disease accessory system based on diffusion tensor technology |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Bron et al. | Standardized evaluation of algorithms for computer-aided diagnosis of dementia based on structural MRI: the CADDementia challenge | |
Alam et al. | Alzheimer disease classification using KPCA, LDA, and multi‐kernel learning SVM | |
Wolz et al. | LEAP: learning embeddings for atlas propagation | |
Choi et al. | Fast and robust segmentation of the striatum using deep convolutional neural networks | |
Bellec et al. | Multi-level bootstrap analysis of stable clusters in resting-state fMRI | |
Challis et al. | Gaussian process classification of Alzheimer's disease and mild cognitive impairment from resting-state fMRI | |
Hu et al. | Multi-scale features extraction from baseline structure MRI for MCI patient classification and AD early diagnosis | |
Pachauri et al. | Topology-based kernels with application to inference problems in Alzheimer's disease | |
Lerch et al. | Cortical thickness measured from MRI in the YAC128 mouse model of Huntington's disease | |
CN104715260B (en) | Based on the multi-modal fusion image classification method of RLS-ELM | |
Mori et al. | Atlas-based neuroinformatics via MRI: harnessing information from past clinical cases and quantitative image analysis for patient care | |
CN110534195B (en) | Alzheimer disease detection method based on data space transformation | |
Tong et al. | Five-class differential diagnostics of neurodegenerative diseases using random undersampling boosting | |
Roy et al. | QuickNAT: segmenting MRI neuroanatomy in 20 seconds | |
CN106096636A (en) | A kind of Advancement Type mild cognition impairment recognition methods based on neuroimaging | |
US10307139B2 (en) | System, method and computer-accessible medium for diffusion imaging acquisition and analysis | |
Kuceyeski et al. | The generation and validation of white matter connectivity importance maps | |
Jongkreangkrai et al. | Computer-aided classification of Alzheimer's disease based on support vector machine with combination of cerebral image features in MRI | |
Yang et al. | Large-scale brain functional network integration for discrimination of autism using a 3-D deep learning model | |
Misiura et al. | Cognitive control, learning, and clinical motor ratings are most highly associated with basal ganglia brain volumes in the premanifest Huntington’s disease phenotype | |
Eskildsen et al. | Detecting Alzheimer’s disease by morphological MRI using hippocampal grading and cortical thickness | |
CN114663445A (en) | Three-dimensional heart image segmentation method based on multi-scale edge perception | |
CN108805181A (en) | A kind of image classification device and sorting technique based on more disaggregated models | |
CN116051545B (en) | Brain age prediction method for bimodal images | |
CN112233805A (en) | Mining method for biomarkers based on multi-map neuroimaging data |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20131106 |