Similarity metrics for groupwise non-rigid registration

Area of Science:

• Medical Imaging
• Computational Anatomy
• Neuroscience

Background:

• Groupwise registration is crucial for constructing average brain atlases.
• Developing effective groupwise similarity metrics is challenging and does not always extend from pairwise approaches.

Purpose of the Study:

• To investigate and evaluate various groupwise similarity metrics for atlas construction.
• To propose and validate a novel groupwise similarity metric based on Normalized Mutual Information.

Main Methods:

• Quantitative evaluation of groupwise metrics on simulated and 3D Magnetic Resonance (MR) datasets.
• Comparison of proposed groupwise metrics against traditional pairwise registration methods.

Main Results:

• The novel Normalized Mutual Information metric demonstrates competitive or superior performance.
• Groupwise metrics show potential for enhancing average atlas construction.

Conclusions:

• The proposed Normalized Mutual Information metric is a viable option for groupwise registration.
• Further development in groupwise similarity metrics can advance the field of computational anatomy and atlas construction.