Automated White Matter Fiber Tract Segmentation for the Brainstem

Area of Science:

• Neuroimaging
• Neuroanatomy
• Medical Image Analysis

Background:

• Manual segmentation of brainstem fiber bundles is labor-intensive and prone to variability.
• Accurate identification of these tracts is crucial for understanding neurological function and disease.

Purpose of the Study:

• To develop and validate an automated method for segmenting major brainstem fiber bundles.
• To create a standardized brainstem fiber cluster atlas for improved tractography analysis.

Main Methods:

• Probabilistic tractography using multishell, multitissue constrained spherical deconvolution on Human Connectome Project (HCP) data.
• Construction of a brainstem fiber cluster atlas by registering and clustering tractography data from 40 subjects.
• Application of cortical parcellation for fiber selection and automatic segmentation in HCP subjects and patients with brainstem cavernous malformations.

Main Results:

• An atlas of 100 identified and annotated fiber clusters was created.
• Eight major fiber bundles, including the corticospinal tract (CST) and medial lemniscus, were successfully segmented.
• High spatial overlap (mean wDice scores > 0.90) was achieved between automatic and manual reconstructions, even in the presence of brainstem lesions.

Conclusions:

• The proposed automated method accurately and reliably segments major brainstem fiber bundles.
• This approach significantly reduces labor costs and interoperator variability compared to manual methods.
• The method demonstrates robustness in segmenting tracts around brainstem cavernous malformations, aiding in clinical assessment.