CN109222902A - Parkinson's prediction technique, system, storage medium and equipment based on nuclear magnetic resonance - Google Patents
Parkinson's prediction technique, system, storage medium and equipment based on nuclear magnetic resonance Download PDFInfo
- Publication number
- CN109222902A CN109222902A CN201810983951.3A CN201810983951A CN109222902A CN 109222902 A CN109222902 A CN 109222902A CN 201810983951 A CN201810983951 A CN 201810983951A CN 109222902 A CN109222902 A CN 109222902A
- Authority
- CN
- China
- Prior art keywords
- parkinson
- nuclear magnetic
- magnetic resonance
- diagnosed
- image
- 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
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/40—Detecting, measuring or recording for evaluating the nervous system
- A61B5/4076—Diagnosing or monitoring particular conditions of the nervous system
- A61B5/4082—Diagnosing or monitoring movement diseases, e.g. Parkinson, Huntington or Tourette
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/0033—Features or image-related aspects of imaging apparatus classified in A61B5/00, e.g. for MRI, optical tomography or impedance tomography apparatus; arrangements of imaging apparatus in a room
- A61B5/004—Features or image-related aspects of imaging apparatus classified in A61B5/00, e.g. for MRI, optical tomography or impedance tomography apparatus; arrangements of imaging apparatus in a room adapted for image acquisition of a particular organ or body part
- A61B5/0042—Features or image-related aspects of imaging apparatus classified in A61B5/00, e.g. for MRI, optical tomography or impedance tomography apparatus; arrangements of imaging apparatus in a room adapted for image acquisition of a particular organ or body part for the brain
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/05—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves
- A61B5/055—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
Abstract
Parkinson's prediction technique based on nuclear magnetic resonance that the invention discloses a kind of, comprising: obtain cranium brain nuclear magnetic resonance image to be diagnosed；The cranium brain nuclear magnetic resonance image to be diagnosed is handled, tractography picture to be diagnosed is obtained；Parkinson is predicted according to the tractography picture to be diagnosed based on Parkinson's prediction model trained in advance.Correspondingly, the invention also discloses a kind of Parkinson's forecasting system, computer readable storage medium and terminal device based on nuclear magnetic resonance.Prediction to Parkinson can be realized using technical solution of the present invention, and improve predictablity rate.
Description
Technical field
The present invention relates to medicine technology field more particularly to a kind of Parkinson's prediction technique based on nuclear magnetic resonance, system,
Computer readable storage medium and terminal device.
Background technique
Parkinson (Parkinson ' s disease, PD) is a kind of common person in middle and old age's nervous system degenerative disease, main
Will to tremble, myotonia, slow movement, the motor symptoms of postural balance obstacle and hyposphresia, constipation, sleep behavior it is abnormal and
The clinical manifestation of the non-motor symptoms such as depression is notable feature, and Parkinson is mostly since substantia nigra dopaminergic neuron progressive moves back
Pathological change caused by the reasons such as change.
Effective precautionary measures there is no to prevent the generation of Parkinson disease, when clinical symptoms occurs in patient, black substance at present
Dopaminergic neuron death is at least 50% or more, still, early detection Preclinical patients, and and take effective precautionary measures
It prevents the denaturation of dopaminergic neuron dead, the generation and progress of Parkinson disease can be slowed down, therefore, how early detection is faced
Patient has become one of the hot spot of Parkinson's disease research field before bed, has very important meaning to the prediction of Parkinson disease
Justice.
Summary of the invention
The technical problem to be solved by the embodiment of the invention is that providing a kind of Parkinson prediction side based on nuclear magnetic resonance
Method, system, computer readable storage medium and terminal device can be realized the prediction to Parkinson, and it is accurate to improve prediction
Rate.
In order to solve the above-mentioned technical problem, the embodiment of the invention provides a kind of Parkinson prediction side based on nuclear magnetic resonance
Method, comprising:
Obtain cranium brain nuclear magnetic resonance image to be diagnosed；
The cranium brain nuclear magnetic resonance image to be diagnosed is handled, tractography picture to be diagnosed is obtained；
Parkinson is predicted according to the tractography picture to be diagnosed based on Parkinson's prediction model trained in advance.
Further, Parkinson's prediction model is the model for having merged convolutional neural networks and fully-connected network；Institute
Parkinson's prediction model is stated including at least activation primitive, convolutional network layer, pond layer and fully connected network network layers.
Further, the method is trained Parkinson's prediction model by following steps:
The cranium brain nuclear magnetic resonance image of disturbances in patients with Parkinson disease and the cranium brain nuclear magnetic resonance image of normal person are obtained respectively；
The cranium brain nuclear magnetic resonance image of cranium brain nuclear magnetic resonance image and the normal person to the disturbances in patients with Parkinson disease carries out
Processing, it is corresponding to obtain disturbances in patients with Parkinson disease tractography picture and normal brain white matter image；
Mould is predicted to the Parkinson according to the disturbances in patients with Parkinson disease tractography picture and the normal brain white matter image
Type is trained and tests.
Further, the method also includes:
The penalty values of prediction are obtained according to test result and actual result；
Parkinson's prediction model is optimized according to the penalty values.
Further, it is described based on Parkinson's prediction model trained in advance tractography picture to be diagnosed according to pa
Jin Sen is predicted, is specifically included:
Feature extraction is carried out according to the tractography picture to be diagnosed based on the convolutional network layer, obtains characteristic pattern；
The characteristic pattern is compressed based on the pond layer, obtains compressed characteristic pattern；
Conversion process is carried out to the compressed characteristic pattern based on the activation primitive；
The result of conversion process is integrated based on the fully connected network network layers, obtains prediction result.
