Fast Automatic Segmentation of White Matter Streamlines Based on a Multi-Subject Bundle Atlas

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Medical Image Analysis

Background:

• Diffusion MRI (dMRI) tractography generates complex datasets of white matter streamlines.
• Accurate segmentation of white matter bundles is crucial for understanding brain connectivity.
• Existing methods can be computationally intensive, limiting interactive analysis of large datasets.

Purpose of the Study:

• To develop a fast and efficient algorithm for segmenting white matter bundles from large dMRI tractography datasets.
• To enable interactive visualization, analysis, and segmentation of complex brain connectomes.
• To improve the speed of white matter tract segmentation without compromising accuracy.

Main Methods:

• A multisubject atlas-based approach using a distance metric to compare streamlines with atlas centroids.
• A two-stage preprocessing strategy employing a simplified distance metric to rapidly discard candidate streamlines.
• Final segmentation using the original metric to ensure accurate labeling and eliminate false positives.

Main Results:

• A single-thread implementation can segment nearly 9 million streamlines in under 6 minutes.
• Parallel implementations achieve segmentation times under 22 seconds (multicore) and 5 seconds (GPU).
• The algorithm guarantees no false negatives and eliminates false positives through its two-stage approach.

Conclusions:

• The developed algorithm significantly accelerates white matter bundle segmentation from dMRI tractography.
• Its high performance supports interactive applications for analyzing large-scale brain connectivity data.
• This method offers a powerful tool for neuroscientists and clinicians studying brain structure and function.