Evaluating brain parcellations using the distance-controlled boundary coefficient.

Area of Science:

• Neuroscience
• Human Brain Mapping
• Computational Neuroscience

Background:

• Defining distinct functional regions in the human brain is key for mapping.
• Existing brain parcellation methods struggle with data smoothness and spatial scale biases.
• Comparing different parcellations requires an unbiased evaluation criterion.

Purpose of the Study:

• To introduce the distance-controlled boundary coefficient (DCBC), an unbiased metric for evaluating discrete brain parcellations.
• To assess the predictive power of various human neocortical parcellations using functional magnetic resonance imaging (fMRI) data.
• To compare task-based, resting-state, and multi-modal parcellations against functional boundaries.

Main Methods:

• Developed the distance-controlled boundary coefficient (DCBC) to overcome limitations in comparing brain parcellations.
• Applied the DCBC to evaluate existing anatomical, functional (fMRI), and multi-modal parcellations of the human neocortex.
• Utilized both task-based and resting-state fMRI datasets to assess functional boundary prediction.

Main Results:

• Common anatomical parcellations showed no better performance than chance in predicting functional boundaries.
• Parcellations derived from resting-state fMRI data demonstrated strong predictive power for functional boundaries.
• Multi-modal parcellations performed significantly worse than functional-data-only parcellations, indicating functional regions can cross anatomical landmarks.

Conclusions:

• The DCBC provides an unbiased method for comparing brain parcellations and advancing functional brain mapping.
• Resting-state fMRI parcellations are superior for predicting functional brain organization compared to anatomical approaches.
• Functional homogeneity often transcends traditional anatomical boundaries, highlighting the importance of functional data in brain mapping.