CN108682009A - A kind of Alzheimer's disease prediction technique, device, equipment and medium - Google Patents
A kind of Alzheimer's disease prediction technique, device, equipment and medium Download PDFInfo
- Publication number
- CN108682009A CN108682009A CN201810431323.4A CN201810431323A CN108682009A CN 108682009 A CN108682009 A CN 108682009A CN 201810431323 A CN201810431323 A CN 201810431323A CN 108682009 A CN108682009 A CN 108682009A
- Authority
- CN
- China
- Prior art keywords
- image
- alzheimer
- disease
- grey matter
- magnetic resonance
- 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
- 238000000034 method Methods 0.000 title claims abstract description 114
- 208000024827 Alzheimer disease Diseases 0.000 title claims abstract description 110
- 210000004884 grey matter Anatomy 0.000 claims abstract description 68
- 210000004556 brain Anatomy 0.000 claims abstract description 58
- 230000008569 process Effects 0.000 claims abstract description 51
- 238000005481 NMR spectroscopy Methods 0.000 claims abstract description 46
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 42
- 230000004913 activation Effects 0.000 claims description 47
- 238000012549 training Methods 0.000 claims description 37
- 238000012545 processing Methods 0.000 claims description 24
- 210000004885 white matter Anatomy 0.000 claims description 22
- 238000004590 computer program Methods 0.000 claims description 20
- 238000012216 screening Methods 0.000 claims description 19
- 230000006870 function Effects 0.000 claims description 18
- 201000010099 disease Diseases 0.000 claims description 14
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims description 14
- 238000006243 chemical reaction Methods 0.000 claims description 12
- 230000006835 compression Effects 0.000 claims description 12
- 238000007906 compression Methods 0.000 claims description 12
- 238000009499 grossing Methods 0.000 claims description 10
- 238000012797 qualification Methods 0.000 claims description 7
- 230000011218 segmentation Effects 0.000 claims description 7
- 238000011156 evaluation Methods 0.000 claims description 5
- 238000001514 detection method Methods 0.000 abstract description 13
- 230000002265 prevention Effects 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 11
- 238000013528 artificial neural network Methods 0.000 description 6
- 238000012360 testing method Methods 0.000 description 5
- 239000013598 vector Substances 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 210000002569 neuron Anatomy 0.000 description 4
- 238000007781 pre-processing Methods 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 210000001175 cerebrospinal fluid Anatomy 0.000 description 3
- 238000003709 image segmentation Methods 0.000 description 3
- 238000002203 pretreatment Methods 0.000 description 3
- 108090000623 proteins and genes Proteins 0.000 description 3
- 210000001519 tissue Anatomy 0.000 description 3
- 101710137189 Amyloid-beta A4 protein Proteins 0.000 description 2
- 101710151993 Amyloid-beta precursor protein Proteins 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 2
- 206010012289 Dementia Diseases 0.000 description 2
- 208000026139 Memory disease Diseases 0.000 description 2
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 2
- 235000003140 Panax quinquefolius Nutrition 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000001037 epileptic effect Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 235000008434 ginseng Nutrition 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000010030 laminating Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 102000004169 proteins and genes Human genes 0.000 description 2
- 238000000547 structure data Methods 0.000 description 2
- 208000006888 Agnosia Diseases 0.000 description 1
- 241001047040 Agnosia Species 0.000 description 1
- 208000000044 Amnesia Diseases 0.000 description 1
- 102100022704 Amyloid-beta precursor protein Human genes 0.000 description 1
- 208000016192 Demyelinating disease Diseases 0.000 description 1
- WQZGKKKJIJFFOK-GASJEMHNSA-N Glucose Natural products OC[C@H]1OC(O)[C@H](O)[C@@H](O)[C@@H]1O WQZGKKKJIJFFOK-GASJEMHNSA-N 0.000 description 1
- 206010027336 Menstruation delayed Diseases 0.000 description 1
- 102000015499 Presenilins Human genes 0.000 description 1
- 108010050254 Presenilins Proteins 0.000 description 1
- 208000024777 Prion disease Diseases 0.000 description 1
- 201000007201 aphasia Diseases 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 210000004369 blood Anatomy 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000002490 cerebral effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 208000010877 cognitive disease Diseases 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 230000004064 dysfunction Effects 0.000 description 1
- 238000001962 electrophoresis Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 239000008103 glucose Substances 0.000 description 1
- 230000002489 hematologic effect Effects 0.000 description 1
- 208000015181 infectious disease Diseases 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000006984 memory degeneration Effects 0.000 description 1
- 208000023060 memory loss Diseases 0.000 description 1
- 208000030159 metabolic disease Diseases 0.000 description 1
- 208000027061 mild cognitive impairment Diseases 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 208000015122 neurodegenerative disease Diseases 0.000 description 1
- 238000002610 neuroimaging Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000007639 printing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000036301 sexual development Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- 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
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing
-
- 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]
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30016—Brain
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Public Health (AREA)
- Data Mining & Analysis (AREA)
- Epidemiology (AREA)
- Quality & Reliability (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Primary Health Care (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention discloses a kind of Alzheimer's disease prediction technique, device, terminal device and computer readable storage medium, methods to include the following steps:The nuclear magnetic resonance brain structure image of acquisition is pre-processed, to generate grey matter image;3D convolutional neural networks models according to the grey matter image and based on Alzheimer's disease carry out operation, to export Alzheimer's disease prediction result, the automatic prediction of Alzheimer's disease is realized with high-accuracy, the universal detection process for realizing the large area of Alzheimer's disease, so that resident carries out timely prevention and treatment.
Description
Technical field
The present invention relates to field of computer technology more particularly to a kind of Alzheimer's disease prediction technique, device, terminal to set
Standby and computer readable storage medium.
Background technology
Alzheimer disease (AD) is a kind of nervous system degenerative disease of the progress sexual development of onset concealment.Clinically
With memory disorders, aphasia, appraxia, agnosia, the damage of visual space technical ability, execute dysfunction and personality and behavior change etc. comprehensively
Property dementia performance be characterized.China is that patients with Alzheimer disease number is at most also the country that speedup ranks forefront, in China, 65 years old
There are more than 900 ten thousand people to suffer from Alzheimer disease in above old man, 40,000,000 people suffer from mild cognitive impairment.Although not having also at present
There is the treatment that can stop or reverse the Alzheimer disease course of disease, but if can be in early detection Alzheimer disease, for prolonging
The development of slow Alzheimer disease is critically important, if patients with Alzheimer disease, which reaches an advanced stage, just carries out diagnoses and treatment, just
It is difficult to delay the development of Alzheimer disease again, and the patients with Alzheimer disease in late period can not usually recognize children, Wu Fazheng
Saying language is easy to get lost, and prodigious pain can be all caused to patients with Alzheimer disease and its household.
