2017 Winter Project Week/DWI Similarity Metrics

From NAMIC Wiki
Revision as of 04:34, 13 January 2017 by Lchauvin (talk | contribs)
Jump to: navigation, search
Home < 2017 Winter Project Week < DWI Similarity Metrics

Key Investigators

  • Laurent Chauvin, ETS Montreal
  • Fan Zhang, BWH
  • Lauren O'Donnell, BWH
  • Matthew Toews, ETS Montreal

Project Description

Objective Approach and Plan Progress and Next Steps
  • Investigate possible metrics for identification of information-rich patches in DWI data
  • Discuss about DWI data structure and acquisition protocol to have a better understanding of the information available in DWI datasets (e.g. single shell / multi shell, gradient orientations, b-values)
  • Identify possible metrics available to measure similarity (e.g. Mutual Information, Correlation) between datasets
  • Investigate the potential of CNN for this type of application
  • Discussed with Lauren and clarified some points on Diffusion Imaging. It also might help me to optimize some of my code.
  • Discussed with Mohsen about machine learning and how I might use it for data classification in my project.
  • Discussed with Henrik about information-rich patches of information in DWI data, and I learned about his LORDWI approach that could be useful to help me identify relevant information in DWI datasets.
  • Attended almost all machine learning events, and got a better understanding of their utility and when it might be useful to use them and when it is not necessary.
  • Participated at the tutorial contest.

Background and References