Motion correction for 4D brain image (DTI and fMRI)

Beginner Questions
, ,
1

Dear ITK members,

We want to perform motion correction in 4D brain image such as DTI and fMRI. Is it fine to register the series of 3D brain volume on the first 3D image or b0 (baseline) MRI image in case DTI? or Any additional step is required to perform motion correction?

For such registration which error check method is good?

Kindly please reply.

Best regards,
Sunita

2

Hi,

Typically registering to the first scan is sufficient. There are heuristics that could be applied to determine the time point with the least motion artifact / highest quality, but we have processed hundreds of CT perfusion time series, and using the first has worked consistently.

If the change in the images is small and sparse (e.g., contrast spreading in blood vessels) then mean squared error is a good metric. To be more “robust” you could instead use a mutual information metric - it might be the better choice for MRI. Both have fast implementations in ITK.

2 Likes