CN109886944B - A kind of white matter high signal intensity detection and localization method based on multichannel chromatogram - Google Patents
A kind of white matter high signal intensity detection and localization method based on multichannel chromatogram Download PDFInfo
- Publication number
- CN109886944B CN109886944B CN201910107496.5A CN201910107496A CN109886944B CN 109886944 B CN109886944 B CN 109886944B CN 201910107496 A CN201910107496 A CN 201910107496A CN 109886944 B CN109886944 B CN 109886944B
- Authority
- CN
- China
- Prior art keywords
- white matter
- voxel
- multichannel chromatogram
- image
- target 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.)
- Active
Links
Landscapes
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The present invention disclose it is a kind of based on multichannel chromatogram white matter high signal intensity detection and localization method.The following steps are included: firstly, the FLAIR data using normal aging people brain establish multichannel chromatogram library.Secondly, the map in spectrum library is registrated to target image one by one, and merged using multichannel chromatogram blending algorithm.Finally, realizing image segmentation according to fusion results, and the voxel with local anomaly intensity is automatically detected as white matter hyperintensities.The present invention also detects and has oriented white matter hyperintensities region while being split to cerebral white matter, and the result of the region and manual delineation is coincide well, accuracy with higher and accuracy, and effect is better than current other white matter hyperintensities detection methods.
Description
Technical field
This application involves brain magnetic resonance field of image processings, more particularly to detection and localization method based on multichannel chromatogram.
Background technique
Occur in T2 weighting or FLAIR (fluid attenuated inversion recovery) magnetic resonance imaging
White matter hyperintensities (White matter hyperintensities, WMH) are a kind of common radiographic features.White matter height letter
Number the degree and distribution of thin vessels disease are mainly reflected, and they are increasingly common with advancing age.Nearest grinds
Study carefully the result shows that, they are also likely to be that there are also one of core features of Gray matter decrease for Alzheimer disease.
When assessment has the patient of memory disorders, consider that the spatial distribution of white matter hyperintensities is of great significance.For example, brain
White matter hyperintensities around room have stronger relevance compared with the white matter hyperintensities of deep layer and cognitive ability；In Alzheimer
In the development of disease, posteriority white matter hyperintensities load is played a very important role.The research group having before is according to priori knowledge
The white matter hyperintensities of specific brain region are detected, testing result will receive the influence of subjective factor, not so as to cause accuracy
It is high and do not have wide applicability.In the past decade, someone has developed full automatic white matter hyperintensities detection calculation
Method, including the various forms of thresholding methods based on intensity, cluster and machine learning method, Outlier Analysis method, form
Operation and bayes method etc. are learned, is detailed in document 1:Admiraal-Behloul et al., 2005；Jack et al.,
2001；Ji et al.,2013；Ong et al.,2012；Simoes et al.,2013；Yoo et al.,2014.Document 2:
Dyrby et al.,2008；Ghafoorian et al.,2017；Ghafoorian et al.,2016；Ithapu et
al.,2014；Lao et al.,2008；Seghier et al.,2008.Document 3:Maldjian et al., 2013；Ong
et al.,2012；Van Leemput et al.,2001；Yang et al.,2010.Document 4:Beare et al., 2009；
Shi et al.,2013.Document 5:Herskovits et al., 2008；Ithapu et al.,2014.Have in these methods
Using list FLAIR compare, some use multiple modalities include T1, T2, proton density or even diffusion tensor imaging data.This
The detection accuracy of a little methods is largely dependent upon application and image modalities.They would generally generate the white matter hyperintensities of full brain
The measured value of load, but be less able to automatically measure the white matter hyperintensities load in specific region, for systematically commenting
Estimate the distribution of white matter hyperintensities.Therefore, develop it is a kind of can be not only used for measuring total white matter hyperintensities, can also systematically comment
The tool for estimating white matter hyperintensities distribution is very important, for example, in different leaflet divisions, deep layer brain and subcortical structures
White matter hyperintensities load.
