CN104537711B - A kind of white matter fiber parameterized model construction method based on healthy population - Google Patents
A kind of white matter fiber parameterized model construction method based on healthy population Download PDFInfo
- Publication number
- CN104537711B CN104537711B CN201410835831.0A CN201410835831A CN104537711B CN 104537711 B CN104537711 B CN 104537711B CN 201410835831 A CN201410835831 A CN 201410835831A CN 104537711 B CN104537711 B CN 104537711B
- Authority
- CN
- China
- Prior art keywords
- fiber
- point
- fibre bundle
- prototype
- distance
- 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
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/50—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for simulation or modelling of medical disorders
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of white matter fiber parameterized model construction method based on healthy population, first, the full brain fibre bundle tracking of deterministic type is carried out on tensor collection of illustrative plates, fibre bundle interested is extracted.Fiber ligature is carried out to fibre bundle according to average minimum distance criterion, prototype fiber is chosen using local density's method of weighting for the fibre bundle after ligature, and discretization is carried out to prototype fiber to set up common coordinate system system according to arc length, it will set up corresponding relation between the point on the point and prototype fiber on fibre bundle on other fibers using optimal point matching algorithm.Finally, individual fibers characteristic is mapped in collection of illustrative plates fibre bundle with reference to trilinear interpolation, so as to establish the corresponding relation of fiber point between Different Individual, the characteristic curve of individual subject is obtained, so as to construct the parameterized model of healthy population.
Description
Technical field
The invention belongs to diffusion tensor imaging field, it is related to a kind of white matter fiber parametrization mould based on healthy population
Type construction method.
Background technology
Diffusion tensor imaging (diffusion tensor imaging, DTI) is that one kind can detect in vivo water
The mr imaging technique of molecule diffusion motion, with non-invasive feature, has been widely used in brain science research.Based on expansion
The cerebral white matter fiber beam imaging for dissipating tensor imaging technology is one of focus of current Neuroscience Research, and it utilizes diffusion tensor
Directional information is tracked to the diffusion motion state of big intracerebral hydrone, can obtain shape, the topology knot of white matter fiber tract
The information such as structure and locus.The tracking display of fiber can be used for the solution formulation before neurosurgery, in operation
Positioning is instructed and postoperative treatment and evaluation, contributes to influence of the correct understanding disease to white matter fiber channel, so as to have
The cerebral white matter disease that the diagnosis of effect ground is caused due to fiber missing.
However, because the difference of brain form between Different Individual make it that white matter fiber tract is without comparable between Different Individual
Property, and the space anatomical position of tissue points is difficult to correspondence on fibre bundle, therefore the statistical analysis of fibre bundle carry out group is relatively stranded
It is difficult.This research is based on brain tensor collection of illustrative plates, the white matter fiber tract information between the perceptual brain area of extraction, between the different fiber points of establishment,
The anatomical position corresponding relation of templatespace and individual space, so as to build the fiber parameters model based on healthy population.Base
White matter fiber tract can be not only accurately positioned in the fiber parameters method of tensor collection of illustrative plates, and can be to white matter fiber tract amount
The statistical analysis of the parameter characteristic group of change, this has great importance for the research of white matter fiber abnormal diseases.
The main stream approach that we handle DTI data sets at present is the spatial statisticses (Tract-Based based on fibre bundle
Spatial Statistics, TBSS) method.This method can be comparing the skeleton white matter fiber tract group of different subjects
Compared with searching has the position of significant difference, that is, abnormal findings of white matter region.Based on abnormal area, we can carry out probabilistic type
Fibre bundle is followed the trail of, so as to estimate the trend of fibre bundle.Probabilistic type is followed the trail of and solves the problems, such as fiber crossovers to a certain extent, and can
Dissection connection between the grey matter brain area relatively low to study FA values.
But TBSS methods can only analyze the white matter fiber tract on skeleton, it is impossible to obtain the damage of white matter elsewhere.Generally
The tracking of rate fiber type can only describe between two brain areas the presence or absence of fibre bundle and probably trend, it is impossible to the local shape of fibre bundle
State is quantified, and causes healthy population and patient to carry out the comparison in group.Therefore, there is office in current mainstream research method
It is sex-limited, it is impossible to the need for meeting clinical research and diagnosis.
The content of the invention
It is an object of the invention to the defect for overcoming above-mentioned technology presence, there is provided a kind of white matter fiber based on healthy population
Parameterized model construction method, it is possible to achieve deploy the analysis to fibre bundle in the range of full brain, and be not limited solely to skeleton,
And fibre bundle bar number can be quantified between brain area and the correspondence put in fibre bundle on different fibers can be solved along fiber beam analysis
Problem, so that the statistical analysis of realization group horizontal fibre beam.This method is based on brain tensor collection of illustrative plates, proposes cerebral white matter fiber beam
Correlation technique is parameterized, so that the corresponding relation for the anatomical position put between establishing Different Individual fiber, constructs healthy population
Fiber parameters model.First, the full brain fibre bundle tracking of deterministic type is carried out on tensor collection of illustrative plates, fiber interested is extracted
Beam.Fiber ligature is carried out to fibre bundle according to average minimum distance criterion, added for the fibre bundle after ligature using local density
Power method chooses prototype fiber, and discretization is carried out to prototype fiber to set up common coordinate system system according to arc length, using optimal
Point matching algorithm will set up corresponding relation between the point on the point and prototype fiber on fibre bundle on other fibers.Finally, with reference to three
Individual fibers characteristic is mapped in collection of illustrative plates fibre bundle by sublinear interpolation, so as to establish the correspondence pass of fiber point between Different Individual
System, obtains the characteristic curve of individual subject, so as to construct the parameterized model of healthy population.
