Dinggang Shen

1
Development and Dissemination of Robust Brain MRI
Measurement Tools (1R01EB006733)

• Dinggang Shen

Department of Radiology and BRIC UNC-Chapel Hill
2
Team

• UNC-Chapel Hill
• - Dinggang Shen
• - Guorong Wu (postdoc)
• - Minjeong Kim (postdoc)

• GE
• - Jim Miller
• - Xiaodong Tao

3
Goal of this project

• To further develop HAMMER registration and white
  matter lesion (WML) segmentation algorithms, for
  improving their robustness and performance.
• To design separate software modules for these two
  algorithms and incorporate them into the 3D
  Slicer.

4
Progress of HAMMER in 2009

• Successfully implemented HAMMER in ITK.
• (Over 2,000 lines of code)
• Integrated HAMMER into Slicer3
• Verified and tested its performance in Slicer3

5
Progress of HAMMER in 2009
Typical Registration Result of HAMMER in Slicer3
Template
Average of 18 aligned images
Subject
Registration result
6
Progress of HAMMER in 2009
RABBIT To speed up our HAMMER registration
algorithm (1.5 hours)
e2
1215 minutes
Template
e1
(1.5 hours)
Subject

• Tang et. al., RABBIT Rapid Alignment of Brains
  by Building Intermediate Templates. Neuroimage,
  47(4)1277-87, Oct 1 2009.

7
Progress of HAMMER in 2009
e2
e1
1215 mins
Subject

• Tang et. al., RABBIT Rapid Alignment of Brains
  by Building Intermediate Templates. Neuroimage,
  47(4)1277-87, Oct 1 2009.

8
Progress of HAMMER in 2009
TPS-HAMMER

• Use soft correspondence detection to robustly
  establish correspondences for the driving voxels
• Use Thin Plate Splines (TPS) to effectively
  interpolate deformation fields, based on those
  estimated at the driving voxels

• Wu et. al., TPS-HAMMER Improving HAMMER
  Registration Algorithm by Soft Correspondence
  Matching and Thin-Plate Splines Based Deformation
  Interpolation. Neuroimage, 49(3)2225-2233, Feb
  2010.

9
Work Plan of HAMMER in 2010

• Further improve HAMMER in Slicer3
• Implement RABBIT to speedup the registration
• Implement TPS-HAMMER in ITK
• Implement intensity-HAMMER in ITK
• Serve HAMMER user community
• To provide training and tutorial
• To provide technical support
• To develop user-friendly interface to the end user

10
WML Segmentation

• Attribute vector for each point v

FLAIR
PD
T2
T1
Neighborhood O (5x5x5mm)

• SVM ? To train a WML segmentation
  classifier.
• Adaboost ? To adaptively weight the training
  samples and improve the generalization of WML
  segmentation method.

• Lao, Shen, et al "Computer-Assisted
  Segmentation of White Matter Lesions in 3D MR
  images Using Support Vector Machine", Academic
  Radiology, 15(3)300-313, March 2008.

11
Progress in 2009

• We have implemented all WML segmentation
  components in ITK

Manual Segmentation
Co-registration
Skull-stripping
Training SVM model via training sample and
Adaboost
Intensity normalization
Pre-processing
Training
Voxel-wise evaluation segmentation
False positive elimination
Testing
Post-processing
12
Progress in 2009

• Have incorporated it into Slicer3
• Developer Tools gtgt White Matter Lesion
  Segmentation

13
Progress in 2009

• User interface of WML segmentation in Slicer3

Training
Segmentation

• Input T1, T2, PD, FLAIR images and lesion
  ROI of n training subjects
• Output SVM model

• Input T1, T2, PD, FLAIR images of test
  subject(s) and trained SVM model
• Output segmented lesion ROI

14
Progress in 2009

• A typical segmentation result

Our result
Ground truth
FLAIR
15
Plan of 2010

• Further development of WML segmentation algorithm
• Improve the robustness of multi-modality image
  registration (for T1/T2/PD/FLAIR) by using a
  novel quantitative and qualitative measurement
  for mutual information
• Design region-adaptive classifiers, in order to
  allow each classifier for capturing relative
  simple WML intensity pattern in each region
• Develop a WML atlas for guiding the WML
  segmentation
• Upgrade of WML lesion segmentation module in
  Slicer3

16
Conclusion

• Further develop HAMMER registration and WML
  segmentation algorithms ? improve their
  robustness and performance

17
Thank you!
http//bric.unc.edu/IDEAgroup/
http//www.med.unc.edu/dgshen/