Schedule

The Preprocessing Overview session is a theoretical lecture focused on the rationale and impact of essential dMRI data cleaning steps. The goals are to understand why each preprocessing step is necessary, learn to identify common data problems (e.g., incorrect b-values/vectors, signal dropouts), and appreciate how these choices impact downstream tractography. The lecture covers the what, why, and how of critical artifacts and corrections, including: data inspection for consistency (using tools like mrinfo), denoising (like N4 bias field correction and Gibbs ringing removal), motion correction, and mitigation of susceptibility- and eddy currents-induced distortions. A recommended order of operations will be presented, emphasizing how processing decisions heavily influence the final quantitative results.

This session focuses on the crucial step of fitting models to the preprocessed DWI signal to estimate white matter fiber orientation. The lecture distinguishes between model-based (which assumes a specific tissue structure) and model-free (which makes fewer or no assumptions on the tissue structure) approaches. The core principles of the Diffusion Tensor Model (DTI) are covered, including how its parameters are fit and how its eigenvalues are used to calculate primary scalar maps like Fractional Anisotropy (FA) and Mean Diffusivity (MD). The session highlights the limitation of DTI in resolving crossing fibers, which motivates the introduction of the Orientation Distribution Function (ODF) and Fiber ODF (fODF). The fODF, which represents the probability of fiber orientations on a sphere, is presented as the primary model used for accurately resolving complex fiber configurations in the remainder of the workshop.

Lunch time at the university canteen

This session is a practical lab that guides participants through a minimal DTI processing pipeline using a clean, controlled dataset. The primary goals are to apply basic processing commands (using tools like FSL and MRtrix) and to practice visualizing and quality-checking the key DTI outputs. Participants will learn commands for brain extraction (skull-stripping) of the B0 image and applying it to the DWI, and performing the DTI fit. A key focus of the session is quality assurance: participants will practice header and data inspection, learn how to detect a good versus bad BET (Brain Extraction Tool) result or DTI fit, and understand the necessary steps to resolve common issues (e.g., flipping b-vectors, adjusting BET parameters). The session emphasizes how to visually inspect the resulting FA and RGB maps and distinguish between processing steps done per subject versus those needed for the entire cohort.

This session is a practical lab where participants apply the fundamental DTI processing workflow from the previous session to their own personal diffusion MRI datasets. The primary goal is to diagnose potential data-specific issues arising from heterogeneous data structures, such as different b-value shells or varying image dimensions and strides. Participants will execute the DTI fit and generate FA and RGB maps, with instructors circulating to help troubleshoot common problems. This includes identifying issues like poor brain extraction requiring adjusted parameters, incorrect RGB coloring suggesting b-vector orientation errors, or unexpected FA values. While this session focuses only on the DTI fit and does not run full preprocessing, instructors will help identify any visible raw data artifacts and provide expert recommendations on the required preprocessing steps and parameters to look for in the subsequent pipeline overview session.

This session is a lecture that introduces the algorithms used to reconstruct macroscopic white matter pathways. The goal of tractography is to generate streamlines (3D polylines) that approximate fiber bundle trajectories by integrating the voxel-wise orientation information. The session covers the fundamental concepts and dichotomies of tractography algorithms, including the differences between local versus global and deterministic versus probabilistic. It also details key practical considerations: seeding strategies (e.g., whole-brain or ROI-based tracking) and termination criteria (e.g., exiting a brain mask or exceeding a maximum curvature angle). Crucially, the lecture highlights the biases and limitations of tractography, emphasizing that it is an estimate and does not represent true axonal density, and that different parameters choices can results in drastically different tractography results.

This session is a practical lab focused on generating and validating a first deterministic tractogram. Building on prior concepts, participants will learn how to generate an anatomical seed region using automatic segmentation (via registration) and isolate a specific structure, such as the Corpus Callosum (CC). Using the DTI eigenvector map, participants will then execute the deterministic tractography algorithm (tckgen) seeding from the CC mask and constraining the tracking to the brain mask. The final, crucial step involves quality control and visualization of the resulting tractogram. Participants will use tools like mrview to overlay the streamlines onto the FA map and visually validate that the reconstructed streamlines accurately follow the known anatomical shape of the target bundle, demonstrating the critical link between image processing and neuroanatomy.

Via S. Giovanni Lupatoto, 24, 37134 Verona VR