Further, described to obtain cranium brain nuclear magnetic resonance image to be diagnosed, it specifically includes:
The cranium brain nuclear magnetic resonance image to be diagnosed is obtained by nuclear magnetic resonane scanne.
Further, described that the cranium brain nuclear magnetic resonance image to be diagnosed is handled, obtain white matter of brain to be diagnosed
Image specifically includes:
Skull removing, image segmentation, non-linear registration, figure are at least carried out to the cranium brain nuclear magnetic resonance image to be diagnosed
As standardization and smoothing processing, the tractography picture to be diagnosed is obtained.
In order to solve the above-mentioned technical problem, Parkinson's prediction based on nuclear magnetic resonance that the embodiment of the invention also provides a kind of
System, comprising:
Image collection module, for obtaining cranium brain nuclear magnetic resonance image to be diagnosed；
Image processing module obtains brain to be diagnosed for handling the cranium brain nuclear magnetic resonance image to be diagnosed
White matter image；And
Prediction module, for based on Parkinson's prediction model trained in advance tractography picture to be diagnosed according to pa
Jin Sen is predicted.
The embodiment of the invention also provides a kind of computer readable storage medium, the computer readable storage medium includes
The computer program of storage；Wherein, where the computer program controls the computer readable storage medium at runtime
Equipment executes Parkinson's prediction technique described in any of the above embodiments based on nuclear magnetic resonance.
The embodiment of the invention also provides a kind of terminal device, including processor, memory and it is stored in the storage
In device and it is configured as the computer program executed by the processor, the processor is real when executing the computer program
Existing Parkinson's prediction technique described in any of the above embodiments based on nuclear magnetic resonance.
Compared with prior art, the Parkinson's prediction technique that the embodiment of the invention provides a kind of based on nuclear magnetic resonance is
System, computer readable storage medium and terminal device, by obtaining cranium brain nuclear magnetic resonance image to be diagnosed, and treat diagnosis
Cranium brain nuclear magnetic resonance image is handled, and tractography picture to be diagnosed is obtained, based on Parkinson's prediction model root trained in advance
Parkinson is predicted according to tractography picture to be diagnosed, can be realized the prediction to Parkinson, and improve predictablity rate.
Detailed description of the invention
Fig. 1 is a kind of stream of a preferred embodiment of Parkinson's prediction technique based on nuclear magnetic resonance provided by the invention
Cheng Tu；
Fig. 2 is a kind of detailed process of a preferred embodiment of Parkinson's prediction model training method provided by the invention
Figure；
Fig. 3 is the schematic diagram of ROC curve provided in an embodiment of the present invention；
Fig. 4 is that one of the step S13 of a kind of Parkinson's prediction technique based on nuclear magnetic resonance provided by the invention is preferred
The specific flow chart of embodiment；
Fig. 5 is a kind of structural schematic diagram of a preferred embodiment of Parkinson's prediction model provided by the invention；
Fig. 6 is a kind of knot of a preferred embodiment of Parkinson's forecasting system based on nuclear magnetic resonance provided by the invention
Structure block diagram；
Fig. 7 is a kind of structural block diagram of a preferred embodiment of terminal device provided by the invention.
Specific embodiment
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 description, 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, those of ordinary skill in the art are obtained all without creative efforts
Other embodiments shall fall within the protection scope of the present invention.
It is shown in Figure 1, it is that one of a kind of Parkinson's prediction technique based on nuclear magnetic resonance provided by the invention is preferred
The flow chart of embodiment, including step S11 to step S13:
Step S11, cranium brain nuclear magnetic resonance image to be diagnosed is obtained；
Step S12, the cranium brain nuclear magnetic resonance image to be diagnosed is handled, obtains tractography picture to be diagnosed；
Step S13, based on Parkinson's prediction model trained in advance according to the tractography picture to be diagnosed to Parkinson
It is predicted.
Specifically, obtaining cranium brain nuclear magnetic resonance image to be diagnosed, and to the cranium brain nuclear magnetic resonance figures to be diagnosed of acquisition
As carrying out respective handling, so that the tractography picture to be diagnosed for being suitable for Parkinson's prediction model is obtained, it will white matter of brain be diagnosed
Image is input to the prediction that Parkinson disease is carried out in Parkinson's prediction model that training is completed in advance.
After the cranium brain nuclear magnetic resonance image for treating diagnosis is handled, tractography picture and brain ash can be accordingly obtained
Matter image, and Parkinson disease is predicted according to tractography picture, prediction result ratio obtained is according to ectocinerea image
The prediction result predicted Parkinson disease is more accurate, and therefore, the embodiment of the present invention is using tractography picture to pa gold
Gloomy disease is predicted, and the prediction result output finally obtained is a probability value (between 0%~100%), table
Show that patient suffers from the probability of Parkinson disease.
It should be noted that the cranium brain nuclear magnetic resonance image in any one embodiment of the invention is cranium brain nuclear magnetic resonance
3D structural images.
