Events:Registration Summit August 2009

Agenda

• Attendees: Casey Goodlett, Steve Pieper, Dominik Meier, Andriy Fedorov, Ron Kikinis, Sylvain Jaume
• Date and place: Friday, August 21, 1249 Boylston, 2nd floor conference room
• Schedule

1. 09:00-12:00 Review of registration in Slicer:
   1. What input is the user asked to provide and in what form.
   2. Does a standard user understand what they are supposed to provide?
   3. What does the user expect registration results to look like?
   4. Bias correction
   5. histogram normalization
   6. handling image distortion (EPI)
   7. capture range and start pose
   8. ROI/VOI
   9. greyscale versus segmentation (binaries, surface models)
   10. inter subject versus intrasubject
   11. List of modules:
       1. transformation module
       2. linear registration
       3. rigid registration
       4. affine registration
       5. b-spline registration
       6. register images
       7. Utah b-spline
       8. VMTK/Python ICP
       9. ACPC registration
2. 12:00-01:00 Lunch
3. 01:00-05:00 Making plans, use case scenarios, sample data sets, plan larger registration summit
   1. Action Items
      1. User specified ROI - mask image and/or box
      2. Test transform is correctly inverted
      3. Prototype user-version RegisterImages (up to affine) (2 organ regions)
         1. Same subject - same subject different modality (grayscale)
         2. same subject different timepoints (grayscale) (user knows nothing about algorithms!)
      4. Try datasets that are now available on this page
      5. document RegisterImages
      6. automated testing
      7. non-rigid once affine is working

Other items:

• Workflow with Wendy
• CUDA Implementation with Yogesh

Needs

• Robust solutions
• Clinical APIs as opposed to engineering APIs
• Good default parameters
• Modality recorded in the image class to automate use case and parameter selection
• Fast techniques for interactive investigations
• Non-interactive techniques can run longer
• Transform IO to all modules (mapping RAS to LPS as needed)
• Ability to apply estimated transforms to other types of data
• Region of interest registration (anything from a brain mask to a structure segmentation)
• Validation datasets
• Regression testing of registration accuracy

References

Bundled Registration and Tests

• https://www.slicer.org/wiki/Documentation-3.4#Registration
• http://www.na-mic.org/Wiki/index.php/Projects:DBP2:Harvard:Registration_Documentation
• http://www.na-mic.org/Wiki/index.php/MeningiomaMRIRegistrationStudy#Limitations_of_the_Slicer_registration_tools

Some Sample data

• Breast Cancer - original data, two time points after some registration and processing
• Prostate image database (intra- and inter-subject registration use cases) http://prostatemrimagedatabase.com/index.html
• liver MR to CT -- will make available
• kidney MR to CT -- will make available
• pet CT to Pet CT -- see here question: what changed between the two time points. Load scene and move the slider in slice views panel.
• knee registration for Simbios project -- http://slicer.spl.harvard.edu/slicerWiki/index.php/Stanford_Simbios_group
• Vervet MRI -- made available to Kitware for testing, will go public once the project is over

Slicer-compatible add-on registration modules

• http://www.bioimagesuite.org/
• http://www.nitrc.org/frs/shownotes.php?release_id=575
• http://www.nitrc.org/projects/multimodereg/
• http://www.nitrc.org/projects/brainsmush/
• http://www.nitrc.org/projects/brainsdemonwarp/
• http://www.nitrc.org/projects/cmtk/

Comprehensive evaluations of registration tools

• http://mindboggle.info/papers/index.php

In-house evaluations of registration tools

• Sophisticated end-user without computer vision background using Slicer modules
• Vervet head MRI atlas-to-subject registration
• Head MRI registration between different timepoints (human subjects, changing pathology present)
• Side by side comparisons in slicer2 and slicer3 with data and images