Projects/Diffusion/2007 Project Week DTI Registration

Return to Project Week Main Page

Key Investigators

• Utah: Ran Tao, Tom Fletcher, Ross Whitaker

Objective

We want to implement the Mutual Information metric using partial voluming for the registration of multi-modal images. The aim is to correct EPI distortions in diffusion-weighted images (DWIs).

Approach, Plan

Our approach is to register the DWIs to the baseline (B0) image using affine transformations and mutual information. We are implementing a version of MI that interpolates both joint histogram and partial voluming. This is smoother than the Mattes MI currently in ITK.

Progress

• finished 3D mutual information metric coding
• tested to register T2 diffusion weighted image to its own baseline T2 image
• with the help of Luis, rewrote the code into ITK filter

Future Work

• test the algorithm using Autism data and data from MIND
• compare the acuracy between this MI metric and Matte Metric

Additional Information

Plot of MI versus xyz translation using PTV (T1 and T2 images)

Same plot using Mattes

Reference

• Wei.M, Liu.Jundong and Liu.Junhong Artifact reduction in mutual-information-based CT-MR image registration in Medical Imaging 2004: Image Processing
• Liu.J Artifacts reduction in mutual information-based image registration using prior information in Image Processing, 2003. ICIP 2003. Proceedings. 2003 International Conference