Optimizing Reproducibility of White Matter Tractography in Longitudinal Imaging of a Large Sample of Developing Adolescents

The best method for automatically producing replicable white matter (WM) tracts from diffusion imaging is not known. We use Automated Fiber Quantification (AFQ) to compare probabilistic and deterministic models’ performance generating 18 WM tracts for each of 1000 adolescents at baseline and 2-year follow up. We gauged performance on the correlation between fractional anisotropy (FA) values and degree of volumetric overlap (DICE score). Based on these measures we found that a probabilistic approach was able to more reliably generate tract profiles over time within individual subjects.