当前位置：文档库 › Symmetric diffeomorphicimage registration

Symmetric diffeomorphicimage registration

Symmetric di?eomorphic image registration with cross-correlation:Evaluating automated labeling

of elderly and neurodegenerative brain

B.B.Avants

a,*

,C.L.Epstein b ,M.Grossman c ,J.C.Gee

Department of Radiology,University of Pennsylvania,3600Market Street,Philadelphia,PA 19104,United States

Department of Mathematics,University of Pennsylvania,Philadelphia,PA 19104,United States c

Department of Neurology,University of Pennsylvania,Philadelphia,PA 19104,United States

Received 11November 2006;received in revised form 23May 2007;accepted 6June 2007

Available online 23June 2007

Abstract

One of the most challenging problems in modern neuroimaging is detailed characterization of neurodegeneration.Quantifying spatial and longitudinal atrophy patterns is an important component of this process.These spatiotemporal signals will aid in discriminating between related diseases,such as frontotemporal dementia (FTD)and Alzheimer’s disease (AD),which manifest themselves in the same at-risk population.Here,we develop a novel symmetric image normalization method (SyN)for maximizing the cross-correlation within the space of di?eomorphic maps and provide the Euler–Lagrange equations necessary for this optimization.We then turn to a careful evaluation of our method.Our evaluation uses gold standard,human cortical segmentation to contrast SyN’s performance with a related elastic method and with the standard ITK implementation of Thirion’s Demons algorithm.The new method compares favorably with both approaches,in particular when the distance between the template brain and the target brain is large.We then report the correlation of volumes gained by algorithmic cortical labelings of FTD and control subjects with those gained by the manual rater.This comparison shows that,of the three methods tested,SyN’s volume measurements are the most strongly correlated with volume measurements gained by expert labeling.This study indicates that SyN,with cross-correlation,is a reliable method for normalizing and making anatomical measurements in volumetric MRI of patients and at-risk elderly individuals.Published by Elsevier B.V.

Keywords:Di?eomorphic;Deformable image registration;Human cortex;Dementia;Morphometry;Cross-correlation

1.Introduction

Frontotemporal dementia (FTD)prevalence may be higher than previously thought and may rival Alzheimer’s disease (AD)in individuals younger than 65years (Ratnav-alli et al.,2002).Because FTD can be challenging to detect clinically,it is important to identify an objective method to support a clinical diagnosis.MRI studies of individual patients are di?cult to interpret because of the wide range of acceptable,age-related atrophy in an older cohort sus-ceptible to dementia.This has prompted MRI studies that

look at both the rate and the anatomic distribution of change (Chan et al.,2001;Fox et al.,2001;Studholme et al.,2004;Kertesz et al.,2004;Avants et al.,2005a;Ball-maier et al.,2004).

Manual,expert delineation of image structures enables in vivo quanti?cation of focal disease e?ects and serves as the basis for important studies of neurodegeneration (Stud-holme et al.,2004).Expert structural measurements from images also provide the gold-standard of anatomical eval-uation.The manual approach remains,however,severely limited by the complexity of labeling 2563or more voxels.Such labor is both time consuming and expensive to sup-port,while the number of individual experts available for such tasks is limited.A third signi?cant di?culty is the

1361-8415/$-see front matter Published by Elsevier B.V.doi:10.1016/j.media.2007.06.004

Corresponding author.

E-mail address:avants@https://www.wendangku.net/doc/e717608575.html, (B.B.Avants).

https://www.wendangku.net/doc/e717608575.html,/locate/media

Available online at https://www.wendangku.net/doc/e717608575.html,

Medical Image Analysis 12(2008)

26–41

problem of inter-rater variability which limits the reliability of manual labeling(Sparks et al.,2002).While rarely avail-able for large-scale data processing,an expert eye remains valuable for limited labeling tasks that give a basis for algo-rithmic evaluation.

Deformable image registration algorithms are capable of functioning e?ectively in time-sensitive clinical applica-tions(Dawant et al.,2003)and high-throughput environ-ments and are used successfully for automated labeling and measurement research tasks.One challenge is reliable performance on non-standard data,as in studies of poten-tially severe neurodegenerative disorders.These types of images violate the basic assumptions of small deformations and/or simple intensity relationships used in many existing image registration methods.

Di?eomorphic image registration algorithms hold the promise of being able to deal successfully with both small (Bajcsy et al.,1983;Gee et al.,1993;Gee and Bajcsy, 1999;Peckar et al.,1998;Rueckert et al.,1999;Rogelj and Kovacic,2006;Ashburner et al.,2000)and large deformation problems(Trouv’e,1998;Christensen et al., 1997;Dupuis et al.,1998;Younes,1998;Joshi and Miller, 2000;Miller et al.,2002;Beg et al.,2005;D’Agostino et al.,2003;Lorenzen et al.,2006;Vaillant et al.,2004). State of the art methods also give full space–time optimi-zations,are symmetric with respect to image inputs and allow probabilistic similarity measures(Avants et al., 2005b).Inverse consistent image registration(ICIR)is an additional common alternative to di?eomorphic map-ping.Inverse consistency was?rst introduced by Thirion as an extension to his Demons algorithm(Thirion,1998) but was popularized by Christensen and Johnson(2001) and others(Shen and Davatzikos,2002).Symmetric meth-ods are distinct from ICIR in that symmetric algorithms,?rst,guarantee that results are identical regardless of the order of input data and,second,use exact inverse trans-formations guaranteed by di?eomorphisms.Inverse con-sistency approximates symmetry by including variational penalties in the normalization optimization algorithm. Depending on the weights of the various data,regulariza-tion and inverse consistency terms,consistency may be satis?ed(or not)at the expense of the other matching cri-terion.Furthermore,inverse consistent algorithms use approximate inverse transformations(Christensen and Johnson,2001).Because the inverse transformations them-selves are approximate,the consistency term,as well,is compromised.

Here,we develop a novel symmetric di?eomorphic opti-mizer for maximizing the cross-correlation in the space of topology preserving maps.The cross-correlation measure has been used in medical image registration before(Bajcsy et al.,1983;Gee et al.,1993;Hermosillo et al.,2002)and more extensively in computer vision.However,this mea-sure has not been investigated for the di?eomorphic case. Furthermore,it has not been used in symmetric normaliza-tion or‘‘inverse consistent’’studies.Applying our novel normalization formulation to cross-correlation provides the advantage(or option)of symmetrizing the cross-corre-lation Euler–Lagrange equations.We show that these sym-metric Euler–Lagrange equations can be computed with only minor additional computational costs.We then give a careful evaluation of the performance of our symmetric di?eomorphic algorithm for high dimensional normaliza-tion of elderly and neurodegenerative cortical anatomy. We compare the method to an elastic cross-correlation optimizer as well as the Demons algorithm which was shown to outperform other methods in a careful evaluation of inter-subject brain registration(Hellier et al.,2003).

2.Registration methods

2.1.Demons

Thirion’s Demons algorithm(Thirion,1996)is known to perform well in inter-subject deformable image registra-tion.The method uses an approximate elastic regularizer to solve an optical?ow problem,where the‘‘moving’’image’s level sets are brought into correspondence with those of a reference or‘‘?xed’’template image.In practice, the algorithm computes an optical?ow term which is added to the total displacement(initially zero).The total displacement is then smoothed with a Gaussian?lter. The process repeats for a set number of iterations for each resolution in a multi-resolution optimization scheme.The method is freely available in the Insight ToolKit and has been optimized by the ITK community(https://www.wendangku.net/doc/e717608575.html,).

Dawant et https://www.wendangku.net/doc/e717608575.html,ed the Demons algorithm for segment-ing the caudate nucleus,the brain and the cerebellum for a morphometric comparison of normal and chronic alcoholic individuals(Dawant et al.,1999).Their evaluation of the algorithm found reasonable agreement between automated and manual labeling.They also showed results on the auto-mated labeling of hippocampus but did not evaluate per-formance.Their comparison used the Dice statistic (overlap ratio):

SeR1;R2T?

2]eR1\R2T

]eR1Tt]eR2T

;e1T

which measures both di?erence in size and location be-tween two segmentations,R1and R2.The](R)operator counts the number of pixels in the region,R.This sensitive measure varies in the range[0,1]where values greater than 0.6for smaller structures and0.8for larger structures are considered good by some authors(Dawant et al.,1999; Sparks et al.,2002).

The range of acceptable values for the Dice statistic are, of course,highly dependent upon the application.Both the amount of certainty that one has in the‘‘gold standard’’dataset and,also,the speci?c use of the segmentations determine a reasonable operating range.Our goal,in this paper,is to use manually segmented structures as a founda-tion for comparing automated normalization methods.In this respect,it is relative performance(measured with respect to the Dice statistic)that is of critical value.

B.B.Avants et al./Medical Image Analysis12(2008)26–4127

2.2.Symmetric di?eomorphisms

A di?eomorphism is a di?erentiable map with a di?eren-tiable inverse(Arnold and Khesin,1992;Arnold,1991). We typically restrict our solutions to the di?eomorphic space Di?0with homogeneous boundary conditions.That is,we assume that rigid and scaling transformations have been factored out and the image border maps to itself.1 Shortest paths between elements in this space are termed geodesic.Di?eomorphic methods were introduced into medical computer vision(Trouv’e,1998)for the purpose of providing a group theoretical,large deformation space–time image registration framework.Current devel-opments in large deformation computational anatomy by Miller,Trouve and Younes extended the methods to include photometric variation and to use Euler–Lagrange equations(Miller et al.,2002).However,the standard ver-sion of these methods,Beg’s Large Deformation Di?eo-morphic Metric Matching(LDDMM)(Beg et al.,2005), do not formulate the transformation optimization symmet-rically.They are only symmetric in theory and their imple-mentation requires parallel computation.Through personal communication,however,we understand that a symmetrization operator,based on the transformation Jacobian,is being included in current developments by Trouve and Younes(2005)and Younes(preprint).How-ever,these methods do not guarantee symmetry for similar-ity metrics other than the intensity di?erence.

Our current work extends the Lagrangian di?eomorphic registration technique described in Avants et al.(2006a). This new formulation has symmetry properties required for a geodesic connecting two images,I and J,in the space of di?eomorphic transformations and guarantees symme-try regardless of the chosen similarity measure.This formu-lation accounts for the natural symmetry in the problem: both images move along the shape(di?eomorphism)man-ifold.Symmetric di?eomorphisms guarantee two proper-ties that are intrinsic to the notion of a geodesic path:the path from I to J is the same as it is when computed from J to I,regardless of similarity metric or optimization parameters.Symmetry is required for distance estimates and makes results independent of arbitrary decisions about which image is‘‘?xed’’or‘‘moving.’’

Our method is also unique in that it guarantees sub-pixel accurate,invertible transformations in the discrete domain by directly including invertibility constraints in the optimi-zation.While di?eomorphisms are theoretically guaranteed to be invertible,interpolation errors can lead to invertibil-ity errors that increase linearly with the number of interpo-lation steps.Our solution,on the other hand,directly minimizes this error by exploiting the invertibility guaran-teed by di?eomorphisms.Finally,the method is e?cient enough to use on single-processor machines and in process-ing large datasets.

We de?ne a di?eomorphism/of domain X,generally, for transforming image I into a new coordinate system by/I=I /(x,t=1)=I(/(x,t=1)),which indicates that I is warped forward by the map de?ned by/(x,1).One may also use a more standard backward warping strategy, through/à1(x,1),to achieve the same deformation.The parameters of these transformations are time,t,a spatial coordinate,x,and a velocity?eld,v(x,t)on X,which is a square-integrable,continuous vector?eld(Arnold,1991). The correspondence maps,/,are gained by integrating the velocity?elds in time,/ex;1T?/ex;0TtR1

ve/ex;tT;tTd t,where v is indexed at/(x,t)=y.The dis-tance is then De/ex;0T;/ex;1TT?

k vex;tTk

d t,wher

e L de?nes the linear operator regularizing the velocity.The functional norm,i?i L,induces regularity on the velocity ?eld via a linear di?erential operator such as L=a$2+b Id (a and b are constants).Such a di?eomorphism gives a dense map in both space and time and is illustrated in Fig.1.

A basic fact of di?eomorphisms allows them to be decom-posed into two parts,/1and/2.We exploit this fact to de?ne a variational energy that explicitly divides the image registra-tion di?eomorphisms into two halves such that I and J con-tribute equally to the path and deformation is divided between them.Assume that x indicates the identity position of some anatomy in image I and z indexes the identity posi-tion of the same anatomy in image J.We assume,also,that the di?eomorphism maps homologous anatomy in these images.The prior knowledge that this di?eomorphic map should apply evenly to both images can be captured by including the constraint D(Id,/1(x,0.5))=D(Id,/2(z,0.5)) directly in the formulation of the problem.The result is a method that?nds correspondences with equal consideration of both images.Note that below we will derive the equations assuming intensity di?erence as a similarity measure,for sim-plicity.However,in actuality,we have a variety of statistical image similarity measures(robust intensity di?erence,cross-correlation,mutual information)at our disposal,as in Her-mosillo et al.(2002),or employ user landmarks as in Avants et al.(2006a).After this introductory section,we will develop our method for the cross-correlation.

De?ne the image registration optimization time, t2[0,1]where t indexes both/1and/2,though in oppo-site directions.The similarity seeks/1such that /1(x,1)I=J.Recalling the basic de?nition of di?eomor-phisms allows us to write any geodesic through composing two parts,each of which is a geodesic(any sub-part of a geodesic is a geodesic).We apply this fact in the second step of our derivation below:

/1ex;1TI?J;

/à1

e/1ex;tT;1àtTI?J;

e/à1

e/1ex;tT;1àtT;1àtTI?/2ez;1àtTJ;

ex;tTI?/2ez;1àtTJ;

e2T

which converts the similarity term from j/1(x,1)IàJ j to j/1(x,t)Ià/2(z,1àt)J j2.A visualization of these

1Note that extending the background space of an image allows almost

any di?eomorphism of the image to be captured in Di?0.

28 B.B.Avants et al./Medical Image Analysis12(2008)26–41

components of /is in Fig.1.The forward and backward optimization problem is then,solving to time t =0.5

E sym eI ;J T?inf /1inf /2

Z 0:5t ?0

k v 1ex ;t Tk 2L tk v 2ex ;t Tk 2

L n o

d t

j I e/1e0:5TTàJ e/2e0:5TTj 2d X :

Subject to each /i 2Di?0the solution of:d /i ex ;t T=d t ?v i e/i ex ;t T;t Twith /i ex ;0T

?Id and /à1i e/i T?Id ;/i e/à1

i T?Id :

e3T

Minimization with respect to /1and /2provides the sym-metric normalization (SyN)solution and also solves a 2-mean https://www.wendangku.net/doc/e717608575.html,ndmarks may also be included,as in

the Lagrangian Push Forward method (Avants et al.,2006a ),by dividing the similarity term,as done with the image match terms above.The constraint D (/1(x ,0.5))=D (/2(x ,0.5))is built into the fact that we integrate the solution from 0to 0.5.Because the velocities are of (approximately)constant arc length and are,at each itera-tion,of exactly the same length,the length of /1,integrated over [0,0.5],is equivalent to the length of /2integrated over [0,0.5].

As noted in Section 1,this method is quite distinct from inverse consistent image registration (ICIR)(Johnson and Christensen,2002).ICIR uses vector ?elds,h and g ,to de?ne correspondence from I to J and J to I ,respectively.Therefore,in total,ICIR uses four vector ?elds in its approach to normalization,h ,g ,h à1and g à1.The inverses of these ?elds are not guaranteed to exist (as the optimiza-tion is not performed in the di?eomorphic space)and no exact method is used to compute the inverses.Instead,judg-ing from the brief discussion in Johnson and Christensen

(2002),the inverse is only guaranteed to be exact at a few points in the domain and the inversion algorithm itself is not well-speci?ed or exact.This inexactness means that the variational term measuring the di?erence between two vector ?elds,i h (x )àg à1(x )i 2,only coarsely estimates con-sistency.Furthermore,this means that ICIR must include two terms to compute consistency,that is,both i h (x )àg à1(x )i 2and i h à1(x )àg (x )i 2.If inverses were exact (or very close to exact)only one of the above terms would be enough as minimizing one would imply minimizing the other.

SyN,alternatively,provides an inverse that is guaran-teed to be everywhere sub-pixel accurate.Furthermore,SyN avoids a basic computational redundancy present in ICIR –the solutions h (x )and g (x )overlap in time.SyN’s pair of solutions,/1and /2,on the other hand,do not overlap in time.They are only two parts of a longer path.This distinction is shown in Fig.2.However,SyN is still able to compute the measure of inverse consistency by sim-ply composing our time 0.5maps together and comparing

/1(x ,1)and /à1

2ex ;1Tin the I domain (a similar computa-tion may be done in the J domain).Recall that,by de?ni-tion and di?eomorphic computations,we are able to

guarantee /à11e/1T?Id and /à1

2e/2T?Id .We therefore have an inverse consistency error of zero,up to the inaccu-racy induced by the single interpolation required in com-puting the time 1maps, e.g./1ex ;1T?/à1

2e/1ex ;0:5T;0:5T.A key to the symmetry of our method is the ability to compute sub-pixel accurate inverses such that,for each i =1,2,we have (to a user-selected numerical precision)/à1i e/i T?Id and /i e/à1i T?Id .Typical precision is 1.0·10à6L 1norm and 0.2L 1norm,measured in terms of voxel/pixel coordinates.Further details on the numerical methods employed in optimizing this energy may be found in Avants et al.(2006a)and will be summarized

below.

Fig.1.An illustration of a SyN geodesic path between images.The images in the top row are the original images,I and J ,at the initialization of the method.After the SyN solution converges (second row),these images deform in time along the series of di?eomorphisms that connect them.The deforming grids associated with these di?eomorphisms are shown in the bottom row.The maps,/1and /2are of equivalent length and map I and J to the mean shape between the images.The full path,/and /à1,are found by joining the paths /1and /2.The symmetric nature of this problem (proven in Avants et al.,2006b )is due to the interchange-ability of the labels I and J in our problem formulation.

B.B.Avants et al./Medical Image Analysis 12(2008)26–4129

2.3.The cross-correlation with symmetric di?eomorphisms

We now propose to symmetrically solve the following image matching problem:Find a spatiotemporal mapping, /2Di?0,such that the cross-correlation between the image pair is maximized.This formulation of the problem repre-sents a change of philosophy when compared to Bajcsy et al.(1983)and,later,Gee et al.’s(1993)elastic matching, which also used cross-correlation.Elastic image registra-tion methods seek to balance a regularization term and a similarity term allowing one to?nd a constrained deform-able solution.The approach recommended here departs from this strategy by allowing an unconstrained optimiza-tion of the similarity term within the space of di?eomor-phisms.This strategy is used when?nding optimal mappings in lower-dimensional(e.g.a?ne)transformation groups.The main advantage of an unconstrained search within the space of di?eomorphisms is its simplicity:one allows the method to maximize the similarity until a local maximum or limit on computation time is reached.2The disadvantage is one requires a similarity metric that,when optimized in Di?,provides a useful solution.Therefore, design and choice of the similarity metric requires great care.

The mutual information(MI)similarity metric garnered signi?cant interest in recent years(Maes et al.,1997;Wells et al.,1997;Rueckert et al.,1999;Studholme et al.,2006). Intuitively speaking,MI estimates the optimal matching between images by inferring how much global information is shared in the image pair,as estimated from the pair’s joint histogram.At the same time,MI prevents over-?tting by penalizing clustering of the marginal image probabilities (Hermosillo et al.,2002).The globality of this approach makes it extremely useful for robust rigid registration but may limit performance in deformable registration,in par-ticular in cases where non-stationary noise patterns or intensity inhomogeneity requires a locally adaptive similar-ity.This problem has been addressed by Studholme et al. (2006)and in our previous work(Yoo,2004).One di?culty with a locally varying estimate to the MI is that joint prob-abilities need a large number of samples for reliable statis-tics.Therefore,as locality in the MI estimate increases,its statistical reliability decreases.

Cross-correlation(CC),on the other hand,adapts natu-rally to situations where locally varying intensities occur and is suitable for some multi-modality problems.The CC depends only on estimates of the local image average and variance which may be accurately/exactly measured with relatively few samples.Furthermore,the cross-corre-lation has shown historically to perform well in many real-world computer vision applications where one requires robustness to unpredictable illumination,re?ectance,etc. An example of the robustness of our method to strong MRI inhomogeneity(bias?eld)is shown in Fig.3.For these reasons,we revisit the classical cross-correlation as a similarity metric for use in our emerging di?eomorphic image registration.

We now provide a symmetric formulation of the di?eo-morphic image registration problem as driven by CC, along with the Euler–Lagrange equations for this problem. First,de?ne I1=I(/1(x,0.5))and J2=J(/2(x,0.5)).We also de?ne a variable to represent each image with its local mean subtracted as IexT?I1exTàl I

exTand JexT?J2à

exT.We compute l over a local n D window centered at each position x,where D is the image dimension.We usu-ally choose n=5.The cross-correlation is then

CC I;J;x

àá

h I;J i2

h I ih J i

?A2=BC;e4T

where the inner product is also taken over a n D window. Note that our CC is implicitly a function of/1and/2 through its dependence on I1and J2.We now are able to de?ne the variational optimization problem in similar fash-ion to Eq.(3),

E CC I;J

àá

?inf

inf

Z0:5

t?0

k v1ex;tTk2

tk v2ex;tTk2

n o

d t

CC I;J;x

àá

d X:

Subject to each/i2Di?0the solution of:

d/iex;tT=d t?v ie/iex;tT;tTwith/iex;0T

?Id and/à1

e/iT?Id;/ie/à1

T?Id:e5TThe problem,here,is the same as before but with the cross-correlation as the driving similarity term.We now take the variation of this function with respect to/1at time0.5and /

2at time0.5.Following Beg’s derivation,adapted for our

Fig.2.The symmetric normalization method is represented,at top,by its

two components,/1and/2,meeting at the middle of the normalization

domain.Note that each sub-path may be traversed either from the middle

to the end or from an end to the middle.Alternatively,the ICIR method is

shown in a schematic at the bottom panel of the?gure.The correspon-

dence de?ning vector?elds associated with ICIR are called h and g.In

ICIR,all four deformation?elds overlap in time and may,in fact,be

di?erent from each other.The inverse of h may not be its true inverse.

Further,the inverse of h may not be equivalent to g.

2The contribution of the regularization term to the total energy is small

compared to the similarity.However,the regularization term in the

di?eomorphic matching problem remains important.It guarantees that

one?nds a path of minimal length.

30 B.B.Avants et al./Medical Image Analysis12(2008)26–41

symmetric normalization and particular similarity metric,we ?nd two Euler–Lagrange equations:

r /1ex ;0:5TE CC ex T;?2L v 1ex ;0:5Tt

BC ?J ex TàA

I ex T

j D /1jr I ex T;

e6T

r /2ex ;0:5TE CC ex T;?2L v 2ex ;0:5Tt

BC ?I ex TàA

J ex T

j D /2jr J ex T:

e7T

The gradients given in Eqs.(6)and (7)include that fact that the velocities exist in the space of smooth vector ?elds given by the linear operator L as well as the determinant of each /i transformation Jacobian,j D /i j .This equation is similar to both that derived for LDDMM (Beg et al.,2005)and the derivative for the cross-correlation given by Hermosillo et al.(2002).This equation di?ers from Beg’s Euler–La-grange equation in that we have an E–L equation for both /1and /2instead of just /1.In addition,Beg used the intensity di?erence metric given in Eq.(3).The derivation of this equation di?ers from that in Hermosillo et al.(2002)by the presence of the derivative of the di?eomor-phism o I =o /i ,which leads to the Jacobian term,instead of a vector ?eld (small deformation)derivative.Further-more,we have represented the CC in a di?erent arrange-ment of terms that suits our own derivation of the CC variation and computational implementation.This arrangement of terms allows one to precompute,for each iteration,the locally varying values of A ,B ,C and store them for use in the local computation of the derivative.

Our novel symmetric formulation,therefore,does not add signi?cant additional cost to the normalization.The main additional cost is that of smoothing the derivative estimate to gain the v i ,not in the actual estimate of the sim-ilarity term.Both the /1and /2derivatives require the terms A ,B and C .Therefore,in estimating Eqs.(6)and (7),we use the procedure detailed in Algorithm 1.Algorithm 1(Computing cross-correlation derivatives ).

(1)Deform I by /1(0.5)and J by /2(0.5).

(2)Calculate I and J from the result of step (1).

(3)Calculate and store images representing A ,B and C .These steps enable us to loop over the image domain to rapidly compute Eqs.(6)and (7)at the same time from these precomputed variables.Note,however,that it could be possible to modify Lewis’s fast normalized correlation measure (Lewis,1995)to speed up this process even further.

Eqs.(6)and (7)gives the update to the velocities at time 0.5and consequently the di?eomorphisms /1and /2.How-ever,we also require the /i to satisfy our o.d.e.and exact invertibility constraints.We satisfy these constraints by,given a new velocity estimate,?rst updating the /i through

the o.d.e.and then integrating backward in time to ?nd /à1

i and verify invertibility.This approach is a standard one and detailed,algorithmically,in the LPF method (Avants et al.,2006a ).For completeness,we include an abbreviated explanation here.

First,assume arbitrary /and v ,related by the stan-dard o.d.e.d /(x ,t )/d t =v (/(x ,t ),t )used to

generate

Fig.3.The local cross-correlation measure allows robust matching of images despite the presence of a strong bias ?eld a?ecting the image quality.The smoothness of the grid is also una?ected by the bias.

B.B.Avants et al./Medical Image Analysis 12(2008)26–4131

di?eomorphisms.The update method for our di?eomor-phism comes from discretization of this o.d.e.,such that:/ex ;t tD t T /ex ;t TtD t v e/ex ;t T;t T:

e8T

This discretization is used to update both /i from time 0to 0.5.Second,Algorithm 2gives a method for generating in-verses when an arbitrary di?eomorphism /is updated by a small time-step through a velocity ?eld,as in the previous equation.The algorithm typically converges within one to a few iterations,particularly if the time step is small.The existence of a solution is guaranteed by the integrability condition established for di?eomorphic image registration (Dupuis et al.,1998),while uniqueness comes from the uniqueness theorem of o.d.e.s (Arnold,1991).

Algorithm 2(Inversion method ).

The algorithm uses a ?xed point method to push the inverse of /forward by a small amount performing a gradient descent on /à1(/)=Id ,enforcing /à1(/)=Id to a sub-pixel level.The same approach was used in the Lagrang-ian Push Forward algorithm (Avants et al.,2006a ).We use a temporary variable,w ,to represent the input diffeomorphism that,on output,will be the inverse of /.Input /ex T?y ;w à1e~y T?x where ~y ?y and output w à1(y )=x =/à1(y ).At convergence,~y ?y and i w à1(/)àId i 1< 2r where r is the image resolution and 2a small constant,typically 0.1.

1:while i w à1(/(x ))àx i 1> 2r do 2:Compute v à1(x )=w à1(/(x ))àx .3:Find scalar c such that i v à1i 1=0.5r .

4:Integrate w à1s.t.w à1e~

y ;t Tt?c v à1ew à1e~y ;t TT.5:end while

We now summarize the SyN method in Algorithm 3.Our implementation essentially updates each /i with the current estimate to the velocity,follows the update to each /i by calling Algorithm 2to generate inverse maps,and then re-estimates the velocity from the new estimate to the /i .This formulation and di?eomorphic representation guarantee SyN’s sub-pixel invertibility and algorithmic independence to input permutations.These methods allow us to symmetrically match images to the degree that dis-crete di?eomorphisms are invertible.Furthermore,while extended here to cross-correlation,similar techniques may be used to e?ciently symmetrize almost any other sim-ilarity measure.We now leverage ITK for a fair implemen-tation and evaluation of SyN.Algorithm 3(Overview

symmetric

normalization

method ).

(1)Initialize /1?Id ?/à11and /2?Id ?/à1

2.(2)Repeat the following steps until

convergence:

Fig.4.The atlas was initially aligned to these FTD images via a rigid plus uniform scaling transformation.The subsequent Demons registration to each image,used for labeling,is in the right column.The SyN result is in the center column,while the corresponding original images are in the left column.The Demons method does a reasonable normalization,but leaves the ventricles and other smaller structures only partly normalized.The quadratic elastic penalty prevents the remaining shape di?erences from being captured.A similar loss of resolution in the mapping is seen in the elastic cross-correlation mappings.These are illustrative images from our previous study,Avants et al.(2006c),and were not used as actual study data in this exposition.

32 B.B.Avants et al./Medical Image Analysis 12(2008)26–41

(3)Compute the cross-correlation as described in Algo-rithm 1.

(4)Compute each v i by smoothing the result of step (3)in

this table.One may also use the modi?ed midpoint method for each velocity,as in the LPF algorithm (Avants et al.,2006a ),to give smoothness in time.(5)Update each /i by v i through the o.d.e.as described in

Eq.(8).This step automatically adjusts the time step-size such that the maximum length of the updates to the /i is sub-pixel and approximately constant over iter-ations.We explicitly guarantee k v 1eá;t Tk ?k v 2eá;t Tk .We also update the estimate to the geodesic distance by trapezoidal rule,as in the LPF method.

(6)Use Algorithm 2to get the inverses of the /i .

(7)Generate the time 1solutions from /1e1T?/à1

e/1ex ;0:5T;0:5Tand /à11e1T?/2e1T?/à1

1e/2ex ;0:5T;0:5T.2.4.Implementation in ITK

The Demons algorithm is freely available in the stan-dard ITK distribution and has been quantitatively evalu-ated by the ITK community.We have implemented SyN within our extended version of the ITK deformable image registration framework,described in Yoo (2003).There-fore,we are in a position to measure performance

gains

Fig.5.Normalization results.All images in each row should look similar to the ?rst image in the row.The atlas was initially aligned to these elderly (top three)and FTD (bottom three)images via a rigid plus uniform scaling transformation.These examples are from our current study.We have highlighted (with a circle)the type of small scale di?erence one may see in the registration https://www.wendangku.net/doc/e717608575.html,rger scale di?erences are also clear.In particular,the Demons and elastic cross-correlation have problems both shrinking and expanding the ventricles (the second elderly image).In addition,the Demons intensity consistency assumption causes signi?cant errors in the ?rst FTD image,a case where this assumption does not hold.

B.B.Avants et al./Medical Image Analysis 12(2008)26–4133

by varying ?rst the similarity metric and then the the transformation model ,keeping an identical underlying code base.

2.4.1.Switching the similarity metric

For this part of the study,we replace the itkDemons-RegistrationFunction,in the itkDemonsRegistrationFilter,with our own itkCrossCorrelationRegistrationFunction.The interface and operation of this modi?ed deformable registration algorithm are identical to the Demons algo-rithm,except that the image forces come from the small deformation version of Eq.(6),as may be found in Hermo-sillo et al.(2002).In addition to switching the itkDemons-

RegistrationFunction for our cross-correlation image forces,we also modi?ed the time step used during this elas-tic registration process.The time-step was increased from 1,for Demons,to 4for cross-correlation.The Demons force,based on optical ?ow,is extremely aggressive com-pared to our analytical cross-correlation derivative.We found,for our dataset,that time-steps beyond 4led to non-positive Jacobians in a signi?cant fraction of our normalizations.

2.4.2.Switching the transformation model

As mentioned previously,we have implemented SyN in our extended Insight ToolKit.For this study,SyN will

use

Fig.6.Normalization results di?erence images.This ?gure shows the same results as Fig.4,but with absolute value of the image di?erence,after normalization.One can observe that small sulci may not be captured by any of the methods.Under the intensity di?erence metric,the Demons method should be expected to perform best,as it explicitly minimizes this error.

34 B.B.Avants et al./Medical Image Analysis 12(2008)26–41

the itkPDEDeformableRegistrationFilter as implemented in ITK for its Gaussian smoothing operator.3This smooth-ing operator will be applied only to the incremental output of our itkCrossCorrelationRegistrationFunction.The Demons and elastic cross-correlation algorithms both

apply Gaussian smoothing to the total displacement ?eld,in accordance with small deformation transformation mod-els.The presence of a large deformation assumption,along with the use of a space–time parameterization of the trans-formation,are the main (not the only)di?erences between di?eomorphic and elastic registration algorithms.

The key contrast between the ?rst pair of methods (Demons,elastic cross-correlation)is the similarity metric.The only di?erence between the second two (elastic cross-correlation,SyN cross-correlation)is in the

transformation

Fig.7.Normalization results label error images.Here,we show three individual images along with the manual,SyN,elastic and Demons labels.This bottom row for each individual shows the di?erence of the algorithmic and manual labels.The majority of the errors are due to a shift or a curvature in the boundary de?nition gained algorithmically,with respect to the manual de?nition.Manually de?ned lobar boundaries tend to be planar.Further,when one compares the (in particular,parietal)boundaries determined by the labeler on these three images,it becomes apparent that there may be some inconsistency.Secondary errors are due to lack of sulcal de?nition in the template labels,with respect to the individual labeled ground truth.

3The Gaussian kernel,G r ,is an estimate to the Green’s kernel for the linear operator L =$2+Id (Bro-Nielsen and Gramkow,1996)and is adequate for providing the velocity ?eld regularity necessary for integrat-ing o.d.e.s.

B.B.Avants et al./Medical Image Analysis 12(2008)26–4135

model.We are therefore investigating,?rst,if cross-correla-tion provides better normalization than Demons optical ?ow,for our neurodegenerative and elderly T1MRI data-set.The second comparison evaluates if changing from the elastic to the symmetric di?eomorphic transformation model yields additional improvement.Our hypothesis was that each change would yield bene?ts.This hypothesis was born out in our experimental data,as suggested by the preliminary study data shown in Fig.4.

3.Data and experiments

3.1.Dataset

Our database consists of20T1MRI images (0.85·0.85·1.5mm,GE Horizon Echospeed1.5T scan-ner)from10normal elderly and10frontotemporal demen-tia patients.The10frontotemporal dementia individuals are a di?erent set than used in our previous study(Avants et al.,2006c).Each of the20images,along with the elderly template,was manually labeled with the protocol described in Sparks et al.(2002).This protocol was shown to be highly reproducible for both small and large structures via six-month intra-rater reliability and inter-rater reliabil-ity measurements.Left hippocampus labeling,for example, showed a0.92intra-rater overlap ratio(Eq.(1))and0.83 average for inter-rater overlap.As the hippocampus is rel-atively small,these values are reasonable.Finally,as the labeling was intended to be only on the cortex(not cerebro-spinal?uid),we masked the label images by the inverse of the cerebrospinal?uid segmentations.This also makes our evaluation more sensitive to di?erences in the accuracy of sulcal and gyral alignment.

3.2.Evaluation

We now use these two comparative pairings of three algorithms to study performance for characterizing the volumetric di?erences between elderly and frontotemporal dementia cortex,hippocampus,amygdala and cerebellum. An example comparison of the SyN and Demons meth-ods’normalization abilities is in Fig.4.This study reveals the ability of these methods,SyN,elastic cross-correla-tion and Demons,to reproduce results gained from an expert user’s labeling of our20image dataset.This test is performed by using each method to map the template labels(an elderly individual also labeled by our expert user)to each individual.Twenty rigid registrations and 60non-rigid registrations were required(20subjects were each deformably registered by three algorithms).The atlas labelings are then warped by the composed rigid and non-rigid transformation into the space of the patient image.We then compute Dice overlap ratios between the manual and automatic structural segmenta-tions for each structure.

Each method was applied in an identical four-level multi-resolution scheme and ran until convergence or a ?xed(maximum)number of iterations was reached.We allowed up to100iterations at the?rst level,100iterations at the second level,100iterations at the third level and20 iterations at the full resolution.The relative running times for each algorithm varied depending upon the particular cases being run.However,the relative average running times(in terms of the average Demons running time)were 1(Demons),4.2(elastic cross-correlation)and5.5(SyN). The mean runtime of the Demons method on a2.0GHz Intel Mobile Pentium processor was20.4min.The relative similarity of the elastic time cost to the cost of SyN also indicates that our symmetric approach does not create sig-ni?cant additional computational cost.

4.Results and discussion

A visualization of six individuals from our dataset and the normalization of the template to these individuals is shown in Fig.5.Associated intensity error images are shown in https://www.wendangku.net/doc/e717608575.html,bel error images are shown in Fig.7. The eight labeled individual structures are shown,along with the summary statistics for our results,in Fig.8.Both cross-correlation algorithms produced segmentation results above the minimum threshold of0.6for all structures,as shown in Fig.8.We also compared the minimum Jacobian of the elastic cross-correlation and SyN cross-correlation methods and did not?nd signi?cant di?erences (T=à1.67725,P<0.101703).This indicates that SyN’s results are not signi?cantly less constrained at a local level. The smallest di?erence in performance between the Demons and elastic cross-correlation methods was on the cerebellum.The highest di?erence was on the frontal lobe. The largest performance di?erence in SyN and elastic cross-correlation was on the temporal and frontal lobe. Performance gains expectedly focus on frontal and tempo-ral lobe due to the known e?ect of FTD on these two areas. The shape changes caused by this complex disease are likely di?cult to capture with a constrained method such as elastic normalization.The smallest di?erences between the elastic and di?eomorphic methods were found in the hippocampus and amygdala.This is likely to indicate that the similarity metric does not provide a rich enough feature space over which to optimize the correspondence of these structures.Therefore,a better optimizer operating with that similarity metric will not be likely to provide an advantage.

A likely improvement in performance would come from using a method such as the STAPLE algorithm(War?eld et al.,2004)to bootstrap results.Similarly,an optimal tem-plate would augment results(Avants and Gee,2004).How-ever,both of these enhancements would increase performance in a consistent manner across all our tested registration methods and would be unlikely to change the relative performance of our algorithms.

The second test we performed was to evaluate whether the volume measurements obtained by the normalization are strongly correlated with the measurements of the expert

36 B.B.Avants et al./Medical Image Analysis12(2008)26–41

user.We performed this analysis only on regions (frontal,temporal,parietal lobes)where one may reasonably expect FTD to induce a di?erence from the normal elderly popu-lation and where labeling performance was good.The pur-pose of this test is to determine if conclusions made by analyzing the output of our automated normalization methods are at all consistent with the expert user’s analysis.As we are using the expert labeling as our gold standard,the ‘‘better’’method should produce volume measures that correlate more strongly with the volume measures gained by the expert.The volume,in cubic centimeters,of each structure was calculated by summing the voxel volumes that were given the appropriate label for that structure.As shown in Table 1,SyN outperformed elastic cross-cor-relation (and the Demons method)according to the degree to which the automated volumes correlate with the manual volumes for each of three structures.SyN also had the least overall discrepancy in the measured volume,that is,P

i j V Alg àV Man j ,where V Man is manually measured vol-ume and V Alg refers to algorithmically measured volume.These results are shown in Table 2.

We also plot the estimated volume versus the manually computed volume for the temporal lobe in Fig.9,for fron-tal lobe in Fig.10and for parietal lobe in Fig.11.We use linear regression to ?t the estimated volume with a line in order to assess the closeness of the method’s estimate to

Table 1

Pearson correlations between manual and algorithmic volume measures Structure Corr(Man,Syn)Corr(Man,Elas)Corr(Man,Demons)Temporal 0.860.690.79Frontal 0.890.670.71Parietal

0.71

0.42

0.66

Table 2

The average absolute volume error between manual volume and algorithmic volume measured over the dataset for each structure Structure VolErr(Man,Syn)VolErr(Man,Elas)VolErr(Man,Demons)Temporal 8.49.28.7Frontal 11.116.115.8Parietal

7.9

9.3

7.9

Fig.8.Performance comparison and average Dice statistic for each method and each structure.Example brain labelings mark the structures over which we evaluated the three algorithms.The ?nal results showed that,overall,SyN >Elastic >Demons for automatically labeling these lobar structures.The P -values for the di?erence in performance were computed using non-parametric permutation testing based on the paired T -test on vectors of Dice statistics.We used 10,000permutations per example,thus limiting our minimum P -value to 0.0001.Extremely small P -values indicate that the method uniformly and strongly outperformed the method with which it was paired.

B.B.Avants et al./Medical Image Analysis 12(2008)26–4137

the manual estimate.The Pearson correlation of SyN vol-umes with the manually measured volumes was 0.86for temporal lobe,0.89for frontal lobe and 0.71for parietal lobe.The correlation of the elastic method volumes with the manually measured volumes was 0.69for temporal lobe,0.67for frontal lobe and 0.42for parietal lobe.The correlation of Demons volumes with the manually mea-sured volumes was 0.79for temporal lobe,0.67for frontal lobe and 0.66for parietal lobe.One interesting ?nding,here,is that the Demons method,compared to the elastic method,produces stronger correlations with manual label-ings (in terms of volume measurements)but produces smal-ler overlap ratios.

Our ?nal analysis tests the signi?cance of the di?erence in volumes between the FTD and elderly members of our image population.The P-values of these results,obtained

by non-parametric permutation testing,are shown in Table 3.None of the structures showed a signi?cant volumetric di?erence between groups.However,SyN appears to re?ect the signi?cance obtained by the manual rater more closely than the other methods.At the same time,we do not ?nd these results to be strong enough to warrant de?nitive https://www.wendangku.net/doc/e717608575.html,paring a larger dataset (or possibly a di?er-ent set of control and subject images)would likely show stronger di?erences than this dataset,as we found in our previous,preliminary study (Avants et al.,2006c ).4.1.Contrasting automated and manual results

Thus,we can see how an apparently small,but consistent di?erence in performance (as measured by overlap ratio)can have an impact on the validity of the study

outcome.

Fig.9.Temporal lobe volume,in cubic centimeters,as measured by each of the three methods.The algorithmic measures are plotted against the manually measured

volume.

Fig.10.Frontal lobe volume,in cubic centimeters,as measured by each of the three methods.The algorithmic measures are plotted against the manually measured volume.

38 B.B.Avants et al./Medical Image Analysis 12(2008)26–41

That is,the performance gains in overlap ratio translate to more accuracy in making clinically meaningful measure-ments,such as volume.Although the quality of our results are reasonable,by some standards,these algorithms,as applied here,cannot claim to accurately reproduce manual labelings.The reasons for this are well-known.The pri-mary reason is that expert knowledge is not directly encoded in these methods.Secondarily,the uncertainty inherent to the problem of neuroanatomical labeling limits the accuracy and reproducibility of both manual and auto-mated segmentation.Finally,it is not yet known the extent to which brains of di?erent individuals,when represented as magnetic resonance images,are di?eomorphic to each other.This problem is even less well understood with elderly and patient brains.Note also that,when measuring the trends in the di?erence of elderly and FTD structure volumes,the Demons,elastic correlation and SyN signi?-cance tend to estimate larger volumes than the manual results.This is likely caused by two things:segmentation bias towards the template and the fact that registration-based segmentations are smoothed (elastic and Demons more than SyN),while the manual segmentations are not.Both of these types of errors may be observed in Fig.7.Fig.7also shows that the labeler may have used di?erent

‘‘styles’’of labeling or changed the decision-making pro-cess over the dataset.All of these confounds impact the overlap ratios and correlations that we ?nd here.We accen-tuate that our goal,here,is not speci?cally to reproduce the expert labels,but to use these labeled structures as a neuro-anatomically based evaluation of relative algorithmic performance.5.Conclusion

We ?rst described the symmetric normalization formu-lation.We then extended this formulation to use the cross-correlation similarity function providing,in addition,Euler–Lagrange equations for the variational problem in the symmetric di?eomorphic context.We contrasted our method with the popular inverse consistent image registra-tion technique,which was outperformed by the Demons method in an unbiased comparative evaluation of brain segmentation and alignment (Hellier et al.,2003).We then provided algorithmic details of the SyN methodology.Finally,we leveraged our ITK implementation to compare the SyN cross-correlation optimizer with the Demons and an elastic cross-correlation optimizer.This enabled us to explore the e?ect of similarity metric and optimizer sepa-rately and purely,as we use an identical linear operator for regularization and an identical code base,the Insight ToolKit.The relative similarity of the elastic computation time to the computational cost of SyN also indicates that our symmetric approach does not require signi?cant addi-tional computational expense.We found,in short,that the cross-correlation behaves well on elderly and neurodegen-erative data.In particular,it outperforms the Demons opti-cal ?ow on our dataset.We also found that the symmetric di?eomorphic optimizer outperforms the elastic optimizer on the same dataset,while guaranteeing that the

topology

Fig.11.Parietal lobe volume,in cubic centimeters,as measured by each of the three methods.The algorithmic measures are plotted against the manually measured volume.

Table 3

Signi?cance of elderly-FTD volume di?erences as measured by manual,SyN,elastic and Demons methods Structure Eld-FTD Man.Eld-FTD SyN Eld-FTD Elas.Eld-FTD Demons Temporal 0.250.350.470.26Frontal 0.180.130.490.52Parietal

0.32

0.46

0.82

0.15

Signi?cance was determined by a permutation test based on the unpaired T -test of elderly and FTD structure volumes.

B.B.Avants et al./Medical Image Analysis 12(2008)26–4139

of the images is preserved and,importantly,that the algo-rithm’s performance does not depend upon the order in which one inputs the images.

This careful comparison establishes the distinct advan-tage of SyN for segmenting elderly and neurodegenerative cerebrum,cerebellum and,in particular,the temporal and frontal lobe.Note that,in addition to better performance, SyN provides a dense space–time map and transformation inverses.The di?erences in performance are consistent,sta-tistically signi?cant and have a major impact on study out-come.One can extrapolate even larger di?erences between SyN and algorithms with lower dimensionality than either Demons or SyN.For this reason,along with the theoretical advantages that translate into practical bene?ts,we pro-mote di?eomorphic algorithms and the cross-correlation similarity metric in neuroimaging research,in particular when studying non-standard datasets,such as FTD and AD.

Acknowledgements

We thank the reviewers for greatly improving the con-tents of this paper.Much of this work was supported by NIH grant R01-EB006266.

References

Arnold,V.I.,1991.Ordinary Di?erential Equations.Springer,Berlin. Arnold,V.I.,Khesin,B.A.,1992.Topological methods in hydrodynamics.

Ann.Rev.Fluid Mech.24,145–166.

Ashburner,J.,Good,C.,Friston,K.,2000.Tensor based morphometry.

Neuroimage11,465.

Avants,B.,Gee,J.C.,2004.Geodesic estimation for large deformation anatomical shape and intensity averaging.Neuroimage(Suppl.1), S139–S150.

Avants,B.,Grossman,M.,Gee,J.C.,2005a.The correlation of cognitive decline with frontotemporal dementia induced annualized gray matter loss using di?eomorphic morphometry.Alz.Dis.Assoc.Dis.19 (Supplement1),S25–S28.

Avants,B.,Epstein,C.L.,Gee,J.C.,2005.Geodesic image interpolation: parameterizing and interpolating spatiotemporal images.In:ICCV Workshop on Variational and Level Set Methods,pp.247–258. Avants, B.,Schoenemann,P.T.,Gee,J.C.,https://www.wendangku.net/doc/e717608575.html,ndmark and intensity-driven Lagrangian frame di?eomorphic image registration: application to structurally and functionally based inter-species com-parison.Med.Image Anal.10,397–412.

Avants,B.,Epstein,C.L.,Gee,J.C.,2006b.Geodesic image normalization in the space of di?https://www.wendangku.net/doc/e717608575.html,put.Anat.,125–133.

Avants,B.,Grossman,M.,Gee,J.C.,2006c.Symmetric di?eomorphic image registration:evaluating automated labeling of elderly and neurodegenerative cortex and frontal lobe.WBIR,50–57.

Bajcsy,R.,Lieberson,R.,Reivich,M.,1983.A computerized system for the elastic matching of deformed radiographic images to idealized atlas https://www.wendangku.net/doc/e717608575.html,put.Assist.Tomogr.5,618–625.

Ballmaier,M.,O’Brien,J.T.,Burton,E.J.,Thompson,P.M.,Rex,D.E., Narr,K.L.,McKeith,I.G.,DeLuca,H.,Toga,A.W.,https://www.wendangku.net/doc/e717608575.html,par-ing gray matter loss pro?les between dementia with Lewy bodies and Alzheimer’s disease using cortical pattern matching:diagnosis and gender e?ects.Neuroimage23,325–335.

Beg,F.,Miller,M.,Trouv’e,A.,Younes,L.,https://www.wendangku.net/doc/e717608575.html,puting large deformation metric mappings via geodesic?ows of di?eomorphisms.

https://www.wendangku.net/doc/e717608575.html,put.Vision61,139–157.Bro-Nielsen,M.,Gramkow,C.,1996.Fast?uid registration of medical images.Proceedings of Visualization in Biomedical Computing.

Springer,Hamburg,pp.267–276.

Chan,D.,Fox,N.C.,Jenkins,R.,Scahill,R.I.,Crum,W.R.,Rossor, M.N.,2001.Rates of global and regional cerebral atrophy in ad and frontotemporal dementia.Neurology57(10),1756–1763. Christensen,G.,Johnson,H.,2001.Consistent image registration.IEEE Trans.Med.Imaging20(7),568–582.

Christensen,G.E.,Joshi,S.C.,Miller,M.I.,1997.Volumetric trans-formation of brain anatomy.IEEE Trans.Med.Imaging16(6), 864–877.

D’Agostino,E.,Maes,F.,Vandermeulen,D.,Suetens,P.,2003.A viscous ?uid model for multimodal non-rigid image registration using mutual information.Med.Image Anal.7(4),565–575.

Dawant,B.,Hartmann,S.,Thirion,J.P.,Maes,F.,Vandermeulen,D., Demaerel,P.,1999.Automatic3-D segmentation of internal structures of the head in MR images using a combination of similarity and free-form transformations,part II:methodology and validation on severely atrophied brains.IEEE Trans.Med.Imaging18,926–971. Dawant,B.,Li,R.,Cetinkaya,E.,Kao,C.,Fitzpatrick,J.,Konrad, P.,https://www.wendangku.net/doc/e717608575.html,puterized atlas-guided positioning of deep brain simulators:a feasibility study.In:Gee,J.,Maintz,J.B.(Eds.), Workshop on Biomedical Image Registration.Philadelphia,July, pp.142–150.

Dupuis,P.,Grenander,U.,Miller,M.I.,1998.Variational problems on ?ows of di?eomorphisms for image matching.Q.Appl.Math.56(3), 587–600.

Fox,N.,Crum,W.,Scahill,R.,Stevens,J.,Janssen,J.,Rossor,M.,2001.

Imaging of onset and progression of Alzheimer’s disease with voxel-compression mapping of serial magnetic resonance https://www.wendangku.net/doc/e717608575.html,ncet358, 201–205.

Gee,J.C.,Bajcsy,R.K.,1999.Elastic matching:continuum mechanical and probabilistic analysis.In:Toga, A.W.(Ed.),Brain Warping.

Academic Press,San Diego,pp.183–197(Chapter11).

Gee,J.C.,Reivich,M.,Bajcsy,R.,1993.Elastically deforming a3D atlas to match anatomical brain https://www.wendangku.net/doc/e717608575.html,put.Assist.Tomogr.17, 225–236.

Hellier,P.,Barillot,C.,Corouge,I.,Gibaud,B.,Le Goualher,G.,Collins,

D.L.,Evans, A.,Malandain,G.,Ayache,N.,Christensen,G.

E.,

Johnson,H.J.,2003.Retrospective evaluation of inter-subject brain registration.IEEE Trans.Med.Imaging22(9),1120–1130. Hermosillo,G.,Chefd’Hotel, C.,Faugeras,O.,2002.A variational approach to multi-modal image https://www.wendangku.net/doc/e717608575.html,put.Vision50

(3),329–343.

Johnson,H.J.,Christensen,G.E.,2002.Consistent landmark and inten-sity-based image registration.IEEE Trans.Med.Imaging21(5),450–461.

Joshi,S.C.,Miller,M.I.,https://www.wendangku.net/doc/e717608575.html,ndmark matching via large deformation di?eomorphisms.IEEE Trans.Image Process.9(8),1357–1370. Kertesz,A.,Martinez-Lage,P.,Davidson,W.,Munoz,D.G.,2004.The corticobasal degeneration syndrome overlaps progressive aphasia and frontotemporal dementia.Neurology55,1368–1375.

Lewis,J.P.,1995.Fast normalized cross-correlation.Vision Interface.

Canadian Image Processing and Pattern Recognition Society,pp.120–123.

Lorenzen,P.,Prastawa,M.,Davis,B.,Gerig,G.,Bullitt,E.,Joshi,S., 2006.Multi-modal image set registration and atlas formation.Med.

Image Anal.19(3),440–451.

Maes,F.,Collignon,A.,Meulen,D.,Marchal,G.,Suetens,P.,1997.

Multi-modality image registration by maximization of mutual infor-mation.IEEE Trans.Med.Imaging16,187–198.

Miller,M.,Trouv’e,A.,Younes,L.,2002.On the metrics and Euler–Lagrange equations of computational anatomy.Annu.Rev.Biomed.

Eng.4,375–405.

Peckar,W.,Schnorr,C.,Rohr,K.,Stiehl,H.S.,Spetzger,U.,1998.Linear and incremental estimation of elastic deformations in medical regis-tration using prescribed displacements.Mach.Graph.Vision7(4), 807–829.

40 B.B.Avants et al./Medical Image Analysis12(2008)26–41

Ratnavalli, E.,Brayne, C.,Dawson,K.,Hodges,J.R.,2002.The prevalence of frontotemporal dementia.Neurology58(11),1585–1586. Rogelj,P.,Kovacic,S.,2006.Symmetric image registration.Med.Image Anal.10(3),484–493.

Rueckert, D.,Sonoda,L.I.,Hayes, C.,Hill, D.L.G.,Leach,M.O., Hawkes,D.J.,1999.Nonrigid registration using free-form deforma-tions:application to breast MR images.IEEE Trans.Med.Imaging18

(8),712–721.

Shen,D.,Davatzikos,C.,2002.Hammer:hierarchical attribute matching mechanism for elastic registration.IEEE Trans.Med.Imaging21(11), 1421–1439.

Sparks,B.,Friedman,S.,Shaw,D.,Aylward,E.,Echelard,D.,Artru,A., Maravilla,K.,Giedd,J.,Munson,J.,Dawson,G.,Dager,S.,2002.

Brain structural abnormalities in young children with autism spectrum disorder.Neurology59,184–192.

Studholme,C.,Cardenas,V.,Blumenfeld,R.,Schu?,N.,Rosen,H.J., Miller,B.,Weiner,M.,2004.Deformation tensor morphometry of semantic dementia with quantitative validation.Neuroimage21(4), 1387–1398.

Studholme,C.,Drapaca,C.,Iordanova,B.,Cardenas,V.,2006.Defor-mation-based mapping of volume change from serial brain MRI in the presence of local tissue contrast change.IEEE Trans.Med.Imaging25

(5),626–639.

Thirion,J.P.,1996.Non-rigid matching using demons,In:IEEE Com-puter Vision and Pattern Recognition.pp.245–251.Thirion,J.,1998.Image matching as a di?usion process:an analogy with Maxwell’s demons.Med.Image Anal.2(3),243–260.

Trouv’e,A.,1998.Di?eomorphism groups and pattern matching in image https://www.wendangku.net/doc/e717608575.html,put.Vision28(3),213–221.

Trouve,A.,Younes,L.,2005.Metamorphoses through lie group action.

https://www.wendangku.net/doc/e717608575.html,put.Math.5(2),173–198.

Vaillant,M.,Miller,M.I.,Younes,L.,Trouv’e,A.,2004.Statistics on di?eomorphisms via tangent space representations.Neuroimage23, S161–S169.

War?eld,S.K.,Zou,K.H.,Wells III,W.M.,2004.Simultaneous truth and performance level estimation(STAPLE):an algorithm for the valida-tion of image segmentation.Med.Image Anal.23(7),903–921. Wells,W.M.,Viola,P.,Atsumi,H.,Nakajima,S.,Kikinis,R.,1997.

Multi-modal volume registration by maximization of mutual infor-mation.Med.Image Anal.1,35–51.

Yoo,T.,2003.Insight Into Images:Principles and Practice for Segmen-tation,Registration and Image Analysis.A.K.Peters Ltd.,Natick, MA.

Yoo,T.(Ed.),2004.Insight Into Images:Theory for Segmentation, Registration and Image Analysis. A.K.Peters Ltd.,Natick,MA, Chapter Avants et al.,Non-rigid registration.

Younes,L.,https://www.wendangku.net/doc/e717608575.html,putable elastic distance between shapes.SIAM J.

Appl.Math.58,565–586.

Younes,L.,preprint.Jacobi?elds in groups of di?eomorphisms and applications,Q.Appl.Math.

B.B.Avants et al./Medical Image Analysis12(2008)26–4141

标题：常用协议对应的端口号由Anonymous 于星期日, 04/01/2007 - 01:28 发表 DHCP：服务器端的端口号是67 DHCP：客户机端的端口号是68 POP3：POP3仅仅是接收协议，POP3客户端使用SMTP向服务器发送邮件。POP3所用的端口号是110。 SMTP：端口号是25。SMTP真正关心的不是邮件如何被传送，而只关心邮件是否能顺利到达目的地。SMTP具有健壮的邮件处理特性，这种特性允许邮件依据一定标准自动路由，SMTP具有当邮件地址不存在时立即通知用户的能力，并且具有在一定时间内将不可传输的邮件返回发送方的特点。 Telnet：端口号是23。Telnet是一种最老的Internet应用，起源于ARPNET。它的名字是“电信网络协议（Telecommunication Network Protocol）”的缩写。 FTP：FTP使用的端口有20和21。20端口用于数据传输，21端口用于控制信令的传输，控制信息和数据能够同时传输，这是FTP的特殊这处。FTP采用的是TCP连接。 TFTP：端口号69，使用的是UDP的连接。端口号的作用及常见端口号用途说明 IP协议是由TCP、UDP、ARP、ICMP等一系列子协议组成的。其中，主要用来做传输数据使用的是TCP和UDP协议。在TCP和UDP协议中，都有端口号的概念存在。端口号的作用，主要是区分服务类别和在同一时间进行多个会话。举例来说，有主机A需要对外提供FTP和WWW两种服务，如果没有端口号存在的话，这两种服务是无法区分的。实际上，当网络上某主机B需要访问A的FTP服务时，就要指定目的端口号为21；当需要访问A的WWW服务时，则需要将目的端口号设为80，这时A根据B访问的端口号，就可以区分B的两种不同请求。这就是端口号区分服务类别的作用。再举个例子：主机A需要同时下载网络上某FTP服务器B上的两个文件，那么A需要与B同时建立两个会话，而这两个传输会话就是靠源端口号来区分的。在这种情况下如果没有源端口号的概念，那么A就无法区分B传回的数据究竟是属于哪个会话，属于哪个文件。而实际上的通信过程是，A使用本机的1025号端口请求B的21号端口上的文件1，同时又使用1026号端口请求文件2。对于返回的数据，发现是传回给1025号端口的，就认为是属于文件1；传回给1026号端口的，则认为是属于文件2。这就是端口号区分多个会话的作用。如果说IP地址让网络上的两个节点之间可以建立点对点的连接，那么端口号则为端到端的连接提供了可能。理解端口号的概念，对于理解TCP/IP协议的通信过程有着至关重要的作用。端口号的范围是从1～65535。其中1～1024是被RFC 3232规定好了的，被称作“众所周知的端口”(Well Known Ports)；从1025～65535的端口被称为动态端口（Dynamic Ports），

最详细协议端口号端口号码/ 名称层 1 5 7 9 11 13 17 18 19 20 21 tcpmux rje echo discard systat daytime qotd msp chargen ftp-data ftp 注释TCP 端口服务多路复用远程作业入口Echo 服务用于连接测试的空服务用于列举连接了的端口的系统状态给请求主机发送日期和时间给连接了的主机发送每日格言消息发送协议字符生成服务；发送无止境的字符流FTP 数据端口文件传输协议端口；有时被文件服务协议使用安全Shell服务Telnet 服务简单邮件传输协议时间协议资源定位协议互联网名称服务WHOIS 目录服务用于基于TCP/IP 验证和访问的终端访问控制器访问控制系统远程邮件检查协议22 23 25 37 39 42 43 49 ssh telnet smtp time rlp nameserver

nicname tacacs 50 re-mail-ck 端口号码/ 名称层53 63 67 domain whois++ bootps 注释域名服务WHOIS++，被扩展了的WHOIS 服务引导协议服务；还被动态主机配置协议服务使用Bootstrap客户；还被动态主机配置协议客户使用小文件传输协议Gopher 互联网文档搜寻和检索远程作业服务远程作业服务远程作业服务远程作业服务用于用户联系信息的Finger 服务用于万维网服务的超文本传输协议Kerberos 网络验证系统Telnet 协议扩展SRI-NIC 机器上的主机名服务ISO 开发环境网络应用邮箱名称服务器；也被CSO 名称服务器使用远程Telnet 邮局协议版本 2 68 bootpc 69 70 71 72 73 73 79 80 tftp gopher netrjs-1 netrjs-2 netrjs-3 netrjs-4 finger http 88 95 101 102 105 kerberos supdup hostname iso-tsap csnet-ns 107 109 rtelnet pop2 端口号码/ 名称层110

常用协议端口号 1813端口使用UDP传输 3306端口使用TCP传输 Tracert 默认使用UDP 数据包来探测路由路径, 端口为33434 TCP协议支持协议名称TCP端口号协议名称解释 ACAP 674 AIM 5190 BEEP 10288 CAST 4224 CMP 829 COPS 3288 PKTCABLE_COPS 2126 PKTCABLE_MM_COPS 3918 DAAP 3689 DHCPFO 519 DIAMETER 3868 DISTCC 3632 DLSW 2065 NP 20000 NS 53

DSI 548 FTPDATA 20 FTP 21 GIFT 1213 CS 1720 HTTP 80 PROXY_HTTP 3128 PROXY_ADMIN_HTTP 3132 HKP 11371 DAAP 3689 SSDP 1900 IB 3050 ICAP 1344 IMAP 143 IRC 6667 ISAKMP 500 JABBER 5222 KERBEROS 88 LAPLINK 1547 LDAP 389 GLOBALCAT_LDAP 3268

PRINTER 515 MBTCP 502 MSNMS 1863 MSRP 0 MySQL 3306 NBSS 139 CIFS 445 NCP 524 NDMP 10000 PA 0x0d44 BROKER 0x0bc6 SRS 0x0bca ENS 0x0bc8 RMS 0x0bcb NOTIFY_LISTENER 0x0bc9 NETSYNC 5253 NNTP 119 NTP 123 POP 110 PPTP 1723 PVFS2 3334

Ethereal支持的常用协议端口号 TCP协议支持协议名称TCP端口号协议名称解释 ACAP 674 AIM 5190 BEEP 10288 CAST 4224 CMP 829 COPS 3288 PKTCABLE_COPS 2126 PKTCABLE_MM_COPS 3918 DAAP 3689 DHCP FO 519 DIAMETER 3868 DISTCC 3632 DLSW 2065 NP 20000 NS 53 DNS5353 DSI 548 FTP DATA 20 FTP21 GIFT 1213 CS 1720 HTTP 80

PROXY_HTTP 3128 PROXY_ADMIN_HTTP 3132 HKP 11371 DAAP 3689 SSDP 1900 IB 3050 ICAP 1344 IMAP 143 IRC 6667 ISAKMP 500 JABBER 5222 KERBEROS 88 LAPLINK 1547 LDAP 389 GLOBALCAT_LDAP 3268 LDP 646 PRINTER 515 MB TCP502 MSNMS 1863 MSRP 0 MySQL 3306 NBSS 139 CIFS 445 NCP 524 NDMP 10000 PA 0x0d44

BROKER 0x0bc6 SRS 0x0bca ENS 0x0bc8 RMS 0x0bcb NOTIFY_LISTENER 0x0bc9 NETSYNC 5253 NNTP 119 NTP 123 POP 110 PPTP 1723 PVFS2 3334 RMI 1099 RSH 514 RSYNC 873 RTSP 554 SIP 5060 SKINNY 2000 SLSK_1 2234 SLSK_2 5534 SLSK_3 2240 SMRSE 4321 SMTP25 SNMP161 SNMP_TRAP 162 SMUX 199 SOCKS 1080

经常用到的网络协议端口号: 用来规定信息格式；数据及控制信息的格式、编码及信号电平等。用来说明通信双方应当怎么做；用于协调与差错处理的控制信息。）详细说明事件的先后顺序；速度匹配和排序等网际层协议：包括：IP 协议、ICMP 协议、ARP 协议、RARP 协议。传输层协议：TCP 协议、UDP 协议。应用层协议：FTP、Telnet、SMTP、HTTP、RIP、NFS、DNS。TCP （1）FTP：定义了文件传输协议，使用21 端口。常说某某计算机开了FTP 服务便是启动了文件传输服务。下载文件，上传主页，都要用到FTP 服务。（2）Telnet：它是一种用于远程登陆的端口，用户可以以自己的身份远程连接到计算机上，通过这种端口可以提供一种基于DOS 模式下的通信服务。如以前的BBS 是纯字符界面的，支持BBS 的服务器将23 端口打开，对外提供服务。（3）SMTP：定义了简单邮件传送协议，现在很多邮件服务器都用的是这个协议，用于发送邮件。如常见的免费邮件服务中用的就是这个邮件服务端口，所以在电子邮件设置中常看到有这么SMTP 端口设置这个栏，服务器开放的是25 号端口。（4）POP3：它是和SMTP 对应，POP3 用于接收邮件。通常情况下，POP3 协议所用的是110 端口。也是说，只要你有相应的使用POP3 协议的程序（例如Foxmail 或Outlook），就可以不以Web 方式登陆进邮箱界面，直接用邮件程序就可以收到邮件（如是163 邮箱就没有必要先进入网易网站，再进入自己的邮箱来收信）。UDP （1）HTTP：这是大家

用得最多的协议，它就是常说的"超文本传输协议"。上网浏览网页时，就得在提供网页资源的计算机上打开80 号端口以提供服务。常说"WWW服务"、"Web 服务器"用的就是这个端口。（2）DNS：用于域名解析服务，这种服务在Windows NT 系统中用得最多的。因特网上的每一台计算机都有一个网络地址与之对应，这个地址是常说的IP 地址，它以纯数字+"."的形式表示。然而这却不便记忆，于是出现了域名，访问计算机的时候只需要知道域名，域名和IP 地址之间的变换由DNS 服务器来完成。DNS 用的是53 号端口。（3）SNMP：简单网络管理协议，使用161 号端口，是用来管理网络设备的。由于网络设备很多，无连接的服务就体现出其优势。（1）. HTTP 协议代理服务器常用端口号：80/8080/3128/8081/9080 （2）. SOCKS 代理协议服务器常用端口号：1080 （3）. FTP 协议代理服务器常用端口号：

各协议的端口号篇一：常用协议对应的端口号标题：常用协议对应的端口号由Anonymous于星期日,04/01/2007-01:28发表 DHCP：服务器端的端口号是67 DHCP：客户机端的端口号是68 POP3：POP3仅仅是接收协议，POP3客户端使用SMTP向服务器发送邮件。POP3所用的端口号是110。 SMTP：端口号是25。SMTP真正关心的不是邮件如何被传送，而只关心邮件是否能顺利到达目的地。SMTP具有健壮的邮件处理特性，这种特性允许邮件依据一定标准自动路由，SMTP具有当邮件地址不存在时立即通知用户的能力，并且具有在一定时间内将不可传输的邮件返回发送方的特点。 Telnet：端口号是23。Telnet是一种最老的Internet应用，起

源于ARPNET。它的名字是“电信网络协议（TelecommunicationNetworkProtocol）”的缩写。 FTP：FTP使用的端口有20和21。20端口用于数据传输，21端口用于控制信令的传输，控制信息和数据能够同时传输，这是FTP的特殊这处。FTP采用的是TCP连接。 TFTP：端口号69，使用的是UDP的连接。端口号的作用及常见端口号用途说明 IP协议是由TCP、UDP、ARP、ICMP等一系列子协议组成的。其中，主要用来做传输数据使用的是TCP和UDP协议。在TCP和UDP协议中，都有端口号的概念存在。端口号的作用，主要是区分服务类别和在同一时间进行多个会话。举例来说，有主机A需要对外提供FTP和WWW两种服务，如果没有端口号存在的话，这两种服务是无法区分的。实际上，当网络上某主机B需要访问A的FTP服务时，就要指定目的端口号为21；当需要访问A的WWW服务时，则需要将目的端口号设为80，这时A根据B

我们常用的协议以及对应端口号

以下内容第一段为端口号，第二段为端口对应的服务名称，第三段为注释信息。 1 tcpmux TCP端口服务多路复用。 18 msp 消息发送协议。 20 ftp-data FTP数据端口。 21 ftp 文件传输协议（FTP）端口，有时候被文件服务协议(FSP)使用。 22 ssh 安全Shell(SSH)服务。 23 telnet Telnet 服务。 25 smtp 简单邮件传输协议（SMTP）。 37 time 时间协议。 42 nameserver互联网名称服务。 53 domain 域名服务（BIND）。 67 bootps 引导协议（BOOTS）服务；还被动态主机配置协议（DHCP）使用。 69 tftp 小文件传输协议（TFTP）。 80 http 用于万维网（WWW）服务的超文本传输协议（HTTP）。 107 rtelnet 远程Telnet。 109 pop2 邮局协议版本2。 110 pop3 邮局协议版本3. 115 sftp 安全文件传输协议（SFTP）服务。 119 nntp 用于USENET讨论系统的网络新闻传输协议（NNTP）。 137 在红帽企业Linux中被Samba使用NETBIOS名称服务。 138在红帽企业Linux中被Samba使用NETBIOS数据报服务。 139在红帽企业Linux中被Samba使用NETBIOS会话服务。 143 imap 互联网消息存取协议（IMAP）。 209 qmtp 快速邮件传输协议（QMTP）。 220 imap3 互联网消息存取协议版本3. 389 idap 轻型目录存取协议（LDAP）。 443 https 安全超文本传输协议。 445 microsoft-ds 通过TCP/IP的服务器消息块（SMB）。 487 saft 简单不对称文件传输SAFT协议。 488 gss-http 用于HTTP的通用安全服务(GSS)。 546 dhcpv6-client动态主机配置协议（DHCP）版本6 客户 547 dhcpv6-client 动态主机配置协议（DHCP）版本6服务。 994 ircs 通过安全套接字层的互联网中继聊天（IRCS）。 995 pop3s 通过安全套接字层的邮局协议版本3(POPS3).

常见的网络协议\端口号一个网络协议至少包括三要素: 语法用来规定信息格式；数据及控制信息的格式、编码及信号电平等。语义用来说明通信双方应当怎么做；用于协调与差错处理的控制信息。时序（定时）详细说明事件的先后顺序；速度匹配和排序等网际层协议：包括：IP协议、ICMP协议、ARP协议、RARP协议。传输层协议：TCP协议、UDP协议。应用层协议：FTP、Telnet、SMTP、HTTP、RIP、NFS、DNS。使用TCP协议的常见端口主要有以下几种：（1） FTP：定义了文件传输协议，使用21端口。常说某某计算机开了FTP服务便是启动了文件传输服务。下载文件，上传主页，都要用到FTP服务。（2） Telnet：它是一种用于远程登陆的端口，用户可以以自己的身份远程连接到计算机上，通过这种端口可以提供一种基于DOS模式下的通信服务。如以前的B BS是纯字符界面的，支持BBS的服务器将23端口打开，对外提供服务。（3） SMTP：定义了简单邮件传送协议，现在很多邮件服务器都用的是这个协议，用于发送邮件。如常见的免费邮件服务中用的就是这个邮件服务端口，所以在电子邮件设置中常看到有这么SMTP端口设置这个栏，服务器开放的是25号端口。（4） POP3：它是和SMTP对应，POP3用于接收邮件。通常情况下，POP3协议所用的是110端口。也是说，只要你有相应的使用POP3协议的程序（例如Fox mail或Outlook），就可以不以Web方式登陆进邮箱界面，直接用邮件程序就可以收到邮件（如是163邮箱就没有必要先进入网易网站，再进入自己的邮箱来收信）。使用UDP协议端口常见的有：（1） HTTP：这是大家用得最多的协议，它就是常说的"超文本传输协议"。上网浏览网页时，就得在提供网页资源的计算机上打开80号端口以提供服务。常说"W WW服务"、"Web服务器"用的就是这个端口。（2） DNS：用于域名解析服务，这种服务在Windows NT系统中用得最多的。因特网上的每一台计算机都有一个网络地址与之对应，这个地址是常说的IP地址，它以纯数字+"."的形式表示。然而这却不便记忆，于是出现了域名，访问计算机的时候只需要知道域名，域名和IP地址之间的变换由DNS服务器来完成。DNS用的是53号端口。（3）

协议端口详解范文计算机"端口"是英文port的译义，可以认为是计算机与外界通讯交流的出口。其中硬件领域的端口又称接口，如USB端口、串行端口等。软件领域的端口一般指网络中面向连接服务和无连接服务的通信协议端口，是一种抽象的软件结构，包括一些数据结构和I/O（基本输入输出）缓冲区。在网络技术中，端口（Port）有好几种意思。集线器、交换机、路由器的端口指的是连接其他网络设备的接口，如RJ-45端口、Serial端口等。我们这里所指的端口不是指物理意义上的端口，而是特指TCP/IP 协议中的端口，是逻辑意义上的端口。那么TCP/IP协议中的端口指的是什么呢？如果把IP地址比作一间房子，端口就是出入这间房子的门。真正的房子只有几个门，但是一个IP地址的端口可以有65536（即256×256）个之多！端口是通过端口号来标记的，端口号只有整数，范围是从0到65535（256×256-1）。在Inter上，各主机间通过TCP/IP协议发送和接收数据包，各个数据包根据其目的主机的ip地址来进行互联网络中的路由选择。可见，把数据包顺利的传送到目的主机是没有问题的。问题出在哪里呢?我们知道大多数操作系统都支持多程序（进程）同时运行，那么目的主机应该把接收到的数据包传送给众多同时运行

的进程中的哪一个呢？显然这个问题有待解决，端口机制便由此被引入进来。本地操作系统会给那些有需求的进程分配协议端口（protocol port，即我们常说的端口），每个协议端口由一个正整数标识，如80，139，445，等等。当目的主机接收到数据包后，将根据报文首部的目的端口号，把数据发送到相应端口，而与此端口相对应的那个进程将会领取数据并等待下一组数据的到来。说到这里，端口的概念似乎仍然抽象，那么继续跟我来，别走开。端口其实就是队，操作系统为各个进程分配了不同的队，数据包按照目的端口被推入相应的队中，等待被进程取用，在极特殊的情况下，这个队也是有可能溢出的，不过操作系统允许各进程指定和调整自己的队的大小。不光接受数据包的进程需要开启它自己的端口，发送数据包的进程也需要开启端口，这样，数据包中将会标识有源端口，以便接受方能顺利的回传数据包到这个端口。端口详解在开始讲什么是端口之前，我们先来聊一聊什么是port 呢？常常在网络上听说『我的主机开了多少的port，会不会被入侵呀！？』或者是说『开那个port会比较安全？又，我的服务应该对应什么port呀！？』呵呵！很神奇吧！怎么一部主机上面有这么多的奇怪的port呢？这个port有什么作用呢？！由于每种网络的服务功能都不相同，因此有必要将不同的封包送给不同的服务来处理，所

常见端口号对应的协议

协议号 ip 0 IP # Interne protocol互联网协议 icmp 1 ICMP # Interne control message protocol ggp 3 GGP # Gateway-gateway protocol tcp 6 TCP # Transmission control protocol egp 8 EGP # Exterio gateway protocol pup 12 PUP # PARC universal packet protocol udp 17 UDP # Use datagram protocol hmp 20 HMP # Hos monitoring protocol xns-idp 22 XNS-IDP # Xerox NS IDP rdp 27 RDP # "reliable datagram" protocol ipv6 41 IPv6 # Interne

protocol IPv6 ipv6-route 43 IPv6-Route # Routing header for IPv6 ipv6-frag 44 IPv6-Frag # Fragment header for IPv6 esp 50 ESP # Encapsulating security payload ah 51 AH # Authentication header ipv6-icmp 58 IPv6-ICMP # ICMP for IPv6 ipv6-nonxt 59 IPv6-NoNxt # No next header for IPv6 ipv6-opts 60 IPv6-Opts # Destination options for IPv6 rvd 66 RVD # MIT remote virtual disk 端口编号 echo 7/tcp echo 7/udp

常见的网络协议\端口号一．端口的分类端口的分类根据其参考对象不同有不同划分方法，如果从端口的性质来分，通常可以分为以下三类：（1）公认端口（Well Known Ports）：这类端口也常称之为"常用端口"。这类端口的端口号从0到1024，它们紧密绑定于一些特定的服务。通常这些端口的通信明确表明了某种服务的协议，这种端口是不可再重新定义它的作用对象。例如：80端口实际上总是HTTP通信所使用的，而23号端口则是Telnet服务专用的。这些端口通常不会像木马这样的黑客程序利用。（2）注册端口（Registered Ports）：端口号从1025到49151。它们松散地绑定于一些服务。也是说有许多服务绑定于这些端口，这些端口同样用于许多其他目的。这些端口多数没有明确的定义服务对象，不同程序可根据实际需要自己定义。（3）动态和/或私有端口（Dynamic and/or Private Ports）：端口号从49152到65535。如果根据所提供的服务方式的不同，端口又可分为"TCP协议端口"和"UDP协议端口"两种。因为计算机之间相互通信一般采用这两种通信协议。上面所介绍的"连接方式"是一种直接与接收方进行的连接，发送信息以后，可以确认信息是否到达，这种方式大多采用TCP协议；另一种是不是直接与接收方进行连接，只管把信息放在网上发出去，而不管信息是否到达，这种方式大多采用UDP协议，IP 协议也是一种无连接方式。二．常见的网络协议网际层协议：包括：IP协议、ICMP协议、ARP协议、RARP协议。传输层协议：TCP协议、UDP协议。应用层协议：FTP、Telnet、SMTP、HTTP、RIP、NFS、DNS。使用TCP协议的常见端口主要有以下几种：（1）FTP：定义了文件传输协议，使用21端口。常说某某计算机开了FTP服务便是启动了文件传输服务。下载文件，上传主页，都要用到FTP服务。（2）Telnet：它是一种用于远程登陆的端口，用户可以以自己的身份远程连接到计算机上，通过这种端口可以提供一种基于DOS模式下的通信服务。如以前的BBS是纯字符界面的，支持BBS的服务器将23端口打开，对外提供服务。（3）SMTP：定义了简单邮件传送协议，现在很多邮件服务器都用的是这个协议，用于发送邮件。如常见的免费邮件服务中用的就是这个邮件服务端口，所以在电子邮件设置中常看到有这么SMTP端口设置这个栏，服务器开放的是25号端口（4）POP3：它是和SMTP对应，POP3用于接收邮件。通常情况下，POP3协议所用的是110端口。使用UDP协议端口常见的有：（1）HTTP：超文本传输协议。上网浏览网页时，就得在提供网页资源的计算机上打开80号端口以提供服务。常说"WWW服务"、"Web服务器"用的就是这个端口。（2）DNS：用于域名解析服务，这种服务在Windows NT系统中用得最多的。DN S用的是53号端口。（3）SNMP：简单网络管理协议，使用161号端口，是用来管理网络设备的。另外代理服务器常用以下端口：（1）. HTTP协议代理服务器常用端口号：80/8080/3128/8081/9080

协议号是存在于IP数据报的首部的20字节的固定部分，占有8bit.该字段是指出此数据报所携带的是数据是使用何种协议，以便目的主机的IP层知道将数据部分上交给哪个处理过程。也就是协议字段告诉IP层应当如何交付数据。端口号存在于UDP和TCP报文的首部，而IP数据报则是将UDP或者TCP报文做为其数据部分，再加上IP数据报首部，封装成IP数据报。而协议号则是存在这个IP数据报的首部. 可以这么理解，协议号是大类标识，端口号是小类标识。比如我们通常说80为HTTP协议，这里的80是端口号，其中HTTP协议是属于TCP协议，其协议号为6。即：TCP协议中对数据报做了更进一步的规定，其中有一个编号用以对应其下各种不同服务，这个编号即端口号。当然UDP也可以这样规定，但是UDP和80和TCP的80就不是一回事了。同样，ICMP也可以规定自己的80端口号，当然这是没有的。这与协议的设计有关。就像学校里1班和2班都可以给本班的同学编号，1班可以有18号，但2班不一定就有18号，或者2班根本就不编号。协议号和端口号的区别网络层-数据包的包格式里面有个很重要的字段叫做协议号。比如在传输层如果是tcp连接，那么在网络层ip包里面的协议号就将会有个值是6，如果是udp的话那个值就是17-----传输层传输层--通过接口关联（端口的字段叫做端口）---应用层，详见RFC 1700 协议号是存在于IP数据报的首部的20字节的固定部分，占有8bit.该字段是指出此数据报所携带的是数据是使用何种协议，以便目的主机的IP层知道将数据部分上交给哪个处理过程。也就是协议字段告诉IP层应当如何交付数据。而端口，则是运输层服务访问点TSAP，端口的作用是让应用层的各种应用进程都能将其数据通过端口向下交付给运输层，以及让运输层知道应当将其报文段中的数据向上通过端口交付给应用层的进程。端口号存在于UDP和TCP报文的首部，而IP数据报则是将UDP或者TCP报文做为其数据部分，再加上IP数据报首部，封装成IP数据报。而协议号则是存在这个IP数据报的首部. 比如，客户端发送一个数据包给ip,然后ip将进来的数据发送给传输协议(tcp或者udp),然后传输协议再根据数据包的第一个报头中的协议号和端口号来决定将此数据包给哪个应用程序(也叫网络服务)。也就是说，协议号+端口号唯一的确定了接收数据包的网络进程。由于标志数据发送进程的'源端口号'和标志数据接受进程的'目的端口号'都包含在每个tcp段和udp段的第一个分组中，系统可以知道到底是哪个客户应用程序同哪个服务器应用程序在通讯，而不会将数据发送到别的进程中。但是要注意的一点是同样的一个端口在不同的协议中的意义是不同的，比如tcp和udp中的端口31指的并不是同一个端口。但是对于同一个协议，端口号确实唯一的。在端口中分为两种，一是'知名端口'，也即小于256的端口号。另一种是'动态分配的端口'，也就是在需要时再将其赋给特定的进程。这类似于nt服务器或者163拨号上网，也就是动态的分配给用户一个目前没有用到的标志。动态分配的端口号都是高于标准端口号范围的。网络服务常用的应用协议和对应的标准端口号。《网络服务通用的应用协议和对应的标准（默认）端口号：》应用协议端口号/协议说明 ftp-data 20/tcp FTP, data

网络常用端口与协议 HTTP:80：www服务。 DHCP：服务器端的端口号是67 DHCP：客户机端的端口号是68 POP3：POP3仅仅是接收协议，POP3客户端使用SMTP向服务器发送邮件。POP3所用的端口号是110。 SMTP：端口号是25。SMTP真正关心的不是邮件如何被传送，而只关心邮件是否能顺利到达目的地。SMTP具有健壮的邮件处理特性，这种特性允许邮件依据一定标准自动路由，SMTP具有当邮件地址不存在时立即通知用户的能力，并且具有在一定时间内将不可传输的邮件返回发送方的特点。Telnet：端口号是23。Telnet是一种最老的Internet应用，起源于ARPNET。它的名字是“电信网络协议（Telecommunication Network Protocol）”的缩写。 FTP：FTP使用的端口有20和21。20端口用于数据传输，21端口用于控制信令的传输，控制信息和数据能够同时传输，这是FTP的特殊这处。FTP采用的是TCP连接。 TFTP：端口号69，使用的是UDP的连接。 DNS:53，名称服务 NetBIOS: 137,138,139,其中137、138是UDP端口，当通过网上邻居传输文件时用这个端口。而139端口：通过这个端口进入的连接试图获得NetBIOS/SMB服务。这个协议被用于windows 文件和打印机共享和SAMBA。还有WINS Regisrtation也用它。 NNTP 网络新闻传输协议:119 SNMP（简单网络管理协议）:161端口 RPC（远程过程调用）服务:135端口 QQ:使用8000(服务端)和4000端口（客户端） 21 端口：21 端口主要用于FTP（File Transfer Protocol，文件传输协议）服务。 23 端口：23 端口主要用于Telnet（远程登录）服务，是Internet上普遍采用的登录和仿真程序,最初设计被用来方便管理员远程管理计算机,可现在真正将其发挥到极致的是"黑客"! 25 端口：25 端口为SMTP（Simple Mail Transfer Protocol，简单邮件传输协议）服务器所开放，主要用于发送邮件，如今绝大多数邮件服务器都使用该协议。 53 端口：53 端口为DNS（Domain Name Server，域名服务器）服务器所开放，主要用于域名解析，DNS 服务在NT 系统中使用的最为广泛。 67、68 端口：67、68 端口分别是为Bootp 服务的Bootstrap Protocol Server（引导程序协议服务端）和Bootstrap Protocol Client（引导程序协议客户端）开放的端口。 69 端口：TFTP 是Cisco 公司开发的一个简单文件传输协议，类似于FTP。 79 端口：79 端口是为Finger 服务开放的，主要用于查询远程主机在线用户、操作系统类型以及是否缓冲区溢出等用户的详细信息。 80 端口：80 端口是为HTTP（HyperText Transport Protocol，超文本传输协议）开放的，这是上网冲浪使用最多的协议，主要用于在WWW（World Wide Web，万维网）服务上传输信息的协议。 99 端口：99 端口是用于一个名为“Metagram Relay”（亚对策延时）的服务，该服务比较少见，一般是用不到的。 109、110 端口：109 端口是为POP2（Post Office Protocol Version2，邮局协议2）服务开放的，110 端口是为POP3（邮件协议3）服务开放的，POP2、POP3 都是主要用于接收邮件的。111 端口：111 端口是SUN 公司的RPC（Remote Procedure Call，远程过程调用）服务所开

竭诚为您提供优质文档/双击可除 mms,协议,端口篇一：常见协议端口号 bgp端口179 Rip,v1,v2都使用udp端口520 eigRp在tcp/ip中使用ip协议号88 224.0.0.9eigRp支持许多 ospFversion2使用ip协议号89 224.0.0.5和224.0.0.6 isis clnsios/osi集成isis 端口号和协议号的概念。 2121端口主要用于Ftp Filetransferprotocol 2323端口主要用于telnet 是internet上普遍采用的登录和仿真程序。 2525端口为smtp simplemailtransferprotocol

务器都使用该协议。 5353端口为dns domainnameserver dns服务在nt系统中使用的最为广泛。 67、6867、68端口分别是为bootp服务的bootstrapprotocolserver bootstrapprotocolclient 69tFtp是cisco公司开发的一个简单文件传输Ftp。 7979端口是为Finger 询远程主机在线用户、操作系统类型以及是否缓冲区溢出等用户的详细信息。 8080端口是为http hypertexttransportprotocol www worldwideweb 输信息的协议。 9999端口是用于一个名为“metagramRelay

109、110109端口是为pop2 postofficeprotocolVersion22 110端口是为pop33 pop2、pop3都是主要用于接收邮件的。 111111端口是sun公司的Rpc Remoteprocedurecall Rpc在多种网络服务中都是很重要的组件。 113113端口主要用于windows的“authenticationservice 119119端口是为“networknewstransferprotocol 简称nntp 135135端口主要用于使用Rpc Remoteprocedurecall 并提供dcom 137137端口主要用于“netbiosnameservice” netbios 139139端口是为“netbiossessionservice”提