Summary of the invention
For overcome the deficiencies in the prior art, the white matter high signal intensity detection that the invention proposes a kind of based on multichannel chromatogram and
Localization method, this method have in conjunction with multichannel chromatogram likelihood fusion method using white matter hyperintensities by multichannel chromatogram FLAIR database
Have do not meet its anomaly intensity of local intensity distribution in the region of interest the characteristics of, cause it to merge in multichannel chromatogram
Journey has lower posterior probability, so that the segmentation of cerebral white matter and the detection of white matter hyperintensities can be achieved at the same time.
In order to achieve the above object, the present invention is achieved by the following scheme:
Based on multichannel chromatogram white matter high signal intensity detection and localization method, it the following steps are included:
Step 1: establishing FLAIR image multichannel chromatogram database, and the multichannel chromatogram database includes several cognitions normally but has
There are the FLAIR spectrum data of the tested individual of different degrees of encephalatrophy anatomical features, the brain region quilt of each spectrum data
Picture is divided into several area-of-interests, and is marked by label；
Step 2: by FLAIR image multichannel chromatogram database each atlas image and its label be registrated to target together
Image；
Step 3: being weighted fusion using multichannel chromatogram likelihood fusion method, obtains each voxel in target image and belongs to
The posterior probability of different labels；
Step 4: using Bayesian MAP estimate obtain target image in each voxel final segmentation tag and
The voxel belongs to the maximum a posteriori probability of final segmentation tag；
Step 5: the voxel by maximum posteriori probability value in target image lower than probability threshold value is judged as white matter hyperintensities body
Element.
Based on the program, each step can also further provide for following preferred implementation.It should be noted that each excellent
Select the technical characteristic in mode that can be combined with each other in case of no collision.These certain preferred embodiments can also lead to
It crosses the mode that other can be realized same technique effect to realize, not constitute a limitation.
Preferably, the method for building up of the FLAIR image multichannel chromatogram database of the step 1 is as follows:
The white matter hyperintensities load of all subjects is calculated, white matter hyperintensities is therefrom selected to be less than 1.8ml and cognition normally
The FLAIR data of individual establish multichannel chromatogram database, and the different maps in multichannel chromatogram database have same age stage crowd
It is middle minimum to moderate encephalatrophy anatomical features；Then FLAIR image is maximized using the SPM software package in Matlab
Mutual information between t1 weighted image, by each FLAIR image registration to corresponding t1 weighted image；Reuse MRICloud
Multichannel chromatogram segmentation is carried out, the brain region in every map is divided into several area-of-interests.
Preferably, the method for registering images of the step 2 is as follows:
Firstly, carrying out global nonuniformity correction to target image using the correction of N4 deviation；It then, will by affine transformation
Every atlas image and its label are registrated to the target image after correction together；Finally, being mapped using big deformation differential metric
LDDMM method iteratively carries out nonlinear transformation to atlas image and its label, is registrated to target image.
Preferably, the multichannel chromatogram likelihood weighting fusion method of the step 3 is as follows:
Calculate the posterior probability of each voxel:
Wherein, N is the total number of map in FLAIR image multichannel chromatogram database,It is target image ITMiddle body
Plain x belongs to the posterior probability of label l, wiIt (l) is weight term,It is the atlas image i obtained by Stochastic Separated Flow Model
Middle voxel x belongs to the priori likelihood of label l.
Preferably, the final segmentation tag calculation formula of target image is as follows in the step 4:
Wherein LTIt (x) is target image ITThe final segmentation tag of middle voxel x；L is the total of label in every atlas image
Number；
The maximum a posteriori probability of each voxel x, the i.e. voxel belong to final segmentation tag LT(x) posterior probability is
Preferably, the probability threshold value is set as 0.02；The voxel that maximum a posteriori probability is less than threshold value is demarcated as white
Matter high RST.
