Retrospective evaluation of intersubject brain registration.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• Numerous methods exist for registering individual brains.
• Objective comparison of nonrigid (elastic) registration methods is lacking.
• Intersubject registration is crucial for anatomical and functional normalization.

Purpose of the Study:

• To propose and validate an evaluation framework for nonrigid brain registration methods.
• To objectively compare the performance of different registration techniques.
• To focus on the matching of cortical areas for intersubject normalization.

Main Methods:

• Development of an evaluation framework using global and local relevance measures.
• Application of the framework to 6 different nonrigid registration methods.
• Experiments conducted on a database of 18 subjects.
• Analysis focused on the matching of cortical areas and sulci.

Main Results:

• Global measures indicated that registration quality correlates with transformation degrees of freedom.
• Local measures, specifically for cortical sulci matching, showed no significant difference between rigid and nonrigid methods.
• The proposed framework allows for objective performance evaluation.

Conclusions:

• The study provides a novel framework for evaluating brain registration methods.
• Findings suggest that for cortical sulci matching, simpler rigid methods may suffice, challenging the necessity of complex nonrigid approaches in all cases.
• Further research is needed to explore the implications for deep brain structures.