Groupwise Registration in NAMICkit - PowerPoint PPT Presentation
1 / 19
Title:
Groupwise Registration in NAMICkit
Description:
Affine. Non-rigid using B-Splines. 3. Registering to the Mean of the Population. 4 ... Before Affine BS 4. BS 8 BS 16. 11. Summary. Implemented group-wise ... – PowerPoint PPT presentation
Number of Views:49
Avg rating:3.0/5.0
Title: Groupwise Registration in NAMICkit
1
Group-wise Registration in NAMIC-kit
- Serdar K Balci (MIT)
- Lilla Zöllei (MGH)
- Kinh Tieu (BWH)
- Mert R Sabuncu (MIT)
- Polina Golland (MIT)
2
Robust Group-wise Registration
- Entropy based group-wise registration
- ITK implementation
- Empirical Evaluation
3
Background Groupwise Registration
- Images Transforms
- Transforms
- Affine
- Non-rigid using B-Splines
4
Registering to the Mean of the Population
5
Groupwise Registration Congealing
L . Zöllei, E. Learned-Miller, E. Grimson, W.M.
Wells III. "Efficient Population Registration of
3D Data."
6
Congealing Intuition
- If Gaussian
- If also
- Registering to the mean with LS metric
7
Implementation
- ITK classes
- Group-wise registration using congealing
- Variance
- Entropy
8
Results
Before Affine BS 4 BS 8
BS 16
Entropy
Variance
9
Overlap Measures
10
Full Term Babies
Before Affine
BS 4
BS 8 BS 16
BS 32
11
Pre Term Babies
Before Affine BS 4
BS 8 BS 16
12
Summary
- Implemented group-wise registration in ITK
- Congealing Entropy based registration
- Affine and BSpline
- Multithreaded implementation
- Bspline optimization
- Initial Evaluation
- A population of 50 subjects
- Used segmentation labels to evaluate
13
Ongoing Work
- Finding optimal parameters
- B-Spline mesh size,
- of hierarchical levels
- Subsampling
- Quantitative comparison to other methods
- Pair-wise registration to the mean using MI
14
(No Transcript)
15
Congealing with Two Images
- Using Parzen windows
- As we only have two images
- Pairwise registration using LS metric
16
Groupwise Reg. using Pairwise Reg.
- If we assume that images are independent given a
subject, representative of the population
17
Registering to the mean
- We assume independence over images
- and draw i.i.d. samples from each image
18
Groupwise Registration using Joint Entropy
- Assume i.i.d over space, but dont make any
assumptions about images
- Estimating entropy of an N-dimensional
distribution is a challenging task
19
Results Congealing with Entropy
Before After
(B-Splines 20mm)
Recommended
CrystalGraphics Presentations