It should be added that nuclear magnetic resonance is the atomic nucleus that magnetic moment is not zero, spin energy level under external magnetic field
Zeeman splitting, the physical process of the radio-frequency radiation of RESONANCE ABSORPTION certain frequency occurs.Magnetic resonance imaging (Nuclear
Magnetic Resonance Imaging, NMRI) be also referred to as magnetic resonance imaging (Magnetic Resonance Imaging,
It MRI), is using nuclear magnetic resonance principle, the decaying different in different structure environment inside the substance according to the energy discharged is led to
It crosses additional gradient magnetic and detects launched electromagnetic wave, it can be learnt that the nuclear position of this object and type are constituted, according to
This can be depicted as the structural images of interior of articles.Cranium brain nuclear magnetic resonance figures seem to be applied RF pulse-to-pulse according to brain different tissues
It is formed by image after punching, with the different relaxation times to distinguish the knot of tissue such as white matter of brain, ectocinerea, cerebrospinal fluid
Structure, in clinical medicine and scientific research field, most common institutional framework is white matter of brain and ectocinerea, both institutional frameworks are to constitute
The main component of human brain.
A kind of Parkinson's prediction technique based on nuclear magnetic resonance, passes through the follow-up to acquisition provided by the embodiment of the present invention
Disconnected cranium brain nuclear magnetic resonance image carries out respective handling, to obtain the tractography to be diagnosed for being suitable for Parkinson's prediction model
Picture, and by Parkinson's prediction model after diagnosing tractography picture and being input to training, it can be realized to Parkinson disease
Prediction, and Parkinson disease is predicted according to tractography picture, improve predictablity rate.
In a further advantageous embodiment, Parkinson's prediction model is to have merged convolutional neural networks and fully connected network
The model of network；Parkinson's prediction model includes at least activation primitive, convolutional network layer, pond layer and fully connected network network layers.
Specifically, Parkinson's prediction model used by the embodiment of the present invention is a kind of prediction based on deep learning algorithm
Model, preferably 3D convolutional neural networks model (3D-CNN), are made of, specific structure convolutional neural networks and fully-connected network
At including at least activation primitive, convolutional network layer, pond layer and fully connected network network layers.
Preferably, Parkinson's prediction model in the embodiment of the present invention includes at least level 2 volume product network layer, 2 layers of pond layer
With 4 layers of fully connected network network layers, and there is before each layer of convolutional network layer an activation primitive be used to judge, in fully connected network
The output layer that is followed by of network layers is used to export prediction result.
It should be added that:
(1) cardinal principle of neural network is to imitate the working principle of cerebral neuron, several neurons are connected into
The signal of input is centainly handled (may be cumulative, filtering or other any modes) by network, each neuron, then will
Input that signal spreads out of as target nerve member that treated, passes through the extensive repetition of single neuron simple rule, shape
At a complication system, it is finally reached the purpose of nonlinear prediction.
(2) in layered structure, each node of preceding layer can connect fully-connected network with each node of later layer
It connects, network structure in total is the quantity of the cartesian product of whole nodes；The embodiment of the present invention uses 4 layers of fully-connected network, often
One layer is 2048 nodes, 512 nodes, 256 nodes and 128 nodes respectively.
(3) convolutional network needs first to be arranged the convolution rule of image, generally uses matrix multiplication as convolution rule.With this reality
For applying Parkinson's prediction model in example, (3D image is 3 rank tensors) uses 3 rank tensors as volume on an image tensor
Product core, is slided, the convolution value of convolution kernel and image is as a result on 3D image；Each volume machine core is sliding on 3D image
It can be corresponded to after the completion of dynamic and generate a characteristic pattern, characteristic pattern is equally 3D tensor；Then characteristic pattern is subjected to pondization operation, obtained
The characteristic pattern of compressed version (characteristic pattern is the corresponding Feature Mapping of convolution kernel)；Characteristic pattern is may be selected behind pond using activation letter
Number carries out conversion process, or is directly entered the training of next layer of convolutional network layer；Wherein, each layer of convolutional network layer is by convolution
Core forms, and the number of the convolution kernel of each layer of convolutional network layer in the embodiment of the present invention is adjustable parameter, default setting 128
× 64；Weight in convolutional network layer is approximately convolution kernel.
(4) convolution kernel: for carrying out the feature unit of convolutional calculation to image, general dimension is identical with target dimension, volume
The size of product core is adjustable ginseng item, and default setting 8*8*8, i.e. length, width and height are 8 pixels.The generation of convolution kernel: machine learning
There are many kinds of the generating modes of convolution kernel, such as artificial customization, random generation, autocoder generation and PAC principal component analysis
Method generation etc..The embodiment of the present invention uses autocoder generating mode, the principle of autocoder be simply interpreted as by
Image is input in trainable neural network, by the feature extraction of network, then is exported one and is inputted identical image, with
Precision is continuously improved by way of backpropagation as penalty values in mean square deviation between input and output, this nerve net
Network as a result, being exactly convolution kernel.By the convolution kernel that autocoder generates can to have in training one it is relatively good
Convolution kernel substantially increases trained precision problem.
(5) autocoder: being divided into two stages of coding and decoding, is the effective means for extracting high dimensional data feature.It compiles
The code stage carries out feature extraction to target data by convolutional network, and convolution kernel at this time is generated by the mode of random initializtion；
After coding stage, one layer of fully-connected network decoding is connect, i.e., full connection is carried out by convolution kernel and generates a data, comparison generates
Data and input data difference, generate a mean square deviation penalty values, be trained, obtained as mesh using optimizing this penalty values
Self-encoding encoder as a result, i.e. classification based training when convolution kernel.
(6) pond: to the compression processing of characteristic pattern, the mode in pond has maximum value pond, minimum value pondization and mean value pond
Change, the embodiment of the present invention uses mean value pond.
(7) activation primitive: for activating the state of node, it is used in the embodiment of the present invention have " relu ", " tanh ",
" liner " three kinds of activation primitives, wherein " relu " and " tanh " is nonlinear activation function.
