Improving Diffusion MRI Tractography Accuracy With Topology-Informed Pruning (TIP).pdf

Summary:

This study introduces Topology-Informed Pruning (TIP), a method to automatically identify and eliminate false connections in diffusion MRI tractography. False connections are a significant challenge in mapping brain connections non-invasively. TIP leverages the topology of the tractogram to pinpoint singular tracts—trajectories with no neighboring tracks—which are likely false continuities.

Key Findings:

1. Improvement in Tracking Accuracy:
- TIP improved tracking accuracy by 11.93% in the single-shell scheme and 3.47% in the grid scheme compared to standard tractography, with statistically significant differences (p < 0.001).

2. Comparison with Neuroanatomist Assessment:
- The diagnostic agreement between TIP and neuroanatomists was comparable to the agreement between neuroanatomists themselves.

3. Methodology:
- TIP constructs a 3D tract density histogram to identify voxels with only one trajectory passing through them, indicating potential false continuities.
- Singular tracts (i.e., those with no neighboring tracks) are pruned from the tractogram.

Background and Motivation:

- Traditional methods struggle with two main causes of false connections: premature termination and false continuity.
- While white matter masks can address premature terminations, there's no effective strategy for detecting false continuities.
- The study's previous atlas work revealed that clusters with fewer neighboring tracts are more likely to be false connections due to overlapping boundaries between fiber bundles.

Conclusion:

TIP demonstrates the potential for automatic cleanup of noisy fibers in deterministic tractography, offering a valuable tool for improving the accuracy of diffusion MRI studies in both basic neuroanatomical research and clinical neurosurgical planning.