Abstract #

Quantitative measures of connectivity are biased by erroneous streamlines produced by tractography algorithms:

• Streamline count (density)
• Avg length
• Spatial extent (volume) Solutions proposed to reduce the biases in the streamline distribution:

1. Optimize tractography params in terms of connectivity
2. Relax stopping criterion with novel probabilistic stopping criterion and particle filtering method (using partial volume estimation maps).

Introduction #

Bundles hard to track:

• Bundles positioned in partial volume with CSF: streamline propagation is more likely to be stopped (e.g. CC, fornix).
• Narrow bundles: they are more likely to be affected by error in the tracking mask => stopping the streamline propagation.
• Curved bundles: noise can make the tracking direction harder to follow in curved regions (since discrete steps are taken in the estimated tangent direction) Length of bundles bias their reconstruction:

1. Seeding from WM increases the density (more seeds positioned in longer bundles overall)
2. Longer bundles are harder to completely recover because of several premature stops => lower density of streamlines. Reconstruction biases: stopping/masking criterions, tractography parameters.