It as shown in connection with fig. 2, is an a kind of preferred embodiment of Parkinson's prediction model training method provided by the invention
Specific flow chart, the method is trained Parkinson's prediction model by step S21 to step S23:
Step S21, the cranium brain nuclear magnetic resonance image of disturbances in patients with Parkinson disease and the cranium brain nuclear magnetic resonance figures of normal person are obtained respectively
Picture；
Step S22, to the cranium brain nuclear magnetic resonance of the cranium brain nuclear magnetic resonance image of the disturbances in patients with Parkinson disease and the normal person
Image is handled, corresponding to obtain disturbances in patients with Parkinson disease tractography picture and normal brain white matter image；
Step S23, according to the disturbances in patients with Parkinson disease tractography picture and the normal brain white matter image to the pa gold
Gloomy prediction model is trained and tests.
Specifically, obtaining the cranium brain nuclear magnetic resonance image of the disturbances in patients with Parkinson disease of preset quantity, and the Parkinson of acquisition is suffered from
The cranium brain nuclear magnetic resonance image of person carries out respective handling, to obtain the brain for being suitable for the disturbances in patients with Parkinson disease of Parkinson's prediction model
The tractography picture of disturbances in patients with Parkinson disease is input in Parkinson's prediction model and is trained and tests, similarly, obtains by white matter image
The cranium brain nuclear magnetic resonance image of the normal person of preset quantity is taken, and phase is carried out to the cranium brain nuclear magnetic resonance image of the normal person of acquisition
It should handle, so that the tractography picture for being suitable for the normal person of Parkinson's prediction model is obtained, by the tractography picture of normal person
It is input in Parkinson's prediction model and is trained and tests, Parkinson's prediction model after training is allowed to distinguish pa gold
The tractography picture of gloomy patient and the tractography picture of normal person, so as to according to tractography as automatic Prediction goes out Parkinson
Patient and normal person.
It should be noted that when being trained to Parkinson's prediction model, the cranium brain core of used disturbances in patients with Parkinson disease
The quantity of the cranium brain nuclear magnetic resonance image of magnetic resonance image and normal person is more, and the prediction of Parkinson's prediction model after training is quasi-
True rate is higher.
It should be added that differentiating two parameters of test result service precision and AUC, wherein precision is specifically criticized
Really prediction number accounts for the ratio of overall test number；AUC refers specifically to ROC curve (receiver operating
Characteristic curve, recipient's operating characteristic curve) under area, between 0.5~1, ROC curve shows value
It is intended to as shown in figure 3, horizontal axis is false positive example rate, the longitudinal axis is real example rate, when judging the quality of a binary classification algorithm, AUC
Good reference role can be played, for the value of AUC closer to 1, algorithm classification effect is better, it is meant that correctly judges ratio
It is higher.
In another preferred embodiment, the method also includes:
The penalty values of prediction are obtained according to test result and actual result；
Parkinson's prediction model is optimized according to the penalty values.
Specifically, when Parkinson's prediction model is trained and is tested according to different tractography pictures, each
Tractography picture has an actual result (to indicate that the tractography seems tractography picture or the normal person of disturbances in patients with Parkinson disease
Tractography picture) and the corresponding test result obtained, according to tractography as corresponding test result and actual result obtain
The penalty values of prediction, so as to be optimized according to penalty values to Parkinson's prediction model, to improve Parkinson's prediction model
Precision.
It should be noted that the embodiment of the present invention is using the mean square deviation of prediction result collection and actual result collection as prediction
Penalty values, it is optimal the result is that mean square deviation is 0.Regard each layer network as a function about input image data,
Variance equally can be regarded as a function of network parameter, by asking mean square deviation functional minimum value that can make the standard of prediction
True rate reaches highest, and mean square deviation function is minimized corresponding parameter then and is the parameter of target network, and solution procedure can pass through
The mode of gradient decline carries out, this process is referred to as backpropagation, by backpropagation, so that penalty values reach minimum to obtain
To optimal network parameter.
After the decline of constantly gradient, the precision of Parkinson's prediction model can be just improved, gradient index meaning
On gradient, regard each layer network as a function, penalty values can regard a function about network parameter, network here as
Parameter includes the weight of fully connected network network layers, the convolution kernel of biasing and convolutional network layer, biasing.It, must be first if function is most worth
Extreme value is taken, the purpose of gradient decline is the value of network parameter when obtaining each extreme point calculating penalty values minimum.
It is one of the step S13 of a kind of Parkinson's prediction technique based on nuclear magnetic resonance provided by the invention referring to fig. 4
The specific flow chart of preferred embodiment, it is described based on Parkinson's prediction model trained in advance tractography to be diagnosed according to
Picture predicts Parkinson, specifically includes step S1301 to step S1304:
Step S1301, feature extraction is carried out according to the tractography picture to be diagnosed based on the convolutional network layer, obtained
Characteristic pattern；
Step S1302, the characteristic pattern is compressed based on the pond layer, obtains compressed characteristic pattern；
Step S1303, conversion process is carried out to the compressed characteristic pattern based on the activation primitive；
Step S1304, the result of conversion process is integrated based on the fully connected network network layers, obtains prediction result.
