Tractography segmentation using a hierarchical Dirichlet processes mixture model.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Biostatistics

Background:

• Accurate clustering of white matter fiber tracts is crucial for understanding brain connectivity.
• Existing methods often require manual specification of cluster numbers or pairwise fiber comparisons.
• Unsupervised learning approaches are needed for large-scale neuroimaging datasets.

Purpose of the Study:

• To propose a new nonparametric Bayesian framework for unsupervised clustering of white matter fiber tracts.
• To develop a method that automatically determines the number of fiber bundles.
• To enable efficient and scalable clustering of large neuroimaging datasets across multiple subjects.

Main Methods:

• Utilized a hierarchical Dirichlet processes mixture (HDPM) model for nonparametric Bayesian clustering.
• Employed a Dirichlet process (DP) prior to automatically learn the number of clusters from data.
• Developed a method for using learned bundle models as priors for new subject data, allowing for novel cluster creation.

Main Results:

• Successfully clustered white matter fiber tracts into bundles using the proposed HDPM model.
• Demonstrated automatic determination of the number of clusters without manual input.
• Showcased the framework's ability to cluster large datasets (over 120,000 fibers) without pairwise distances or subsampling.
• Validated the approach on multiple datasets, including a large-scale one.

Conclusions:

• The proposed nonparametric Bayesian framework offers an effective and scalable solution for unsupervised white matter tract clustering.
• The method automates cluster number determination and facilitates the analysis of large, multi-subject neuroimaging datasets.
• This approach advances the field of computational neuroscience by providing a robust tool for brain connectivity analysis.