Cortical sulci recognition and spatial normalization.

Area of Science:

• Neuroimaging
• Computational anatomy
• Medical image analysis

Background:

• Brain mapping requires aligning anatomical information across individuals, often using spatial normalization to reduce inter-subject variability.
• Accurate anatomical landmarks are crucial for driving registration processes in neuroimaging.
• Current automatic labeling methods often depend heavily on the accuracy of normalization, creating a dependency loop.

Purpose of the Study:

• To develop a unified Bayesian framework for simultaneous automatic sulcal labeling and spatial normalization.
• To improve the accuracy and reliability of brain mapping by enhancing landmark identification and registration.
• To integrate global and local labeling strategies for robust sulcal structure analysis.

Main Methods:

• A coherent Bayesian framework utilizing a probabilistic atlas (Statistical Probabilistic Anatomy Map) to estimate normalization parameters and identify sulcal labels.
• Simultaneous estimation of normalization parameters and approximately 60 sulcal labels per hemisphere.
• Application to global affine and piecewise affine registration, with a focus on a novel global affine approach.

Main Results:

• The proposed global affine approach significantly outperforms standard intensity-based affine normalization in sulci alignment.
• A combined global and local joint labeling strategy achieved a mean recognition rate of 86%.
• The method provides more reliable labeling posterior probabilities without additional computational cost.

Conclusions:

• The developed Bayesian framework offers an effective and integrated solution for automatic sulcal labeling and spatial normalization in brain mapping.
• This approach enhances the precision of anatomical comparisons and registration in neuroimaging studies.
• The methods have been implemented in the BrainVISA software platform, demonstrating practical applicability.