Schedule

This session is a theoretical lecture dedicated to the concept of tractometry, which involves profiling microstructural metrics along a fiber bundle's trajectory for quantitative analysis. The session distinguishes between calculating a single mean value for a bundle and generating a full tract profile, which investigates metric variations along the tract's length for greater specificity in studying disease or functional gradients. The lecture breaks down the comprehensive Tractometry workflow: isolating the tract, defining a single "centerline," subdividing it into fixed nodes, and then sampling scalar map values (such as FA, MD, or AFD) from the underlying streamlines at each node. Finally, the session highlights crucial challenges and considerations, including the need to ensure accurate point-to-point correspondence across subjects for group analysis, the use of more advanced microstructural metrics (like NODDI or DKI), and the importance of appropriate statistical handling to account for multiple comparisons.

This session introduces the network neuroscience paradigm, which models the brain as a graph consisting of nodes (gray matter regions defined by a parcellation) and edges (white matter bundles). The main focus is on the process of constructing a structural connectivity matrix. This involves first choosing an appropriate anatomical atlas to define the network's nodes, and then populating the matrix based on the streamlines connecting every pair of nodes. The session explains the different ways to weight the connections (edges), contrasting simple unweighted (binary) matrices with weighted connectomes, where edges can be defined by the raw streamline count or by a microstructural metric (e.g., mean FA or AFD) along the connecting streamlines. Crucially, the lecture addresses a major challenge: that raw streamline counts are biased and non-quantitative. It introduces modern streamline filtering methods such as SIFT and SIFT2 as essential techniques for generating biologically meaningful and quantitative connectomes that are suitable for robust statistical analysis.

This session synthesizes the entire workshop content into a practical framework for planning and executing a dMRI study. This guided discussion emphasizes that the research question dictates the choice of methods, forcing crucial decisions regarding the analysis pipeline: whether DTI is sufficient or if CSD is necessary for resolving complex fibers, and whether to use tractometry (profiling specific pathways) or connectomics (mapping the whole network). The session stresses the need to address variability and confounds, covering issues like scan-rescan variability in tractography, the challenges of data harmonization across multiple sites, and the importance of adapting default processing parameters when working with special populations (e.g., infants, non-human primates) or pathologies that can disrupt standard processing. Finally, it guides participants on the responsible interpretation of results, reinforcing that dMRI metrics are indirect markers of microstructure and that structural connectivity does not imply direct synaptic function.

Lunch time at the university canteen.

This session is a practical lab focused on generating the final quantitative outputs from the processed dMRI data: a structural connectome matrix and tract profiles. This session demonstrates two key analysis branches. In the Connectomics Branch, participants will use a tool like MRtrix's tck2connectome to map their whole-brain tractogram onto an anatomical parcellation (e.g., FreeSurfer's aparc+aseg.nii.gz), generating a connectivity matrix (e.g., connectome.csv) weighted by streamline count and other metrics. In the Tractometry Branch, participants will be shown how to use an automated tool like pyAFQ on their BIDS-formatted data. A single command can leverage their segmented bundles to generate detailed tract profile CSV files (tract_profiles.csv) for multiple bundles and microstructural metrics. The goal is for participants to leave the workshop with the essential final quantitative files ready for statistical analysis.