Groupwise track filtering via iterative message passing and pruning.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Medical Image Analysis

Background:

• Diffusion MRI-based tractography is crucial for in vivo brain connectivity analysis.
• Existing tractography methods suffer from false positives and negatives, limiting accurate neuroanatomy reconstruction.
• Anatomical priors like regions of interest (ROIs) have limitations in improving tractography fidelity.

Purpose of the Study:

• To develop a novel track filtering method to enhance the anatomical fidelity of reconstructed fiber bundles.
• To leverage groupwise consistency across subjects for more robust tractography.
• To address limitations in current tractography methods, particularly false positives and negatives.

Main Methods:

• Proposed a novel filtering method based on groupwise consistency of fiber bundles across subjects.
• Formalized groupwise consistency using degree, affinity, and proximity measures.
• Developed an iterative message-passing algorithm to update streamline consistency and prune outliers.

Main Results:

• Successfully applied the method to diffusion MRI data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) and Human Connectome Project (HCP).
• Achieved consistent reconstruction of key human brain fiber bundles: fornix, locus coeruleus pathways, and corticospinal tract.
• Demonstrated significant improvements in anatomical fidelity through qualitative and quantitative evaluations.

Conclusions:

• The proposed groupwise consistency filtering method effectively enhances the anatomical accuracy of tractography.
• This approach offers a significant advancement over traditional methods, especially when dealing with complex neuroanatomy.
• The method shows promise for reliable brain connectivity analysis in research and clinical settings.