Progress Report

February 8th, 2005

Overview

• Mini-presentation: Registration and Models in Depth
• VXL:
  • Progress on tools has been limited
  • Few experiments performed
  • Implementation of points selection for triangulation
• MATLAB Meeting
• Gio - similarities in work
• Mike Rogers - Registration

Progress Reports Template

• Based on S5 (A Simple Standards-Based Slide Show System)
  • Makes agenda easier to step through
  • Can view entire report as marked-up text
  • Or move step-by-step, point-by-point
  • Images and links are well-integrated
  • Searchable

MATLAB Meetup

• Turn-out was poor
• Main points raised:
  • People must begin to use and submit code using CVS
  • Functions need to be generalised and documented/annotated
  • Only 2-3 people use the repository
  • We could identify about 15 MATLAB users in ISBE
  • Documentation tool can/should be extended
  • Ought to be in frequent use first

Mini-Presentation

• Mini-presentation: Registration and Models in Depth
• Informal
• Collection of thoughts
• Introduces:
  • Registration
  • Models
  • Algorithms

Contrast-enhanced MR Registration (Gio)

• Registration of volume time-series
• Currently just translation (3 parameters)
• Using correlation ratio
• C library from Oxford
• No multi-resolution approach
• Model of expected intensity versus time
• Re-iterate registration until results fit model

Contrast-enhanced MR Registration (Gio) - ctd.

• Results are accepted based on deviation threshold
• Among the suggestions made:
  • Incorporate model in the OF
  • Attempting more degrees of freedom
  • Using multi-scale approach, e.g. by using VXL
• In summary: discussion appeared helpful to Gio's work

Mike Rogers - Registration

• Registeration protein gel images
• Using CPS warps
• under MATLAB with some C
• Looked at some early results which VXL code produced
• Wish to try more knot-points, larger set
• Interest in porting/interfacing warp data

Information-Theoretic OF in 3-D

• Intent: seeing if the OF is stable at the solution
• Description of the experiment:
  • 2 identical data volumes
  • Registration is based on the two histograms and joint histogram
  • Rigid, affine and non-rigid registration stages
• Result: No warps (rigid, affine, or non-rigid) are being applied
• Conclusion: the minimum is found at the correct solution (not necessarily in theory)

VXL

• Programming of point selection strategy
• Points are used for triangulation
• Tools:
  • Synthesis of 3-D data
  • Metric of 3-D registration performance
  • The second depends on the first
  • At lower priority at this stage
• Vlad and Tim are fairly busy

Next Steps

• Work on VXL
• Another MATLAB meeting is being considered

Miscellany

• ISBE-related banner for CS front page
• Financial issues including the MATLAB documentation project