Schedule

This session is a lecture that advocates for using automated dMRI preprocessing tools in modern research. It highlights the benefits of pipelines, such as reproducibility, time-saving, and integrated Quality Assurance (QA), arguing against manual, error-prone, step-by-step processing. The session provides an overview of leading pipelines, categorizing them by scope: for example, dwifslpreproc for robust distortion correction using FSL's tools, and comprehensive, end-to-end solutions like Tractoflow and QSI-Prep that handle preprocessing all the way through to advanced outputs like connectomics. Crucially, the lecture introduces the Brain Imaging Data Structure (BIDS) standard, explaining its importance for organizing neuroimaging data to ensure maximum portability and research reproducibility.

This session provides a detailed theoretical understanding of modern tractography, moving beyond the simple tensor model. The first concept covered is Constrained Spherical Deconvolution (CSD) as an input for tractography (SSST or MSMT). Then it covers an explanation of deterministic (DET) versus probabilistic (PROB) tractography and addresses limitations, biases, and pitfalls such as density bias, gyral bias, length bias, coverage, and fanning. Furthermore, it introduces more complex variants, like anatomical constraints, parallel transport tractography, particle filtering tractography, surface-enhanced tractography, and bundle-specific tractography, focusing on the problems these methods aim to solve and their potential usefulness. The curriculum also includes an anatomy of a command line and explaining common parameters like step size, choice of seeding and tracking mask, number of seeds (per voxel, total), curvature and step-size or angle, maximum/minimum length, sf_threshold, and Runge-Kutta integration.

This session is a practical lab that takes the provided dataset through a complete analysis pipeline and visualizing the resulting fiber Orientation Distribution Functions (fODFs), allowing for a direct comparison with the simpler DTI tensors from Day 2. Finally, participants will execute probabilistic tractography using the CSD results and the same Corpus Callosum seed from the previous day. The session concludes by comparing the resulting CSD-based probabilistic tractogram with the DTI-based deterministic result, highlighting the advantages of CSD in resolving complex fiber architecture.

Lunch time at the university canteen.

This session is a practical lab dedicated to applying the advanced modeling and tractography pipeline to each participant's personal diffusion MRI data. Building directly on the previous session, the primary goal is to generate a high-quality, whole-brain probabilistic tractogram suitable for quantitative analysis. Participants will execute the same key steps—estimating the Fiber Response Function (FRF), fitting the CSD model, and generating both deterministic and probabilistic tractograms—but will be challenged to adapt the workflow to account for their data's specific characteristics (e.g., single-shell vs. multi-shell). The session will also demonstrate the use of different software packages to perform the identical task, showcasing the common underlying principles of the methodology. Instructors will provide personalized guidance to help troubleshoot data-specific issues, ensuring participants leave with a high-quality, anatomically constrained whole-brain tractogram (wb_100k.tck) that is ready for subsequent analysis.

This session is a theoretical lecture dedicated to the crucial process of converting a whole-brain tractogram (the "spaghetti" of streamlines) into meaningful, anatomically plausible fiber bundles. The session covers different philosophies for this dissection, starting with the manual ROI-based "virtual dissection," which involves drawing inclusion and exclusion masks on anatomical images to select streamlines based on specific criteria. The lecture then introduces automated methods that increase efficiency and reproducibility. These include atlas-based methods, which register a pre-defined bundle atlas (like in pyAFQ) to the individual's space for streamline assignment, and clustering-based methods (like Recobundles or BundleSeg), which are data-driven approaches that group streamlines based on their geometric similarity. The goal is to provide a conceptual understanding of the different approaches that can be taken to reliably extract and analyze specific white matter bundles.

This session is a practical lab focused on transforming a whole-brain tractogram into meaningful, specific fiber bundles using both manual and automated techniques. Participants will start with basic filtering (e.g., removing anatomically implausible streamlines by length). The session then guides them through the ROI-based segmentation logic, where they will use an anatomical parcellation (like from mri_synthseg) to define and extract a simple bundle using inclusion/exclusion masks to mimic manual dissection. Finally, participants will be introduced to automated segmentation by running a tool like BundleSeg (part of Scilpy) to segment multiple bundles from their data efficiently. Successful completion of this lab, including the generation of essential bundle files, provides the inputs required for all of Day 4's quantitative analyses.

This session is a practical lab where participants apply the manual (ROI-based) and automated bundle segmentation techniques to their own whole-brain tractograms. Participants will first replicate the steps for both atlas-based and shape-based segmentation, adapting the process to their specific data. The session emphasizes troubleshooting and provides personalized instructor guidance on determining the best segmentation approach for their specific bundles of interest, preparing their data for subsequent quantitative analysis.

Via Pallone, 10A, 37121 Verona VR