Its concrete technical scheme is：
A kind of white matter fiber parameterized model construction method based on healthy population, comprises the following steps：
Step one：The structure of collection of illustrative plates is carried out using magnetic resonance imaging means acquisition brain function data and to the data of acquisition
Build, full brain fibre bundle tracking is carried out on tensor collection of illustrative plates, and extract fibre bundle interested；
Step 2：Fiber ligature processing is carried out based on obtained fibre bundle is extracted in step one, to remove abnormal fibrous；
Fiber is a continuous curve in three dimensions, is represented by a series of three-dimensional coordinate point, in fibre bundle
Any two fiber FiAnd Fj(i ≠ j), d represents the distance between they, and t represents distance threshold, as d (Fi, Fj) ＜ t when,
Then think that this two fibers belong to same cluster fiber；
When the number put on fiber is more, 15 points, F are taken at equal intervals on every fiberiTo FjDistance not
Equal to FjTo FiDistance, be averaged using to the value at symmetric position in matrix come symmetrization processing array；
If pkIt is FiOn point, plIt is FjOn point, then fiber FiAnd FjThe distance between using nearest average distance carry out
Calculate, specific formula is as follows：
dM(Fi, Fj)=mean (dm(Fi, Fj), dm(Fj, Fi)) (1)
Wherein,||·||being the Euclidean norm；
Nearest average distance dMThe global similarity of fiber is described, shape is more similar, spatially closer to dMValue is more
It is small；
Selection for distance threshold t is obtained using histogrammic method, seeks distance two-by-two between fiber first, right
In each fiber, first it is averaged with the distance of other all fibres, then judges whether the histogram meets normal state
Distribution, if normal probability plot is shown as linear pattern form, meets normal distribution, a point position is asked by setting significance
Number, as distance threshold t, so that the fiber whole piece that average distance value is more than quantile be removed, reaches and fibre bundle is tied up
The purpose of beam；
Step 3：Prototype fiber is extracted using local fiber Density Weighted most long method to the fibre bundle after ligature, to build
Found public fibre bundle coordinate system；
Fiber is normalized by setting up a common coordinate system system, selection can most represent other all in whole fibre bundle
One of fiber, namely prototype fiber, set up common coordinate system by the discretization based on arc length to prototype fiber and unite, then
The corresponding relation put on point and prototype fiber on other fibers is determined according to certain algorithm；
Select in fibre bundle that local fiber Density Weighted most long fiber is as prototype fiber, fibre density is by each
The number of the fiber track of voxel, it is necessary first to calculate the density of point equidistant on fiber, will then along every fiber
Density is integrated, so as to obtain the fibre density of every fiber, that maximum fiber of selective value is used as prototype fiber；
Step 4：Discretization is carried out to prototype fiber according to arc length, public fibre bundle coordinate system is set up；
Used during tracking straight line on the streamline form tracing algorithm based on Euler's method, fiber between two consecutive points away from
It is equal to step value from size, the point on fiber is than comparatively dense, and the air line distance between two consecutive points is approximately equal to it along fiber
Arc length distance, arc length is approximately replaced using step-length come to prototype fiber sampling discretization, the public affairs defined based on prototype fiber arc length
Coordinate system has built up altogether；
Step 5：Arc length reference coordinate is passed to by other all fibres by fiber point matching algorithm, realizes that collection of illustrative plates is fine
Tie up the normalization of beam；
Fiber Point matching process is transmitted the arc length reference coordinate set up according to prototype fiber by optimum point matching process
Other all fibres are given, to determine itself and the corresponding relation put on prototype fiber, so as to set up arc length coordinate and each collection of illustrative plates
The corresponding relation put on fiber, so that collection of illustrative plates fibre bundle be normalized；
OP methods are matched using Hungarian algorithms to each point on prototype fiber, along being approximately perpendicular to prototype
The direction finding match point of fiber, this method carries out rewards and punishments using cost function, it is assumed that i is the point on prototype fiber, and i is fiber
On point, this distance definition is measure distance D between point i and point jij, formula is as follows：
Wherein, vijIt is the vector between point i and point j；Estimate tensorvijIt is cutting at i points on prototype fiber
Line vector, the effect of this measure distance be make along perpendicular to the direction of prototype fiber carry out zero cost matching, now away from
From tangent vector is perpendicular to, as angle is by vertically changing, cost increase is only measured along prototype fiber tangent line side
To distance, set a threshold value to limit match point within the 40% of prototype fiber point spacing, so along positive and negative tangent line
The search box size in direction is the 80% of spacing, it is ensured that non-optimal match point is not matched；
By being matched with prototype fiber point, each fiber can all produce a series of arc length coordinate；
Step 6：The diffusion properties being tested under individual space are mapped on collection of illustrative plates fibre bundle, the fibre of personal subject is built
Tie up parameterized model；
Detailed process is：DTI images using individual calculate total during map construction non-linear registration as reference picture
The inversion carry over of Deformation Field, then, using individual space as object space, utilizes inverse transformation using tensor collection of illustrative plates as place space respectively
Field calculates coordinate position corresponding under individual space, root according to corresponding collection of illustrative plates fiber coordinate position at each arc length coordinate points
Coordinate position under the individual space obtained according to calculating, the seat is obtained in individual space FA images using tri-linear interpolation methods
Diffusion properties value at diffusion properties value at cursor position, namely correspondence arc length coordinate points, so as to obtain individual subject diffusion category
Change curve of the property on arc length coordinate.