Preferably, after step 1 to five processing, then post-processed to reduce false positive rate, the post-processing side
Method is as follows: firstly generating target image deutocerebrum white matter area mask (mask), is handled using exposure mask target image, gone
White matter hyperintensities voxel in addition to masked areas；Then voxel intensities in target image are removed again is less than the dark of minimum intensity threshold
Black matrix element；The white matter hyperintensities cluster that volume in target image is less than minimum volume threshold value is finally removed, is obtained in target image most
Whole white matter hyperintensities voxel.
Further, the generation method of the cerebral white matter area mask is as follows:
FLAIR image multichannel chromatogram database is established according to the step 1, wherein the brain region quilt of each spectrum data
It is divided into the area-of-interest of several simplification；The area-of-interest of the simplification includes cerebral white matter region and cerebral white matter region
Other outer several functional areas；It is then based on the FLAIR image multichannel chromatogram database, repeating said steps two arrive step 4,
The final segmentation tag that each voxel in target image corresponds to simplified area-of-interest is obtained, target figure is determined according to label
As the distribution in deutocerebrum white matter region, cerebral white matter area mask is obtained.
Further, the minimum intensity threshold is that the voxel intensities mean value of area-of-interest subtracts its standard deviation
1.5 again.
Further, the minimum volume threshold value is 50mm3。
Compared with the existing technology, the invention has the characteristics that: the present invention is based on multichannel chromatogram blending algorithm and multichannel chromatograms
FLAIR database, traditional multichannel chromatogram algorithm is usually used in image segmentation, and the algorithm is used for white matter hyperintensities and examined by the present invention
It surveys.The present invention can regard a special case based on rejecting outliers as, i.e., go to distinguish have anomaly intensity with normal image
Voxel, the difference is that the strength characteristic that previous method is based only on patient image is not involved with reference to map, and
The present invention is to detect exceptional value based on the Strength mis match with Normal brain.It is detected in addition, being composed compared to free hand drawing, the present invention
Age, the anatomical features etc. between atlas image and target image can be preferably matched using multichannel chromatogram.Importantly,
If atlas includes structure predetermined, so that it may systematically identify the anatomical position of white matter hyperintensities.With it is existing
Detection method is compared, and white matter hyperintensities detection and location algorithm based on multichannel chromatogram have preferable detection accuracy, after maximum
Testing has stronger robustness in the selection of probability threshold value.In addition, the present invention has important clinical value, for example analyze
The variation of white matter hyperintensities load with age.
Detailed description of the invention
Fig. 1 is the part map of selection, these maps have from slightly to the encephalatrophy anatomical structure of moderate.
Fig. 2 is the flow chart of white matter high signal intensity detection and localization method based on multichannel chromatogram.
Fig. 3 be based on multichannel chromatogram white matter high signal intensity detection and localization method detection full white matter high signal intensity with manually
The Comparative result delineated.
Fig. 4 is the different degrees of white matter hyperintensities of white matter high signal intensity detection and localization method detection based on multichannel chromatogram
Load test result.
Specific embodiment
It is flow chart of the invention such as attached drawing 2.Show that it is specific in conjunction with the embodiments below based on method proposed by the present invention
Technical effect, so that those skilled in the art more fully understand essence of the invention.
In a kind of more excellent implementation of the invention, white matter high signal intensity detection and localization method packet based on multichannel chromatogram
Include following steps:
Step 1: establishing FLAIR image multichannel chromatogram database, which includes several cognitions normally but have
The FLAIR spectrum data of the tested individual of different degrees of encephalatrophy anatomical features, the brain region of each spectrum data are refreshing
It is drawn through anatomy expert and is divided into several area-of-interests, and be marked by label.The FLAIR image multichannel chromatogram database
Method for building up is as follows:
The white matter hyperintensities load in the map of all subjects is calculated, therefrom there are minimum white matter hyperintensities (to be less than for selection
1.8ml) and the FLAIR data of the normal tested individual of cognition establish multichannel chromatogram database, the different figures in multichannel chromatogram database
Spectrum represents minimum to moderate encephalatrophy anatomical features in same age stage crowd.Then using in Matlab
SPM software package maximizes the mutual information between FLAIR image map and t1 weighted image, by each FLAIR image registration to pair
The high-resolution t1 weighted image answered.It reuses MRICloud and identical more figures is carried out to t1 weighted image and FLAIR image
Spectrum segmentation, is divided into several area-of-interests (regions of interest, ROIs) for the brain region in every map,
And each region is marked by label.The dividing number of area-of-interest, which can according to need, to be adjusted, in this reality
It applies in example, in order to be split to entire brain region, every map can be divided into 143 ROI regions, this 143 areas
There are 24 white matter regions, others are grey matter regions, celiolymph region etc., or are non-brain areas in domain.Each
The label in region can be used for being split target image by subsequent step, determine the position in the regions such as white matter of brain in target image
It sets.
Step 2: by FLAIR image multichannel chromatogram database each atlas image and its label be registrated to target together
Image (image to be split).Method for registering images is as follows:
Firstly, carrying out global nonuniformity correction to target image using the correction of N4 deviation.It then, will by affine transformation
Every atlas image and its label are registrated to the target image after correction together.Finally, being mapped using big deformation differential metric
(LDDMM) method iteratively carries out nonlinear transformation to atlas image and its label, is registrated to target image.Specific the number of iterations
It can be set as needed, it is iteration 3 times generally optional.
Step 3: being weighted fusion using multichannel chromatogram likelihood fusion method, obtains each voxel in target image and belongs to
The posterior probability of different labels.Wherein in multichannel chromatogram likelihood weighting fusion method, need to calculate the posterior probability of each voxel:
Wherein, N is the total number of map in FLAIR image multichannel chromatogram database；It is target image ITMiddle body
Plain x belongs to the posterior probability of label l；wiIt (l) is weight term, the best value of the weight term can be in pre-trial by multiple
Iteration is determined；It is that voxel x belongs to the priori of label l seemingly in the atlas image i obtained by Stochastic Separated Flow Model
So.Voxel with anomaly intensity has lower priori likelihood about corresponding label.
Step 4: using Bayesian MAP estimate obtain target image in each voxel final segmentation tag and
The voxel belongs to the maximum a posteriori probability of final segmentation tag.The final segmentation tag calculation formula of target image is as follows:
Wherein LTIt (x) is target image ITThe final segmentation tag of middle voxel x；L is the total of label in every atlas image
Number.
The final segmentation tag of each voxel x can be transferred through estimating to be determined in target image as a result, can be according to this
A little labels judge the position belongs to which kind of area type, such as white matter, grey matter, celiolymph etc..
And rule needs the maximum a posteriori probability (maximized based on each voxel x to white matter hyperintensities voxel really
Posterior probability, MPP) it is determined, the MPP i.e. voxel belongs to final segmentation tag LT(x) posteriority is general
Rate, value are
Step 5: maximum posteriori probability value MPP in target image is judged as that white matter height is believed lower than the voxel of probability threshold value
Number voxel.In this step, the probability threshold value for judgement can be according to actually being set, and usually experience optimizes,
Preferably 0.02.
By above-mentioned steps one to five, target image can be divided into several ROIs, and realize brain region point
The positioning of white matter hyperintensities voxel is gone back while realized while cutting.It is however noted that obtained by above-mentioned steps
There may be false positive phenomenons for white matter hyperintensities voxel, i.e., the voxel that part not should belong to white matter hyperintensities originally is identified as
White matter hyperintensities voxels.Therefore, it can be post-processed based on the 6th following steps to reduce false positive rate.
Step 6: post-processing to target image, false positive is reduced, the specific method is as follows:
Target image deutocerebrum white matter area mask is firstly generated, target image is handled using exposure mask, positioned at covering
Voxel outside diaphragm area is not belonging to white matter of brain, therefore even if it is identified as white matter hyperintensities voxel, it should also be picked
It removes.
The generation method of cerebral white matter area mask can be in the following way:
Establish another FLAIR image multichannel chromatogram database in the way of described in abovementioned steps one, but the database
Difference is that the brain region of wherein each spectrum data is divided into the area-of-interest of several simplification.In original database
In, every map is divided into 143 ROIs, but every map is only divided into 7 ROIs in the spectrum library of this step, including
7 brain major regions such as cerebral white matter region, cerebral gray matter region, celiolymph region.It is then based on FLAIR figure
As multichannel chromatogram database, step 2 is repeated to step 4 to target image again, obtain can into target image each voxel
Corresponding to the final segmentation tag of simplified area-of-interest, the segmentation tag in this step is also simplified segmentation tag.
It is assured that the distribution in target image deutocerebrum white matter region according to label, thus obtains cerebral white matter area mask.
After mask process, there is also the black dull voxel in part in target image, need to be removed it.The present invention
In, minimum intensity threshold is set as the voxel intensities mean value of area-of-interest and subtracts 1.5 times of its standard deviation, then by target figure
The voxel that voxel intensities are less than minimum intensity threshold as in is accordingly to be regarded as black dull voxel and is rejected.
Finally, in target image there is likely to be part dispersion small size white matter hyperintensities cluster, this part be also required into
Row removal.Therefore, the minimum volume threshold value of white matter hyperintensities cluster is set as 50mm3, then remove volume in target image and be less than
The white matter hyperintensities cluster of minimum volume threshold value.
By above-mentioned post-processing, white matter hyperintensities voxel final in target image is obtained.
The step of below based on the above method one to six, is in conjunction with the embodiments shown its technical effect, so as to ability
Field technique personnel more fully understand essence of the invention.
Embodiment
By the above-mentioned white matter high signal intensity detection based on multichannel chromatogram and localization method in 135 old subject FLAIR data
In be tested, these participants are to recognize normal (n=113) or have mild cognitive impairment (n=22).15 are chosen to recognize
The FLAIR data of normal individual are known as map, and remaining 120 are used as algorithm evaluation and analysis.As shown in Figure 1,15 subjects
The FLAIR spectrum data of individual (part subject is only shown in figure) has different degree encephalatrophy anatomical features.Specifically
Way is referring to above-mentioned steps one, and details are not described herein again, only introduces design parameter herein below.MRI scan is sharp with flying
What Pu Achieva 3.0T scanner carried out；FLAIR data use multilayer fast acquisition interleaved spin echo, inversion recovery pulse reversion
Time (TI)/echo time (TE)/repetition time (TR)=2800/100/11000ms, the visual field (FOV)=256 × 256mm, altogether
69 layers slice, each slice with a thickness of 2mm；T1 weighted image is to prepare double echo steady state using three-dimensional magnetization
(MPRAGE), using TI/TE/TR=800/3/7ms, flip angle is 8 degree, and the visual field is 256 × 256 × 204mm, resolution ratio 1
×1×1.2mm。
Technical effect of the invention is shown in order to compare simultaneously, and the present embodiment is also to complete in such a way that expert manually delineates
White matter high signal intensity region is marked.The present embodiment and the comparing result manually delineated are as shown in Fig. 3:
By attached drawing 3, it can be seen that, the full white matter high signal intensity load of method detection proposed by the present invention delineates knot with artificial
Fruit correlation is good, and correlation reaches 0.97 between class.
Further, different degrees of white matter hyperintensities load is detected.Highlight regions are shown to be examined by method proposed by the present invention
The white matter hyperintensities label measured, experimental result are as shown in Fig. 4:
By attached drawing 4, it can be seen that, the biggish testing result of white matter hyperintensities load is preferable.