Specifically, Parkinson predicts mould by Parkinson's prediction model after diagnosing tractography picture and being input to training
Type treats diagnosis tractography picture by the convolution kernel of convolutional network layer and carries out feature extraction, obtains tractography picture pair to be diagnosed
The characteristic pattern answered is compressed by characteristic pattern of the pond layer to acquisition, corresponding to obtain compressed characteristic pattern, by activating letter
The compressed characteristic pattern of several pairs of acquisitions carries out conversion process, then cumulative to activation primitive by the judgement of fully connected network network layers
Conversion process result is integrated, and the prediction result of Parkinson disease is finally obtained, to realize the prediction to Parkinson disease.
It should be understood that passing through Parkinson's prediction model pair after training according to wait diagnose tractography picture in the present embodiment
The basic principle that Parkinson is predicted is trained and tests to Parkinson's prediction model from according to different tractography pictures
Basic principle it is identical.
It as shown in connection with fig. 5, is that the structure of a preferred embodiment of Parkinson's prediction model provided by the invention a kind of is shown
It is intended to, each layer of convolutional network layer is all made of several convolution kernels, and has an activation before each layer of convolutional network layer
Function；Convolution kernel in convolutional network layer 1 carries out feature extraction to the MRI image received, obtains corresponding characteristic pattern,
In, the corresponding characteristic pattern of each convolution kernel, the characteristic pattern of 1 pair of pond layer acquisition compresses, corresponding to obtain compressed spy
Sign figure, characteristic pattern carries out conversion process using activation primitive behind pond, and is carried out by convolutional network layer 2 and pond layer 2
It handles again, obtains prediction result finally by fully connected network network layers, and prediction result is exported by output layer.
It is described to obtain cranium brain nuclear magnetic resonance image to be diagnosed in another preferred embodiment, it specifically includes:
The cranium brain nuclear magnetic resonance image to be diagnosed is obtained by nuclear magnetic resonane scanne.
It should be understood that the cranium brain nuclear magnetic resonance image of people can be obtained by nuclear magnetic resonane scanne.
It should be noted that obtaining method and the training Parkinson of cranium brain nuclear magnetic resonance image to be diagnosed in the present embodiment
The method phase of the cranium brain nuclear magnetic resonance image of disturbances in patients with Parkinson disease and the cranium brain nuclear magnetic resonance image of normal person is obtained when prediction model
Together.
It is described that the cranium brain nuclear magnetic resonance image to be diagnosed is handled in another preferred embodiment, it obtains
Tractography picture to be diagnosed, specifically includes:
Skull removing, image segmentation, non-linear registration, figure are at least carried out to the cranium brain nuclear magnetic resonance image to be diagnosed
As standardization and smoothing processing, the tractography picture to be diagnosed is obtained.
Specifically, brain structure image original number to be diagnosed can be obtained according to cranium brain nuclear magnetic resonance image to be diagnosed
According to by carrying out skull removing, white matter, grey matter segmentation, non-linear registration, image standardization to brain structure original image data
With smooth etc. reason, the tractography picture to be diagnosed predicted for Parkinson's prediction model is obtained.
The analysis software that the cranium brain nuclear magnetic resonance image for treating diagnosis that the present embodiment uses is handled is to be based on
The SPM software package of MATLAB, the software package are absorbed in the processing and analysis of brain phantom data, and associated process steps are established
On the basis of DARTEL instrumental function in SPM software, the processing of traditional nuclear magnetic resonance brain structure original image data is built
Environment and SPM software package in MATLAB exploitation are found, and the embodiment of the present invention will be in SPM software using MATLAB runtime environment
The whole processing steps being related to are integrated into code, automatic to read brain structure original image data and run in order all
Processing step, all treatment processes are run in Cent OS system.
The first step of DARTEL tool processing based on SPM software package is to carry out skull separation to cranium brain nuclear magnetic resonance image
Processing, and is partitioned into corresponding tractography picture and ectocinerea image, divide tractography picture and ectocinerea image uses at
Year people's template is the white people's brain template defaulted in SPM software package.
The second step of DARTEL tool processing based on SPM software package is drawing template establishment, and the effect of drawing template establishment is, examines
Brain structure original image data is considered from Different Individual and different scanning instrument type, and needing to create a reference template will not
Cranium brain nuclear magnetic resonance image with individual is mapped in the template, to achieve the purpose that image information fusion.The first step generates
Be tractography picture and ectocinerea image by rigid registration, second step is used on the basis of first step rigid registration
Non-linear registration, in second step, DARTEL tool generates one first with the image after handled subject rigid registration
A template then again in the image registration to the template of each subject, and then has been registrated to using these image weight of template
A newly-generated new template, it is best until obtaining repeatedly by the image registration of each subject to this new template
Registration effect.
The third step of DARTEL tool processing based on SPM software package is standardization and smooth, the brain that second step is generated
(MNI coordinate space is most common human brain coordinate body in the world to MNI coordinate space for white matter image and ectocinerea image standardization
One of system), Gaussian smoothing then is carried out to tractography picture and ectocinerea image again, used smoothing kernel is 8mm.
The whole process that the present embodiment handles the cranium brain nuclear magnetic resonance image for treating diagnosis is by way of coding
It realizes and is fully automated, it is automatic to read brain structure original image data, it is entire to be then based on the completion of MATLAB runtime environment
Treatment process, does not need the MATLAB of commercial version, does not need to be handled manually according to each step of DARTEL tool yet.
It should be noted that obtaining the method and training Parkinson's prediction model of tractography picture to be diagnosed in the present embodiment
When obtain disturbances in patients with Parkinson disease tractography picture it is identical with the method for normal brain white matter image.
