Robustness of brain structural network analysis with respect to diffusion fiber tractography parameter settings in patients with dementia
Summary
Whole-brain tractography has shown promising results distinguishing patients with Alzheimer’s disease from individuals without dementia. To assess whether the analytical results differ based on the parameter settings, the difference between healthy controls and dementia patients will be analysed when changing the different parameters and measurements of the connectomic analysis.
The data employed during this research was obtained from the study performed by Reijmer et al., 2013, and consisted of a 3-T MRI scan and cognitive assessment of 59 dementia patients and 47 control subjects. DTI and CSD tractographies were obtained from every patient while changing the FA or FOD threshold, respectively, and the angle deviation. This data was then employed to perform two different network analysis approaches: based on the registration to the atlas or based on individual parcellation. Finally, the analysis were quantified by measuring the clustering coefficient, global and local efficiency, and distance from the binary connectivity matrices, and connectivity matrices weighted by the FA values and percentage of tracts.
The students t-tests (alpha 0.05) performed and the Cohen’s D values calculated showed that the measurement decreased while the FA threshold was increased, been anatomically implausible a threshold higher than 0.3. With respect to the angle deviation, the measures were increased together with the threshold, and the analysis showed that, for angles higher than 15º, a change of the angle deviation does not have a significant change in the results. Moreover, DTI tractography, network analysis based on individual parcellations, FA connectivity matrix and global efficiency have shown the higher discrimination performance when differentiating between healthy controls and dementia patients.