The existing detection method for Alzheimer disease has following methods:The first:Hematological examination, mainly
For finding existing adjoint disease or complication, the potential risk factor of discovery, excluding dementia caused by other causes of disease;Second
Kind:Neuroimaging inspection, for excluding other potential diseases and finding the specific Radiologic imaging of AD;The third:Brain electricity
Scheme (EEG), the early stage evidence of prion disease be provided, or prompt there may be poisoning-metabolic disorder, temporary epileptic memory loss or
Other epileptic conditions;4th kind:Cerebrospinal fluid detects, cerebrospinal fluid cell count, protein, glucose and protein electrophoresis analysis, blood
Guan Yan, infection or the doubtful person of demyelinating disease should be detected;5th kind:Genetic test, amyloid precursor protein gene
(APP), presenilin 1,2 genes (PS1, PS2) detection.
During realizing the embodiment of the present invention, inventor has found:The equipment that above method is required to profession is examined
The process of survey, detection is expensive and the consuming time is longer, it has not been convenient to which the resident of big quantity carries out Alzheimer disease screening
Process, efficiency is more low, while diagnostic result depends on the diagnostic analysis of medical practitioner, and the level of doctor, which also directly affects, examines
It is disconnected as a result, and medical staff time and efforts it is all limited, above method cannot achieve the universal detection process of large area.
Invention content
In view of the above-mentioned problems, the purpose of the present invention is to provide a kind of Alzheimer's disease prediction technique, device, terminals to set
Standby and computer readable storage medium realizes the automatic prediction of Alzheimer's disease with high-accuracy, realizes Alzheimer
The universal detection process of the large area of disease, so that resident carries out timely prevention and treatment.
In a first aspect, an embodiment of the present invention provides a kind of Alzheimer's disease prediction technique, include the following steps:
The nuclear magnetic resonance brain structure image of acquisition is pre-processed, to generate grey matter image;
3D convolutional neural networks models according to the grey matter image and based on Alzheimer's disease carry out operation, with output
Alzheimer's disease prediction result.
In the first realization method of first aspect, further include:
Picture quality screening is carried out to the nuclear magnetic resonance brain structure original image of input;
Data Format Transform is carried out to the nuclear magnetic resonance brain structure original image of screening qualification, can be analyzed with obtaining
Nuclear magnetic resonance brain structure image.
According to the first realization method of first aspect, in second of realization method of first aspect, described pair of acquisition
Nuclear magnetic resonance brain structure image pre-processed, to generate grey matter image, specially:
The segmentation that grey matter image and white matter image are carried out to the nuclear magnetic resonance brain structure image of acquisition, to generate initial ash
Matter image and initial white matter image;
Non-linear registration is carried out to the initial grey matter image and initial white matter image, to generate registration image;
The registration image is standardized and smoothing processing, with the grey matter image that generates that treated.
It is described according to institute in the third realization method of first aspect according to second of realization method of first aspect
It states grey matter image and the 3D convolutional neural networks models based on Alzheimer's disease carries out operation, it is pre- to export Alzheimer's disease
It surveys as a result, being specially:
Using the grey matter image as the input value of the 3D convolutional neural networks models based on Alzheimer's disease;
In convolutional layer, the characteristic value of the grey matter image is extracted by self-encoding encoder, to generate at least one characteristic pattern
Picture;
Compression processing is carried out at least one characteristic image by pond layer, at least one compression spy is generated with corresponding
Levy image;
First activation primitive conversion is carried out at least one compressive features image, it is corresponding at least one non-to generate
Linear character image;
Feature superposition is carried out at least one nonlinear characteristic image by full articulamentum and the second activation primitive turns
It changes, to generate Superposition Characteristics image;
The Superposition Characteristics image is handled by output layer, to export Alzheimer's disease prediction result.
According to the third realization method of first aspect, in the 4th kind of realization method of first aspect, at the compression
The mode of reason is any one in maximum value pond, minimum value pond or mean value pond.
According to the third realization method of first aspect, in the 5th kind of realization method of first aspect, described first swashs
Function living is any one in ReLU activation primitives, TanH activation primitives and liner activation primitives;Second activation primitive
For any one in ReLU activation primitives, TanH activation primitives and liner activation primitives.
According to any of the above realization method of first aspect, in the 6th kind of realization method of first aspect, based on
In the training process of the 3D convolutional neural networks models of Alzheimer's disease, convolution kernel is trained by autocoder;
In the training process of the 3D convolutional neural networks models based on Alzheimer's disease, will training prediction result collection with
The mean square deviation of legitimate reading collection is as the penalty values predicted in training process;
In the training process of the 3D convolutional neural networks models based on Alzheimer's disease, using accuracy rate and AUC as
The index of the evaluation 3D convolutional neural networks models based on Alzheimer's disease.
Second aspect, an embodiment of the present invention provides a kind of Alzheimer's disease prediction meanss, including:
Pretreatment unit is pre-processed for the nuclear magnetic resonance brain structure image to acquisition, to generate grey matter image;
Model prediction unit, for the 3D convolutional neural networks moulds according to the grey matter image and based on Alzheimer's disease
Type carries out operation, to export Alzheimer's disease prediction result.
In the first realization method of second aspect, further include:
Quality screening unit carries out picture quality screening for the nuclear magnetic resonance brain structure original image to input;
Format conversion unit carries out data format for the nuclear magnetic resonance brain structure original image to screening qualification
Conversion, to obtain analyzable nuclear magnetic resonance brain structure image.
According to the first realization method of second aspect, in second of realization method of second aspect, the pretreatment
Unit specifically includes:
Image segmentation module, for carrying out grey matter image and white matter image to the nuclear magnetic resonance brain structure image of acquisition
Segmentation, to generate initial grey matter image and initial white matter image;
Non-linear registration module, for carrying out non-linear registration to the initial grey matter image and initial white matter image, with
Generate registration image;
Standardization and Leveling Block, for being standardized to the registration image and smoothing processing, after being handled with generation
Grey matter image.
According to second of realization method of second aspect, in the third realization method of second aspect, the model is pre-
Surveying unit is specially:
Input module, for using the grey matter image as the 3D convolutional neural networks moulds based on Alzheimer's disease
The input value of type;
Characteristic extracting module, in convolutional layer, the characteristic value of the grey matter image being extracted by self-encoding encoder, with life
At at least one characteristic image;
Characteristic image compression module, for carrying out compression processing at least one characteristic image by pond layer, with
It is corresponding to generate at least one compressive features image;
First conversion module, for carrying out the first activation primitive conversion at least one compressive features image, with life
At corresponding at least one nonlinear characteristic image;
Feature laminating module, for carrying out feature superposition at least one nonlinear characteristic image by full articulamentum
And second activation primitive conversion, to generate Superposition Characteristics image;
Prediction result output unit is handled the Superposition Characteristics image for passing through output layer, to export A Er
Ci Haimo disease prediction results.
