Short association bundle atlas based on inter-subject clustering from HARDI data.

Area of Science:

• Neuroscience
• Medical Imaging
• Computational Biology

Background:

• Short association fibers are crucial for brain connectivity but challenging to map.
• Existing methods for white matter tractography lack automated, robust anatomical labeling.
• Inter-subject variability in white matter necessitates advanced clustering techniques.

Purpose of the Study:

• To develop and validate an automated method for identifying and labeling short association fiber bundles.
• To create a reproducible, anatomically informed white matter atlas.
• To assess the efficacy of non-linear registration in fiber bundle analysis.

Main Methods:

• Inter-subject hierarchical clustering of superficial white matter fibers.
• Distance-based fiber similarity measurement.
• Automatic anatomical labeling of stable fiber connections.
• Application and comparison of linear and non-linear registration techniques.

Main Results:

• Successful identification and clustering of short association fibers across two independent subject groups.
• Non-linear registration yielded significantly superior results compared to linear registration.
• Creation of a comprehensive white matter atlas with 35 left and 27 right hemisphere bundles.
• Validation of the atlas through successful segmentation of new subjects from a HARDI database.

Conclusions:

• The automated hierarchical clustering method effectively identifies and labels short association fibers.
• The resulting white matter atlas is reproducible and valuable for neuroimaging analysis.
• Non-linear registration is essential for accurate inter-subject white matter analysis and atlas creation.