Table 1 is white matter high signal intensity detection and localization method and other two kinds of advanced white matter hyperintensities based on multichannel chromatogram
The comparing result of detection algorithm, in remaining two kinds of algorithm, one is brain intensity anomaly classification algorithm, (BIANCA is detailed in
Griffanti et al., 2016), another kind is lesion segmentation tool box (LST, is detailed in Schmidt, 2017).It can be with by table 1
See, method proposed by the present invention has highest intra class correlation, and three has similar Dice coefficient, and BIANCA has most
High false positive rate and minimum false negative rate, LST have minimum false positive rate and highest false negative rate.
The comparing result of 1 three kinds of algorithms of table
It should be pointed out that above-mentioned embodiment is only a preferred solution of the present invention, so its not to
The limitation present invention.Those of ordinary skill in related technical field may be used also without departing from the spirit and scope of the present invention
To make a variety of changes and modification.Therefore all mode technical solutions obtained for taking equivalent substitution or equivalent transformation, fall
Within the scope of the present invention.
Claims (10)
1. a kind of white matter high signal intensity detection and localization method based on multichannel chromatogram, which comprises the following steps:
Step 1: establishing FLAIR image multichannel chromatogram database, and the multichannel chromatogram database includes several cognitions normally but has not
With the FLAIR spectrum data of the tested individual of degree encephalatrophy anatomical features, the brain region of each spectrum data is divided
For several area-of-interests, and it is marked by label；
Step 2: by FLAIR image multichannel chromatogram database each atlas image and its label be registrated to target image together；
Step 3: being weighted fusion using multichannel chromatogram likelihood fusion method, obtains each voxel in target image and belongs to difference
The posterior probability of label；
Step 4: estimate to obtain the final segmentation tag and the body of each voxel in target image using Bayesian MAP
Element belongs to the maximum a posteriori probability of final segmentation tag；
Step 5: the voxel by maximum posteriori probability value in target image lower than probability threshold value is judged as white matter hyperintensities voxel.
2. the white matter high signal intensity detection and localization method according to claim 1 based on multichannel chromatogram, which is characterized in that institute
The method for building up for stating the FLAIR image multichannel chromatogram database of step 1 is as follows:
The white matter hyperintensities load of all tested individuals is calculated, white matter hyperintensities is therefrom selected to be less than 1.8ml and cognition normally
The FLAIR data of individual establish multichannel chromatogram database, and the different maps in multichannel chromatogram database have same age stage crowd
It is middle minimum to moderate encephalatrophy anatomical features；Then FLAIR image is maximized using the SPM software package in Matlab
Mutual information between t1 weighted image, by each FLAIR image registration to corresponding t1 weighted image；Reuse MRICloud
Multichannel chromatogram segmentation is carried out, the brain region in every map is divided into several area-of-interests.
3. the white matter high signal intensity detection and localization method according to claim 1 based on multichannel chromatogram, which is characterized in that institute
The method for registering images for stating step 2 is as follows:
Firstly, carrying out global nonuniformity correction to target image using the correction of N4 deviation；Then, by affine transformation by every
Atlas image and its label are registrated to the target image after correction together；Finally, mapping the side LDDMM using big deformation differential metric
Method iteratively carries out nonlinear transformation to atlas image and its label, is registrated to target image.
4. the white matter high signal intensity detection and localization method according to claim 1 based on multichannel chromatogram, which is characterized in that institute
The multichannel chromatogram likelihood weighting fusion method for stating step 3 is as follows:
Calculate the posterior probability of each voxel:
Wherein, N is the total number of map in FLAIR image multichannel chromatogram database,It is target image ITMiddle voxel x belongs to
In the posterior probability of label l, wiIt (l) is weight term,It is body in the atlas image i obtained by Stochastic Separated Flow Model
Plain x belongs to the priori likelihood of label l.
5. the white matter high signal intensity detection and localization method according to claim 1 based on multichannel chromatogram, which is characterized in that institute
The final segmentation tag calculation formula for stating target image in step 4 is as follows:
Wherein LTIt (x) is target image ITThe final segmentation tag of middle voxel x；L is the sum of label in every atlas image；Often
The maximum a posteriori probability of a voxel x, the i.e. voxel belong to final segmentation tag LT(x) posterior probability is
6. the white matter high signal intensity detection and localization method according to claim 1 based on multichannel chromatogram, which is characterized in that institute
It states probability threshold value and is set as 0.02；The voxel that maximum a posteriori probability is less than threshold value is demarcated as white matter hyperintensities.
7. the white matter high signal intensity detection and localization method according to claim 1 based on multichannel chromatogram, which is characterized in that warp
It after crossing step 1 to five processing, then is post-processed to reduce false positive rate, the post-processing approach is as follows: firstly generating target
Image deutocerebrum white matter area mask, is handled target image using exposure mask, removes the white matter hyperintensities outside masked areas
Voxel；Then the black dull voxel that voxel intensities in target image are less than minimum intensity threshold is removed again；Finally remove target image
Middle volume is less than the white matter hyperintensities cluster of minimum volume threshold value, obtains white matter hyperintensities voxel final in target image.
8. the white matter high signal intensity detection based on multichannel chromatogram and localization method as claimed in claim 7, which is characterized in that described
Cerebral white matter area mask generation method it is as follows:
FLAIR image multichannel chromatogram database is established according to the step 1, wherein the brain region of each spectrum data is divided
For the area-of-interest of several simplification；The area-of-interest of the simplification includes outside cerebral white matter region and cerebral white matter region
Other several functional areas；It is then based on the FLAIR image multichannel chromatogram database, repeating said steps two arrive step 4, obtain
Each voxel corresponds to the final segmentation tag of simplified area-of-interest in target image, is determined in target image according to label
The distribution in cerebral white matter region obtains cerebral white matter area mask.
9. the white matter high signal intensity detection based on multichannel chromatogram and localization method as claimed in claim 7, which is characterized in that described
Minimum intensity threshold be that the voxel intensities mean value of area-of-interest subtracts 1.5 times of its standard deviation.
10. the white matter high signal intensity detection based on multichannel chromatogram and localization method as claimed in claim 7, which is characterized in that institute
The minimum volume threshold value stated is 50mm3。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910107496.5A CN109886944B (en) | 2019-02-02 | 2019-02-02 | A kind of white matter high signal intensity detection and localization method based on multichannel chromatogram |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910107496.5A CN109886944B (en) | 2019-02-02 | 2019-02-02 | A kind of white matter high signal intensity detection and localization method based on multichannel chromatogram |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109886944A CN109886944A (en) | 2019-06-14 |
| CN109886944B true CN109886944B (en) | 2019-11-22 |
Family
ID=66927888
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910107496.5A Active CN109886944B (en) | 2019-02-02 | 2019-02-02 | A kind of white matter high signal intensity detection and localization method based on multichannel chromatogram |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109886944B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110517222A (en) * | 2019-07-10 | 2019-11-29 | 杭州电子科技大学 | A kind of brain image dividing method based on multichannel chromatogram |
| CN110503655B (en) * | 2019-08-21 | 2022-04-01 | 山东师范大学 | MRI glioma segmentation method and system based on sparse Bayesian model and multi-map fusion |
| CN111681230A (en) * | 2020-06-10 | 2020-09-18 | 华中科技大学同济医学院附属同济医院 | System and method for scoring high-signal of white matter of brain |
| CN113777546B (en) * | 2021-09-13 | 2023-04-25 | 西安交通大学医学院第一附属医院 | Magnetic resonance multi-parameter white matter high signal quantification method based on three-dimensional mapping |
| CN113902724B (en) * | 2021-10-18 | 2022-07-01 | 广州医科大学附属肿瘤医院 | Method, device, equipment and storage medium for classifying tumor cell images |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107146228A (en) * | 2017-03-22 | 2017-09-08 | 东南大学 | A kind of super voxel generation method of brain magnetic resonance image based on priori |
| CN108010048A (en) * | 2017-12-05 | 2018-05-08 | 华中科技大学 | A kind of hippocampus dividing method of the automatic brain MRI image based on multichannel chromatogram |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101540047A (en) * | 2009-04-30 | 2009-09-23 | 西安电子科技大学 | Texture image segmentation method based on independent Gaussian hybrid model |
| CN103295236B (en) * | 2013-05-29 | 2015-12-23 | 湘潭大学 | Markov multiple features random field models construction method and brain MR image cutting techniques thereof |
| CN104599270B (en) * | 2015-01-18 | 2017-10-10 | 北京工业大学 | A kind of Ultrasound Image of Breast Tumor dividing method based on improvement level set algorithm |
| CN107679566A (en) * | 2017-09-22 | 2018-02-09 | 西安电子科技大学 | A kind of Bayesian network parameters learning method for merging expert's priori |
-
2019
- 2019-02-02 CN CN201910107496.5A patent/CN109886944B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107146228A (en) * | 2017-03-22 | 2017-09-08 | 东南大学 | A kind of super voxel generation method of brain magnetic resonance image based on priori |
| CN108010048A (en) * | 2017-12-05 | 2018-05-08 | 华中科技大学 | A kind of hippocampus dividing method of the automatic brain MRI image based on multichannel chromatogram |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109886944A (en) | 2019-06-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109886944B (en) | A kind of white matter high signal intensity detection and localization method based on multichannel chromatogram | |
| Fahmy et al. | Automated analysis of cardiovascular magnetic resonance myocardial native T1 mapping images using fully convolutional neural networks | |
| García-Lorenzo et al. | Review of automatic segmentation methods of multiple sclerosis white matter lesions on conventional magnetic resonance imaging | |
| Bouix et al. | On evaluating brain tissue classifiers without a ground truth | |
| Chen et al. | Tissue-specific compartmental analysis for dynamic contrast-enhanced MR imaging of complex tumors | |
| EP3712849B1 (en) | Automated uncertainty estimation of lesion segmentation | |
| Vasilic et al. | A novel local thresholding algorithm for trabecular bone volume fraction mapping in the limited spatial resolution regime of in vivo MRI | |
| Fantini et al. | Automatic detection of motion artifacts on MRI using Deep CNN | |
| Dugas-Phocion et al. | Hierarchical segmentation of multiple sclerosis lesions in multi-sequence MRI | |
| Tan et al. | Cardiac left ventricle segmentation using convolutional neural network regression | |
| Shahzad et al. | Fully-automatic left ventricular segmentation from long-axis cardiac cine MR scans | |
| Xue et al. | Unsupervised inline analysis of cardiac perfusion MRI | |
| CN111415361B (en) | Method and device for estimating brain age of fetus and detecting abnormality based on deep learning | |
| Tian et al. | Improving in vivo human cerebral cortical surface reconstruction using data-driven super-resolution | |
| Wang et al. | Detection of spectral signatures in multispectral MR images for classification | |
| Vakli et al. | Predicting body mass index from structural MRI brain images using a deep convolutional neural network | |
| Bartoli et al. | Deep learning–based automated segmentation of left ventricular trabeculations and myocardium on cardiac MR images: a feasibility study | |
| Zhang et al. | Learning from multiple annotators for medical image segmentation | |
| US8315451B2 (en) | Method for segmentation of an MRI image of a tissue in presence of partial volume effects and computer program implementing the method | |
| Hann et al. | Ensemble of deep convolutional neural networks with monte carlo dropout sampling for automated image segmentation quality control and robust deep learning using small datasets | |
| Wang et al. | An automatic interpretable deep learning pipeline for accurate Parkinson's disease diagnosis using quantitative susceptibility mapping and T1‐weighted images | |
| Roy et al. | Synthesizing MR contrast and resolution through a patch matching technique | |
| CN112785540A (en) | Generation system and method of diffusion weighted image | |
| Arzhaeva et al. | Automated estimation of progression of interstitial lung disease in CT images | |
| CN115861464A (en) | Pseudo CT (computed tomography) synthesis method based on multimode MRI (magnetic resonance imaging) synchronous generation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |