White matter fiber tractography based on diffusion-weighted magnetic resonance imaging is a promising method for non-invasive investigation of anatomical connectivity in the human brain. Knowledge of the white matter connections linking functional brain areas can inform interpretation of functional imaging results and allow the construction of biologically informed computational and statistical models. However, relatively little attention has been paid to the reproducibility and external validity of tractography results, even as the user base of this technology continues to grow, and as tractography research is applied to cognitive neuroscience research in novel ways. In this investigation, we addressed the reliability and validity of deterministic tractography results based on diffusion spectrum imaging (DSI). Reliability was evaluated both in terms of the presence/absence of fiber connections across sessions and the correlation of fiber density values. Validity was assessed by comparing tractography results to findings from invasive studies of the macaque monkey: we focused on the cortical and subcortical connections of the frontal eye fields (FEF). Results indicated significant variability in tractography: on average, intercortical connections present in one session had only a 75% likelihood of being detected in a second session from the same individual. However, the fiber density of repeatedly-detected connections was highly reliable, with an average between-session correlation coefficient of 0.94. Next, we investigated how global vs. targeted tractography approaches affected reliability and detection power. We found that a targeted approach, involving the use of region-of interest (ROI) constraints, yielded a large advantage in detection power and modest improvements in reliability. Finally, fiber connections of the human FEF were broadly consistent with hypotheses derived from a meta-analysis of macaque findings: we found reliable projections to the supplementary eye fields (SEF), striatum, thalamus, and parietal cortex. In contrast, we found lesser connectivity to a set of foil regions. The combined results of this study validate the use of DSI-based fiber tractography to address hypotheses relating to human brain connectivity. However, widespread noise in tractography results highlights the need for conservative approaches to fiber tracking research. We especially emphasize the benefits of collecting multiple data samples per participant and of addressing targeted hypotheses.
展开▼