According to the third realization method of second aspect, in the 4th kind of realization method of second aspect, at the compression
The mode of reason is any one in maximum value pond, minimum value pond or mean value pond.
According to the third realization method of second aspect, in the 5th kind of realization method of second aspect, described first swashs
Function living is any one in ReLU activation primitives, TanH activation primitives and liner activation primitives;Second activation primitive
For any one in ReLU activation primitives, TanH activation primitives and liner activation primitives.
According to any of the above realization method of second aspect, in the 6th kind of realization method of second aspect, based on
In the training process of the 3D convolutional neural networks models of Alzheimer's disease, convolution kernel is trained by autocoder;
In the training process of the 3D convolutional neural networks models based on Alzheimer's disease, will training prediction result collection with
The mean square deviation of legitimate reading collection is as the penalty values predicted in training process;
In the training process of the 3D convolutional neural networks models based on Alzheimer's disease, by accuracy rate and AUC curves
Index as the evaluation 3D convolutional neural networks models based on Alzheimer's disease.
The third aspect, an embodiment of the present invention provides a kind of Alzheimer's diseases to predict terminal device, including processor, deposits
Reservoir and it is stored in the memory and is configured as the computer program executed by the processor, the processor is held
Row the computer program when realize it is any one of above-mentioned described in Alzheimer's disease prediction technique.
Fourth aspect, an embodiment of the present invention provides a kind of computer readable storage medium, the computer-readable storage
Medium includes the computer program of storage, wherein controls the computer-readable storage medium when the computer program is run
Equipment where matter execute it is any one of above-mentioned described in Alzheimer's disease prediction technique.
An embodiment of the present invention provides a kind of Alzheimer's disease prediction technique, device, terminal devices and computer-readable
Storage medium has the advantages that:
It is pre-processed by the nuclear magnetic resonance brain structure image to acquisition, to generate grey matter image, then according to institute
It states grey matter image and the 3D convolutional neural networks models based on Alzheimer's disease carries out operation, it is pre- to export Alzheimer's disease
It surveys as a result, realizing that the automatic prediction of Alzheimer's disease, detection process need not rely upon the professional water of doctor with high-accuracy
It is flat, and detection process high-efficient simple, the universal detection process of the large area of Alzheimer's disease can be realized, so that resident carries out
Timely prevention and treatment.
Description of the drawings
In order to illustrate more clearly of technical scheme of the present invention, attached drawing needed in embodiment will be made below
Simply introduce, it should be apparent that, the accompanying drawings in the following description is only some embodiments of the present invention, general for this field
For logical technical staff, without creative efforts, other drawings may also be obtained based on these drawings.
Fig. 1 is the flow diagram for the Alzheimer's disease prediction technique that first embodiment of the invention provides.
Fig. 2 is the schematic diagram for the nuclear magnetic resonance brain structure image preprocessing flow that first embodiment of the invention provides.
Fig. 3 is the structure for the 3D convolutional neural networks models based on Alzheimer's disease that first embodiment of the invention provides
Schematic diagram.
Fig. 4 is the image processing flow schematic diagram of the convolutional layer that first embodiment of the invention provides and pond layer.
Fig. 5 is the characteristic image processing flow schematic diagram for the full articulamentum that first embodiment of the invention provides.
Fig. 6 is the training for the 3D convolutional neural networks models based on Alzheimer's disease that third embodiment of the invention provides
The schematic diagram of flow.
Fig. 7 is the structural schematic diagram of the encoding and decoding process for the autocoder that third embodiment of the invention provides.
Fig. 8 is the structural schematic diagram for the Alzheimer's disease prediction meanss that fourth embodiment of the invention provides.
Specific implementation mode
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation describes, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Referring to Fig. 1, first embodiment of the invention provides a kind of Alzheimer's disease prediction technique, can be set by terminal
It is standby to execute, and include the following steps:
S11 pre-processes the nuclear magnetic resonance brain structure image of acquisition, to generate grey matter image.
In embodiments of the present invention, the terminal device can be realized by way of software and/or hardware, the terminal
Equipment can be that two or more physical entities are constituted, and can also be that a physical entity is constituted.The annotation shows that equipment can be with
It is the computing devices such as desktop PC, notebook, palm PC, mobile terminal, Intelligent flat and cloud server.
In embodiments of the present invention, the analysis software that preprocessing process uses is the SPM software packages based on MATLAB, this is soft
Part packet is absorbed in the processing and analysis of brain phantom data, and pre-treatment step establishes the DARTEL functions in SPM softwares, passes
The nuclear magnetic resonance brain structure data prediction of system establishes environment and SPM software packages when MATLAB is developed, and this system utilizes
MATLAB runtime environments by involved in SPM softwares to whole pre-treatment steps be integrated into code and read initial data automatically
And all pre-treatment steps are run in order, all preprocessing process are run in Cent OS systems.
In embodiments of the present invention, referring to Fig. 2, the terminal device to the nuclear magnetic resonance brain structure image of acquisition into
The segmentation of row grey matter image and white matter image, to generate initial grey matter image and initial white matter image, nuclear magnetic resonance brain structure
Imaging is to be applied in after radio-frequency pulse to have the different relaxation times white to distinguish ectocinerea, brain according to brain different tissues
The tissues such as matter, cerebrospinal fluid, in clinical medicine and scientific research field, most common structure organization is grey matter and white matter, both tissues
Structure is to constitute the main component of human brain, and the pretreated first step is the nuclear magnetic resonance brain structure figure according to acquisition
Picture selects the brain template in SPM softwares to carry out rigid registration, to be partitioned into original grey matter image and original white matter image;So
The terminal device carries out non-linear registration to the initial grey matter image and initial white matter image afterwards, to generate registration image,
Specifically, it is according to the original grey matter image and original white matter image creation template, the effect of drawing template establishment:In view of big
Brain structured data comes from Different Individual and different scanning type, needs the image mapping for creating a reference template by Different Individual
To achieve the purpose that image information fusion on to the template, then the original grey matter image and original white matter image are matched again
In standard to the template, and then the image for being registrated to using these template regenerates a new template, the original grey matter
It is last described repeatedly until obtaining best registration effect on image and original white matter image registration to this new template
Terminal device is standardized to the registration image and smoothing processing, to generate treated grey matter image and white matter image,
Specifically, by the registration image standardization to MNI coordinate spaces, (MNI coordinate spaces are most common human brain coordinates in the world
One of system), Gaussian smoothing then is carried out to grey matter again, used smoothing kernel is 8mm, it should be noted that entire pre- place
Reason process realizes that full automation, system read original brain structure data and then depend on automatically by the form of coding
Environment when MATLAB is run completes MATLAB of the entire preprocessing process without commercial version, while also need not basis
Each step of DARTEL tools carries out manual handle, uses manpower and material resources sparingly, and is carried out to nuclear magnetic resonance brain structure image pre-
The process of processing can obtain higher accuracy rate in the pre- plan process of subsequent model, to reduce the error in clinical application
Rate.
S12, the 3D convolutional neural networks models according to the grey matter image and based on Alzheimer's disease carry out operation, with
Export Alzheimer's disease prediction result.
In embodiments of the present invention, the cardinal principle of neural network is to imitate the operation principle of cerebral neuron, by one
A neuron connects networking, and each neuron performs some processing the signal of input, then will treated signal spreads out of work
For the input of target nerve member, processing mode may be cumulative, it may be possible to which filtering or other modes pass through the letter of single neuron
The extensive repetition of single gauge then forms a complication system, is finally reached the purpose of nonlinear prediction.Referring to Fig. 3, Fig. 3 is
The structural schematic diagram of 3D convolutional neural networks models based on Alzheimer's disease, with level 2 volume accumulate network, 2 layers of pond layer, 4 layers
Fully-connected network, fully-connected network be followed by an output layer be used for export differentiate result for, it should be noted that this hair
The bright number of plies for convolutional network, pond layer and fully-connected network does not do any restrictions, and the fully-connected network is to be layered
Each node of preceding layer can be connect with each node of later layer in structure, and network structure in total is whole nodes
The quantity of cartesian product, the convolutional network need the convolution rule that image is first arranged, matrix multiplication are generally used to be advised as convolution
Then.
In embodiments of the present invention, the terminal device using the grey matter image as described based on Alzheimer's disease
The input value of 3D convolutional neural networks models, referring to Fig. 4, Fig. 4 is the image processing flow schematic diagram of convolutional layer and pond layer,
In convolutional layer, the characteristic value of the grey matter image is extracted by self-encoding encoder, to generate at least one characteristic image, in convolution
In layer, convolution kernel is generated according to the image of self-encoding encoder training input, it may also be said to, the grey matter is extracted by convolution kernel
Then the characteristic value of image carries out at least one characteristic image by pond layer with generating at least one characteristic image
Compression processing generates at least one compressive features image with corresponding, then carries out first at least one compressive features image
Activation primitive is converted, to generate corresponding at least one nonlinear characteristic image, referring to Fig. 5, Fig. 5 is the feature of full articulamentum
Image procossing schematic diagram carries out feature superposition and the second activation by full articulamentum at least one nonlinear characteristic image
Function is converted, and to generate Superposition Characteristics image, is handled the Superposition Characteristics image finally by output layer, to export Ah
Er Cihaimo disease prediction results, wherein the mode of the compression processing is in maximum value pond, minimum value pond or mean value pond
Any one, first activation primitive be ReLU activation primitives, TanH activation primitives and liner activation primitives in it is arbitrary
It is a kind of;Second activation primitive is any one in ReLU activation primitives, TanH activation primitives and liner activation primitives.
In conclusion first embodiment of the invention provides a kind of Alzheimer's disease prediction technique, by acquisition
Nuclear magnetic resonance brain structure image is pre-processed, to generate grey matter image, then according to the grey matter image and based on A Er
The 3D convolutional neural networks models of Ci Haimo diseases carry out operation, to export Alzheimer's disease prediction result, with high-accuracy reality
The automatic prediction of existing Alzheimer's disease, detection process needs not rely upon the professional standards of doctor, and detection process is efficiently simple
Just, the universal detection process that can realize the large area of Alzheimer's disease, so that resident carries out timely prevention and treatment.
In order to facilitate the understanding of the present invention, some currently preferred embodiments of the present invention will be done and will further be retouched below
It states.
Second embodiment of the invention:
On the basis of first embodiment of the invention, further include:
Picture quality screening is carried out to the nuclear magnetic resonance brain structure original image of input.
Data Format Transform is carried out to the nuclear magnetic resonance brain structure original image of screening qualification, can be analyzed with obtaining
Nuclear magnetic resonance brain structure image.
In embodiments of the present invention, the nuclear magnetic resonance brain structure obtained by different types of nuclear magnetic resonane scanne is former
Beginning image, format are DICOM format, and the nuclear magnetic resonane scanne includes but not limited to Siemens, the machines such as Philip and GE
Type requires it to be scanned under the setting of following sweep parameter the nuclear magnetic resonance brain structure original image of input and obtains in of the invention
:Slicce thickness is 1mm or 1.2mm, and field strength is 1.5T or 3.0T, and resolution ratio is 1 × 1 or 1.2 × 1.2, and the terminal is set
After the standby nuclear magnetic resonance brain structure original image to input carries out picture quality screening, to the nuclear magnetic resonance of screening qualification
Brain structure original image carries out Data Format Transform, and the nuclear magnetic resonance brain structure original image of DICOM format is turned
NIFTI formats are changed to, are follow-up pretreatment process to obtain the analyzable nuclear magnetic resonance brain structure image of NIFTI formats
Standard picture is provided, the scheme of abandoning is taken for screening underproof nuclear magnetic resonance brain structure original image.
Third embodiment of the invention:
Referring to Fig. 6, on the basis of above example of the present invention, the 3D convolutional Neurals based on Alzheimer's disease
The training process of network model includes the following steps:
S21 carries out picture quality screening to the nuclear magnetic resonance brain structure sample image of input, and to the institute of screening qualification
It states nuclear magnetic resonance brain structure sample image and carries out Data Format Transform, to obtain sample image.
In embodiments of the present invention, the nuclear magnetic resonance brain knot that the terminal device screening is obtained by parameter preset scanning
Structure sample image, the parameter preset are:Slicce thickness is 1mm or 1.2mm, and field strength is 1.5T or 3.0T, and resolution ratio is 1 × 1
Or 1.2 × 1.2;Then the nuclear magnetic resonance brain structure sample image for being DICOM to the qualified format of screening is converted to format
For the sample image of NIFTI formats.
S22 pre-processes the sample image, to generate grey matter sample image.
In embodiments of the present invention, the terminal device described first is partitioned into grey matter image and white matter according to the sample image
Image;Then the grey matter image of segmentation is subjected to non-linear registration, specifically, first with the grey matter image of segmentation
A template is generated, in the grey matter image registration to the template after then each sample image is divided again, and then is utilized
These images for being registrated to template regenerate a new template, and the grey matter image after each sample image segmentation is matched
In standard to this new template, repeatedly until obtaining best registration effect;Finally to the grey matter image after non-linear registration
It is standardized and smoothing processing, specifically, by the grey matter image standardization after non-linear registration to MNI coordinate spaces, (MNI is sat
Mark space is one of most common human brain coordinate-system in the world), Gaussian smoothing then is carried out to standardized grey matter image again,
Smoothing kernel is 8mm.
S23 is based on Alzheimer according to the legitimate reading of the sample image and the grey matter sample image to described
The 3D convolutional neural networks models of disease are trained to export trained prediction result.
In embodiments of the present invention, there are many kinds of the generations of machine learning convolution kernel, there is artificial customization, generate at random, certainly
Dynamic encoder generates and PAC Principal Component Analysis generates.The present invention is illustrated by taking autocoder training convolutional core as an example,
The principle of autocoder is interpreted as inputting an image into trainable neural network, by the feature extraction of network, then it is defeated
Go out one and makes precision by way of backpropagation using the difference of input and output as penalty values with identical image is inputted
Be continuously improved, this neural network as a result, being exactly convolution kernel.The convolution kernel trained by autocoder can instructing
There are one relatively good convolution kernels when practicing, and trained precision problem are substantially increased, referring to Fig. 7, autocoder is divided into volume
Code and two stages of decoding, are the effective means for extracting high dimensional data feature, coding stage is by convolutional network to target data
Feature extraction is carried out, convolution kernel at this time is generated by the mode of random initializtion, after coding stage, connects one layer of full connection solution
Code carries out full connection by convolution kernel and generates a data, compares the difference of the data of generation and the data of input, generate one
Mean square deviation penalty values are trained for the purpose of optimizing this penalty values, obtain self-encoding encoder as a result, i.e. classification based training when volume
Product core, it should be noted that the convolution kernel is used to carry out convolutional calculation feature unit, general dimension and target dimension to image
Identical, convolution kernel size of the invention is adjustable ginseng item, and for being 8,8,8 with default setting, i.e., length, width and height are all 8 pixels, often
A convolution kernel corresponds to a characteristic pattern, and characteristic pattern is the Feature Mapping of the convolution kernel;What the convolutional layer was made of convolution kernel, be
For the feature set of one layer of neural network, the adjustable ginseng item of convolution kernel numerical digit of each layer of the present invention, with default setting for 128 Hes
The process being trained for 64;In the full articulamentum, each node of last layer and next layer each node
Connection, the process that the present invention is trained by taking totally four layers full articulamentum as an example is 2048 nodes respectively, 512 nodes, 256
A node, 128 nodes;Weight is approximately volume machine core in convolutional network, and in fully-connected network, weight is a vector.
In embodiments of the present invention, described the 3D convolutional neural networks models based on Alzheimer's disease training process
In, a trained prediction result collection is will produce, using the mean square deviation of training prediction result collection and legitimate reading collection as training process
The penalty values of middle prediction, optimal result mean square deviation are 0, regard each layer network as a letter about input image data
Number, mean square deviation equally can be regarded as a function of network parameter, and the standard of prediction can be made by the minimum value found a function
True property reaches highest, and it is the parameter of target network that function, which is minimized corresponding parameter then, and solution procedure can be by under gradient
The mode of drop carries out, and after the decline of constantly gradient, the precision of model can be just improved, the ladder in gradient index meaning
Degree, regards each layer network as a function, penalty values can regard a function about network parameter as, network parameter packet here
Include the weight of full articulamentum, the convolution kernel of biasing and convolutional layer, biasing must first take extreme value if function is most worth, under gradient
The purpose of drop is the value of neural network parameter when obtaining each extreme point counting loss value minimum, to obtain an essence
Highest network structure is spent, this process is referred to as backpropagation, passes through backpropagation so that penalty values reach minimum to obtain
Optimal neural network parameter, it should be noted that the main training of the 3D convolutional neural networks models based on Alzheimer's disease
Program uses predicted value and the mean square deviation of label value, autocoder to generate convolution kernel and use output image and input
The mean square deviation of image.
In embodiments of the present invention, in the training process of the 3D convolutional neural networks models based on Alzheimer's disease,
It is described accurate using accuracy rate and AUC as the index of the evaluation 3D convolutional neural networks models based on Alzheimer's disease
Rate is the ratio that correctly predicted number accounts for overall test number, and the AUC (Area Under Curve) indicates under ROC curve
Area, value is between 0.5-1, and when judging the quality of a binary classification algorithm, AUC can play good reference, AUC
Value closer to 1, algorithm classification effect is better, it is meant that correctly judges that ratio is high.
In embodiments of the present invention, to the entire training flow of the 3D convolutional neural networks models based on Alzheimer's disease
It illustrates, including:
Step L1, the automation pretreatment of brain structure sample image;Step L2 divides file;Step L3 loads data;
Step L4,3D convolutional neural networks model training;Step L5, fine tuning;Step L6, cross validation.
Wherein, step L2 includes:
L21 reads file;File number is arranged in L22;L23, traverse folder:Ergodic process includes establishment file folder,
Write test set filename, write verification collection filename;L24 discharges resource.
Step L3 includes:
L31, load document catalogue;L32 filters out length, width and height minimum value;L33 traverses non-illness sample image file:
Ergodic process includes creating empty list, load document, stacked documents;L34 creates X0 vectors, and is assigned a value of the non-illness sample
The set of this image creates Y0 vectors, and is assigned a value of 0;L35 traverses illness sample image file:Ergodic process includes creating
Empty list, load document, stacked documents;L36 creates X1 vectors, and is assigned a value of the set of the illness sample image, creates Y1
Vector, and it is assigned a value of 1;L37 returns to X0, Y0;X1, Y1.
Step L4 includes:
L41 obtains X0, Y0;X1, Y1;L42 samples length, width and height;L43 obtains training round;L44, traversal instruction
Practice round:Ergodic process includes obtaining minimum lot data;Training autocoder, obtains penalty values;Often train 10 printings
Penalty values;L45 preserves autocoder destination file.
Step L5 includes:
L51 obtains X0, Y0;X1, Y1;L52 reads autocoder destination file;L53 obtains training round;L54,
Traversal training round:Ergodic process includes obtaining minimum lot data;Training grader, obtains penalty values;It often trains 10 times and beats
Print a penalty values;L45 preserves sorter model data file.
Step L6 includes:
L61:Traverse ready-portioned file:Ergodic process includes:L611 reads the test set under this document folder;L612,
Obtain X0, Y0;X1, Y1;L613 samples length, width and height;L614 reads trained sorter model data file;
L615 obtains verification round;L616, traversal verification round:Ergodic process includes obtaining minimum lot data;Testing classification device,
Obtain classification results;It will determine that correct number is cumulative;L617, using correct number divided by verification collection size as accuracy rate essence
Degree;L618 prints accuracy rate precision.
Referring to Fig. 8, fourth embodiment of the invention provides fourth embodiment, an embodiment of the present invention provides a kind of A Er
Ci Haimo disease prediction meanss, including:
Pretreatment unit 11 is pre-processed for the nuclear magnetic resonance brain structure image to acquisition, to generate grey matter figure
Picture.
Model prediction unit 12, for according to the grey matter image and the 3D convolutional neural networks based on Alzheimer's disease
Model carries out operation, to export Alzheimer's disease prediction result.
In the first realization method of fourth embodiment, further include:
Quality screening unit carries out picture quality screening for the nuclear magnetic resonance brain structure original image to input.
Format conversion unit carries out data format for the nuclear magnetic resonance brain structure original image to screening qualification
Conversion, to obtain analyzable nuclear magnetic resonance brain structure image.
It is described pre- in second of realization method of fourth embodiment according to the first realization method of fourth embodiment
Processing unit 11 specifically includes:
Image segmentation module, for carrying out grey matter image and white matter image to the nuclear magnetic resonance brain structure image of acquisition
Segmentation, to generate initial grey matter image and initial white matter image.
Non-linear registration module, for carrying out non-linear registration to the initial grey matter image and initial white matter image, with
Generate registration image.
Standardization and Leveling Block, for being standardized to the registration image and smoothing processing, after being handled with generation
Grey matter image.
According to second of realization method of fourth embodiment, in the third realization method of fourth embodiment, the mould
Type predicting unit 12 is specially:
Input module, for using the grey matter image as the 3D convolutional neural networks moulds based on Alzheimer's disease
The input value of type.
Characteristic extracting module, in convolutional layer, the characteristic value of the grey matter image being extracted by self-encoding encoder, with life
At at least one characteristic image.
Characteristic image compression module, for carrying out compression processing at least one characteristic image by pond layer, with
It is corresponding to generate at least one compressive features image.
First conversion module, for carrying out the first activation primitive conversion at least one compressive features image, with life
At corresponding at least one nonlinear characteristic image.
Feature laminating module, for carrying out feature superposition at least one nonlinear characteristic image by full articulamentum
And second activation primitive conversion, to generate Superposition Characteristics image.
Prediction result output unit is handled the Superposition Characteristics image for passing through output layer, to export A Er
Ci Haimo disease prediction results.
According to the third realization method of fourth embodiment, in the 4th kind of realization method of fourth embodiment, the pressure
The mode of contracting processing is any one in maximum value pond, minimum value pond or mean value pond.
According to the third realization method of fourth embodiment, in the 5th kind of realization method of fourth embodiment, described
One activation primitive is any one in ReLU activation primitives, TanH activation primitives and liner activation primitives.Second activation
Function is any one in ReLU activation primitives, TanH activation primitives and liner activation primitives.
According to any of the above realization method of fourth embodiment, in the 6th kind of realization method of fourth embodiment,
In the training process of 3D convolutional neural networks models based on Alzheimer's disease, convolution kernel is trained by autocoder.
In the training process of the 3D convolutional neural networks models based on Alzheimer's disease, will training prediction result collection with
The mean square deviation of legitimate reading collection is as the penalty values predicted in training process.
In the training process of the 3D convolutional neural networks models based on Alzheimer's disease, by accuracy rate and AUC curves
Index as the evaluation 3D convolutional neural networks models based on Alzheimer's disease.
Fifth embodiment of the invention provides a kind of Alzheimer's disease prediction terminal device.The alzheimer ' of the embodiment
Silent disease predicts that terminal device includes:It processor, memory and is stored in the memory and can transport on the processor
Capable computer program, such as Alzheimer's disease Prediction program.The processor is realized when executing the computer program
State the step in each Alzheimer's disease prediction technique embodiment, such as step S11 shown in FIG. 1.Alternatively, the processor
The function of each module/unit in above-mentioned each device embodiment, such as pretreatment unit are realized when executing the computer program.
Illustratively, the computer program can be divided into one or more module/units, one or more
A module/unit is stored in the memory, and is executed by the processor, to complete the present invention.It is one or more
A module/unit can be the series of computation machine program instruction section that can complete specific function, and the instruction segment is for describing institute
State implementation procedure of the computer program in the Alzheimer's disease predicts terminal device.
The Alzheimer's disease prediction terminal device can be desktop PC, notebook, palm PC and high in the clouds
The computing devices such as server.The Alzheimer's disease prediction terminal device may include, but be not limited only to, processor, memory.
It will be understood by those skilled in the art that the component is only the example of Alzheimer's disease prediction terminal device, do not constitute
The restriction for predicting Alzheimer's disease terminal device, may include component more more or fewer than above-mentioned component, or combination
Certain components or different components, such as Alzheimer's disease prediction terminal device can also be set including input and output
Standby, network access equipment, bus etc..
Alleged processor can be central processing unit (Central Processing Unit, CPU), can also be it
His general processor, digital signal processor (Digital Signal Processor, DSP), application-specific integrated circuit
(Application Specific Integrated Circuit, ASIC), ready-made programmable gate array (Field-
Programmable Gate Array, FPGA) either other programmable logic device, discrete gate or transistor logic,
Discrete hardware components etc..General processor can be microprocessor or the processor can also be any conventional processor
It is the control centre of the Alzheimer's disease prediction terminal device Deng, the processor, utilizes various interfaces and connection
The various pieces of entire Alzheimer's disease prediction terminal device.
The memory can be used for storing the computer program and/or module, and the processor is by running or executing
Computer program in the memory and/or module are stored, and calls the data being stored in memory, described in realization
Alzheimer's disease predicts the various functions of terminal device.The memory can include mainly storing program area and storage data
Area, wherein storing program area can storage program area, needed at least one function application program (such as sound-playing function,
Image player function etc.) etc.;Storage data field can be stored uses created data (such as audio data, electricity according to mobile phone
Script for story-telling etc.) etc..In addition, memory may include high-speed random access memory, can also include nonvolatile memory, such as
Hard disk, memory, plug-in type hard disk, intelligent memory card (Smart Media Card, SMC), secure digital (Secure
Digital, SD) card, flash card (Flash Card), at least one disk memory, flush memory device or other volatibility are solid
State memory device.
Wherein, if the Alzheimer's disease predicts the integrated module/unit of terminal device with SFU software functional unit
Form is realized and when sold or used as an independent product, can be stored in a computer read/write memory medium.Base
In such understanding, the present invention realizes all or part of flow in above-described embodiment method, can also pass through computer program
It is completed to instruct relevant hardware, the computer program can be stored in a computer readable storage medium, the calculating
Machine program is when being executed by processor, it can be achieved that the step of above-mentioned each embodiment of the method.Wherein, the computer program includes
Computer program code, the computer program code can be source code form, object identification code form, executable file or certain
A little intermediate forms etc..The computer-readable medium may include:Any entity of the computer program code can be carried
Or device, recording medium, USB flash disk, mobile hard disk, magnetic disc, CD, computer storage, read-only memory (ROM, Read-Only
Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software
Distribution medium etc..It should be noted that the content that the computer-readable medium includes can be according to making laws in jurisdiction
Requirement with patent practice carries out increase and decrease appropriate, such as in certain jurisdictions, according to legislation and patent practice, computer
Readable medium does not include electric carrier signal and telecommunication signal.
It should be noted that the apparatus embodiments described above are merely exemplary, wherein described be used as separating component
The unit of explanation may or may not be physically separated, and the component shown as unit can be or can also
It is not physical unit, you can be located at a place, or may be distributed over multiple network units.It can be according to actual
It needs that some or all of module therein is selected to achieve the purpose of the solution of this embodiment.In addition, device provided by the invention
In embodiment attached drawing, the connection relation between module indicates there is communication connection between them, specifically can be implemented as one or
A plurality of communication bus or signal wire.Those of ordinary skill in the art are without creative efforts, you can to understand
And implement.
The above is the preferred embodiment of the present invention, it is noted that for those skilled in the art
For, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also considered as
Protection scope of the present invention.
Claims (10)
1. a kind of Alzheimer's disease prediction technique, which is characterized in that include the following steps:
The nuclear magnetic resonance brain structure image of acquisition is pre-processed, to generate grey matter image;
3D convolutional neural networks models according to the grey matter image and based on Alzheimer's disease carry out operation, to export A Er
Ci Haimo disease prediction results.
2. Alzheimer's disease prediction technique according to claim 1, which is characterized in that further include:
Picture quality screening is carried out to the nuclear magnetic resonance brain structure original image of input;
Data Format Transform is carried out to the nuclear magnetic resonance brain structure original image of screening qualification, to obtain analyzable core
Magnetic resonance brain structural images.
3. Alzheimer's disease prediction technique according to claim 2, which is characterized in that the nuclear magnetic resonance of described pair of acquisition
Brain structure image is pre-processed, to generate grey matter image, specially:
The segmentation that grey matter image and white matter image are carried out to the nuclear magnetic resonance brain structure image of acquisition, to generate initial grey matter figure
Picture and initial white matter image;
Non-linear registration is carried out to the initial grey matter image and initial white matter image, to generate registration image;
The registration image is standardized and smoothing processing, with the grey matter image that generates that treated.
4. Alzheimer's disease prediction technique according to claim 3, which is characterized in that described according to the grey matter image
And the 3D convolutional neural networks models based on Alzheimer's disease carry out operation, and to export Alzheimer's disease prediction result, tool
Body is:
Using the grey matter image as the input value of the 3D convolutional neural networks models based on Alzheimer's disease;
In convolutional layer, the characteristic value of the grey matter image is extracted by self-encoding encoder, to generate at least one characteristic image;
Compression processing is carried out at least one characteristic image by pond layer, at least one compressive features figure is generated with corresponding
Picture;
First activation primitive conversion is carried out at least one compressive features image, it is corresponding at least one non-linear to generate
Characteristic image;
Feature superposition is carried out at least one nonlinear characteristic image by full articulamentum and the second activation primitive is converted, with
Generate Superposition Characteristics image;
The Superposition Characteristics image is handled by output layer, to export Alzheimer's disease prediction result.
5. Alzheimer's disease prediction technique according to claim 4, which is characterized in that the mode of the compression processing is
Any one in maximum value pond, minimum value pond or mean value pond.
6. Alzheimer's disease prediction technique according to claim 4, which is characterized in that first activation primitive is
Any one in ReLU activation primitives, TanH activation primitives and liner activation primitives;Second activation primitive swashs for ReLU
Any one in living function, TanH activation primitives and liner activation primitives.
7. the Alzheimer's disease prediction technique according to claim 1 to 6 any one, which is characterized in that based on Ah
In the training process of the 3D convolutional neural networks models of Er Cihaimo diseases, convolution kernel is trained by autocoder;
In the training process of the 3D convolutional neural networks models based on Alzheimer's disease, by training prediction result collection and really
The mean square deviation of result set is as the penalty values predicted in training process;
In the training process of the 3D convolutional neural networks models based on Alzheimer's disease, using accuracy rate and AUC curves as
The index of the evaluation 3D convolutional neural networks models based on Alzheimer's disease.
8. a kind of Alzheimer's disease prediction meanss, which is characterized in that including:
Pretreatment unit is pre-processed for the nuclear magnetic resonance brain structure image to acquisition, to generate grey matter image;
Model prediction unit, for according to the grey matter image and the 3D convolutional neural networks model based on Alzheimer's disease into
Row operation, to export Alzheimer's disease prediction result.
9. a kind of Alzheimer's disease predicts terminal device, including processor, memory and be stored in the memory and
It is configured as the computer program executed by the processor, the processor realizes such as right when executing the computer program
It is required that the Alzheimer's disease prediction technique described in any one of 1 to 7.
10. a kind of computer readable storage medium, which is characterized in that the computer readable storage medium includes the calculating of storage
Machine program, wherein equipment where controlling the computer readable storage medium when the computer program is run is executed as weighed
Profit requires the Alzheimer's disease prediction technique described in any one of 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810431323.4A CN108682009A (en) | 2018-05-08 | 2018-05-08 | A kind of Alzheimer's disease prediction technique, device, equipment and medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810431323.4A CN108682009A (en) | 2018-05-08 | 2018-05-08 | A kind of Alzheimer's disease prediction technique, device, equipment and medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108682009A true CN108682009A (en) | 2018-10-19 |
Family
ID=63801969
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810431323.4A Pending CN108682009A (en) | 2018-05-08 | 2018-05-08 | A kind of Alzheimer's disease prediction technique, device, equipment and medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108682009A (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109871838A (en) * | 2019-04-18 | 2019-06-11 | 电子科技大学 | The extracting method of Alzheimer's disease lesion brain area based on cascade deep network |
CN110246566A (en) * | 2019-04-24 | 2019-09-17 | 中南大学湘雅二医院 | Method, system and storage medium are determined based on the conduct disorder of convolutional neural networks |
CN110674773A (en) * | 2019-09-29 | 2020-01-10 | 燧人(上海)医疗科技有限公司 | Dementia recognition system, device and storage medium |
CN110866909A (en) * | 2019-11-13 | 2020-03-06 | 上海联影智能医疗科技有限公司 | Training method of image generation network, image prediction method and computer equipment |
CN111126624A (en) * | 2019-12-20 | 2020-05-08 | 北京首汽智行科技有限公司 | Method for judging validity of model prediction result |
CN112465900A (en) * | 2020-12-09 | 2021-03-09 | 中国科学院深圳先进技术研究院 | Stereo electroencephalogram electrode positioning method |
CN113440107A (en) * | 2021-07-06 | 2021-09-28 | 浙江大学 | Alzheimer's symptom diagnosis device based on voice signal analysis |
CN114373095A (en) * | 2021-12-09 | 2022-04-19 | 山东师范大学 | Alzheimer disease classification system and method based on image information |
CN114418966A (en) * | 2021-12-29 | 2022-04-29 | 澄影科技(北京)有限公司 | Alzheimer disease risk assessment method and device |
CN114758774A (en) * | 2022-02-23 | 2022-07-15 | 北京科技大学 | Alzheimer disease evolution trend analysis based on thalamocortical nerve calculation model |
CN115578387A (en) * | 2022-12-06 | 2023-01-06 | 中南大学 | Multimodal-based Alzheimer disease medical image classification method and system |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107506797A (en) * | 2017-08-25 | 2017-12-22 | 电子科技大学 | One kind is based on deep neural network and multi-modal image alzheimer disease sorting technique |
-
2018
- 2018-05-08 CN CN201810431323.4A patent/CN108682009A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107506797A (en) * | 2017-08-25 | 2017-12-22 | 电子科技大学 | One kind is based on deep neural network and multi-modal image alzheimer disease sorting technique |
Non-Patent Citations (2)
Title |
---|
ADRIEN PAYAN: "Predicting Alzheimer’s disease: a neuroimaging study with 3D convolutional neural networks", 《COMPUTER SCIENCE》 * |
SUNGDEN: "脑部MRI图像灰白质分割以及体积测量方法研究综述", 《CSDNHTTPS://BLOG.CSDN.NET/SUNGDEN/ARTICLE/DETAILS/79007590》 * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109871838A (en) * | 2019-04-18 | 2019-06-11 | 电子科技大学 | The extracting method of Alzheimer's disease lesion brain area based on cascade deep network |
CN109871838B (en) * | 2019-04-18 | 2019-11-05 | 电子科技大学 | The extracting method of Alzheimer's disease lesion brain area based on cascade deep network |
CN110246566A (en) * | 2019-04-24 | 2019-09-17 | 中南大学湘雅二医院 | Method, system and storage medium are determined based on the conduct disorder of convolutional neural networks |
CN110674773A (en) * | 2019-09-29 | 2020-01-10 | 燧人(上海)医疗科技有限公司 | Dementia recognition system, device and storage medium |
CN110866909A (en) * | 2019-11-13 | 2020-03-06 | 上海联影智能医疗科技有限公司 | Training method of image generation network, image prediction method and computer equipment |
CN111126624A (en) * | 2019-12-20 | 2020-05-08 | 北京首汽智行科技有限公司 | Method for judging validity of model prediction result |
CN112465900A (en) * | 2020-12-09 | 2021-03-09 | 中国科学院深圳先进技术研究院 | Stereo electroencephalogram electrode positioning method |
CN113440107A (en) * | 2021-07-06 | 2021-09-28 | 浙江大学 | Alzheimer's symptom diagnosis device based on voice signal analysis |
CN114373095A (en) * | 2021-12-09 | 2022-04-19 | 山东师范大学 | Alzheimer disease classification system and method based on image information |
CN114418966A (en) * | 2021-12-29 | 2022-04-29 | 澄影科技(北京)有限公司 | Alzheimer disease risk assessment method and device |
CN114758774A (en) * | 2022-02-23 | 2022-07-15 | 北京科技大学 | Alzheimer disease evolution trend analysis based on thalamocortical nerve calculation model |
CN115578387A (en) * | 2022-12-06 | 2023-01-06 | 中南大学 | Multimodal-based Alzheimer disease medical image classification method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108682009A (en) | A kind of Alzheimer's disease prediction technique, device, equipment and medium | |
Sarraf et al. | DeepAD: Alzheimer’s disease classification via deep convolutional neural networks using MRI and fMRI | |
DE102018128581A1 (en) | DETECTION AND CLASSIFICATION OF DAMAGE SOFTWARE USING AN ARTIFICIAL NEURAL NETWORK | |
CN110910982A (en) | Self-coding model training method, device, equipment and storage medium | |
CN104881683A (en) | Cataract eye fundus image classification method based on combined classifier and classification apparatus | |
CN110859624A (en) | Brain age deep learning prediction system based on structural magnetic resonance image | |
CN108363979A (en) | Neonatal pain expression recognition method based on binary channels Three dimensional convolution neural network | |
CN112037179B (en) | Method, system and equipment for generating brain disease diagnosis model | |
CN111584021A (en) | Medical record information verification method and device, electronic equipment and storage medium | |
CN109376766A (en) | A kind of portrait prediction classification method, device and equipment | |
CN109222902A (en) | Parkinson's prediction technique, system, storage medium and equipment based on nuclear magnetic resonance | |
CN110046622A (en) | A kind of attack sample generating method, device, equipment and storage medium having target | |
Wu et al. | 3D transfer learning network for classification of Alzheimer’s disease with MRI | |
CN114748053A (en) | fMRI high-dimensional time sequence-based signal classification method and device | |
CN112926332A (en) | Entity relationship joint extraction method and device | |
CN114782394A (en) | Cataract postoperative vision prediction system based on multi-mode fusion network | |
Cinar et al. | A novel convolutional neural network‐based approach for brain tumor classification using magnetic resonance images | |
Baskar et al. | An Accurate Prediction and Diagnosis of Alzheimer’s Disease using Deep Learning | |
CN116452592B (en) | Method, device and system for constructing brain vascular disease AI cognitive function evaluation model | |
CN116628161A (en) | Answer generation method, device, equipment and storage medium | |
Wang et al. | A disentangled representation based brain image fusion via group lasso penalty | |
Sarraf et al. | The Alzheimer’s Disease Neuroimaging Initiative OViTAD: Optimized Vision Transformer to Predict Various Stages of Alzheimer’s Disease Using Resting-State fMRI and Structural MRI Data | |
CN109223001A (en) | Hyperactivity prediction technique, system, storage medium and equipment based on nuclear magnetic resonance | |
CN116309615A (en) | Multi-mode MRI brain tumor image segmentation method | |
CN110096145A (en) | Psychological condition display methods and device based on mixed reality and neural network |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181019 |
|
RJ01 | Rejection of invention patent application after publication |