Parkinson's forecasting system based on nuclear magnetic resonance that the embodiment of the invention also provides a kind of, can be realized any of the above-described
All processes of Parkinson's prediction technique based on nuclear magnetic resonance provided by embodiment, modules, unit in system
Effect and realize technical effect respectively with Parkinson's prediction technique provided by above-described embodiment based on nuclear magnetic resonance
Effect and the technical effect realized correspond to identical, and which is not described herein again.
It is shown in Figure 6, it is that one of a kind of Parkinson's forecasting system based on nuclear magnetic resonance provided by the invention is preferred
The structural block diagram of embodiment, comprising:
Image collection module 11, for obtaining cranium brain nuclear magnetic resonance image to be diagnosed；
Image processing module 12 is obtained for handling the cranium brain nuclear magnetic resonance image to be diagnosed wait diagnose
Tractography picture；And
Prediction module 13, for based on Parkinson's prediction model trained in advance tractography picture pair to be diagnosed according to
Parkinson predicts.
Preferably, Parkinson's prediction model is the model for having merged convolutional neural networks and fully-connected network；It is described
Parkinson's prediction model includes at least activation primitive, convolutional network layer, pond layer and fully connected network network layers.
Preferably, the system is trained Parkinson's prediction model by following steps:
The cranium brain nuclear magnetic resonance image of disturbances in patients with Parkinson disease and the cranium brain nuclear magnetic resonance image of normal person are obtained respectively；
The cranium brain nuclear magnetic resonance image of cranium brain nuclear magnetic resonance image and the normal person to the disturbances in patients with Parkinson disease carries out
Processing, it is corresponding to obtain disturbances in patients with Parkinson disease tractography picture and normal brain white matter image；
Mould is predicted to the Parkinson according to the disturbances in patients with Parkinson disease tractography picture and the normal brain white matter image
Type is trained and tests.
Preferably, the step of Parkinson's prediction model being trained further include:
The penalty values of prediction are obtained according to test result and actual result；
Parkinson's prediction model is optimized according to the penalty values.
Preferably, the prediction module specifically includes:
Feature extraction unit is mentioned for carrying out feature according to the tractography picture to be diagnosed based on the convolutional network layer
It takes, obtains characteristic pattern；
Compression unit obtains compressed characteristic pattern for compressing based on the pond layer to the characteristic pattern；
Converting unit, for carrying out conversion process to the compressed characteristic pattern based on the activation primitive；And
Predicting unit obtains prediction knot for integrating based on the fully connected network network layers to the result of conversion process
Fruit.
Preferably, described image obtains module and specifically includes:
Image acquisition unit, for obtaining the cranium brain nuclear magnetic resonance image to be diagnosed by nuclear magnetic resonane scanne.
Preferably, described image processing module specifically includes:
Image processing unit, at least carrying out skull removing, image to the cranium brain nuclear magnetic resonance image to be diagnosed
Segmentation, non-linear registration, image standardization and smoothing processing obtain the tractography picture to be diagnosed.
The embodiment of the invention also provides a kind of computer readable storage medium, the computer readable storage medium includes
The computer program of storage；Wherein, where the computer program controls the computer readable storage medium at runtime
Equipment executes Parkinson's prediction technique described in any of the above-described embodiment based on nuclear magnetic resonance.
It is shown in Figure 7 the embodiment of the invention also provides a kind of terminal device, it is that a kind of terminal provided by the invention is set
The structural block diagram of a standby preferred embodiment, including processor 10, memory 20 and be stored in the memory 20 and
It is configured as the computer program executed by the processor 10, the processor 10 is realized when executing the computer program
Parkinson's prediction technique described in any of the above-described embodiment based on nuclear magnetic resonance.
Preferably, the computer program can be divided into one or more module/units (such as computer program 1, meter
Calculation machine program 2), one or more of module/units are stored in the memory 20, and by
The processor 10 executes, to complete the present invention.One or more of module/units, which can be, can complete specific function
Series of computation machine program instruction section, the instruction segment is for describing execution of the computer program in the terminal device
Journey.
The processor 10 can be central processing unit (Central Processing Unit, CPU), can also be
Other general processors, digital signal processor (Digital Signal Processor, DSP), 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 10 is also possible to any conventional place
Device is managed, the processor 10 is the control centre of the terminal device, utilizes terminal device described in various interfaces and connection
Various pieces.
The memory 20 mainly includes program storage area and data storage area, wherein program storage area can store operation
Application program needed for system, at least one function etc., data storage area can store related data etc..In addition, the memory
20 can be high-speed random access memory, can also be nonvolatile memory, such as plug-in type hard disk, intelligent memory card
(Smart Media Card, SMC), secure digital (Secure Digital, SD) card and flash card (Flash Card) etc., or
The memory 20 is also possible to other volatile solid-state parts.
It should be noted that above-mentioned terminal device may include, but it is not limited only to, processor, memory, those skilled in the art
Member is appreciated that Fig. 7 structural block diagram is only the example of terminal device, does not constitute the restriction to terminal device, may include
Than illustrating more or fewer components, certain components or different components are perhaps combined.
To sum up, a kind of Parkinson's prediction technique based on nuclear magnetic resonance, system, computer provided by the embodiment of the present invention
Readable storage medium storing program for executing and terminal device carry out respective handling by the cranium brain nuclear magnetic resonance image to be diagnosed to acquisition, to obtain
Obtain the tractography picture to be diagnosed suitable for Parkinson's prediction model, and the pa that will be input to after training wait diagnose tractography picture
In the gloomy prediction model of gold, the prediction to Parkinson disease can be realized, and carry out to Parkinson disease according to tractography picture
Prediction, improves predictablity rate.