Preferably, the diffusion properties being tested under individual space are mapped to the mapping on collection of illustrative plates fibre bundle described in step 6
During used in tri-linear interpolation methods algorithm it is as follows：
Postulated point V (x, y, z) is known point, and f (x, y, z) is to need to calculate the value at the V points asked for.For V points,
X, y, z is its coordinate value respectively, and x+, y+, z+ are integer coordinate values nearest with V points in positive direction, and x-, y-, z- are negative directions
The upper and nearest integer coordinate values of V points, if
First, along y-axis interpolation, the point that interpolation is obtained be respectively V1, V2, V3 and V4 (correspond to Fig. 3 in black point),
Assuming that the functional value at these points is respectively a1, a2, b1 and b2, then obtained according to linear interpolation formula：
V1(x-, y, z-)：a1=f (x-, y, z-)=yd*f(x-, y+, z-)+(1-yd)*f(x-, y-, z-) (5)
Same has：
V2(x+, y, z-)：a2=f (x+, y, z-)=yd*f(x+, y+, z-)+(1-yd)*f(x+, y-, z-)
(4-4)V3(x-, y, z+)：b1=f (x-, y, z+)=yd*f(x-, y+, z+)+(1-yd)*f(x-, y-, z+)
(4-5)V4(x+, y, z+)：b2=f (x+, y, z+)=yd*f(x+, y+, z+)+(1-yd)*f(x+, y-, z+) (6)
Then, along x-axis interpolation, the point that interpolation is obtained is respectively Q1 and Q2, it is assumed that the functional value at these points is respectively
C1 and c2, then it is same to be obtained according to linear interpolation formula：
Q1(x, y, z-)：c1=f (x, y, z-)=xd*f(x+, y, z-)+(1-xd)*f(x-, y, z-)
=xd*a2+(1-xd)*a1 (7)
Q2(x, y, z+)：c2=f (x, y, z+)=xd*f(x+, y, z+)+(1-xd)*f(x-, y, z+)
=xd*b2+(1-xd)*b1 (8)
Finally, along z-axis interpolation, the point that interpolation is obtained is known point V, the functional value f (x, y, z) at V points obtained
As required interpolation result：
In order to program conveniently, a series of above-mentioned formula can be merged into a formula, i.e.,
So far, Tri linear interpolation process terminates.
Compared with prior art, beneficial effects of the present invention are：
The present invention is studied based on self-built tensor collection of illustrative plates and realizes fibre bundle parametric method, reduces individual morphology
The influence of difference, establishes the dissection corresponding relation put between fiber.It is then based on fibre bundle parametric method and constructs Healthy People
The fibre bundle parameterized model of group.
(1) present invention carries out the research of fibre bundle parametric method using the self-built tensor collection of illustrative plates of team.Self-built tensor
The structure of collection of illustrative plates uses the method based on group, and it has merged the information of all subjects, thus reduces the influence of interindividual variation,
So that the fiber parameters model construction based on collection of illustrative plates is more accurate.In addition, the collection of illustrative plates is believed according to the data of research object
Breath structure is obtained, compared with international standard template, in the research saliency for specific crowd very big advantage.
(2) present invention establishes the corresponding relation put between fiber using the parametric method based on collection of illustrative plates.This method is used
Local density's method of weighting extracts prototype fiber, and prototype fiber is carried out into discretization to set up common coordinate system system according to arc length,
The corresponding relation put in collection of illustrative plates fibre bundle on all fibres is established using Point matching method, individual is tested with reference to Tri linear interpolation
Fiber properties value be mapped on collection of illustrative plates fibre bundle, so as to construct the fibre bundle parameterized model of healthy population, obtain individual
The characteristic curve of subject so that interindividual statistical analysis is possibly realized.
Brief description of the drawings
Fig. 1 is distance signal between fiber；
Fig. 2 is optimal point matching algorithm signal 1；
Fig. 3 is optimal point matching algorithm signal 2；
Fig. 4 is the signal of Tri linear interpolation algorithm；
Fig. 5 is the flow chart of the present invention；
Fig. 6 is the tractus corticospinalis fibre bundle extracted；
Fig. 7 is the result after tractus corticospinalis ligature；
Fig. 8 is the prototype fiber extracted；
Fig. 9 is the discretization of prototype fiber；
Figure 10 is the parameterized results of collection of illustrative plates tractus corticospinalis；
Figure 11 is the fibre bundle parameterized results of subject 1；
Figure 12 is the fibre bundle parameterized results of subject 2；
Figure 13 is the fibre bundle parameterized results of subject 3；
Figure 14 is the anisotropy of subject 1；
Figure 15 is the anisotropy of subject 2；
Figure 16 is the anisotropy of subject 3；
Figure 17 is to be tested 1 to axially diffuse rate curve.
Figure 18 is the Mean diffusivity curve of subject 1；
Figure 19 is the average anisotropy curve of subject 1；
Figure 20 is that being averaged for subject 1 axially diffuses rate curve；
Figure 21 is the average Mean diffusivity curve of subject 1；
Embodiment
Technical scheme is described in more detail with reference to specific drawings and Examples.
A kind of white matter fiber parameterized model construction method based on healthy population, is concretely comprised the following steps：
Step one：The structure of collection of illustrative plates is carried out using magnetic resonance imaging means acquisition brain function data and to the data of acquisition
Build.Full brain fibre bundle tracking is carried out on tensor collection of illustrative plates, and extracts fibre bundle interested.
Step 2：Fiber ligature processing is carried out based on obtained fibre bundle is extracted in upper step, to remove abnormal fibrous.
At fiber intersection points and the low region of noise vs' degree, because the robustness that fibre bundle is followed the trail of is limited, fiber
There can be abnormal fiber in beam.Therefore we need to remove these abnormal fibrous, it is ensured that all fibers all belong in fibre bundle
In same structure.
Fiber is a continuous curve in three dimensions, is represented by a series of three-dimensional coordinate point.For in fibre bundle
Any two fiber FiAnd Fj(i ≠ j), d represents the distance between they, and t represents distance threshold, as shown in Figure 1.As d (Fi,
Fj) ＜ t when, then it is assumed that this two fibers belong to same cluster fiber.
When the number put on fiber is more, we can take 15 points (including two at equal intervals on every fiber
Individual end points), it can thus greatly reduce the calculating time.It is to be noted here that FiTo FjDistance and be not equal to FjTo Fi
Distance, therefore it is asymmetric to calculate obtained distance matrix.In most cases matrix is intended in data handling procedure
It is symmetrical, therefore we are averaged come symmetrization processing array using to the value at symmetric position in matrix herein.
If pkIt is FiOn point, plIt is FjOn point, then fiber FiAnd FjThe distance between use nearest average distance
(closest mean distance) is calculated, and specific formula is as follows：
dM(Fi, Fj)=mean (dm(Fi, Fj), dm(Fj, Fi)) (1)
Wherein,||·||being the Euclidean norm；
Nearest average distance dMDescribe the global similarity of fiber.Shape is more similar, spatially closer to dMValue is more
It is small.
Selection for distance threshold t is obtained using histogrammic method.Seek distance between fiber two-by-two first, it is right
In each fiber, we first average to it with the distance of other all fibres, then judge whether the histogram meets
Normal distribution.If normal probability plot is shown as linear pattern form, meet normal distribution.Asked point by setting significance
Digit, as distance threshold t, so that the fiber whole piece that average distance value is more than quantile be removed, reach and fibre bundle are carried out
The purpose of ligature.
Step 3：Prototype fiber is extracted using local fiber Density Weighted most long method to the fibre bundle after ligature, to build
Found public fibre bundle coordinate system.
Had differences in fibre bundle between different fibers, therefore the corresponding relation put between fiber is unknown, it is impossible to directly carry out fine
The quantitative analysis of dimension.Therefore, we need to normalize fiber by setting up a common coordinate system system, but due to each between fiber
The corresponding relation of point is unclear, and setting up for this common coordinate system system can not be directly using the point on fiber.In order to greatest extent
Ground reduces error, and we, which choose, can most represent one of other all fibres in whole fibre bundle, namely prototype fiber, by right
The discretization based on arc length of prototype fiber sets up common coordinate system system, is then determined according to certain algorithm on other fibers
Point and the corresponding relation put on prototype fiber.
We select the most long fiber of local fiber Density Weighted in fibre bundle as prototype fiber.Fibre density is to pass through
The number of the fiber track of each voxel.Firstly the need of the density for calculating point equidistant on fiber, then along every fibre
Dimension is integrated density, so as to obtain the fibre density of every fiber, that maximum fiber of selective value is used as prototype fiber.
Step 4：Discretization is carried out to prototype fiber according to arc length, public fibre bundle coordinate system is set up.
Due to using the streamline form tracing algorithm based on Euler's method when following the trail of, therefore on fiber between two consecutive points
Air line distance size be equal to step value.Again because the point on fiber is than comparatively dense, the air line distance between two consecutive points is approximate
Equal to its arc length distance along fiber, therefore here, we approximately replace arc length using step-length come to prototype fiber sample variance
Change.
So far, had built up based on the common coordinate system system that prototype fiber arc length is defined.
Step 5：Arc length reference coordinate is passed to by other all fibres by fiber point matching algorithm, realizes that collection of illustrative plates is fine
Tie up the normalization of beam.
Fiber Point matching process is by optimum point matching process (Optimal Point match method, OP) by basis
The arc length reference coordinate that prototype fiber is set up passes to other all fibres, to determine itself and the corresponding pass put on prototype fiber
System, so that arc length coordinate and the corresponding relation put on each collection of illustrative plates fiber are set up, so that collection of illustrative plates fibre bundle be normalized.
OP methods are matched using Hungarian algorithms to each point on prototype fiber, along being approximately perpendicular to prototype
The direction finding match point of fiber, this method carries out rewards and punishments using cost (distance) function.Assuming that i is the point on prototype fiber, j
It is the point on fiber, this distance definition is the measure distance D between point i and point jij, formula is as follows：
Wherein, vijIt is the vector between point i and point j；Estimate tensorvijIt is cutting at i points on prototype fiber
Line vector.The effect of this measure distance be make along perpendicular to the direction of prototype fiber carry out zero cost matching, now away from
From being perpendicular to tangent vector.As angle is by vertically changing, cost increase.Effectively, we are only measured along original
The distance of fiber type tangential direction.In addition, setting a threshold value to limit match point in prototype fiber point spacing (mm)
Within 40%, so along positive and negative tangential direction search box size be spacing 80%, it is ensured that non-optimal match point not by
Matching, it is to avoid Hungarian algorithms are matched all possible point.
By being matched with prototype fiber point, each fiber can all produce a series of arc length coordinate, such as Fig. 2 and Fig. 3
It is shown.
Fig. 2 describes specific Point matching process.First, by the point on prototype fiber with it is closest on other fibers
Point matched, will be in ensuing research for the point (point on the fiber in figure outside matching area) not matched in fiber
In be not considered.Then, the Point matching that fiber is skipped in matching area is to point nearest on prototype fiber.Black in Fig. 3
Track p represent prototype fiber, dotted line represents point and matching for being put on other fibers on prototype fiber.Can from figure
Go out, OP methods have handled the torsional deformation of fiber well.
Step 6：The diffusion properties being tested under individual space are mapped on collection of illustrative plates fibre bundle, the fibre of personal subject is built
Tie up parameterized model.
So far, we have been obtained for the parameterized model of collection of illustrative plates fibre bundle, and next we are by under individual space
The diffusion properties of subject are mapped on collection of illustrative plates fibre bundle, and the parameterized model for carrying out individual subject is built.Detailed process is：With individual
The DTI images of body are reference picture, calculate the inversion carry over of total Deformation Field during map construction non-linear registration, Ran Houfen
Not using tensor collection of illustrative plates as place space, using individual space as object space, using inversion carry over according to each arc length coordinate points at
Corresponding collection of illustrative plates fiber coordinate position calculates coordinate position corresponding under individual space.Under the individual space obtained according to calculating
Coordinate position, the diffusion properties value at the coordinate position is obtained using tri-linear interpolation methods in individual space FA images,
Namely the diffusion properties value at correspondence arc length coordinate points, so that the change for obtaining individual subject diffusion properties on arc length coordinate is bent
Line.
The algorithm of tri-linear interpolation methods used in mapping process is illustrated as shown in figure 4, Interpolation Principle is as follows：
Postulated point V (x, y, z) is known point, and f (x, y, z) is to need to calculate the value at the V points asked for.For V points,
X, y, z is its coordinate value respectively, and x+, y+, z+ are integer coordinate values nearest with V points in positive direction, and x-, y-, z- are negative directions
The upper and nearest integer coordinate values of V points.If
First, along y-axis interpolation, the point that interpolation is obtained is respectively V1, V2, V3 and V4, it is assumed that the functional value minute at these points
Not Wei a1, a2, b1 and b2, then obtained according to linear interpolation formula：
V1(x-, y, z-)：a1=f (x-, y, z-)=yd*f(x-, y+, z-)+(1-yd)*f(x-, y-, z-) (5)
Same has：
V2(x+, y, z-)：a2=f (x+, y, z-)=yd*f(x+, y+, z-)+(1-yd)*f(x+, y-, z-) (6)
V3(x-, y, z+)：b1=f (x-, y, z+)=yd*f(x-, y+, z+)+(1-yd)*f(x-, y-, z+) (7)
V4(x+, y, z+)：b2=f (x+, y, z+)=yd*f(x+, y+, z+)+(1-yd)*f(x+, y-, z+) (8)
Then, along x-axis interpolation, the point that interpolation is obtained is respectively Q1 and Q2, it is assumed that the functional value at these points is respectively
C1 and c2, then it is same to be obtained according to linear interpolation formula：
Q1(x, y, z-)：c1=f (x, y, z-)=xd*f(x+, y, z-)+(1-xd)*f(x-, y, z-)
=xd*a2+(1-xd)*a1 (9)
Q2(x, y, z+)：c2=f (x, y, z+)=xd*f(x+, y, z+)+(1-xd)*f(x-, y, z+)
=xd*b2+(1-xd)*b1 (10)
Finally, along z-axis interpolation, the point that interpolation is obtained is known point V, the functional value f (x, y, z) at V points obtained
As required interpolation result：
In order to program conveniently, a series of above-mentioned formula can be merged into a formula, i.e.,
So far, Tri linear interpolation process terminates.
By carrying out the processing of above-mentioned steps to fibre bundle, we can obtain the characteristic curve of individual subject.Due to this
It is corresponding that a little curves, which use point on public coordinate system, the fiber corresponding to each arc length coordinate points, therefore based on this hair
Bright obtained characteristic curve can carry out the statistical analysis of fiber interfascicular.
The structure of tensor collection of illustrative plates is carried out using the DTI data sets of 37 normal adults in this research, and with corticospinal
Exemplified by beam (Corticospinal tract, CST), with reference to accompanying drawing, 5 couples of present invention do further operating instruction.
Step one：The tensor map construction method proposed according to team carries out the structure of collection of illustrative plates to all normal subjects, so
Afterwards using the deterministic type fibre bundle tracking based on Euler's method the obtained tensor collection of illustrative plates of structure is carried out full brain fibre bundle with
Track, and CST beams are therefrom extracted for follow-up fiber parameters model construction.Wherein, the step-length during tracking is 1mm.
The CST fibre bundles extracted from the full brain fibre bundle tracking result of tensor collection of illustrative plates are as shown in Figure 6.
Step 2：CST beams are carried out fiber ligature to remove abnormal fibrous.Set aobvious according to interfibrous distance distribution histogram
Work property level is 0.05, distance threshold is obtained, so as to remove abnormal fibrous.Result after ligature is as shown in fig. 7, can be with from figure
Find out and improved by ligature fibre bundle global consistency, it is compacter.
Step 3：Prototype fiber is extracted using local fiber Density Weighted most long method to the fibre bundle after ligature, is used for
The foundation of follow-up public fibre bundle coordinate system.Prototype fiber extracts result as shown in figure 8, it can represent most fibers
Track.
Step 4：Arc length is approximately replaced using step-length, to prototype fiber sampling discretization, to set up public fibre bundle and sit
Mark system.The voxel size of constructed tensor collection of illustrative plates is 2mm herein, and ordinary circumstance down-sampled values are approximately that voxel is big
It is small, therefore dis-crete sample values size is 2mm.The step value used when being followed the trail of by the full brain of tensor collection of illustrative plates is fine to prototype for 1mm
The sampling of dimension needs to carry out dot interlace value to the point on fiber, to that is to say and establish a public coordinate points every a point.Fig. 9 is
Display to prototype fiber, color represents arc length coordinate, and the value of color is 100 times of diminutions of arc length value.Here we are by arc length
The origin of coordinates (i.e. arc length=0) has been defined on CST brain precentral motor area.
Step 5：Arc length reference coordinate is passed to by other all fibres by fiber point matching algorithm, realizes that collection of illustrative plates is fine
The normalization of beam is tieed up, so that it is determined that the one-to-one relationship put between the lower fiber of common coordinate system system, the collection of illustrative plates parameterized is fine
Dimension, as shown in Figure 10.OP methods can be very good to handle the torsional deformation of fibre bundle, the characteristic curve of obtained collection of illustrative plates fibre bundle
Uniformity it is higher,
Step 6：The diffusion properties being tested under individual space are mapped on collection of illustrative plates fibre bundle, the fibre of personal subject is built
Tie up parameterized model.
So far, we have been obtained for the parameterized model of collection of illustrative plates fibre bundle, and next we are by under individual space
The diffusion properties of subject are mapped on collection of illustrative plates fibre bundle, and the parameterized model for carrying out individual subject is built.Detailed process is：With individual
The DTI images of body are reference picture, calculate total Deformation Field during map construction non-linear registrationInverse transformation
, then respectively using tensor collection of illustrative plates as place space, using individual space as object space, using inversion carry over according to each arc length
Corresponding collection of illustrative plates fiber coordinate position calculates coordinate position corresponding under individual space at coordinate points.Obtained according to calculating
Coordinate position under body space, obtains the diffusion at the coordinate position in individual space FA images using tri-linear interpolation methods
Diffusion properties value at property value, namely correspondence arc length coordinate points, so as to obtain individual subject diffusion properties on arc length coordinate
Change curve.Individual is returned to fiber properties so that three are tested as an example below to illustrate：Figure 11, Figure 12 and Figure 13 are respectively
The parametrization fiber of three different tested individuals, wherein color represents anisotropy value size.It can be seen that from this three figures
The fibre bundle of tested individual is several obtained from by the way that the diffusion properties being tested under individual space are mapped on collection of illustrative plates fibre bundle
What it is consistent in form with collection of illustrative plates fibre bundle, simply diffusion properties value is different.Figure 14, Figure 15 and Figure 16 are Figure 11, figure respectively
Anisotropy in 12 and Figure 13 corresponding to individual fibers beam.Each of which curve represents a fiber.Figure 17 and figure
18 be to be tested 1 to axially diffuse rate and Mean diffusivity curve respectively.For one is tested, all fibres are in each arc
It is matching at long coordinate points, therefore directly the progress pointwise of fiber properties value can be averaged, so as to obtains average characteristics curve.
Figure 19, Figure 20 and Figure 21 are respectively the average anisotropy of subject, average axially diffuse rate curve and average
Mean diffusivity curve.
The foregoing is intended to be a preferred embodiment of the present invention, protection scope of the present invention not limited to this, any ripe
Those skilled in the art are known in the technical scope of present disclosure, the letter for the technical scheme that can be become apparent to
Altered or equivalence replacement are each fallen within protection scope of the present invention.
Claims (2)
1. a kind of white matter fiber parameterized model construction method based on healthy population, it is characterised in that comprise the following steps：
Step one：The structure of collection of illustrative plates is carried out using magnetic resonance imaging means acquisition brain function data and to the data of acquisition,
Full brain fibre bundle tracking is carried out on tensor collection of illustrative plates, and extracts fibre bundle interested；
Step 2：Fiber ligature processing is carried out based on obtained fibre bundle is extracted in step one, to remove abnormal fibrous；
Fiber is a continuous curve in three dimensions, is represented by a series of three-dimensional coordinate point, for appointing in fibre bundle
Anticipate two fiber FiAnd Fj, i ≠ j, d represents the distance between they, and t represents distance threshold, as d (Fi, Fj) ＜ t when, then it is assumed that
This two fibers belong to same cluster fiber；
When the number put on fiber is more, 15 points, F are taken at equal intervals on every fiberiTo FjDistance and be not equal to
FjTo FiDistance, be averaged using to the value at symmetric position in matrix come symmetrization processing array；
If pkIt is FiOn point, plIt is FjOn point, then fiber FiAnd FjThe distance between calculated using nearest average distance,
Specific formula is as follows：
dM(Fi, Fj)=mean (dm(Fi, Fj), dm(Fj, Fi))
Wherein,| | | | it is Euclid norm；
Nearest average distance dMThe global similarity of fiber is described, shape is more similar, spatially closer to dMValue is just smaller；
Selection for distance threshold t is obtained using histogrammic method, distance is sought two-by-two between fiber first, for every
One fiber, first averages to it with the distance of other all fibres, then judges whether the histogram meets normal distribution,
If normal probability plot is shown as linear pattern form, meet normal distribution, quantile is sought by setting significance, be
Distance threshold t, so that the fiber whole piece that average distance value is more than quantile be removed, reaches the mesh that ligature is carried out to fibre bundle
's；
Step 3：Prototype fiber is extracted using local fiber Density Weighted most long method to the fibre bundle after ligature, to set up public affairs
Common fibre bundle coordinate system；
Fiber is normalized by setting up a common coordinate system system, selection can most represent other all fibres in whole fibre bundle
One, namely prototype fiber is set up common coordinate system by the discretization based on arc length to prototype fiber and united, then basis
Certain algorithm determines the corresponding relation put on point and prototype fiber on other fibers；
Select in fibre bundle that local fiber Density Weighted most long fiber is as prototype fiber, fibre density is by each voxel
Fiber track number, it is necessary first to the density of point equidistant on fiber is calculated, then along every fiber by density
Integrated, so as to obtain the fibre density of every fiber, that maximum fiber of selective value is used as prototype fiber；
Step 4：Discretization is carried out to prototype fiber according to arc length, public fibre bundle coordinate system is set up；
Use the air line distance on the streamline form tracing algorithm based on Euler's method, fiber between two consecutive points big during tracking
Small to be equal to step value, point on fiber is than comparatively dense, and the air line distance between two consecutive points is approximately equal to its arc length along fiber
Distance, arc length is approximately replaced using step-length come to prototype fiber sampling discretization, the public seat defined based on prototype fiber arc length
Mark system has built up；
Step 5：Arc length reference coordinate is passed to by other all fibres by fiber point matching algorithm, collection of illustrative plates fibre bundle is realized
Normalization；
The arc length reference coordinate set up according to prototype fiber is passed to it by fiber Point matching process by optimum point matching process
His all fibres, to determine itself and the corresponding relation put on prototype fiber, so as to set up arc length coordinate and each collection of illustrative plates fiber
The corresponding relation of upper point, so that collection of illustrative plates fibre bundle be normalized；
OP methods are matched using Hungarian algorithms to each point on prototype fiber, along being approximately perpendicular to prototype fiber
Direction finding match point, this method using cost function carry out rewards and punishments, it is assumed that i is the point on prototype fiber, and j is on fiber
Point, this distance definition is the measure distance D between point i and point jij, formula is as follows：
Wherein, vijIt is the vector between point i and point j；Estimate tensortiTangent line on prototype fiber at i points to
Amount, the effect of this measure distance is to make to carry out zero cost matching along perpendicular to the direction of prototype fiber, and distance now is
Perpendicular to tangent vector, as angle is by vertically changing, cost increase is only measured along prototype fiber tangential direction
Distance, sets a threshold value to limit match point within the 40% of prototype fiber point spacing, so along positive and negative tangential direction
Search box size be spacing 80%, it is ensured that non-optimal match point is not matched；
By being matched with prototype fiber point, each fiber can all produce a series of arc length coordinate；
Step 6：The diffusion properties being tested under individual space are mapped on collection of illustrative plates fibre bundle, the fiber ginseng of personal subject is built
Numberization model；
Detailed process is：DTI images using individual is reference pictures, total deformation during calculating map construction non-linear registration
The inversion carry over of field, then respectively using tensor collection of illustrative plates as place space, using individual space as object space, utilizes inversion carry over root
Coordinate position corresponding under individual space is calculated according to corresponding collection of illustrative plates fiber coordinate position at each arc length coordinate points, according to meter
Coordinate position under obtained individual space, the coordinate bit is obtained in individual space FA images using tri-linear interpolation methods
The diffusion properties value at the diffusion properties value at place, namely correspondence arc length coordinate points is put, is closed so as to obtain individual subject diffusion properties
In the change curve of arc length coordinate.
2. the white matter fiber parameterized model construction method as claimed in claim 1 based on healthy population, it is characterised in that
The diffusion properties being tested under individual space are mapped to used in the mapping process on collection of illustrative plates fibre bundle described in step 6
The algorithm of tri-linear interpolation methods is as follows：
Postulated point V (x, y, z) is known point, and f (x, y, z) is to need to calculate the value at the V points asked for；For V points, x, y,
Z is its coordinate value respectively, and x+, y+, z+ are integer coordinate values nearest with V points in positive direction, x-, y-, z- be in negative direction with V
The nearest integer coordinate values of point, if
First, along y-axis interpolation, the point respectively V that interpolation is obtained1、V2、V3And V4, it is assumed that the functional value at these points is respectively a1、
a2、b1And b2, then obtained according to linear interpolation formula：
V1(x-, y, z-)：a1=f (x-, y, z-)=yd*f(x-, y+, z-)+(1-yd)*f(x-, y-, z-)
Same has：
V2(x+, y, z-)：a2=f (x+, y, z-)=yd*f(x+, y+, z-)+(1-yd)*f(x+, y-, z-)
(4-4)V3(x-, y, z+)：b1=f (x-, y, z+)=yd*f(x-, y+, z+)+(1-yd)*f(x-, y-, z+)
(4-5)V4(x+, y, z+)：b2=f (x+, y, z+)=yd*f(x+, y+, z+)+(1-yd)*f(x+, y-, z+)
Then, along x-axis interpolation, the point respectively Q that interpolation is obtained1And Q2, it is assumed that the functional value at these points is respectively c1And c2,
It is then same to be obtained according to linear interpolation formula：
Q1(x, y, z-)：c1=f (x, y, z-)=xd*f(x+, y, z-)+(1-xd)*f(x-, y, z-)
=xd*a2+(1-xd)*a1
Q2(x, y, z+)：c2=f (x, y, z+)=xd*f(x+, y, z+)+(1-xd)*f(x-, y, z+)
=xd*b2+(1-xd)*b1
Finally, along z-axis interpolation, the point that interpolation is obtained is known point V, and the functional value f (x, y, z) at V points obtained is
Required interpolation result：
V (x, y, z)：F (x, y, z)=zd* f (x, y, z+)+(1-zd) * f (x, y, z-)
=zd*c2+(1-zd)*c1
In order to program conveniently, a series of above-mentioned formula are merged into a formula, i.e.,
F (x, y, z)=(1-zd){(1-yd)[(1-xd)f(x-, y-, z-)+xdf(x+, y-, z-)]+
yd[(1-xd)f(x-, y+, z-)+xdf(x+, y+, z-)]}
+zd{(1-yd)[(1-xd)f(x-, y-, z+)+xdf(x+, y-, z+)]+
yd[(1-xd)f(x-, y+, z+)+xdf(x+, y+, z+)]}。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410835831.0A CN104537711B (en) | 2014-12-25 | 2014-12-25 | A kind of white matter fiber parameterized model construction method based on healthy population |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410835831.0A CN104537711B (en) | 2014-12-25 | 2014-12-25 | A kind of white matter fiber parameterized model construction method based on healthy population |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104537711A CN104537711A (en) | 2015-04-22 |
| CN104537711B true CN104537711B (en) | 2017-07-11 |
Family
ID=52853230
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410835831.0A Active CN104537711B (en) | 2014-12-25 | 2014-12-25 | A kind of white matter fiber parameterized model construction method based on healthy population |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104537711B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107316292A (en) * | 2017-06-22 | 2017-11-03 | 电子科技大学 | A kind of method that big brain morphology symmetry is calculated based on structure covariant |
| CN109978872B (en) * | 2019-03-30 | 2022-12-06 | 西安电子科技大学 | White matter microstructure characteristic screening system and method based on white matter fiber tracts |
| CN110599560B (en) * | 2019-08-05 | 2023-07-25 | 上海联影医疗科技股份有限公司 | Magnetic resonance imaging method, device, storage medium and computer equipment |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101763638A (en) * | 2009-12-14 | 2010-06-30 | 西北工业大学 | Method for classifying cerebral white matter fiber tracts in diffusion tensor nuclear magnetic resonance image |
| CN103279633A (en) * | 2013-03-26 | 2013-09-04 | 浙江工业大学 | Brain fiber three-dimensional display method based on diffusion-weighted magnetic resonance data |
-
2014
- 2014-12-25 CN CN201410835831.0A patent/CN104537711B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101763638A (en) * | 2009-12-14 | 2010-06-30 | 西北工业大学 | Method for classifying cerebral white matter fiber tracts in diffusion tensor nuclear magnetic resonance image |
| CN103279633A (en) * | 2013-03-26 | 2013-09-04 | 浙江工业大学 | Brain fiber three-dimensional display method based on diffusion-weighted magnetic resonance data |
Non-Patent Citations (5)
| Title |
|---|
| 《A structural equation modeling investigation of age-related variance in executive function and DTI measured white matter damage》;R.A. Charlton等;《Neurobiology of Aging》;20081231(第29期);全文 * |
| 《In Vivo Demonstration of Microstructural Brain Pathology in Progressive Supranuclear Palsy: A DTI Study Using TBSS》;Susanne Knake等;《Movement Disorders》;20100930;第25卷(第9期);全文 * |
| 《Progressive White Matter Microstructure Damage in Male Chronic Heroin Dependent Individuals: A DTI and TBSS Study》;Yingwei Qiu等;《Progressive WM Microstructure Damage in HDI》;20130531;第8卷(第5期);全文 * |
| 《基于扩散张量成像的人脑三维张量模板构建研究》;熊诗威;《中国优秀硕士学位论文全文数据库》;20141215(第12期);全文 * |
| 《磁共振扩散高阶张量成像的脑白质纤维结构模型及特征提取算法》;李蓉等;《中国生物医学工程学报》;20120630;第31卷(第3期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104537711A (en) | 2015-04-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wismüller et al. | Cluster analysis of biomedical image time-series | |
| Corouge et al. | Towards a shape model of white matter fiber bundles using diffusion tensor MRI | |
| CN103345749B (en) | A kind of brain network function connectivity lateralization detection method based on modality fusion | |
| CN104361318B (en) | A kind of medical diagnosis on disease accessory system based on diffusion tensor technology | |
| US9402562B2 (en) | Systems and methods for improved tractographic processing | |
| JP2013506478A (en) | Medical image processing method and apparatus | |
| CN110232332A (en) | Extraction and brain state classification method and system for dynamic function connection local linear embedded features | |
| CN110838173A (en) | Three-dimensional texture feature-based individual brain covariant network construction method | |
| Jia et al. | Intermediate templates guided groupwise registration of diffusion tensor images | |
| CN104537711B (en) | A kind of white matter fiber parameterized model construction method based on healthy population | |
| WO2024083058A1 (en) | System for accurate positioning of abnormal regions of brain fiber tracts | |
| CN102508184B (en) | Brain function active region detection method based on moving average time series models | |
| US20150363951A1 (en) | Systems and methods for objective tractographic processing using features of grid structures of the brain | |
| CN114065825A (en) | Brain magnetic MEG source positioning method based on structural similarity | |
| CN107330948B (en) | fMRI data two-dimensional visualization method based on popular learning algorithm | |
| CN112950600A (en) | Brain identification extraction method and device, computer equipment and storage medium | |
| CN115169067A (en) | Brain network model construction method and device, electronic equipment and medium | |
| Pai et al. | Evaluation of fiber bundles across subjects through brain mapping and registration of diffusion tensor data | |
| CN114693928A (en) | Blood vessel segmentation method and imaging method of OCTA image | |
| Giorgi et al. | Morphological Analysis of 3D Faces for Weight Gain Assessment. | |
| Styner et al. | Correspondence evaluation in local shape analysis and structural subdivision | |
| Katyal et al. | Gaussian intensity model with neighborhood cues for fluid-tissue categorization of multisequence MR brain images | |
| CN113379249B (en) | Fiber bundle imaging evaluation method based on atlas | |
| Vardhan et al. | Modeling longitudinal MRI changes in populations using a localized, information-theoretic measure of contrast | |
| CN115251889B (en) | Method for describing characteristics of dynamic connection network of functional magnetic resonance image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |