experiments on a new inter- subject registration method john ashburner 2007
TRANSCRIPT
![Page 1: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/1.jpg)
Experiments on a New Inter-Subject Registration Method
John Ashburner
2007
![Page 2: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/2.jpg)
Abstract• The objective of this work was to devise a more precise method of inter-
subject brain image registration than those currently available in the SPM software. This involved a model with many more degrees of freedom, but which still enforces a one-to-one mapping. Speed considerations were also important. The result is an approach that models each warp by single velocity field. These are converted to deformations by a scaling and squaring procedure, and the inverses can be generated in a similar way. Registration is via a Levenberg-Marquardt optimization strategy, which uses a full multi-grid algorithm to rapidly solve the necessary equations.
• The method has been used for warping images of 471 subjects. This involved simultaneously matching grey matter with a grey matter template, and white matter with a white matter template. After every few iterations, the templates were re-generated from the means of the warped individual images. Evaluations involved applying pattern recognition procedures to the resulting deformations, in order to assess how well information such as the ages and sexes of the subjects could be predicted from the encoded deformations. A slight improvement in prediction accuracy was obtained when compared to a similar procedure using a small deformation model.
![Page 3: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/3.jpg)
Overview
• Motivation– Dimensionality– Inverse-consistency
• Principles
• Geeky stuff
• Example
• Validation
• Future directions
![Page 4: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/4.jpg)
Motivation
• More precise inter-subject alignment– Improved fMRI data analysis
• Better group analysis• More accurate localization
– Improve computational anatomy• More easily interpreted VBM• Better parameterization of brain shapes
– Other applications• Tissue segmentation• Structure labeling
![Page 5: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/5.jpg)
Image Registration
• Figure out how to warp one image to match another
• Normally, all subjects’ scans are matched with a common template
![Page 6: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/6.jpg)
Current SPM approach
• Only about 1000 parameters.– Unable model detailed
deformations
![Page 7: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/7.jpg)
A simple 2D example
Individual brain
Warped Individual
Reference
![Page 8: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/8.jpg)
Residual Differences
Individual brain
Warped Individual
![Page 9: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/9.jpg)
Expansion and
contraction
• Relative volumes encoded by Jacobian determinants of deformation
![Page 10: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/10.jpg)
Tissue volume comparisons
Warped grey matter
Jacobian determinants
Absolute grey matter
volumes
![Page 11: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/11.jpg)
A one-to-one mapping
• Many models simply add a smooth displacement to an identity transform– One-to-one mapping not enforced
• Inverses approximately obtained by subtracting the displacement– Not a real inverse Small deformation
approximation
![Page 12: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/12.jpg)
Overview
• Motivation
• Principles
• Geeky stuff
• Example
• Validation
• Future directions
![Page 13: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/13.jpg)
Principles
DiffeomorphicAnatomicalRegistrationThroughExponentiatedLie Algebra
Deformations parameterized by a single flow field, which is considered to be constant in time.
![Page 14: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/14.jpg)
DARTEL
• Parameterizing the deformation
• φ(0)(x) = x
• φ(1)(x) = ∫ u(φ(t)(x))dt• u is a flow field to be estimated
t=0
1
![Page 15: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/15.jpg)
Euler integration
• The differential equation is
dφ(x)/dt = u(φ(t)(x))• By Euler integration
φ(t+h) = φ(t) + hu(φ(t))• Equivalent to
φ(t+h) = (x + hu) o φ(t)
![Page 16: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/16.jpg)
Flow Field
![Page 17: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/17.jpg)
For (e.g) 8 time steps
Simple integration• φ(1/8) = x + u/8• φ(2/8) = φ(1/8) o φ(1/8) • φ(3/8) = φ(1/8) o φ(2/8) • φ(4/8) = φ(1/8) o φ(3/8) • φ(5/8) = φ(1/8) o φ(4/8) • φ(6/8) = φ(1/8) o φ(5/8) • φ(7/8) = φ(1/8) o φ(6/8) • φ(8/8) = φ(1/8) o φ(7/8)
7 compositions
Scaling and squaring• φ(1/8) = x + u/8• φ(2/8) = φ(1/8) o φ(1/8)
• φ(4/8) = φ(2/8) o φ(2/8)
• φ(8/8) = φ(4/8) o φ(4/8)
3 compositions
• Similar procedure used for the inverse.Starts withφ(-1/8) = x - u/8
![Page 18: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/18.jpg)
Scaling and squaring example
![Page 19: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/19.jpg)
DARTEL
![Page 20: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/20.jpg)
Jacobian determinants remain positive
![Page 21: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/21.jpg)
Overview
• Motivation
• Principles
• Geeky stuff– Feel free to sleep
• Example
• Validation
• Future directions
![Page 22: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/22.jpg)
Registration objective function
• Simultaneously minimize the sum of – Likelihood component
• From the sum of squares difference
• ½∑i(g(xi) – f(φ(1)(xi)))2
• φ(1) parameterized by u
– Prior component• A measure of deformation roughness
• ½uTHu
![Page 23: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/23.jpg)
Regularization model
• DARTEL has three different models for H– Membrane energy– Linear elasticity– Bending energy
• H is very sparse
An example H for 2D registration of 6x6 images (linear elasticity)
![Page 24: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/24.jpg)
Regularization models
![Page 25: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/25.jpg)
Optimisation
• Uses Levenberg-Marquardt– Requires a matrix solution to a very large set
of equations at each iteration
u(k+1) = u(k) - (H+A)-1 b
– b are the first derivatives of objective function– A is a sparse matrix of second derivatives– Computed efficiently, making use of scaling
and squaring
![Page 26: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/26.jpg)
Relaxation
• To solve Mx = cSplit M into E and F, where
• E is easy to invert• F is more difficult
• Sometimes: x(k+1) = E-1(c – F x(k))• Otherwise: x(k+1) = x(k) + (E+sI)-1(c – M x(k))
• Gauss-Siedel when done in place.• Jacobi’s method if not
• Fits high frequencies quickly, but low frequencies slowly
![Page 27: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/27.jpg)
H+A = E+F
![Page 28: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/28.jpg)
Highest resolution
Lowest resolution
Full Multi-Grid
![Page 29: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/29.jpg)
Overview
• Motivation
• Principles
• Geeky stuff
• Example– Simultaneous registration of GM & WM– Tissue probability map creation
• Validation
• Future directions
![Page 30: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/30.jpg)
Simultaneous registration of GM to GM and WM to WM
Grey matter
White matter
Grey matter
White matter
Grey matter
White matter
Grey matter
White matter
Grey matter
White matterTemplate
Subject 1
Subject 2
Subject 3
Subject 4
![Page 31: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/31.jpg)
TemplateInitial
Average
After a few iterations
Final template
Iteratively generated from 471 subjects
Began with rigidly aligned tissue probability maps
Used an inverse consistent formulation
![Page 32: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/32.jpg)
Grey matter average of 452 subjects – affine
![Page 33: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/33.jpg)
Grey matter average of 471 subjects
![Page 34: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/34.jpg)
White matter average of 471 subjects
![Page 35: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/35.jpg)
Initial GM images
![Page 36: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/36.jpg)
Warped GM images
![Page 37: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/37.jpg)
Overview
• Motivation
• Principles
• Geeky stuff
• Example
• Validation– Sex classification– Age regression
• Future directions
![Page 38: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/38.jpg)
Validation
• There is no “ground truth”• Looked at predictive accuracy
– Can information encoded by the method make predictions?
• Registration method blind to the predicted information• Could have used an overlap of fMRI results
– Chose to see whether ages and sexes of subjects could be predicted from the deformations
• Comparison with small deformation model
![Page 39: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/39.jpg)
Training and Classifying
ControlTraining Data
PatientTraining Data
?
?
??
![Page 40: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/40.jpg)
Classifying
Controls
Patients
?
?
??
y=f(aTx+b)
![Page 41: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/41.jpg)
Support Vector Classifier
![Page 42: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/42.jpg)
Support Vector Classifier (SVC)
SupportVector
SupportVector
Support
Vector
a is a weighted linear combination of the support vectors
![Page 43: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/43.jpg)
Some Equations
• Linear classification is by y = f(aTx + b)– where a is a weighting vector, x is the test data, b is an
offset, and f(.) is a thresholding operation
• a is a linear combination of SVs a = i wi xi
• So y = f(i wi xiTx + b)
![Page 44: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/44.jpg)
Going Nonlinear
• Nonlinear classification is by
y = f(i wi (xi,x))– where (xi,x) is some function of xi and x.
• e.g. RBF classification (xi,x) = exp(-||xi-x||2/(22))
• Requires a matrix of distance measures (metrics) between each pair of images.
![Page 45: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/45.jpg)
Nonlinear SVC
![Page 46: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/46.jpg)
Cross-validation
• Methods must be able to generalise to new data• Various control parameters
– More complexity -> better separation of training data– Less complexity -> better generalisation
• Optimal control parameters determined by cross-validation– Test with data not used for training– Use control parameters that work best for these data
![Page 47: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/47.jpg)
Two-fold Cross-validation
Use half the data for training.
and the other half for testing.
![Page 48: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/48.jpg)
Two-fold Cross-validation
Then swap around the training and test data.
![Page 49: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/49.jpg)
Leave One Out Cross-validation
Use all data except one point for training.
The one that was left out is used for testing.
![Page 50: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/50.jpg)
Leave One Out Cross-validation
Then leave another point out.
And so on...
![Page 51: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/51.jpg)
Support-vector classification
• Guess sexes of 471 subjects from brain shapes – 207 Females / 264 Males
• Use a random sample of 400 for training.
• Test on the remaining 71.
• Repeat 50 times.
![Page 52: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/52.jpg)
Sex classification results
• Small Deformation– Linear classifier
• 87.0% correct• Kappa = 0.736
– RBF classifier• 87.1% correct• Kappa = 0.737
• DARTEL– Linear classifier
• 87.7% correct• Kappa = 0.749
– RBF classifier• 87.6% correct• Kappa = 0.748
An unconvincing improvement
![Page 53: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/53.jpg)
Regression
23
26
30
29
18
32
40
![Page 54: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/54.jpg)
Relevance-vector regression
• A Bayesian method, related to SVMs– Developed by Mike Tipping
• Guess ages of 471 subjects from brain shapes.
• Use a random sample of 400 for training.
• Test on the remaining 71.
• Repeat 50 times.
![Page 55: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/55.jpg)
Age regression results
• Small deformation– Linear regression
• RMS error = 7.55• Correlation = 0.836
– RBF regression• RMS error = 6.68• Correlation = 0.856
• DARTEL– Linear regression
• RMS error = 7.90• Correlation = 0.813
– RBF regression• RMS error = 6.50• Correlation = 0.867
An unconvincing improvement(slightly worse for linear regression)
![Page 56: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/56.jpg)
![Page 57: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/57.jpg)
Overview
• Motivation
• Principles
• Geeky stuff
• Example
• Validation
• Future directions
![Page 58: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/58.jpg)
Future directions
• Compare with variable velocity methods– Beg’s LDDMM algorithm
• Classification/regression from “initial momentum”
• Combine with SPM5 segmentation model– Similar to Emiliano D’Agostino’s method
• Develop a proper EM framework for generating tissue probability maps
![Page 59: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/59.jpg)
![Page 60: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/60.jpg)
u
Hu
![Page 61: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/61.jpg)
“Initial momentum”
Variable velocity framework
![Page 62: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/62.jpg)
“Initial momentum”
Variable velocity framework
![Page 63: Experiments on a New Inter- Subject Registration Method John Ashburner 2007](https://reader036.vdocuments.mx/reader036/viewer/2022081602/5515db8555034638038b49a8/html5/thumbnails/63.jpg)
Thank you