The above is only a preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
For member, without departing from the technical principles of the invention, several improvement and deformations can also be made, these improvement and deformations
Also it should be regarded as protection scope of the present invention.
Claims (10)
1. a kind of Parkinson's prediction technique based on nuclear magnetic resonance characterized by comprising
Obtain cranium brain nuclear magnetic resonance image to be diagnosed；
The cranium brain nuclear magnetic resonance image to be diagnosed is handled, tractography picture to be diagnosed is obtained；
Parkinson is predicted according to the tractography picture to be diagnosed based on Parkinson's prediction model trained in advance.
2. Parkinson's prediction technique based on nuclear magnetic resonance as described in claim 1, which is characterized in that Parkinson's prediction
Model is the model for having merged convolutional neural networks and fully-connected network；Parkinson's prediction model includes at least activation letter
Number, convolutional network layer, pond layer and fully connected network network layers.
3. Parkinson's prediction technique based on nuclear magnetic resonance as claimed in claim 1 or 2, which is characterized in that the method is logical
Following steps are crossed to be trained Parkinson's prediction model:
The cranium brain nuclear magnetic resonance image of disturbances in patients with Parkinson disease and the cranium brain nuclear magnetic resonance image of normal person are obtained respectively；
The cranium brain nuclear magnetic resonance image of cranium brain nuclear magnetic resonance image and the normal person to the disturbances in patients with Parkinson disease is handled,
It is corresponding to obtain disturbances in patients with Parkinson disease tractography picture and normal brain white matter image；
According to the disturbances in patients with Parkinson disease tractography picture and the normal brain white matter image to Parkinson's prediction model into
Row training and test.
4. Parkinson's prediction technique based on nuclear magnetic resonance as claimed in claim 3, which is characterized in that the method is also wrapped
It includes:
The penalty values of prediction are obtained according to test result and actual result；
Parkinson's prediction model is optimized according to the penalty values.
5. Parkinson's prediction technique based on nuclear magnetic resonance as claimed in claim 2, which is characterized in that described based on instruction in advance
Experienced Parkinson's prediction model predicts Parkinson according to the tractography picture to be diagnosed, and specifically includes:
Feature extraction is carried out according to the tractography picture to be diagnosed based on the convolutional network layer, obtains characteristic pattern；
The characteristic pattern is compressed based on the pond layer, obtains compressed characteristic pattern；
Conversion process is carried out to the compressed characteristic pattern based on the activation primitive；
The result of conversion process is integrated based on the fully connected network network layers, obtains prediction result.
6. Parkinson's prediction technique based on nuclear magnetic resonance as described in claim 1, which is characterized in that described to obtain wait diagnose
Cranium brain nuclear magnetic resonance image, specifically include:
The cranium brain nuclear magnetic resonance image to be diagnosed is obtained by nuclear magnetic resonane scanne.
7. Parkinson's prediction technique based on nuclear magnetic resonance as described in claim 1, which is characterized in that described to the follow-up
Disconnected cranium brain nuclear magnetic resonance image is handled, and is obtained tractography picture to be diagnosed, is specifically included:
Skull removing, image segmentation, non-linear registration, image mark are at least carried out to the cranium brain nuclear magnetic resonance image to be diagnosed
Standardization and smoothing processing obtain the tractography picture to be diagnosed.
8. a kind of Parkinson's forecasting system based on nuclear magnetic resonance characterized by comprising
Image collection module, for obtaining cranium brain nuclear magnetic resonance image to be diagnosed；
Image processing module obtains white matter of brain to be diagnosed for handling the cranium brain nuclear magnetic resonance image to be diagnosed
Image；And
Prediction module, for based on Parkinson's prediction model trained in advance tractography picture to be diagnosed according to Parkinson
It is predicted.
9. 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, the equipment where the computer program controls the computer readable storage medium at runtime executes such as
Parkinson's prediction technique described in any one of claims 1 to 7 based on nuclear magnetic resonance.
10. a kind of terminal device, which is characterized in that including processor, memory and store in the memory and matched
It is set to the computer program executed by the processor, the processor is realized when executing the computer program as right is wanted
Parkinson's prediction technique described in asking any one of 1 to 7 based on nuclear magnetic resonance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810983951.3A CN109222902A (en) | 2018-08-27 | 2018-08-27 | Parkinson's prediction technique, system, storage medium and equipment based on nuclear magnetic resonance |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810983951.3A CN109222902A (en) | 2018-08-27 | 2018-08-27 | Parkinson's prediction technique, system, storage medium and equipment based on nuclear magnetic resonance |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109222902A true CN109222902A (en) | 2019-01-18 |
Family
ID=65069639
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810983951.3A Pending CN109222902A (en) | 2018-08-27 | 2018-08-27 | Parkinson's prediction technique, system, storage medium and equipment based on nuclear magnetic resonance |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109222902A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110991408A (en) * | 2019-12-19 | 2020-04-10 | 北京航空航天大学 | Method and device for segmenting white matter high signal based on deep learning method |
| CN111543994A (en) * | 2020-04-24 | 2020-08-18 | 天津大学 | Epilepsy auxiliary detection system based on white matter connection diagram and parallel convolution neural network |
| WO2021115084A1 (en) * | 2019-12-11 | 2021-06-17 | 北京航空航天大学 | Structural magnetic resonance image-based brain age deep learning prediction system |
| CN114159071A (en) * | 2021-12-22 | 2022-03-11 | 南昌大学 | Parkinson prediction intelligent method and system based on electrocardiogram image |
| CN115040147A (en) * | 2022-06-01 | 2022-09-13 | 上海全景医学影像诊断中心有限公司 | Parkinson's disease prediction method based on 18F-FDG PET metabolic network |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107909588A (en) * | 2017-07-26 | 2018-04-13 | 广州慧扬健康科技有限公司 | Partition system under MRI cortex based on three-dimensional full convolutional neural networks |
-
2018
- 2018-08-27 CN CN201810983951.3A patent/CN109222902A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107909588A (en) * | 2017-07-26 | 2018-04-13 | 广州慧扬健康科技有限公司 | Partition system under MRI cortex based on three-dimensional full convolutional neural networks |
Non-Patent Citations (7)
| Title |
|---|
| SOHEIL ESMAEILZADEH等: "End-to-End Parkinson Disease Diagnosis using Brain MR-Images by 3D-CNN", 《HTTPS://ARXIV.ORG/ABS/1806.05233》 * |
| 丁淑贞等主编: "《神经内科临床护理》", 31 July 2016, 中国协和医科大学出版社 * |
| 俞一云等: "基于卷积神经网络的ADHD的判别分析", 《微型机与应用》 * |
| 庞清华: "《首都医科大学硕士学位论文》", 15 November 2016 * |
| 张巧丽等: "基于深度学习的医学影像诊断综述", 《计算机科学》 * |
| 张春良等主编: "《临床神经内科学》", 30 April 2014, 科学技术文献出版社 * |
| 朱莉等: "基于卷积神经网络的注意缺陷多动障碍分类研究", 《生物医学工程学杂志》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021115084A1 (en) * | 2019-12-11 | 2021-06-17 | 北京航空航天大学 | Structural magnetic resonance image-based brain age deep learning prediction system |
| CN110991408A (en) * | 2019-12-19 | 2020-04-10 | 北京航空航天大学 | Method and device for segmenting white matter high signal based on deep learning method |
| CN110991408B (en) * | 2019-12-19 | 2022-09-06 | 北京航空航天大学 | Method and device for segmenting white matter high signal based on deep learning method |
| CN111543994A (en) * | 2020-04-24 | 2020-08-18 | 天津大学 | Epilepsy auxiliary detection system based on white matter connection diagram and parallel convolution neural network |
| CN111543994B (en) * | 2020-04-24 | 2023-04-07 | 天津大学 | Epilepsy auxiliary detection system based on white matter connection diagram and parallel convolution neural network |
| CN114159071A (en) * | 2021-12-22 | 2022-03-11 | 南昌大学 | Parkinson prediction intelligent method and system based on electrocardiogram image |
| CN115040147A (en) * | 2022-06-01 | 2022-09-13 | 上海全景医学影像诊断中心有限公司 | Parkinson's disease prediction method based on 18F-FDG PET metabolic network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109222902A (en) | Parkinson's prediction technique, system, storage medium and equipment based on nuclear magnetic resonance | |
| Ahmadi et al. | QAIS-DSNN: tumor area segmentation of MRI image with optimized quantum matched-filter technique and deep spiking neural network | |
| Deng et al. | Deep learning-based HCNN and CRF-RRNN model for brain tumor segmentation | |
| Zhang et al. | LU-NET: An improved U-Net for ventricular segmentation | |
| JP6643771B2 (en) | Brain activity analyzer, brain activity analysis method, and brain activity analysis program | |
| CN108682009A (en) | A kind of Alzheimer's disease prediction technique, device, equipment and medium | |
| CN115205300B (en) | Fundus blood vessel image segmentation method and system based on cavity convolution and semantic fusion | |
| CN109223002A (en) | Self-closing disease illness prediction technique, device, equipment and storage medium | |
| CN110246109A (en) | Merge analysis system, method, apparatus and the medium of CT images and customized information | |
| Guo et al. | MMFGAN: A novel multimodal brain medical image fusion based on the improvement of generative adversarial network | |
| Goel et al. | Multimodal neuroimaging based Alzheimer's disease diagnosis using evolutionary RVFL classifier | |
| CN115147600A (en) | GBM multi-mode MR image segmentation method based on classifier weight converter | |
| Hou et al. | Cross attention densely connected networks for multiple sclerosis lesion segmentation | |
| CN109223001A (en) | Hyperactivity prediction technique, system, storage medium and equipment based on nuclear magnetic resonance | |
| Tomassini et al. | Brain-on-Cloud for automatic diagnosis of Alzheimer’s disease from 3D structural magnetic resonance whole-brain scans | |
| Pallawi et al. | Study of Alzheimer’s disease brain impairment and methods for its early diagnosis: a comprehensive survey | |
| Suganyadevi et al. | Alzheimer’s Disease Diagnosis using Deep Learning Approach | |
| Vincent et al. | Detection of hyperperfusion on arterial spin labeling using deep learning | |
| Chen et al. | HybridGAN: hybrid generative adversarial networks for MR image synthesis | |
| Sun | Empirical analysis for earlier diagnosis of Alzheimer’s disease using deep learning | |
| Rao et al. | Grouping and decoupling mechanism for diabetic retinopathy image grading | |
| Sadoon | Classification of medical images based on deep learning network (CNN) for both brain tumors and covid-19 | |
| Unal et al. | A comparison of feature extraction techniques for diagnosis of lumbar intervertebral degenerative disc disease | |
| Lu et al. | An Alzheimer's disease classification method based on ConvNeXt | |
| Kaur et al. | Diagnosis of Brain Tumor using Automated Deep Learning Model |
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: 20190118 |
|
| RJ01 | Rejection of invention patent application after publication |