Group-Wise Cortical Surface Parcellation Based on Inter-Subject Fiber Clustering.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Brain Mapping

Background:

• Cortical surface parcellation is crucial for understanding brain organization.
• Existing methods often lack group-wise consistency or rely on limited connectivity information.
• Developing automated, data-driven parcellation methods is an ongoing challenge.

Purpose of the Study:

• To present an automated, group-wise algorithm for cortical surface parcellation.
• To leverage whole-brain structural connectivity derived from inter-subject fiber clustering.
• To quantitatively evaluate the resulting parcellation against established brain atlases.

Main Methods:

• Structural connectivity was computed from representative brain fiber clusters using an inter-subject clustering scheme.
• Preliminary cortical regions were defined by intersecting fiber clusters with the cortical surface mesh.
• Final parcellation utilized parcel probability maps and graph-based overlap analysis across subjects.

Main Results:

• Two distinct inter-subject clustering schemes yielded parcellations with 171 and 109 parcels.
• The best parcellation generated 69 parcels with a Dice similarity coefficient exceeding 0.5 when compared to state-of-the-art atlases.
• This represents the first diffusion-based cortical parcellation method utilizing whole-brain inter-subject fiber clustering.

Conclusions:

• The proposed algorithm offers an automated and robust approach to group-wise cortical parcellation.
• The method effectively integrates structural connectivity information for precise brain mapping.
• This diffusion-based technique advances the field of computational neuroanatomy and brain atlas development.